This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
yielding further technological advances and further 86. It 'felt' harder because the exhausted soil was yielding less, so there was less plant cover for 87. e. Feathers are heavier and cheaper. Foam is less yielding, more springy, doesn't last as long as do
SENSE AND STRUCTURE IN LEXIS 88. inevitable. In the meantime, dolphin research is yielding new data on the ways in which man's senso 89. t rewriting of his tax program, Mr. Reagan is not yielding on his demands for 25 percent cut in pers 90. y is to drive through a car cleaner. The bliss of yielding self and car to be soaped, washed and bru 91. 3,200 kilos of wheat per hectare. By 1966, highyielding strains of wheat had been developed that 92. old breeze across the oceanic warmths of the ever yielding sub-continent. Even his shabby clothes 93. ishments in St. James's began the slow process of yielding to clubs and commerce. I have chosen to 94. qual fervour, Premier Sagasta was progressively yielding to American demands and the diplomacy of 95. should at least in part be conceded, or else of yielding to extremism what earlier was refused to 96. of this ditch. But all the while, though he kept yielding to these invasions of sleep, he could not yields 97. of new land into cultivation, and an increase in yields. But there has been practically no expans 98. an perhaps make this protein in relatively large yields. For example, at the moment there are some 99. al more fertilizer than Europe to achieve similar yields. In 1975, developed countries used an aver 100. oot zone where they stunt plant growth and reduce yields. In the Punjab seepage has raised the wate 101. en, economics supplies only one— whether a thing yields a money profit to those who undertake it 102. an activity carried on by a group within society yields a profit to society as a whole. Even natio 103. the rents set by a city control board. This board yields, about once a year, to the landlords' plea 104. lure to damp down inflationary expectations. Bond yields almost everywhere are higher than before 105. oured areas — which were already enjoying higher yields and greater wealth than the rest. So the 106. reducing only 80 per cent of the average regional yields, and these are low because landlords under 107. ere exported to Turkey, India and Pakistan. Wheat yields doubled in India between 1964 and 1972, and 108. ency to permanent cultivation. Where this occurs, yields drop to an abysmally low level. As popula109. ng to the same patch of ground too frequently, so yields fall and the soil is exhausted, sometimes
95
96
JOHN McH. SINCLAIR 110. is the spur: poverty may be shared, but when food yields fall too low, some of the people on the 111. paddy has gone on producing the same or increased yields for centuries, even mill 112. 87 in the cropped area over the last decade, and yields have risen by 5 per cent while populat 113. of the soil quickly falls off after clearance — yields in the third year may be down to only one 114. ine and more hospitable to cultivated plants. The yields in the first year of shifting cultivation 115. wall, a group of Barrier Canyon paintings, still yields intriguing detail. A visitor points to 116. of Brompton Road and Thurloe Place Knightsbridge yields insensibly to South Kensington. The pave 117. lation grows, or at best a permanent decline in yields, irrigated paddy has gone on producing 118. ed acute proportions. At Bangladesh's low average yields of half a ton of rice per acre, a man 119. adolescents. Free acting out and talking through yields satisfaction. At the same time it 120. tion of components and vastly increased explosive yields (such as from fuel air weapons — the 121. hey are allowed to charge. The rent control board yields them an extra seven and a half per cent 122. lable means to prevent it. Ovid recommends: 'Love yields to business, be employed, you're safe' 123. of nature at the first touch of spirit, before it yields to spirit's cosmic venture that we call 124. come self-sufficient by 1956. In 1970 her average yields were four times the 1940 levels, reaching 125. ins are late, the growing season is cut short and yields will be slashed. Famine and disease follow
REFERENCES Bolivar, A. 1986. Interaction through written text: a discourse analysis of newspaper editorials. Ph.D. thesis. University of Birmingham. Butler, C.S. 1985. Systemic linguistics. London: Batsford. Carter, R.A. 1987. Vocabulary: an applied linguistic guide. London: Allen and Unwin. Halliday, M.A.K. and R. Hasan. 1976. Cohesion in English. London: Longman.
SENSE AND STRUCTURE IN LEXIS
97
Hanks, P.,et al. 1979. Collins English dictionary. London and Glasgow: Collins. Johansson, S. 1985. A survey of computer-based English language research. IAME News 9 (May). McCarthy, M.J. (Forthcoming). Some vocabulary patterns in conversation. In R.A. Carter and M.J. McCarthy, eds., Vocabulary in language learn ing. Harlow: Longman. Moon, R.E. 1987. Monosemous words and the dictionary. In A.P. Cowie, ed., The dictionary and the language learner. (Lexicographica Series Maior). Tubingen: Max Niemeyer Verlag. Phillips, M. 1985. Aspects of text structure. Elseviers: North-Holland. ------. (Forthcoming). Lexical structure of text. (Discourse Analysis Mono graphs, 12: English Language Research). Birmingham: University of Birmingham. Renouf, A.J. 1984. Corpus development at Birmingham University. In J. Aarts and W. Meijs, eds., Corpus linguistics: recent developments in the use of computer corpora in English language research. Amsterdam: Rodopi. 3-39. Stock, P.F. 1984. Polysemy. In R.R.K. Hartmann, ed., Lexeter '83 pro ceedings. (Lexicographica Series Maior). Tubingen: Max Niemeyer Ver lag. 131-140. Yang, H-Z. 1986. A new technique for identifying scientific/technical terms and describing science texts. Literary and Linguistic Computing 1, No. 2:93-103.
TEXT, TERMS AND MEANINGS: SOME PRINCIPLES OF ANALYSIS
M.K. Phillips English Language Services Department The British Council In 1944 Jorge Luis Borges published a short story called 'Pierre Menard, Autor del Quijote'. In it Borges purports to describe a literary pro ject devised by a modern French symbolist poet. The poet in question, Pierre Menard, had set himself the task of writing Don Quixote. That is to say, his intention was not, trivially, to copy out the original but to under take an entirely new task of composition the outcome of which would be identical, word for word, with Don Quixote. Borges first surveys some of the implications of this project. He discus ses what is involved in attempting, as a twentieth-century poet of French mother-tongue and nationality, to create a seventeenth century Castilian masterpiece. He then proceeds to develop a critique in which he evaluates an extract from the two versions of the novel, that of Cervantes and that of Menard. I offer here my translation of the crucial passage: Cervantes' text and that of Menard are verbally identical but the latter is almost infinitely richer. (Its detractors would say more ambiguous; but ambiguity is a kind of riches). It is a revelation to compare Menard's Don Quixote with that of Cer vantes. The latter, for example, wrote (Don Quixote, first part, chapter nine): . . . truth, whose mother is history, the rival of time, storehouse of actions, witness to the past, example and admonition for the present, warning for the future. Written in the seventeenth century, written by the ingenio lego Cer vantes, this list is nothing more than a rhetorical eulogy of history. Menard, on the other hand, writes: . . . truth, whose mother is history, the rival of time, storehouse of actions, witness to the past, example and admonition for the present,
100
M.K. PHILLIPS warning for the future. History, mother of truth; the idea is startling. Menard, a contempor ary of William James, does not define history as an investigation of reality but as its origin. Historical truth, for him, is not what happened; it is what we judge to have happened. The concluding phrases — 'example and admonition for the present, warning for the future'—are unashamedly pragmatic. The contrast between the two styles is also striking. Menard's style, which is deliberately archaic and essentially foreign, smacks of a certain affectation. It is otherwise with his precursor, who handles with ease the normal Spanish of his day (Borges 1980: 431-432).
Borges' conceit, whilst fantastic, is not fanciful. I have quoted it at some length because it goes, I believe, to the very heart of the problem of how texts mean. It is the purpose of this paper to explore some of the lin guistic foundations of Borges' ingenious exercise in comparative literary criticism. There is one ultimate fact about text. This is that it consists of elements of linguistic substance juxtaposed in linear sequence. In the case of written text the skilled reader somehow internalises from the encounter with graphic substance a model of some aspect of 'reality'. This is what consti tutes the meaning of the text. Unless it stands in some describable relation to the world of phenomena, a text has no meaning for it cannot be related to experience. The fundamental problem posed by text, then, is to eluci date the nature of the relationship between text and reality which allows meanings to be created in this way. How does it come about that complex non-linear conceptual structures are realised through the ultimately linear organisation of language substance? That such non-linear conceptual structures are elaborated is widely accepted. Text linguists argue that this process underlies the reader's ability to summarise, paraphrase and generally to state what a text is about. This ability raises some interesting problems. Van Dijk has pointed out that to be able to state what a book is about depends on the processing of thousands of sentences which cannot normally be memorised individually by the reader (Van Dijk 1977a). In general, it is the 'gist' of a discourse which is recalled rather than its wording. Studies such as those of Sachs or Clark and Clark (Sachs 1967; Clark and Clark 1977) attest the relative sub sistence of semantic memory as opposed to the uncertainty of recall of syn tactic structure. The appreciation of textual meaning is thus a large scale phenomenon which does not depend directly on particularities of linguistic
TEXT, TERMS AND MEANINGS
101
form. It has been suggested that meaning becomes available to the reader through the activation by the text of conceptual 'schemata'. This is the notion which underlies the work on the cognitive representation of narra tives by such investigators as Rumelhart, Thorndyke, Van Dijk and Kintsch (Rumelhart 1975, 1977; Thorndyke 1977; Van Dijk 1977a, b; Kintsch 1977). The difficulty with this approach is that 'schemata' are postulated to fit the text to be analysed which are then used to explain the text. There is a fundamental petitio principii. It will become clearer later that the basic insight is in essence correct, but it is clearly pointless to look to this work for sound methodological guidance. Indications as to the nature of the problem and how to tackle it must be sought elsewhere. Saussure furnishes two important clues to exploring these issues. He was one of the first to see clearly that if reality is to be represented in lan guage, then there can be no necessary relationship between the 'structure' of reality and the categories of language. This is the point of his first princi ple of the arbitrariness of the linguistic sign and this provides the first clue to the problem of textual meaning. He postulated that the linguistic sign consists of an arbitrary relationship between a 'signifié' and a 'signifiant' obtaining within language. He stated his first principle in the following terms: The bond between the signifier and the signified is arbitrary. Since I mean by sign the whole that results from the associating of the signifier with the signified, I can simply say: the linguistic sign is 'arbitrary' (Saussure 1974: 67).
It is usually assumed that the point of this principle was to assert that there is no systematic correspondence between form and meaning. But this was well known long before Saussure's time. In the seventeenth century John Wilkins (1688) addressed himself precisely to the problem of designing a linguistic system in which the relationship would not be arbitrary. There is a more important aspect to Saussure's postulate. This is that both 'signifiant' and 'signifié' are categories of language. I take the follow ing: 'There are no pre-existing ideas, and nothing is distinct before the appearance of language' (1974: 112) as a disavowal of the notion that what is signified can exist apart from the signifier. This position has one crucial consequence. The categories of reality are the categories of meaning fur nished by language. In this theoretical sense, reality is articulated by lan guage. It is thus possible for text to project 'reality' and for different texts
102
M.K. PHILLIPS
to project different 'realities'. How can language be constructive of reality? The notion seems to be counter-intuitive suggesting, for example, that a chair is a fundamentally linguistic rather than a phenomenological object. But so it is; certainly some chairs have physical existence (in some philosophically acceptable meaning of these terms!) but what counts as a chair is essentially a linguistic question. The pages of Scientific American abound in pertinent examples, such as the following: For the atom, the nucleus and the proton, then, the mass of the system is at least as large as the kinetic energy of the constituents and in some cases is much larger. If quarks and leptons are composite, however, the relation of energy to mass must be quite different. Since the prequarks have ener gies well above 100 GeV, one would guess that they would form compo sites with masses of hundreds of GeV or more. Actually the known quarks and leptons have masses that are much smaller; in the case of the lepton and the neutrinos the mass is smaller by at least six orders of magnitude. The whole is much less than the sum of its parts (Harari 1983: 58).
Here a model for phenomenological reality at its most fundamental level is described. It should be noted, however, that there is no direct linguistic evi dence in this extract that this is simply postulated. Many of the sentences have the linguistic form of assertions of fact and the use of 'known' to mod ify 'quarks and leptons' is interesting. It is either elsewhere in the discourse or in the context of shared knowledge that an indication of the hypotheticality of the content must be sought. In other words, quarks are created by talking about them. Perhaps this accounts for the proliferation of funda mental particles. This view of language finds support in contemporary sociology. Berger and Luckman (1967: 30) enquire 'How is it possible that subjective mean ings become objective facticities?' and argue: We have seen how language objectifies the world, transforming the panta rhei of experience into a cohesive order. In the establishment of this order language realises a world, in the double sense of apprehending it and pro ducing it (1967: 173).
Thus the implications of Saussure's first principle provide a clue to the nature of the relationship between text and reality. The perspective may appear unusual. After all, vocabulary is commonly considered to refer to real world entities whilst the view described above suggests that the notion of reference is an illusion. Yet Saussure (1974: 68) was clear about the sig nificance of this principle:
TEXT, TERMS AND MEANINGS
103
No one disputes the principle of the arbitrary nature of the sign, but it is often easier to discover a truth than to assign to it its proper place. Princi ple I dominates all the linguistics of language; its consequences are num berless. It is true that not all of them are equally obvious at first glance; only after many detours does one discover them, and with them the primordial importance of the principle.
The apparent unconventionality of the position thus reached is so only because of a failure within the Anglo-American linguistic tradition to assign to Saussure's principle 'its proper place'. The issues involved would seem less unusual within the French 'post-structuralist' perspective. The use, for example, of the term 'discours' by Foucault (1970: 311) reveals that he con ceives of a discoursal function of language which imposes structure on real ity: In other words, language in so far as it represents—language that names, patterns, combines, and connects and disconnects things as it makes them visible in the transparency of words. In this role, language transforms the sequence of perceptions with a table and cuts up the continuum of beings into a pattern of characters.
It is perhaps not surprising that such a conception of language should stem from within a philosophical tradition rooted more in rationalism than empiricism. The second clue that Saussure provides arises as an immediate conse quence of the principle of the arbitrariness of the linguistic sign. That notion accepted, the rest of linguistics can be seen as a specification of the limits to arbitrariness. Arbitrariness does not and cannot imply mutual substitutability of signs: Since one vocal image is no better suited than the next for what it is com missioned to express, it is evident, even a priori, that a segment of lan guage can never in the final analysis be based on anything except its noncoincidence with the rest. Arbitrary and differential are two correlative qualities (Saussure 1974: 118).
Linguistics then becomes a specification of the limits to arbitrariness. The distinction among signs is thus crucial and language can be seen as a system in which the value of an item is determined by its position in the system, This is what Saussure meant by the term 'valeur'. Again, these notions are usually given a particular interpretation. 'Valeur' is held to arise from the systematicity of 'langue'. Hence it is possi ble to claim, for example, that the 'meaning' of nominative in a five-case system of declension is different from its 'meaning' in, say, a six-case sys-
104
M.K. PHILLIPS
tem. But it is also arguable that the notion can profitably be applied to the analysis of 'parole'. The most fundamental characteristic of 'parole' is, as Saussure pointed out, its linearity. As Sinclair (1980a: 111) puts it: 'succession is the only ulti mate relationship of elements'. This means that 'valeur' can also be viewed as a property of the syntagm arising from the distribution throughout the linear stream of language substance of distinctive segments. Even syntax can be seen as a system of inferences from patterning in linear succession, although as Sinclair (1980a: 112) again points out: 'Syntax is rarely pre sented as a set of limitations upon the free combination of those elements that require to be distinguished'. This suggests that an investigation along these lines might provide information about the way in which conceptual structures are elaborated on the basis of critical juxtaposition of elements. In other words, how does textual sequence lead to cognitive order? In such an investigation it would be inappropriate to invoke semantic criteria, since it is the semantics of text which it would be the purpose of the study to reveal. The second clue provided by Saussure is thus to the methodology for exploring the relationship between text and reality. It conveys an important message. It suggests strongly that words can be viewed as units of textual organisation and that a knowledge-free distributional analysis of the terms used in a text might furnish insights into areas of vocabulary which fix the relationship between form and meaning. Again, however, this approach to text through the analysis of the dis tributional properties of its linguistic substance may seem a perversely unconventional way of proceeding. It implies a deliberate refusal to be seduced by the attractions of discourse analysis, text grammar or syntactic structure. Indeed, it could be considered an outmoded technique. Harris' (1952) view of discourse analysis was essentially distributional but his methods no longer find favour. There is, then, a requirement to justify the approach to text which reflection upon Saussure's principles has suggested. Here a number of per tinent considerations suggest themselves. Linguistic substance itself has its own meaning-bearing properties which can be exploited to enhance the conceptual meaning. In the spoken language use can be made of 'special effects' such as comical accents or unusual pitch. Manipulation of graphic layout and type faces contributes importantly to the creation of meaning in the written language. In languages such as Arabic where cultural develop-
TEXT, TERMS AND MEANINGS
105
ment has engendered a strong calligraphic tradition, the linguistic sign often acquires a multiple symbolism by virtue of the interplay of sense and form. This aspect of language has been elegantly exploited in English by Scott Kim (Kim 1981). Many types of word play depend upon the possibility of exploiting the medium to add to the message. Crosswords, palindromes, word squares and acrostics are all manifestations of the use of language relying heavily on language substance. Another familiar example is onomatopaeia. Poetry in general places great value on the properties of substance and employs often to good effect the meaning possibilities inherent in rhyme, assonance, metre and alliteration. The bridle of Chaucer's monk once heard 'gynglen in a whistlynge wynd' can still be heard today. A rose by any other name could not rhyme with 'those' or 'enclose', but Will Shakespeare knows that as things are it always will, just as he knows the power of a pun. On the other hand, the tension between meaning and substance accounts for the kind of paradox reported in Hofstader (1981) such as: 'This sentence no verb'. Such phenomena should not be considered as mere eccentricities of the language system operating on the periphery of meaningfulness. They are, on the contrary, manifestations of a potential central to the nature of lan guage. This is the possibility of exploiting the medium in which language must necessarily be realised to meaningful effect. It is this property which renders much literary creation possible. Thus an advantage of adopting an approach to text which focuses on the substance of language is that litera ture can be viewed, not as a sophisticated excrescence on an assumed more fundamental communicative function, but as rooted in the very core of lan guage and basic to its use. Sinclair (1981) has pointed out that to be able to accommodate literature naturally within its description is an important goal for any linguistic theory. The realisation of this potential is not restricted only to the more obvi ous manifestations of artistic activity. In expository prose the substantial nature of the linguistic sign may be somewhat suppressed, but its influence can still be felt. Thus it is possible to speak of prose having a rhythm or cadence and of aesthetic effect. Less august examples of language use also point in the same direction. The phenomenon of phatic communion illus trates that meaning can sometimes reside less in the semantics of the dis course as in its substance, the simple fact that something rather than nothing is said. The converse is also true and the absence of substance, that
106
M.K. PHILLIPS
is, silence, also has a role to play as Basso (1972) has pointed out. Likewise, with taboo language expletives, for example, the significance is in the act of utterance, not in the denotation, which, indeed, may command a physical impossibility. To summarise the argument so far; first, it seems likely that the prob lem of textual meaning can fruitfully be thought of in terms of the projec tion by the text of a reality. Secondly, this projection is brought about by a mechanism which must be discernable as a patterning of textual substance. In other words, a study which seeks evidence of such patterning might be expected to throw light on how texts relate to reality and hence on how they mean. Significant methodological consequences stem from such a position. The study would be addressing itself to a large-scale phenomenon in two senses. It has already been observed that perception of textual meaning is a high order process not directly dependent on the organisation of text at the local level. The implication of this is that information relating to such local structuring is irrelevant to the task of discerning latent patterning in textual substance. This means that information of a morphological and syntactic nature can largely be ignored. Secondly, the distribution of distinctive elements throughout the linearity of text may be expected to lead to patterns which are only detecta ble over considerable extents of language. Most lexical items occur infrequently. Even in a highly specialised expository text of some 50,000 running words where particular technical terms may be expected to achieve unusually high frequencies, it is exceptional that the occurrence of a single term exceeds 1% of the total. The repetition of most lexical items upon which any kind of patterning must be based is an extremely rare event. Consequently, as Sinclair (1980b: 15) points out: Even a glance at the statistics of word occurrence suggests that to gain access to the characteristics of language one requires texts of length that puts them well out of scale of direct human observation.
Thus observation of such patterning is difficult and requires the assistance of computers. Techniques for studying texts in this way are in their infancy. It was also noted earlier that if the object of such a study is to elucidate the semantics of text, then semantic information cannot be assumed. In other words, the analysis must be knowledge-free. Thus to the rejection of morphological and syntactic meaning can be added the exclusion of infor mation relating both to discourse organisation and lexical homonymy. If
TEXT, TERMS AND MEANINGS
107
large-scale patterning of the type that theory suggests in fact exists in text, then it should be recognisable formally from the distributional elements in text. Evidence is now available from an initial study of a restricted number of texts that such an approach to textual analysis can indeed throw light on the issues raised here. Five undergraduate science textbooks, two novels and a work of popular science were analysed for evidence of lexical pattern ing (Phillips 1985)1. From the findings of this study it is possible to draw a number of conclusions about the way in which different classes of text create meanings. At an initial level of analysis it appears that the processes at work are somewhat different in the case of the five science textbooks from those operating in the three non-science texts. As can be seen from the table in the notes, the type-token ratios are considerably higher for the science texts. This is because the number of distinct lexical types in these texts is smaller than in the other texts. Moreover, since the type-token ratio is an increasing function of the number of types, a text such as the chemical engineering sample, which is shorter than the others, has a type-token ratio which, although already quite high, would be even higher if the sample were as long as the rest. In other words, the vocabulary of the non-science texts is more diverse, that of the science texts more restricted and hence the average frequency of occurrence of any lexical item tends to be greater in the science texts. The higher density of a relatively small number of lexical items in sci ence text means that they tend to cooccur in recognisable patterns. By exa mining the frequency of such cooccurrences it is possible to distinguish net works of lexical items which regularly articulate critical areas in the cogni tive content of the texts. A network is distinguished as a distinct pattern of organisation by virtue of the relative strength of association obtaining among its member lexical items. Obviously the text as a whole constitutes a vast network, since every word in it over the span of the whole text cooccurs with every other. By focusing on significantly high frequencies of cooc currence, it is possible to isolate individual networks. The sense of 'signific antly' in this context is fully discussed elsewhere (Phillips 1985). In brief, it refers to a threshold level beyond which the cluster analysis procedure used to identify networks fails to distinguish distinct groupings of lexical items. These networks can conveniently be represented in graphical form where the nodes stand for lexical items and the edges represent the exis-
M.K. PHILLIPS
108
tence of significantly frequent cooccurrence. Thus in one of the science texts the following pattern of association occurs:
Figure 1 This indicates the importance of these notions in the text in question, which deals with chemical reaction engineering. The phenomena designated 'temperature', 'composition' and 'pressure' are brought into consistent rela tion and thus an aspect of reality is consciously structured by the text. The text is organised through the association of certain lexical items and thereby a particular conceptualisation of real-world phenomena is presented to the reader. These lexical items constitute what may be called the 'terminology' of the text and may themselves be referred to as 'terms'. Support for this interpretation is provided by Benson and Greaves (1983). In a small-scale study of the descriptive and instructional literature accompanying a hi-fi system they demonstrated that lexical sets can be determined syntagmatically. This is to say, on the basis of the regular cooc currence of items separated from each other by relatively small numbers of intervening words, groupings of lexical items can be discerned which clearly articulate the principal cognitive content and functional purposes of their texts. The existence of such networks provides, then, a major clue to the way semantic structures are derivable from the linear sequence of text. In other parts of the chemical reaction engineering text referred to ear lier, the following networks are observed: compositiondependency temperature given
pressure Figure 2
In this network it can be seen that the node corresponding to the term 'temperature' tends to occupy a central position, where 'central' is defined
TEXT, TERMS AND MEANINGS
109
as attracting significant coocurrence with the largest number of the terms forming the network. The network exists largely because of the critical occurrence of this particular term. In this sense, the term 'temperature' can be said to 'organise' the network. The concept for which it is the label is crucial to the structuring of our understanding of this aspect of reality. A finding of the study referred to earlier is that there is a relatively restricted set of these 'central terms' (or 'nuclear nodes' as they are called in the study) in science texts which have a particularly important organising func tion to perform in projecting the writer's view of his subject matter. Empir ical investigation suggests that they rarely account for more than 20% of the total lexical vocabulary of a text. The organising function of such terms appears to extend beyond their immediate environment and to contribute to the overall structure of the text. This gives rise to a perception of what may be called the 'macrostruc ture' of the text. It can be argued that the following network is, in a crucial way, 'similar' to the network illustrated in Figure 2, where similarity is defined as having the same central term and at least one other in common. variation temperature pressure drop Figure 3 Figures 2 and 3 represent networks found in chapters two and five respec tively of the same text. The recognition that similar networks thus exist in widely separated locations in the text indicates that semantic relationships extend over long stretches of text and create large-scale patterning. These patternings have been determined for the texts in the study and it has proved possible to provide an interpretation for them consistent with the organisation of the subject matter in the text. In other words, it is argued that large-scale patterns of lexical organisation are responsible for the struc ture of the subject matter as projected by the text. At this level of analysis, the five science texts appear to be quite shar ply distinguished from the non-science texts. There is very little evidence for similar patterns of association in the non-science texts. In the two novels the networks that can be observed tend either to reflect common col locations such as 'sort' and 'thing' or 'telephone' and 'box' or represent the
110
.K. PHILLIPS
juxtaposition of the names of the principal characters. It can be surmised that the latter associations arise in part through the structuring of dialogue in the novels. Even in the popular science text, which speculates on the future impact of computers on society, there are relatively few networks which clearly indicate the kind of subject matter structuring which was observed to be the case with the chemical reaction engineering text. A further point of difference between the two classes of text is that in the non-science texts the proportion of central terms is much higher, reach ing betv/een 50% and 70% of the total. In other words, there is almost no evidence for a restricted set of organising terms responsible for the structur ing of the cognitive content of these texts. It is thus not surprising that the evidence for lexical macrostructure is correspondingly weak. Thus there is little indication of regularity of lexical patterning in the non-science texts either at the local level or on the macro-scale. It is perhaps to be expected in the non-science texts, and particularly the novels, that what the texts are 'about' should not be straight-forwardly revealed by clues provided by patterns of lexical association. The truth of literature is literary rather than literal. And yet the inability to distinguish significant patterns of lexical association in these texts has certain crucial consequences. The first inference that can be drawn is that there is no reason to ascribe more significance to one sequence of signs rather than to another. None would appear to be more probable than any other and thus, it could be argued, a purely random sequence would carry as much meaning as the sequence chosen by the author. In non-science text, then, it is dif ficult to see that there is any area of vocabulary which helps to fix a definite correspondence between the text and reality. This conclusion appears so improbable, shocking even, that the suspi cion arises that the reasoning must be flawed. It can perhaps be countered by pointing out that it assumes that the totality of a word's meaning arises from its contextual use. It could be argued that if the notion of 'sense' is accepted, then it becomes clear that certain senses are more likely than others to be capable of juxtaposition with each other. The price to be paid for this argument is that it is necessary to invoke extratextual knowledge to explain the semantics of text and that meaning cannot be sought wholly on the basis of clues provided by the text. I shall return to this point below. In the meantime it is necessary to ask how far 'sense' is itself a fixed and immutable property of the word. For the problem with sense is that it can only be derived from our experience of language; in other words, it has to
TEXT, TERMS AND MEANINGS
111
be generated by some form of regularity in the use of words. In order to examine the question further it is appropriate to return to consideration of the science texts. For in these texts it was seen that certain regularities of patterning do exist. Perhaps, then, these texts can provide some clue as to the genesis of sense and thereby reinstate the text as the location of its own semantics. But in fact it seems that if this new level of analysis is adopted, the distinctions between the two classes of text begin to blur just where they appeared to be most dissimilar. I shall take as an example the history of a term which is very com mon in one of the science texts in the corpus. The text deals with classical mechanics and the term to be investigated is 'force'. As a noun 'force' has a number of potential meanings available to it. For example, the Oxford English Dictionary lists the following eleven primary meanings: I.
Strength, power 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.
Physical strength As an attribute of physical action or movement Power or might (of a ruler, realm or the like) A body of armed men, an army Physical strength or power exerted upon an object Mental or moral strength Of things (in non-material or moral relations): Power to influence, affect or control Of a law etc.: Binding power, validity The real impact or significance (of a document, statement or the like) (Without article prefixed) A large quantity or number, plenty Physics, etc. Used in various senses developed from the older popular uses
These eleven sub-headings themselves comprise thirty-two sub-categories of which four correspond to meaning eleven. In addition a further four derived meanings are listed as well as a number of phrasal uses. Although the reader is likely to be predisposed to take occurrences of the term 'force' in the mechanics texts in the sense of I.11 of the OED, it is clear that there is scope for interpretation even with such an apparently obvious technical term and that its 'scientific' use may be fairly elastic. In other words, a final decision as to the precise import of the term can only be made by taking account of the delimitation of its semantic scope provided
112
M.K. PHILLIPS
by the textual context. Thus the 'instantial' meaning of a term depends on its relationship with other terms in the text. It is perhaps with some relief, then, that the reader encounters as the first significant patterning of the term 'force' in the mechanics text the fol lowing network: force periodic
applied Figure 4
In a chapter entitled 'Linear Motion' and containing sections dealing with oscillators, it is comfortingly clear that the pertinent meaning is that of 'periodic force'. But this certainty is short-lived. In the next chapter the rel evant network is found to be: force central
conservative Figure 5
The 'meaning' of the term has undergone modification and is being rede fined by a new verbal context. The reader's faith in the supposed precision of scientific terminology may not yet be shaken, but it is clear that s/he has to keep an open mind and be prepared to modify the meanings s/he has assigned to terms as the text unfolds. In the present text, the new pattern just established is reinforced in the following chapter, but the next significant occurrence forces, if I may be permitted the term, a further reinterpretation. In chapter eight a new pat tern of association emerges: work
force
sum
subject external
moment Figure 6
113
TEXT, TERMS AND MEANINGS
The meaning of 'force' is now defined by association with such terms as 'external', 'moment' and 'work'. The view of the phenomenological world projected by this grouping is somewhat different from what has been set up previously. There is some uncertainty, however, as to whether this view will predominate in chapter ten of the text (Figure 7), but it appears to establish itself more firmly in chapter eleven (Figure 8), although the occurrence of 'conservative' in the relevant network recalls an earlier pattern.
acting
corresponding Figure 8
Figure 7
The final two chapters, however, abandon this newly established use in order to revive earlier meanings: subjected
problem
force
solve
central
force periodic Figure 9
Figure 10
It seems, then, that even in scientific text, far from a particular term unequivocally structuring a unique cognitive area as had earlier been suggested, it must rather be concluded that each word in each text is con tinually defined and redefined by its use throughout the text. Firth (1955: 46-47), who first systematically developed the contextual theory of mean ing, had already observed that such variability is characteristic of word meaning:
114
M.K. PHILLIPS There is need to recognise indeterminacy, not only in the restricted techni cal language of description but also in the language under description.
Thus if there appears to be an element of randomness in the non-science texts, it also appears that a degree of semantic indeterminacy must be expected for the terms which contribute to the structuring of science text. Meanings are not static and in consequence the reality projected by a text is constantly threatening by some Protean transformation to elude the reader's perception. It is now time to draw together the threads of the discussion and to see what conclusions are suggested by the findings of textual analyses based on the theoretical principles discussed earlier. The first conclusion that can be drawn is that it is not possible to view word meaning in isolation from the function of terms in a given text. Not only is the meaning of a term deter mined by the patterns of significant association with its fellows in text but, reciprocally, these patterns are essentially concerned with structuring the text. As Sinclair (1982:1) puts it: 'terms are by no means independent of text: in fact they help to create it'. It is the behaviour of terms in a given text which creates meanings. At the same time, however, terms cannot be considered fixed entities. This is the second main conclusion. Text can be seen as a dynamic network of shifting relations among terms which thereby undergo continual semantic modification. This is perhaps not a new insight. Hjemslev argued forty -one years ago that language has to be viewed as a network of relations and correctly perceived that this view stemmed from Saussure, whose ideas formed a point of departure in this paper and who, according to Hjemslev (1947: 69) was the first to call for a structural approach to language, i.e. a scientific description of language in terms of relations between units irrespective of any properties which may be displayed by these units but which are not rel evant to the relations or deducible from the relations.
What is new and provided by the present study is evidence for the validity of this view. The study thus furnishes evidence for the contextual view of term meaning proposed by Firth. Meaning arises as a function of the behaviour of terms in text, and each fresh occurrence of a term is located by reference to previous but provisional patterns of use. Thus the tangibility of sense as enshrined in the dictionary and perceived as a psychological reality is, as was discovered earlier with its companion concept reference, an illu sion. Sense is a derivative of textual relations which are constantly fluctuat-
TEXT, TERMS AND MEANINGS
115
ing and the reader applies a kind of semantic calculus to fix the meaning of a term at any one point in the text. This position has one very important consequence which is the third and final conclusion of this discussion. If the notion of fixed meaning, of sense, is simply a convenient fiction, then the model of reality projected by a text is only available to the reader if he reads into it his own perceptions of structure. In other words, the content of the text depends on what the reader brings to it. Thus it is that a text can mean different things to differ ent readers. In conclusion, I should like to return to the extract from Borges with which I began this discussion. It is now possible to see how Borges could develop two totally different views of what are at one level identical texts. It is the meanings available to a particular reader, which depend on the idiosyncracies of individual experience, which determine the meaning of a text. The reality projected by a text is not preexistent, as Saussure rightly observed, but is constructed through the interaction of readers and text. Thus, despite their literal identity, the Don Quixote of Cervantes and the Don Quixote of Pierre Menard are distinct literary creations. The reader's knowledge that one is a product of a seventeenth-century Spanish ex-soldier and the other a turn-of-the-century work by a French symbolist will result in the perception of quite different meanings, some of which Borges has, tongue in cheek perhaps but nonetheless validly, suggested. The particular meaning of each work cannot be attributed to the text, for at this level the two works are indistinguishable. As with the phenomenon of lexical structure in the corpus discussed here, it is not the terms used which give rise to meaning but the value given them by the reader. Thus, to adapt the letter of Borges' thesis but without distorting its spirit, we can say: we cannot define a text as an investigation of reality but as its origin. Text does not mean what it says; it means what we have judged it to have said.
NOTES 1.
The data on which the discussion in this paper is based were obtained in the course of my doctoral research at Birmingham University. This was a computer-assisted investigation of lexical patterning in a corpus of science texts. The corpus consisted of the following five first year undergraduate texts: Ahmed, H. & P.J. Spreadbury. 1973. Electronics for engineers. Cambridge: Cambridge University Press.
116
M.K. PHILLIPS
Kibble, T.W.B. 1973. Classical mechanics. 2nd ed. Maidenhead: McGraw-Hill. Levenspiel, O. 1972. Chemical reaction engineering. 2nd ed. New York: Wiley Morris, J. Gareth 1974. A biologist's physical chemistry. 2nd ed. London: Edward Arnold. Sonntag, R.E. & G.J. Van Wylen. 1971 Introduction to thermodynamics: classical and statistical. New York: Wiley. In addition, for purposes of comparison, analyses were undertaken of two novels, Mrs Dalloway by Virginia Woolf and The Human Factor by Graham Greene, as well as of one popular science text, The Mighty Micro by Christopher Evans, published by Gollancz. The analyses of these texts consisted in selecting a subset of the lexical items in each text according to a well defined procedure, each lexical item being lemmatized. Each sub set was then investigated for evidence of regularity of cooccurrence of its members and syntagmatic sets were formed using a cluster analysis technique. The resulting sets, which I have called 'networks', were represented graphically as illustrated in this paper. These networks formed the basis for a study of the contribution made by lexical organisation to text structure. 2.
Frequency distribution characteristics of the texts in the corpus:
Electronics Mechanics Chem. Eng. Physical Chem. Thermodynamics Human Factor Mrs Dalloway Mighty Micro
Numberof tokens (N)
Number of types (V)
Mean type frequency (x)
60363 63068 48124 62001 66644 62003 65730 60383
3252 3309 3235 4555 3321 6767 7834 7291
18.6 19.1 14.9 13.6 20.1 9.2 8.4 8.0
REFERENCES Basso, K.H. 1972. T o give up on words': silence in Western Apache cul ture, In P.P. Giglioli, Language and social context. Harmondsworth: Penguin. 67-86. Benson, J.D. and W.S. Greaves. 1983. Field of discourse: a theoretical vantage point enabling more effective use of computer assisted colloca tional analysis. Paper presented at Association for Literary and Linguis tic Computing (ALLC) Conference, San Francisco. Berger, P. and T. Luckman. 1967. The social construction of reality. Har mondsworth: Penguin. Borges, J.L. 1980. Pierre Menard, autor del Quijote (Ficciones (1944)). In Prosa Completa Vol. 1. Barcelona: Bruguera. 425-433.
TEXT, TERMS AND MEANINGS
117
Clark, H.H. and E.V. Clark. 1977. Psychology and language. New York: Harcourt Brace Jovanovich. Firth, J.R. 1955. Structural linguistics. In F.R. Palmer, ed., Selected papers of J.R. Firth 1952-1959. London and Harlow: Longman, 1968. 35-52. Foucault, M. 1970. The order of things. London: Tavistock. Harari, H. 1983. The structure of quarks and leptons. Scientific American 248,4:48-60. Harris, Z.H. 1952. Discourse analysis. Language 28:1-30. Hjemslev, L. 1947. Structural analysis of language. Studia Linguistica 1:6978. Hofstadter, D.R. 1981. Metamagical themas. Scientific American 244,1:3441. Kim, S. 1981. Inversions: a catalog of calligraphic cartwheels. Peter borough, N.H.: Byte Books. Kintsch, W. 1977. On comprehending stories. In M.A. Just and P.A. Car penter, eds., Cognitive processes in comprehension. Hillsdale, N.J.: Lawrence Erlbaum Associates. 33-62. Phillips, M.K. 1985. Aspects of text analysis. Amsterdam: North-Holland. Rumelhart, D.E. 1975. Notes on a schema for stories. In D.G. Bobrow and A. Collins, eds., Representation and understanding: studies in cognitive science. London: Academic Press. 211-236. . 1977. Understanding and summarizing brief stories. In D. Laberge and S. Jay Samuels, eds., Basic processes in reading: perception and comprehension. Hillsdale, N.J.: Lawrence Erlbaum Associates. 265303. Sachs, J.S. 1967. Recognition memory for syntactic and semantic aspects of connected discourse. Perception and Psychophysics 2,9:437-442. Saussure, F. 1974. Course in general linguistics, trans. by W. Baskin. Lon don: Collins. Sinclair, J.McH. 1980a. Discourse in relation to language structure and semiotics. In S. Greenbaum, G. Leech and J. Svartvik, eds., Studies in English linguistics for Randolph Quirk. London: Longman. 110-124. . 1980b Computational text analysis at the University of Birmingham. ÏCAME News 4:13-16. . 1981. Planes of discourse. Mimeo. . 1982. Remarks at ASLIB Conference. Mimeo. Thorndyke, P.W. 1977. Cognitive structures in comprehension and mem ory of narrative discourse. Cognitive Psychology 9,1:77-110.
118
M.K. PHILLIPS
van Dijk, T.A. 1977a. Text and context: explorations in the semantics and pragmatics of discourse. London: Longman. . 1977b. Semantic macrostructures and knowledge frames in discourse comprehension. In M.A. Just and P.A. Carpenter, eds., Cognitive pro cesses in comprehension. Hillsdale, N.J.: Lawrence Erlbaum Associates. 3-32. Wilkins, J. 1688. An essay towards a real character and a philosophical lan guage. London.
POLITENESS AND THE SEMANTICS OF MODALISED DIRECTIVES IN ENGLISH
Christopher S. Butler University of Nottingham Nottingham, England
1.0 Congruence and the Analysis of Directives The work of Halliday since about 1970 has been characterised by the postulation of system networks at each of several levels, options at any of the 'lower' levels being interpreted as a recoding of options at the next higher level. In a recent paper (Halliday 1984), three non-phonological levels are proposed. At the supra-linguistic level of 'social context', distinc tions are made between the exchange of information and the exchange of goods and services; in either case, the speaker may give, demand or accept the commodity exchanged, or may give on demand. At the semantic level, such choices are reinterpreted in terms of initiations and responses by means of the speech functions of statement, offer, question and command. Further recoding occurs at the grammatical level, in terms of the mood categories, basically declarative, interrogative and imperative. I have attempted to assess the strengths and weaknesses of the model in detail elsewhere (Butler 1986): here, I wish to take up just one of Halliday's points. Halliday points out that although the possible mappings between options at successive levels are numerous (indeed, if there were a one-toone relationship there would be no point in postulating separate levels), there are certain 'congruent' realisation patterns which occur in the absence of any good reason to the contrary. For instance, if the speaker wishes to make an initiating move which gives information, the congruent realisation of this set of features at the semantic (speech function) level is a statement,
120
CHRISTOPHER S. BUTLER
and the congruent grammatical realisation of this semantic choice is a declarative clause. Similarly, demands for information are congruently realised as questions, then as interrogative clauses. Later in the paper, Halliday hints that the concept of congruence may be rather less clearly relatable to the exchange of goods and services. In the case of offers, the problem is that languages do not seem to have clearly defined grammatical patterns for the congruent realisation of this class of act. In the case of commands, the problem is rather different: there is a form, the imperative, in English and many other languages, which is claimed to be the unmarked realisa tion, and yet its use is rather restricted, the function of securing goods and services more often being performed by a variety of non-congruent realisa tions. The following are just a few of the many ways in which a speaker might try to get an addressee to open a window: (1) (2) (3) (4) (5) (6) (7) (8) (9)
I order you to open the window. (explicit performative) Open the window. (imperative) Open the window, could you? (imperative + modalised tag) You must open the window. (modalised declarative) Can you open the window? (modalised interrogative) Could I ask you to open the window? (performative embedded inside a modalised interrogative construction) Haven't you opened the window? (non-modalised interrogative) The window's still closed. (declarative with related, but not iden tical, propositional content) It's hot in here. (declarative with more distantly related proposi tional content)
It is the purpose of this paper to investigate non-congruent realisations of directives1 which contain a modal verb. I shall first review work which shows that the motivation for the selection of one realisation of an illocutionary act rather than another is related to politeness phenomena. I shall then present briefly a semantic analysis of mood and of the 'root' modals in Eng lish, and use the semantic features to predict (i) the acceptability, or other wise, of a given combination of mood and modal verb as a possible direc tive, (ii) the speech act classification of each potentially directive form, (iii) the relative politeness, in a given social context, of the various possible forms. Finally, I shall present the results of extensive informant testing of these predictions.
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
121
2.0 Politeness in Linguistics Some linguists have been sceptical of any attempt to relate linguistic forms to politeness. Sadock (1974) points out that certain requests are inhe rently rude because of their lexical content, whatever their syntactic form. Davison (1975) offers a number of arguments against politeness as a key factor in the analysis of indirect speech acts: it is, she claims, hard to define politeness; it is a property of individual speakers; polite people do not always use indirect speech acts, and not all utterances in polite discourse are indirect; utterances can be polite without being syntactically marked as such; and so on. These criticisms appear to insist that if the concept of politeness is to be of any value in linguistics, we must be able to treat it as an invariant property of sentences, attributable directly to their lexicosyntactic form. But the relationships involved here are, in fact, much more complex than this. It would be quite erroneous to think of politeness as hav ing a simple, unidimensional relationship to the form of sentences. If the concept of politeness is to have any explanatory value at all, we must expand our linguistic horizons beyond the consideration of formal properties of sentences, to include the effects of the context of utterance. It is not surprising, then, that the most enlightening discussions of politeness have been in work in the areas of speech act theory, implicature, and the like, which would be considered by many linguists as part of pragmatics. Heringer (1972) differentiates requests from orders in terms of a politeness condition attaching to requests, namely that the speaker believes the addressee would not object to whatever he is being asked to do. Lakoff (1973, 1974) proposes two general rules of 'pragmatic competence' — 'be clear' and 'be polite' — and relates these to Grice's (1975) conversational maxims. The politeness rule subsumes three more specific constraints: don't impose; give options; make the addressee feel good. Even the clarity rule can be seen as a special case of the politeness rule, since clarity avoids wasting the addressee's time, and so imposing on him. Mohan (1974) also uses the concept of politeness to build on Grice's conversational principles. He argues that for the explication of indirect speech acts, principles addi tional to conversational postulates are required in order to justify the con veying rather than the stating of what is being put across (see also Leech's work, reviewed below). Such justifying principles take the form of second ary implicatures, some of which are based on politeness. Other writers who have recognised the importance of politeness and
122
CHRISTOPHER S. BUTLER
related phenomena in the study of directives include: Fraser (1973), who proposes a 'mitigation marker' morpheme reflecting the degree of polite ness of a sentence (though see my remarks above concerning the danger of such an approach); Searle (1975), who claims that politeness is the main motivation for indirectness in requests; Lee (1975), who suggests that defer ence conditions play a part in determining the sentential realisation of a particular illocutionary force; Ney (1976), who invokes politeness in distin guishing various request forms; Fraser and Nolen (1981), who distinguish between politeness (interpreted as abiding by the rules of a conversational 'contract') and deference (according respect to the hearer). The most detailed and illuminating work on politeness, however, has come from Brown and Levinson (1978) and from Leech (1980, 1983). Brown and Levinson (1978: 60) observe that the utterances with which people choose to express themselves frequently diverge markedly from a rational, efficient mode of communication in which the Gricean maxims of quality, quantity, manner and relation are strictly and overtly adhered to in an uncomplicated way. They also demonstrate that such divergences show a remarkable degree of similarity across diverse languages and cultures. To account for this, they propose that conversationalists apply politeness stragegies whose function is to pay attention to the 'face' of their inter locutors, that is 'the public self-image that every member [of a society] wants to claim for himself (Brown and Levinson 1978: 66; see also Goffman 1967). Face consists of two aspects: negative and positive. Nega tive face is concerned with freedom to act, and the right not to be imposed upon by others; positive face is to do with the desire for positive appraisal of one's self-image. Some acts, both verbal and non-verbal, are inherently face-threatening. Occasionally, we may wish to carry out such an act in the most efficient way, regardless of the effect on the face of our conversa tional partner: in such cases, we produce what Brown and Levinson call a 'bald on record' act. But more usually, conversationalists employ strategies, of varying complexity, for the minimisation of face threats. The more threatening the act, the more the speaker will attempt to minimise the threat, in general. However, speakers do not usually choose strategies which are less risky than necessary, because to do so might be construed as indicating that the act is more threatening than is really the case. Directives are inherently face-threatening because they represent impositions on the addressee. Redress strategies are thus likely to favour the minimisation of the hurt to the addressee's negative face. Brown and
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
123
Levinson discuss in detail the various strategies involved and some of their realisations in English and other languages. The basic strategies are: the avoidance of presumption/assumptions about the hearer's wants; giving the hearer the option not to act; and communicating the speaker's desire not to impose on the hearer (compare Lakoff's rather similar suggestions, men tioned earlier). Although all directives are inherently face-threatening, the extent to which redress is necessary depends on other social factors. Brown and Levinson (1978: 79ff.) isolate three such factors: the 'social distance' between the speaker and hearer; their relative 'power' (see also Brown and Gilman 1960); and the degree to which the action being requested is regarded, in the culture concerned, as threatening to the hearer's negative face. See also the work of Ervin-Tripp (1976), showing that in the selection of an appropriate form of directive, familiarity, authority, rights and duties, also the difficulty of the task, all have a part to play. The greater the social distance and the severity of the task, and the smaller the power difference, the more need there will be for redress of face. Leech (1980, 1983) regards politeness phenomena as part of a set of rhetorical principles which guide the management of conversational interaction. Borrowing, and considerably stretching, the use of terms from Hallidayan linguistics, Leech distinguishes between 'interpersonal' and 'tex tual' rhetorics, each of which consists of a set of principles. The textual rhetoric, with which I shall not be concerned here, subsumes the principles of processibility (concerned with focus and end-weight), clarity, economy and expressivity. The interpersonal rhetoric subsumes the Cooperative Principle (CP) of Grice (1975, 1978), the Irony Principle and the Politeness Principle (PP). Leech's (1983: 81) formulation of the PP is as follows: In its negative form, the PP might be formulated in a general way: 'minimise (other things being equal) the expression of impolite beliefs', and there is a corresponding positive version ('Maximise (other things being equal) the expression of polite beliefs') which is somewhat less important.
The link with Brown and Levinson's negative and positive politeness is obvious. By polite and impolite beliefs Leech means 'respectively beliefs which are "favourable" and "unfavourable" to the hearer or to a third party, where "favourable" and "unfavourable" are measured on some rele vant scale of values' (Leech 1983: 81). Leech points out that in one sense the PP has a higher regulative function even than the CP, in that in many situations of communication, unless the social equilibrium is maintained,
124
CHRISTOPHER S. BUTLER
largely by the exercise of politeness, the channel of communication may well break down, at least to the extent that cooperativeness can no longer be assumed. Politeness, in Leech's (1983: 123ff.) view, is to be seen as a set of scales on which illocutions can be graded between inherently polite and inherently impolite: (i)
the cost-benefit scale, concerned with the cost or benefit, to the speaker and/or hearer, of the action, A, which is proposed; (ii) the optionality scale, concerned with the degree of choice allowed to the hearer (h) by the speaker (s); (iii) the indirectness scale, concerned with the number of steps, in a means/end type of analysis, which link the illocutionary act to the goal the speaker is trying to achieve; (iv) the authority scale, concerned with the power of one participant over another; (v) the social distance between participants (which can be seen as inversely correlated with Brown and Gilman's (1960) 'solidarity'). The values on each of scales (i), (iv) and (v) for a particular social interaction will determine the degree of indirectness employed by the speaker in order to observe the 'Tact Maxim' of the PP, namely that the cost to the hearer should be minimised. Leech (1983: 127) expresses these relations succinctly as follows: (i) (ii) (iii) (iv)
the greater the cost of A to h, the greater the horizontal social distance of h from s, the greater the authoritative status of h with respect to s, the greater will be the need for optionality, and correspondingly for indirectness, in the expression of an impositive, if s is to observe the Tact Maxim.
Leech also suggests, more tentatively, further maxims, of generosity, approbation, modesty, agreement and sympathy, subsumed under the PP. Summarising this all too brief account of recent work, we may say that politeness plays a crucial role in maintaining social relations during every day conversational interaction, and that the linguistic strategies employed by speakers will be conditioned by a number of socially determined factors, including the relations (of status, authority, solidarity and the like) between the interactants, and the cost, to speaker and hearer, of the action being
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
125
negotiated. We may expect that an utterance which is appropriate as a directive in one social context may be seriously inappropriate in another. For instance, (10) below would be inappropriate in the context of drill sergeant/recruit interaction, because the highly asymmetric authority rela tions make the redress of negative face unnecessary, and a bald command such as (11) is perfectly acceptable. (10) Could you possibly stand at ease? (11) Stand at ease. On the other hand, between adult strangers, or even acquaintances, (12) would be much more appropriate than (13), largely because of the shift in authority relations. (12) Could you possibly pass me that book? (13) Pass me that book. A corollary of this, however, is that in a given social situation, defined in terms of social distance, authority and cost/benefit relations, the range of possible forms of a given illocutionary act will be perceived as differing in their relative politeness. In order to test this hypothesis, it seems sensible to start with the most neutral possible situation, where the cost of the act to the hearer and the benefit to the speaker are minimal, and the participants of equal status and at a medium to large social distance in terms of acquain tanceship and solidarity.
3.0 Politeness, Speech Act Classification and the Semantics of Mood and Modalisation in English In this section, I shall show that we can predict, from the semantics of mood and of the modal auxiliaries, certain properties of the use of modalised sentences as directives. Firstly, we can predict which of the syntacti cally possible combinations of grammatical mood and modal verb will be available as potential directives and which will not. Secondly, we can pre dict the speech act classification of the potential directives as orders, requests or suggestions. Thirdly, we may hypothesise relative politeness orderings for the various available directives in the 'neutral' social context defined earlier. Reformulating this in terms of congruence, what I hope to achieve is an account of the constraints on the range of possible non-con gruent as well as congruent realisations of directives, and also an account of
126
CHRISTOPHER S. BUTLER
the politeness differences between these realisations, as used in a particular social context. Lack of space forces me to give only a rather brief and over-simplified account of this complex area. For instance, I shall not discuss the precise theoretical status of the term 'directive': in fact, I regard it as a category at the level of discourse organisation. Furthermore, I shall not be able to jus tify the semantic networks I propose, or to give formal realisation rules. And I shall be able to discuss only a subset of the predictions which can be made about politeness. Full details of all these important matters can be found in Butler (1982 and 1987). 3.1 Semantic Force Options
Figure 1: simplified network for semantic force
Figure 1 shows a simplified network for 'semantic force' (that is, the semantics underlying grammatical mood) in English. It is based on refine ment and extension of the ideas put forward by Hudson (1975), who argues for the recognition of semantic 'force markers', which are associated with sentence structures independently of their context of use, and which, together with the knowledge of coversational participants about the context and the cotext, allow the inference of possible illocutionary forces for utter ances. For futher discussion of the relationship between semantics, speech acts and discourse in this area, see Butler (1987). Only those parts of the network relevant to our immediate concerns will be discussed here. Predications with the feature [+performative] are those in which the speaker actually performs the act specified by the lexical verb. They could be further classified according to the type of verb (see
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
127
Leech 1983), and a full account would have to examine the relationship between these subclasses and the non-performative expression of state ments, questions, etc. Semantic statements are defined in terms of the fea tures [-question, -exclamation], and have as their defining property the fact that the speaker is encoding his (actual or purported) belief in the truth of the proposition. Questions are defined in terms of the feature [+question], glossed as the encoding by the speaker of his (actual or purported) belief that the addressee knows at least as much as he does himself about the truth of the proposition (or, in the case of a wh-question, about the value of the questioned variable which would make the proposition true). The underly ing semantics of imperative-form sentences is captured in the feature [—informational], for which the gloss is that the predication is concerned with the performance of a hypothetical future act by the addressee. The feature [+question tag modification] is present for those sentences in which the speaker modifies his basic choice of semantic force by indicating his (actual or purported) belief that the addressee knows at least as well as he does himself whether the proposition is true (or, in the case of [-informational] predications, whether it will be made true). The various types of semantic force relevant to our discussion of potentially directive function are illus trated in (14)-(18) below. (14) I tell you to open the window. [+performative] (15) Open the window. [-informational] (16) Open the window, will you? [-informational, +question tag mod ification] (17) You will open the window. [-question, -exclamation, -question tag modification] (18) Will you open the window? [+question, -exclamation] 3.2 The Semantics of the 'Root' Modals Let us now turn to the semantic analysis of modal verbs. This owes a great deal to Halliday (1970a), but also to Palmer (1979) and several others. In the present account, modals are seen as representing semantic predicates in their own right. So, for example, one semantic interpretation of (19) is as shown informally in (20), where dots separate arguments and predicates, and brackets indicate a 'rankshifted' predication. (19) You can open the window. (20) Addressee . be able . (addressee . open . the window)
128
CHRISTOPHER S. BUTLER
The entry conditions for distinction within modal meaning are fairly com plex (see Butler 1982, Chapter 8), and will not be discussed further here. Rather, I shall use the term 'modal predicate' as a cover label for the com plex set of defining features. Furthermore, only the 'root' meanings of the modals (Halliday's 'modulation') are involved in the potentially directive forms under discussion here. The distinctions in root modal meaning rele vant to our discussion are shown in Figure 2.
root modal predicate
Figure 2: A semantic network f or modal predicates
A notational convention not previously used in systemic linguistics has been introduced here. It so happens that the pattern of interaction of the four systems is very regular, in that nearly all the combinations of features allowed by treating the systems as simultaneous are in fact well-formed. There are, however, a few exceptions. These could be handled using the normal 'and' and 'or' bracketing notations, but the networks would then become complex, and the underlying regularities would be obscured. The notation
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
129
indicates that the [b/c] choice is available except where [a] has been chosen (note that [a] can itself be a conjunction or disjunction of features). Some times, this means that the distinction is neutralised altogether: for instance, where the features [nec, modality-] are chosen (corresponding to the form needn't — see below), there is complete neutralisation of the [+/—tenta tive] distinction. In other cases, it may be necessary to associate one of the two features [b] and [c] with any selection expression containing [a]. For instance, if [+discourse participant involvement, vol] are selected (corres ponding to shall), the selection expression must also contain the features [-tentative, modality+]; as shown by the asterisks attached to the 'un marked' options in the network. I shall now say a little about each semantic system. The [poss/nec/vol] system represents the three basic types of root modal meanings: possibility (including ability and permission), necessity (including obligation) and vol ition. The [+/—tentative] system distinguishes semantically between the pairs can and could, may and might, will and would, must and should/ought. The [+/—discourse participant involvement] system reflects the distinction, recognised by many writers on modal semantics (see, for example, Palmer 1974, 1979; Leech 1969, 1971; Antinucci and Parisi 1971; Mitchell 1974; Lakoff 1972a; Lodge 1974), between senses of the modals which involve a discourse participant (the speaker in statements, the hearer in questions) as the immediate source of constraint, and those where the source of con straint is more general. Discourse participant oriented modal uses are characterised by their incompatibility with past time (the speaker cannot impose a constraint on a past event), and by the fact that the recipient of the constraint must be a person rather than a non-personal animate entity or an inanimate object. The [modality+/—] system is concerned with whether the modal meaning itself has a negative element combined with it — that is, the distinction between possible and not-possible, necessary and not-necessary, willing and not-willing. In addition, the predicate rep resented by the main verb in the sentence may or may not have a negative
-
-
+
+ + +
± discourse participant involvement
Ø
+
+
-
+
-
+
-
+
-
± tentative
+ + + -
+ + + +
+ + -
+ + -
modality +/—
willingness
volition/
participant
discourse
guarantee by
obligation
¡general
participant
by discourse
obligation
ability
permission
semantic gloss
will would won't wouldn't
shall
must should/ought needn't must/have to should/ought needn't/don't have to
can could can't couldn't
may/can might/could may not/can't mightn't/couldn't
-
Ø
+
Table 1 : Realisations of selection expressions from the root modal semantic network
won't/will not wouldn't/would not won't not wouldn't not
shan't
needn't not
mustn't shouldn't/oughtn't needn't not mustn't/haven't to shouldn't/oughtn't
may/can not might/could not can't not couldn't/ mightn't not can not could not can't not couldn't not
realisation main predicate + main predicate —
Note: In the final column, not indicates a stressed negative, often accompanied by a falling-rising tone -
vol
nec
poss
poss/nec/vol
130 CHRISTOPHER S. BUTLER
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
131
element combined with it. The various combinations of features and their realisations are shown in Table 1. For evidence for the claims made here (for instance, that should/ought and needn't can be either [+discourse par ticipant involvement] or [-discourse participant involvement]) see Butler (1982, Chapter 8). 3.3 Predictions from the Semantics 3.31 Acceptability I shall confine my remarks here to showing why certain combinations of mood and modal verb result in sentences which are not readily available for use as directives, while other, closely related forms are. First, compare (21) and (22): (21) You may open the window. (22) May you open the window? The root meaning of may, according to the network in Figure 2, has the fea ture [+discourse participant involvement], so that (21) is roughly equiva lent to (23): (23) I permit you to open the window. Applying the same reasoning to (22), we arrive at an interpretation in which the speaker is asking whether he grants permission to the addressee. This is clearly an extremely odd thing to do, and we may expect (22) to be unavailable as a directive. Now consider (24) and (25): (24) I may ask you to open the window. (25) May I ask you to open the window? Although (24) could be interpreted, with an epistemic meaning for may, as a statement of the possibility that the speaker will ask the hearer to open the window, as a directive with root meaning for may, it is odd, because it is strange to state that one has one's own permission to do something. But (25) is perfectly acceptable, since the speaker appears to be asking the addressee for permission to request the action. A rather different kind of explanation for unacceptability can be seen if we compare (26) and (27): (26) You will open the window. (27) You would open the window. In (26) the speaker adopts a very strong position by asserting either that the
U/S U/S
s U/S
u/s
u
U/S
U/S
U/S
might
Key: U unacceptable order R request S suggestion
ought
should
must
U U
U R
U
R U
U U
u u
U
U
U/R
u u
R
U
u
R
U
u
U U
u
u/o u u
U
U
U
R U
U
U
o u o
R
U
U
R
U
U
U/S
U
TELL
R
U
ASK
QUESTION PERFORMATIVE
U
o/s
R
Table 2: Hypotheses regarding acceptability and speech act classification
X/Y some informants will give X, others Y, both in considerable proportions
u u u u u u
u u
u u
u/o o
shall
o s s
R
U/R
R
U
would
may
R
R
U/S
R
R
O/S
R
R
R
O
s
o/s
NON-QUES POSITIVE NEGATIVE NON-INFO NON-INFO STATEMENT PERFOR MATIVE NON-EXCL QUESTION QUESTION POS. TAG NEG. TAG (=state ASK TELL ment)
will
could
can
Ø
MODAL
SEMANTIC FORCE / POLARITY
132 CHRISTOPHER S. BUTLER
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
133
addressee will definitely carry out the act, or that he is willing to do so (de pending on whether will is interpreted as a future or a volitional modal). Both interpretations contravene the negative politeness strategy of giving options, and not presuming upon the addressee's goodwill. We therefore expect (26) to be an inherently impolite, but possible, form of directive. In (27), however, the modal would has the added feature [+tentative], and this tentativity clashes with the strong speaker position. We might therefore predict that sentences of the form shown in (27) would be unavailable as directives. The full set of forms predicted as unacceptable on the basis of these and other arguments can be seen in Table 2. Detailed arguments for indi vidual cases can be found in Butler (1982, Chapter 9). 3.32 Speech Act Classification As a general rule, we might predict that those directively available forms which appear to refuse the addressee any options, and/or represent the imposition of the speaker's authority and/or will, will be classified as orders, while forms which explicitly encode the offering of options will be perceived as requests. Straight imperatives, also sentences involving the performative tell (as in I (must) tell you to open the window), and modalised statements with will (see (26)) and must (expressing the imposition of an obligation), leave no options apart from outright refusal to comply, and so would be predicted to be orders. Sentences with question semantic force (that is, those with inter rogative form), or with question tag modification, explicitly encode the offering of options, and so are predicted to be requests, as are sentences with the requestive performative ask. We shall see below, however, that in certain cases the meaning of the modal verb itself interacts with the seman tic force to create a complex meaning which is most readily interpretable as a suggestion, the defining characteristic of suggestions being that, unlike orders and requests, they are made for the benefit of the hearer rather than for that of the speaker, and therefore do not call for negative politeness strategies. The complete set of predictions for speech act classification can be seen in Table 2. 3.33 Politeness The more work the speaker does in attempting to minimise impositions on the addressee, and to maximise the options afforded, the more polite we should expect the directive to be in our 'neutral' social context, though it is
134
CHRISTOPHER .S. BUTLER
possible that at the top end of the scale there may be a 'saturation' point beyond which further linguistic mechanisms achieve no additional increase in perceived politeness. (As I noted earlier, this simple prediction would not hold in all social contexts: for instance, if the authority relations were highly asymmetric, or the social distance very small, the piling up of polite ness strategies could be interpreted as ironic or sarcastic.) We might expect, then, that modalised declaratives will be treated as less polite than the corresponding interrogatives (a claim frequently made in the literature: see, for example, Heringer 1972: 43; Lakoff 1974: 44; Mohan 1974: 454; Forman 1974: 166; Brown and Levinson 1978: 140; Fraser and Nolen 1981: 102-3; Leech 1983: 119ff.); that bare imperatives will be perceived as less polite than imperatives with tags; and that forms with performative tell will be less polite than the corresponding forms with ask. Since the performatives are the most transparent of all forms of a speech act, we should expect the bare command performative I tell you . . . to be less polite even than the bare imperative. A further general prediction is that sentences with [+tentative] modals will be regarded as more polite than the corresponding [-tentative] forms, since the expression of tentativeness is in agreement with the negative politeness requirement of minimising imposition: as Brown and Levinson (1978: 178) and Leech (1983: 121) point out, the past tense modals indicate a hypothetical action, and no assumption is made that the hypothetical world is closely related to the real one. For discussion of tentatives in rela tion to Could you . . . ? vs. Can you . . . ?, Would you . . . ? vs. Will you . . . ?, etc., see also Twaddell 1965: 15; Leech 1969: 236, 1971: 120; Palmer 1974: 127, 1979: 87, 135; Ney 1976: 15; Fraser 1973: 301; Fraser and Nolen 1981: 102-3; Heringer 1972: 43. It might be expected that the same relation ships would be shown also in question tags on imperatives. Furthermore, we can predict that in directive statements [+tentative] should/ought will be more polite than [-tentative] must, might more polite than may, and could more polite than can (see, for example, Leech 1969: 237, 267 and 1971: 121; Diver 1964: 345). Let us now turn to the effects of individual modal lexemes (will/would vs. can/could, and so on). In statements, we should expect modals with the feature [+discourse participant involvement] (that is, shall, may, one interpretation of must) to be impolite, since the speaker, by his choice of modal, overtly indicates that he himself is the immediate source of con straint, so contravening the requirements of negative politeness. Will is also impolite, because it indicates the speaker's assumption either that the
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
135
hearer will in fact do the act, or that he is willing to do it. Statements with can are interesting, because of the polysemous nature of the modal. Where can has the feature [+discourse participant involvement], and so is equiva lent to may, it should be impolite, for the reasons given above. Where it is [-discourse participant involvement], referring literally to ability, the speaker is simply pointing out what it is possible for the hearer to do, and this interpretation does not necessarily involve any imposition on the hearer. We might therefore expect politeness ratings for can statements to be somewhat mixed. Negative modalised statements without any tag are, of course, normally used as prohibitions rather than to get someone to do something, and so are not relevant to our concerns. It is generally agreed that will/would/can/could occur standardly with directive function in the interrogative and also in imperative plus tag con structions. At the time when this research was conducted, I saw no good reason for expecting will and can (or would and could) to differ in polite ness (see the hypothesis set out in Section 4.223). Since then, however, Leech (1983: 120) has pointed out that if, in reply to a Will you . . . ? request, the hearer answers No, I won't, he is placing his own wishes above those of the requester, whereas if, in reply to a Can you . . . ? request, he replies No, I can't, he has a get-out, since no-one can be held blameworthy for not doing something if he is unable to do it. Leech advances this as an explanation for his claim that Can you . . . ? is seen as more polite than Will you . . . ? The effect of negativity on the politeness of interrogative directives is rather complex. Let us consider won't and can't. It has been pointed out (see Close 1975: 264, Zandvoort 1975: 74) that Won't you . . . ? signals an invitation. Green (1973: 73) claims that Won't you . . . ? is more polite than Will you . . . ?, but Fraser (1973: 303) places won't below will on a scale of politeness, and Fraser and Nolen (1981) predict that the positive modals will generally be more polite than their negative counterparts. For Forman (1974: 167) Can't you . . . ? is a suggestion, while for Fraser (1973: 303) it is only slightly more polite than a bare imperative. Green (1973: 73) points out that Can't you . . . ? can be either a suggestion, as in (28), or an impo lite order, as in (29), according to the propositional content. (28) (= Green's 69a) Can't you put the meat on first? (29) (= Green's 69b) Can't you be a little quieter? The reasons for this are fairly clear. Since a question of the form Can't you X? is a question about a negative proposition, it can be glossed as 'My
136
CHRISTOPHER S. BUTLER
assumption was that not-X, but I am now questioning it'. In the case of (28), where there might well be circumstances, unknown to the speaker, which affect the feasibility of the action, both the initial assumption and the new questioning of it are reasonable, and there is no necessary implication of impoliteness. But in the case of (29), where the hearer's ability to do the act cannot be in doubt, the initial assumption is clearly untenable, so that the question is inevitably ironic, and heavily biased towards a positive response, and hence towards the hearer's compliance, so infringing the requirements of negative politeness. For a similar analysis see Leech (1983: 122-3). We might therefore expect informants to show mixed reactions to Can t you . . . ? forms according to how they view the act concerned. The position with regard to negative modals in tags is unclear. Lyons (1977: 761) suggests that a negative tag is used when the speaker has reason to believe that the hearer may not comply, and that it is frequently accom panied by paralinguistic features indicating annoyance or impatience. We may also note, however, that it is polite for a speaker to assume that the hearer will not be willing to demean himself by doing what is required of him, in which case a negative tag should be more polite than a positive one. This fits in with Lakoff's (1972b: 914) claims for invitations, namely that (30) is more polite than (31). (30) (= Lakoff s 13) Come in, won't you? (31) (= Lakoff s 16) Come in, will you? In view of the conflicting factors involved in the interaction of negativ ity with modal meaning, it seems wise to make no predictions of relative politeness (except that can't might be expected to be rated rather lower than can by some people), but to see what emerges from informant tests. Finally, I shall say just a few words about modal plus performative combinations (see Fraser 1975). We should expect forms with ask to be more polite than the corresponding forms with tell. We might also expect I must tell/ask you to. . . to be more polite than the bare / tell/ask you to. . ., since it purports to indicate that the speaker is under some obligation to make the order or request, and so mitigates to some extent the imposition on the hearer. We should expect the interrogatives May/might/can/could/I ask you to . . . ? to be extremely polite, since they can be interpreted in terms of the speaker asking the hearer for permission to make the request, so maximising the options allowed. 3.34 Relationships between Speech Act Classification and Politeness Since the status of a directive as a request or as an order is based on the
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
137
authority relations perceived to be encoded by the directive, we may pre dict that there will be a correlation between speech act classification and politeness, such that those directives classified predominantly as requests will be polite, while those classified predominantly as orders will tend to be impolite. Since suggestions do not involve authority relations, they are not expected to show any significant correlation with politeness.
4.0 Hypothesis Testing 4.1 Methodology Arguments such as those in Section 3 led to the formulation of a large number of hypotheses concerning the relationships between semantic fea tures (and hence, ultimately, lexicosyntactic form) and the acceptability, speech act classification and politeness of various modalised forms. An informant testing program was devised in order to obtain evidence which would support or disconfirm these hypotheses. A set of test items was generated by combining each of the 9 relevant semantic force/polarity types (statement, positive question, negative ques tion, non-informational (that is, imperative syntax) with positive tag, noninformational with negative tag, statement with request performative, ques tion with request performative, statement with command performative, question with command performative) with each of the 10 modals will, would, can, could, may, might, must, shall, should, ought, in sentences con cerned with the act of opening a window. In addition, 3 non-modalised forms were included — the bare imperative, bare request performative and bare command performative — giving a total of 93 test items. After a pilot project to test the methodology, and consequent revisions in test proce dures (see Butler 1982, Chapter 10), these 93 items were presented to a group of 112 first year undergraduates, both on paper and in spoken form on tape. The taped version used standard, unmarked intonation and stress patterns for each formal type of sentence (see Halliday 1970b).2 informants were asked to judge whether each sentence, as spoken on tape, was a possi ble way of getting someone to open a window. If the sentence was judged to be a possible directive, the informant was asked to classify it as an order, request or suggestion. From the results of this test (see Section 4.21) a set of 35 directives with at least 80% acceptability was isolated, and used in politeness rating tests. 97 informants were asked to imagine that they were trying to get an
138
CHRISTOPHER S. BUTLER
acquaintance of the same age and sex to open a window. The sentences were presented on paper and on tape, and informants were asked to rate each one on a politeness scale ranging from 1 (very impolite) to 7 (very polite). A computer program was written to compute the median politeness rating for each directive, the median being a more suitable measure of cen tral tendency than the mean, since it would be unjustifiable to assume that equal intervals on the scale represented equal jumps in politeness. The pro gram also performed the sign test to assess the statistical significance of dif ferences between sets of ratings for each possible pair of directives. A dis cussion of these statistical techniques can be found in Butler (1985). Dif ferences were regarded as significant if the probability of obtaining them by chance was 5% or less; in fact, many of the differences proved to be signif icant even at the 0.1% level. 38 informants were available for re-testing on politeness ratings after 3-4 weeks. Sign tests revealed that for all but one of the test items there was no significant difference between ratings on the ini tial test and the re-test. The testing method thus showed a very high degree of reliability. Two informant studies of politeness in directives previously reported in the literature differed in important ways from the investigation reported here. In Mohan's (1974) study, which originally suggested the lines along which my own testing might proceed, 80 American informants aged 18-35 were presented with the same hypothetical social context used in the pre sent study (acquaintances of the same age and sex, one trying to get the other to open a window), and asked to rate each of 19 directives (only 10 of which were modalised) on a scale from 1 (at the least polite end) to 5 (at the most polite end). Medians were calculated, and the significance of differ ences between median ratings assessed statistically. Fraser and Nolen (1981) took 25 directives (15 of them modalised) and generated pairs of them randomly, then presented 45-50 pairs to each of 40 American college students, and asked them to state which member of each pair indicated the greater degree of deference, defined as respect for the hearer. A computer program was then used to obtain a rank ordering for the 25 directives. One major difference from both Mohan's study and my own was that the direc tives were made absolutely neutral as to content (for example Will you do that?). Furthermore, no social context was specified, even though Fraser and Nolen state, in their discussion of deference, that it is associated with activities or utterances, and not with sentences as such. The significance of differences between ratings was not assessed statistically. In both of these previous studies, the directives were presented to informants in the written
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
139
form only, and no tests of acceptability or speech act classification were undertaken. The findings of both investigations will be compared with those of the present study at relevant points in the following discussion. 4.2 Hypotheses and Results of Testing 4.21 Acceptability and Speech Act Classification Table 2 summarises the detailed hypotheses made concerning the acceptability, as a directive, of each of the 93 items tested, and the speech act classification predicted for the acceptable items. Table 3 shows the results of the acceptability test: it can be seen that for only 3 of the 93 forms is there a clash between predictions and findings. All 3 cases concern the embedded performative type: I shall/'will/'tell you to . . . were accepted by only 67% of the informants, while Must I tell you to . . .? was accepted by 85%, a higher proportion than expected. It is perhaps significant that the only discrepancies were concerned with a form of directive which is almost certainly less common than the other types studied. In Table 4, for each of the 35 directives accepted by 80% or more of the informants, the predicted speech act classification is compared with the predominant classification given by the informants. A test item was regarded as being predominantly of one speech act type if 60% or more of the informants classifed it as such; and items which did not achieve a score of 60% for any one speech act were classified as 'mixed'. The results con firm our predictions to a very high degree, although some directives had a rather more mixed classification than had been predicted. 4.22 Politeness Rating 4.221 Hypotheses Relating Politeness to Speech Act Classification Table 4 also contains data relevant to the following hypotheses: H1: Directives classified mainly as orders will be relatively impolite. H2: Directives classified mainly as requests will be relatively polite. H3: Since politeness is not really an issue in suggestions, directives classified mainly as suggestions will be neither particularly polite nor impolite. Median politeness ratings for orders range from 1.04 to 3.18; all but one (I must tell you to . . . ) having values no greater than 1.62. The median of the median ratings for orders is 1.32, which clearly supports H1 above. Ratings for requests vary from just below the mid-point of the scale to the
1
should might
might
shall
must should ought will may can could
70-79
80-89
90-100
will would can could
ought
may shall
60-69
50-59
40-49
30-39
won't couldn't shouldn't oughtn't
can't
wouldn't mightn't
shan't
will would can could
0t ag
can't won't
couldn't
wouldn't
must
will 0 modal shall
would might ought
may
can could should
ASK
must
0 modal
shall* will*
can could may might
ought must
shall
should ought,can may, could
would
will
ASK
should
|
might
would
TELL
must*
might
can could may
shall ought
should
will would
TELL
STATEMENT + PERFORM. | QUESTION + PERFORM.
Note: Asterisked items are those at variance with predictions
mightn't oughtn't shouldn't
might
mayn't mustn't
20-29
must
shan't mayn't
may, shall should ought
10-19
NON-INFO. NEG.TAG
SEMANTIC FORCE / POLARITY NON-INFO. POS. TAG mustn't
would
NEGATIVE QUESTION must
0-9
POSITIVE ACCEPTABILITY STATEMENT QUESTION
%
Table 3: Degree of acceptability for the 93 forms tested
CHRISTOPHER S. BUTL
140
MIXED
≥ 60% SUGGESTION
≥ 60% REQUEST
≥ 60% ORDER
CLASSIFICATION
should S>0[S] 2.29 may S>0[0] 2.19 can S>0[0/S] 1.82
ought [S] 2.15
could [S] 2.85
will [O] 1.04 shall [] 1.09 must [O] 1.32
STATEMENT
NEG. QUEST.
can't R>S>0 [O/S] 2.12
couldn't R>S[U/S] 3.77
shouldn't [S] 3.37 oughtn't [U/S] 3.02
will [R] 4.73 would [R] 5.62 won't [R] can [R] 3.65 4.94 could [R] 5.64
POS. QUEST.
NON-INFO. NEG.TAG
will [R] 4.20 would [R] 4.70 can [R] 3.67 could [R] 3.99
can't R[0/S] 2.64
won't R>0[R] 2.83
0 tag [O] 1.36
NON-INFO. POS.TAG TELL
0 modal [O] 1.12
must [U/O] 1.62
may [R] 6.74 might [R] 6.62 can [R] 5.96 could [R] 6.46
ASK
TELL
QUESTION PERF.
2.15 to 3.37 (=1.22)
3.65 to 6.74 (=3.09)
(=2.14)
to 3.18
1.04
RANGE OF MEDIANS
2.94
4.94
1.32
MEDIAN OF MEDIANS
will R>0[R] 2.74 must 0>S R shall 0 > R SS[R] [U] EVITCERID DESILADOM FO SCITNAMES DNA SSENETILOP 2.66 1.08 0modal 0>R[R] 141 1.64
must [R] 3.18
ASK
STATEMENT PERF.
SEMANTIC FORCE / POLARITY
Table 4: Predominant speech act classification and median politeness ratings of the 35 'acceptable' directives
142
CHRISTOPHER S. BUTLER
very top, with a median of 4,94, on the polite side of the mid-point, so sup porting H2. Suggestions received rather lower ratings than expected (2.15 to 3.37; median 2.94), but do fall clearly between orders and requests. A second way in which we can test the relationship between speech act classification and politeness is to calculate the Pearson product-moment coefficient (r) for the correlation between the median politeness rating of a directive and its percentage classification as a particular speech act type. We should expect a strongly negative correlation for orders, a strongly posi tive correlation for requests, and a near-zero value for suggestions if, as I have suggested, politeness is not an issue here. The values calculated clearly support the hypotheses: orders requests suggestions
: r = —0.78 : r = +0.87 : r = -0.09
The values for orders and requests are significant at the 0.1% level, whereas that for suggestions is clearly non-significant. 4.222 Hypotheses Relating Politeness to Semantic Force The following hypothesis subsumes a number of detailed claims about the relationship between politeness and semantic force: H4: Within the limits imposed by the acceptability of particular modals in combination with particular semantic forces, the politeness ordering of the semantic force types for a given modal form will be: increasing statement
statement + command performative
statement + request performative
politeness positive question
question + request performative
non-informational (imperative) + tag
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
143
144
CHRISTOPHER S. BUTLER
The relevant data are summarised in Table 5, which shows that every one of the predictions made in H4 is confirmed, for each modal where appropriate pairs can be tested. The results agree with those of Mohan (1974), who showed that for can the order of increasing politeness was statement, state ment with ask, question, question with ask; for will the statement was less polite than the question; and for may the statement was less polite than the question with ask. There is also agreement with Fraser and Nolen's (1981) finding that for can, can't, could, couldn't, will and won't, the imperative . plus tag form is less polite than the corresponding interrogative. Table 5 also shows data relevant to the following hypothesis concern ing non-modalised forms: H5:
The politeness ordering for the non-modalised directives studied will be:
Again, the two sub-hypotheses contained in H5 are supported by the data. The data in Table 5 also allow us to make observations on comparisons where no direction of difference was predicted (and which therefore require a 'non-directional' sign test, as opposed to the 'directional' test used where the direction of difference was predicted). The most interesting and impor tant finding is that in all cases negative modals are less polite than the cor responding positive modals, when combined with any given semantic force. As far as the interrogatives are concerned, this confirms the findings of Fraser and Nolen, who showed that won't was less polite than will, can't less polite than can, and couldn't less polite than could. In imperative plus tag constructions, Fraser and Nolen found that couldn't was less polite than could, but that will was rated below won't. 4.223 Hypotheses Relating Politeness to Modal Semantics Data relevant to the following hypothesis are incorporated into Table 5: H6: For any given modal lexical item, and any given semantic force, the [+tentative] modal (if available) will be more polite than the corresponding [-tentative] modal.
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
145
146
CHRISTOPHER S. BUTLER
All predictions but one are confirmed, the exception being the pair Might/ may I ask you to . . . ?', which shows a non-significant difference. The most likely explanation of this is in terms of a saturation phenomenon (see Sec tion 3.33): the interrogative form with may is already so polite that further signals of politeness have no added effect. The could/can pair in this con struction does show a significant difference; however, the median rating for the can form is lower than that for may (5.96 as against 6.74), and is pre sumably below the saturation level. My results confirm Fraser and Nolen's to a large degree: they found could > can, couldn't > can't for questions, and wouldn't > won't, couldn't > can't for tags. But they also found, inexplicably, that will was more polite than would in questions. However, as mentioned earlier, no tests were carried out to assess the significance of the difference observed. Data for comparison of individual modals in directive statements are given in Table 6, and allow us to test the following hypothesis: H7: In modalised statements, the modals will show the following politeness ordering: 3
All the predictions made by H7 are supported, except for comparisons involving may. You may . . . achieves a significantly higher politeness rating than You can . . . , and shows no significant difference from You should . . . or You ought . . . . The explanation is probably that may, through its use in the polite May I . . . ?, has acquired an overall connotation of politeness. Table 6 also shows that must is more polite than will and shall in state ments, and should more polite than can; the differences between will and shall, should and ought, can and ought, are non-significant. Mohan found the order may > can > will for statements. Table 7 shows comparisons between the modalised statements with various modals, and the bare imperative. It can be seen that You/can/could/ may/should/ought . . . are significantly more polite than the bare impera-
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
147
Table 7: Comparison of politeness of bare imperative and modalised statements tive, You must... is not significantly different, while You will/shall. . . are less polite than the imperative. Hypothesis H8 concerns politeness in interrogative directives: H8:
In modalised questions the modals will show the following polite ness ordering:
The relevant data are given in Table 8, which shows that all predictions con cerning the acceptable forms in the above series are confirmed, with the exception that Won't you . . . ? and Couldn't you . . . ? show no significant
148
CHRISTOPHER S. BUTLER
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
149
difference, the reason not being immediately obvious. Table 8 also allows comparisons to be made between modals grouped together in H8. The most interesting of these findings are the non-significant differences between will and can, would and could (which go against Leech's claim - Section 3.33), and the fact that shouldn't is more polite than oughtn't, whereas in modalised statements should and ought showed no significant difference. Mohan, also Fraser and Nolen, found that will was more polite than can in questions (a finding which is at odds both with the results of the present study and with Leech's claims), though Fraser and Nolen also found that would and could had very similar ratings. In Fraser and Nolen's study, couldn't was marginally more polite than shouldn't, though of course we have no way of knowing whether the difference was statistically significant. Predictions involving tagged directives are given in H9: H9:
In directives with non-informational force (imperative syntax) plus a question tag modification, the modals will show the following politeness ordering: increasing politeness
150
C H R I S T O P H E R S. B U T L E R
Table 9 shows that all but one of the comparisons support the hypothesis, the only discrepancy being that could and will show no significant differ ence. The finding that will is more polite than can, and would more polite than could, is particularly interesting in view of the fact that these pairs show no significant difference in interrogatives. These findings, though not those on questions, also support Leech's claims about these modals. Fraser and Nolen's findings were compatible with my own, except that their infor mants rated won't above will.
5.0 Conclusion The very high degree to which the hypotheses are supported shows quite clearly that it is possible to correlate the relative politeness and speech act classification of modalised directives with the semantics of mood and modalisation, as defined in terms of systemic choice. The study reported here, though revealing some important properties of the large range of noncongruent realisations of directives, needs to be supplemented by further studies testing the politeness of certain directives in different social con texts, in which factors relating to authority, 'horizontal' social distance and the cost/benefit scale are systematically varied.
NOTES 1.
I shall prefer the term 'directive' to Halliday's 'command' on the grounds that the latter is often taken as equivalent to 'order', whereas 'directive' can be used to cover all kinds of action-seeking acts.
2.
Although this meant that the important and interesting effects of stress and intonation could not be studied, it was clearly essential to control as closely as possible for all vari ables other than the lexicogrammatical features under test. The inclusion of prosodic var iables would, in any case, have given a battery of test items which would have been far too long.
3.
Brackets around a modal indicate that the form was predicted to be unacceptable to some informants. Curly braces are used for sets of modals for which no directional predictions regarding relative politeness were made.
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
151
REFERENCES Antinucci, F. and D. Parisi. 1971. On English modal verbs. Papers from the 7th Regional Meeting, Chicago Linguistic Society. 28-39. Brown, R. and A. Gilman. 1960. The pronouns of power and solidarity. In T.A. Sebeok, ed., Style in language. Cambridge, Mass.: MIT Press. 25376. (Reprinted in P.P Giglioli, ed., Language and social context. Harmondsworth: Penguin, 1972. 252- 82.) Brown, P. and S. Levinson. 1978. Universals in language usage: politeness phenomena. In E.N. Goody, ed., Questions and politeness. Cambridge: Cambridge University Press. 56-324. Butler, C.S. 1982. The directive function of the English modals. Ph.D. Thesis, University of Nottingham. ___ 1986. What has systemic functional linguistics contributed to our understanding of spoken text? Proceedings of the 1984 Working Confer ence on Language in Education, Brisbane. Brisbane: Brisbane College of Advanced Education. . 1985. Statistics in linguistics. Oxford: Basil Blackwell. . 1987. Communicative function and semantics. In R.P. Fawcett and M.A.K. Halliday, eds., New developments in systemic linguistics. Lon don: Frances Pinter. 212-29. Close, R.A. 1975. A reference grammar for students of English. London: Longman. Cole, P. and J.L. Morgan, eds. 1975. Syntax and semantics, 3: Speech acts. New York: Academic Press. Davison, A. 1975. Indirect speech acts and what to do with them. In Cole and Morgan 1975: 143-85. Diver, W. 1964. The modal system of the English verb. Word 20:322-52. Ervin-Tripp, S. 1976. Is Sybil there? The structure of some American Eng lish directives. Language in Society 5:25-66. Forman, D. 1974. The speaker knows best principle, or why complicated facts about indirect speech acts are really obvious facts about questions and declaratives. Papers from the 10th Regional Meeting, Chicago Lin guistic Society. 162-76. Fraser, . 1973. On accounting for illocutionary forces. In S.R. Anderson and P. Kiparsky, eds., A Festschrift for Morris Halle. New York: Holt, Rinehart and Winston. 287-307. . 1975. Hedged performatives. In Cole and Morgan 1975: 187-210.
152
CHRISTOPHER S. BUTLER
Fraser, . and Wm. Nolen. 1981. The association of deference with linguis tic form. International Journal of the Sociology of Language 27:93-109. Goffman, E. 1967. Interaction ritual: essays on face to face behaviour. Gar den City, New York: Doubleday. Green, G. 1973. How to get people to do things with words. In R. Shuy, ed., Some new directions in sociolinguistics. Washington, D.C.: Georgetown University Press. 51-81. Grice, H.P. 1975. Logic and conversation. In Cole and Morgan 1975: 4158. . 1978. Further notes on logic and conversation. In P. Cole, ed., Syn tax and semantics, 9: Pragmatics. New York and London: Academic Press. 113-27. Halliday, M.A.K. 1970a. Functional diversity in language, as seen from a consideration of modality and mood in English. Foundations of Lan guage 6:322-61. . 1970b. A course in spoken English: grammar. London: Oxford Uni versity Press. . 1984. Language as code and language as behaviour: a systemic-func tional interpretation of the nature and ontogenesis of dialogue. In R.P. Fawcett, M.A.K. Halliday, S.M. Lamb, and A. Makkai, eds., The semiotics of culture and language, 1: Language as social semiotic. Lon don: Frances Pinter. 3-35. Heringer, J. 1972. Some grammatical correlates of felicity conditions and presuppositions. Working Papers in Linguistics (Ohio State University) 11:1-110. Hudson, R.A. 1975. The meaning of questions. Language 51: 1-31. Lakoff, R. 1972a. The pragmatics of modality. Papers from the 8th Reg ional Meeting, Chicago Linguistic Society. 229-46. . 1972b. Language in context. Language 48:907-27. . 1973. The logic of politeness; or, minding your P's and Q's. Papers from the 9th Regional Meeting, Chicago Linguistic Society. 292-305. . 1974. What you can do with words: politeness, pragmatics and per formatives. Berkeley Studies in Syntax and Semantics 1. 16:1-55. Lee, 1975. Embedded performatives. Language 51:105- 8. Leech, G.N. 1969. Towards a semantic description of English. London: Longman. . 1971. Meaning and the English verb. London: Longman. . 1980. Language and tact. In G.N. Leech, Explorations in semantics
POLITENESS AND SEMANTICS OF MODALISED DIRECTIVES
153
and pragmatics. Amsterdam: John Benjamins. 79-117. . 1983. Principles of pragmatics. London: Longman. Lodge, K.R. 1974. Modality and modal verbs in English and German. Ph.D. Thesis, University of East Anglia. Lyons, J. 1977. Semantics, Vol. 2. Cambridge: Cambridge University Press. Mitchell, K. 1974. Making sense of English modals. In Annales du Centre Regional de Documentation Pédagogique de Caen, 24- 25 octobre 1973. Institut National de Recherches et de Documentation Pédagogiques. 1324. Mohan, B.A. 1974. Principles, postulates, politeness. Papers from the 10th Regional Meeting, Chicago Linguistic Society. 446-59. Ney, J.W. 1976. The modals in English: a floating semantic feature analysis. Journal of English Linguistics 10:8-20. Palmer, F.R. 1974. The English verb. London: Longman. ___. 1979. Modality and the English modals. London: Longman. Sadock, J.M. 1974. Towards a linguistic theory of speech acts. New York: Academic Press. Searle, J.R. 1975. Indirect speech acts. In Cole and Morgan 1975: 60-82. Twaddell, W.F. 1965. The English verb auxiliaries. 2nd ed. Providence: Brown University Press. Zandvoort, R.W. 1975. A handbook of English grammar. 7th ed. London: Longman.
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
Eirian C. Davies University of London London, England
One of the major contributions which Systemic Grammar has to offer in the study of syntax relates to the concept of multiple components of the gram mar (Halliday 1969; 1970a, b; 1973 et passim). In the area of modality this multi-dimensional approach has interesting connections with distinctions made between epistemic and deontic logic (cf. Von Wright 1957: 58; Haack 1978: 4-8). Halliday's (1970b) discussion of modality and modulation in Eng lish, in which he noted distinctions in potentialities of tense, may be linked with Haack's comments (1978: 195-7) on the relevance of considerations of tense to the development of more satisfactory formal modal logics. In taking 'possibility' as the topic for this paper, I want to confine discus sion to a rather narrow area of modality: that which is neither deontic nor epistemic, but has to do with probability judgements in relation to the occur rence of events.1 I hope to show that an element of 'discourse meaning', relat ing to the textual component, is realized by modal verbs of 'prediction' here, in addition to a component of probability meaning relating to the ideational component. In the course of the discussion I offer a somewhat revised and extended version of earlier proposals (Davies 1979: 139-145) with respect to the latter area of meaning, and suggest that two different kinds of possibility are realized, by MAY and MIGHT respectively.2 I want to begin outside the area of modality by considering, informally, some basic factors underlying everyday, rule of thumb, notions of what makes certain remarks 'relevant' things to say, and others not. Suppose that the London Evening paper I bought had a headline which read (1)
The Government has not resigned.
EIRIAN . DAVIES
156
It is 11th April 1985. There has been no question of Mrs. Thatcher's govern ment resigning. Such a headline would be very difficult to interpret, except as a 'lead-in' to the disclosure of a previously unsuspected political crisis. As a statement of fact, it would be no more, and no less, true today than on any other day since the last general election; but it could nevertheless be expected to be treated as 'hot news'. On a (somewhat) more muted scale, gossip col umns can sometimes be accused of containing examples of what could be cal led 'innuendo by denial'. To deny a proposition not previously entertained as at all likely by the addressee(s) to whom it is made may have the marked prag matic significance of suggesting the relevance of the corresponding affirma tion. I have illustrated, with a negative statement, what I take to be a more gen eral phenomenon of the potentially marked significance of 'stating the obvi ous'. Givón's (1978) illuminating discussion of the greater 'markedness' of negatives, as compared with affirmatives, in terms of discourse-pragmatic presuppositions, is relevant here. He claims that while affirmatives 'are used to convey new information on the presumption of ignorance of the hearer, negatives are used to correct misguided belief on the assumption of the hearer's error'. This approach would certainly do much to explain the signifi cance of (1); but, while negatives may give the clearest cases of such 'correc tion', and may be predominantly found in this use, affirmatives can be used at times to similar effect. That is, what is at issue in what follows, is perhaps better seen as 'contradiction' (discourse negation) which can also be achieved by an affirmation rather than 'denial' (propositional negation) as such. Suppose, for example, I solemnly assure you that (2)
Snow is white.
One quite natural response would be for you to query what might have arisen to make me think otherwise, or to make me think that you (/anyone) thought otherwise. That is, by affirming a proposition regarded as self-evidently true, I am interpreted as having brought forward for attention the possibility of its contradictory. The basis for such an interpretation, I suggest, is a natural tendency to believe that our conversational partners normally abide by the 'conversa tional maxims' as outlined by Grice (1975-1979), or something very like them. Perhaps, then, it is an assumption that what is said is 'relevant' which leads us to consider the contradictory proposition in the case of statements of the obvious, whether affirmations or denials. Statements of the obvious are
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
157
interpreted as not themselves relevant, per se, at least in certain contexts, preeminently those of the news media.3 I suggest that there is a tacit conven tional division in any speech community which underlies judgements of 'rele vance' of the type we have been considering. That is, a given set of proposi tions is regarded as 'known'. The content of this set of known propositions will vary among different sub-groups in society, in different places and at dif ferent times. It may, in fact, be used to define group-membership, both by outsiders and by those belonging. The stock of common knowledge rep resented by this set of propositions establishes which assertions will count as 'statements of the obvious' and which can count as conveying some new infor mation. That is, the concepts of new information and of common knowledge can be seen as mutually defining complementary sets of propositions. The question of which propositions belong to which of these two divisions of the universe of discourse is not a feature of language but of social group, and of time and place. But the division itself, the fact of there being a distinction, can be seen as underlying part of the way in which a natural language works in use. I shall follow Kartunnen and Peters (1979) in referring below to the set of propositions of common knowledge at any given point in a text/linguistic interaction as the Common Ground (CG). I have used the term 'relevance' above, but in the framework of Grice's (1975) analysis, (1) and (2) violate rather his first maxim of quantity than the maxim of relation. The examples he gives of floutings of these maxims could be seen, however, to proceed from a single source, though with different effects. To say (3)
The weather has been quite delightful this summer, hasn't it?
in response to the remark 'Mrs. X is an old bag' at a genteel tea-party, is to impute 'inappropriateness' to the preceding remark by uttering a comment ostentatiously unconnected with it. This involves the introduction of a new topic into the discourse.4 To write (4)
Dear Sir, Mr. X's command of English is excellent, and his atten dance at tutorials has been regular. Yours, etc.
as a testimonial, where what is needed and expected is an opinion on his suita bility to teach philosophy, is 'to fail to answer the (implicit) question relev antly'. That is, such a testimonial would be interpreted as failing to say explicitly what you thought of Mr. X. as a potential philosophy teacher. But to utter (3) in the context given could also be interpreted as a failure to say what you thought (about Mrs. X). The different 'effects' in context surely
158
EIRIAN . DAVIES
have something to do with the discourse feature of whether or not your opin ion on the topic is directly sought. The point I wish to suggest is that failure to 'make your contribution as informative as is required (for the current purposes of the exchange)' is one way, among several, in which a speaker may flout the maxim 'Be relevant'. Example (3), by introducing a topic not 'led up to' in preceding discourse, provides too much of what is new in relation to what has come immediately before; (4) fails to follow on by providing too little of what has been specifi cally prepared for; but both are examples of what might be called 'principled inappropriateness' to immediate needs at the given stage of the transaction. It seems allowable then to think of (discourse) relevance in terms wide enough to include the first maxim of quantity, (so perhaps thinking of the maxim of relation as a 'supermaxim' 5 ). My suggestions so far have run as follows: 1. Given the presence of a group (G) of propositions in the CG opera tive in Context a , it is possible to assess the relevance of uttering a state ment, S, of a proposition p in Ca where either, (a): p is itself a member of G, or (b): p is the contradictory of a member of G. 2. Given that the major purpose of the speech event taking place in Ca is to convey information, there is a prima facie case for saying that it will be non-relevant to utter S if (la) holds true; and relevant (although perhaps not 'appropriate' 6 ) to do so if (lb) is the case. Given, further, that the utterance of a sentence S stating that p in Ca is judged non-rele vant on the grounds of (la), this in itself can sometimes be taken to pro duce a conversational implicature amounting to a proposal for a revision of the CG such that the possibility of the contradictory of p be included in it (if the speaker is assumed to be abiding by the co-operative principle). Grice (1975: 49) distinguishes between the 'violation' and the 'flouting' of a conversational maxim. 'Stating the obvious' must always be the latter, since of its nature it cannot quietly mislead, but there is a proviso here in the form of 'obvious to whom?'. The addressee in cases where he perceives S as a statement of a proposition within the CG has two lines of reasoning open to him (given that he believes that the speaker is abiding by the CP): one is of the kind outlined by Grice and leads to the deduction of one kind of conver sational implicature; but the second leads to a deduction about what the speaker believes to be present in the CG, and thence, more indirectly to a dif-
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
159
ferent kind of implicature.7 That is, the addressee may conclude that if I state that p, I do so on the (normal) basis that I believe he does not already know that p ; that is, I do not believe p to be within the area of common knowledge. The significance accorded to this conclusion will vary with the speaker's status. In the case of (2) above, the conclusion that a newspaper editor had indicated that he did not believe his readers knew that the government had not resigned would carry the suggestion that the matter of its resignation was in doubt. 8 In discussing the CG in this way, we are in the area of 'pragmatic presup position' (cf. Stalnaker 1974: 199-201). Two approaches to the relation between CG and what is actually said appear to be: (i) Given that it is relevant to state 'that p' in Context C, what other proposition(s) (q, (r...)) are required to be present in the CG to make this the case? (ii) Given the pre sence of certain propositions (q,(r...)) in the CG, is it relevant to state 'that p' in context (that context in which the given CG is operative)? Of these the second approach has been less explored, and it is this which is of concern in the present discussion. It raises the fundamental problem of how to assess or 'discover' what is present in the CG at any given stage in a particular speech event (other than by examining preceding 'co-text'), if not by implicit references to the content of CG through what is uttered. One par tial solution to this difficulty would be to consider contexts which themselves clearly supply information, which will be automatically part of the CG for all participants in a speech event taking place in them. (This approach rests, not so much on a type of context, as on a class of features commonly known in a large range of cases such as those to do with location in time and place). In exploring whether certain features of 'meaningfulness' in a sentence should be accounted for in the (grammatical) semantics or within pragmatics it seems possible to argue as follows. We might expect that, if the utterance of a sentence, S, is judged 'bizarre' in context C, this must arise from some kind of 'mis-match' between what S conveys in itself (its meaning) and some feature(s) in C. If the nature of the mis-match is that of contrast (or, in the clearest cases, outright clash) we should be able to infer that the features in what S conveys which are involved in this mis-match must belong to its context-independent meaning, since they contrast with features in C; and that they should be accounted for within the semantics. But what we have been discussing so far, under 'statements of the obvious', are cases where the proposition stated in the utterance of S is iden tical with one present in CG: that is, the inappropriateness of 'total match-
160
EIRIAN . DAVIES
ing'. The point here seems to be that the infelicity involved arises from the violation of a pragmatic principle (a conversational maxim), but that this could no more be perceived without the sentence meaning just what it does than could cases where inappropriateness arises from a 'clash'. I would claim that by no means all 'clashes' give rise to inappropriateness (Davies 1985: 242-6); and one burden of the present discussion so far is that, further, a contrast of propositional content between S and CG is a necessary element for appropriateness in an information-conveying exchange. How ever, clearly not all propositional identity, either between S and CG, or between consecutive or neighbouring sentences within a text (especially if occurring in separate turns), is inappropriate, though it may be redundant. Pragmatic principles can be taken as varying in their application according to the type of context, and, in particular, according to the purpose of the exchange, and the relations of the participants. What is at issue, then, is the operation of some 'lower order' pragmatic rules which give different necessary conditions for the satisfaction of the con versational maxim 'Be relevant' according to different types of language events. In some types, such as those where the main business of the exchange is phatic communion, it may be relevant to utter sentences affirming proposi tions present in CG; that is, it may well be relevant to make informationally redundant utterances. This could be thought of as 'affective relevance': the utterance of S is relevant because of the reassurance of agreement/common assumptions which it conveys. Here again, however, it would be by virtue of the context-independent meaning of S, that its utterance could have that pragmatic effect. The present discussion is narrowly confined to language events in which the over-riding purpose is to convey information. In these, I am suggesting, the applicable pragmatic rules are such that they assign a marked status to redundancy. That is, in this class of events the utterance of a sentence stating an informationally redundant proposition (one present in CG or in preceding co-text) counts as a violation of the maxim 'Be relevant'. So far, we have considered only relations between the propositional con tent of an indicative sentence and the CG. In terms of systemic grammar, propositional content falls under the ideational component. It is a truism that sentences of the kind illustrated (major finite) can have a context-indepen dent ideational meaning. What is more controversial is whether they also have a 'relevance meaning' falling under the textual component. On the basis of the discussion so far, the 'relevance' status assigned to the utterance of S
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
161
in context is accountable for in terms of a relation between the ideational meaning of S and the contents of CG. On this approach, relevance is a prag matic, and not a semantic, phenomenon, since it varies according to contex tual factors. However on Givon's (1978) approach, this is not the case. If it is accepted that negation in indicatives always means 'this sentence corrects a previous error of the hearer', then this is to ascribe a semantic feature of 'con trary to previous supposition' to the grammatical feature of sentence nega tion. That is, Givon's analysis associates an element of discourse meaning with sentence negation: this is equivalent to claiming that the relation between the hearer's knowledge (the set of propositions known to the hearer/ assumed by him to be true) and the propositional content (ideational mean ing) of S is realized in English syntax. Understood in this way, sentence nega tion realises both textual and ideational meaning in systemic grammar terms. It is a device for emending the CG by deletion, or rather for challenging another's view of what belongs to it, or his proposal of what should be added to it. It is, in fact, easy to refute this interpretation of negation in cases where a single proposition is denied successively by two different speakers. That is, in English we can agree by uttering a negative statement as in Speaker A: SpeakerB:
They haven t taken that factor into account yet. No they haven't.
While the first utterance could be thought of as correcting another's misap prehension, the primary significance of the second is to convey agreement. If it 'corrects' it does so only indirectly, by virtue of agreeing with the previous utterance. (This example suggests that the 'other' whose misapprehension is corrected by a negative indicative need not be the addressee.) Not unexpec tedly, it is not negation as such which realizes this feature of discourse mean ing, but negation of a particular proposition. Where this is done twice, the 'correction' does not apply from the second to the first; this element of the meaning of negation is tied to sentences, not utterances of them, and is neu tral to repetitions by different speakers. I have suggested earlier that positive indicative declaratives can also be used to emend CG. Givón's analysis of negation could be taken to imply that the positives should be seen as adding to CG, as opposed to their negative equivalents which are marked as deleting from CG: they both emend it, but in opposing fashions. On this interpretation, the grammatical polarity system with sentence scope realizes a semantic distinction within the textual compo nent. This suggestion in no way conflicts with the accepted view that polarity
EIRIAN . DAVIES
162
realizes a semantic distinction within the ideational component, given a multi-component model. The test cases for this view will consist of examples such as (1) and (2) above. Where the relevant contents of CG can be checked independently of the given S, it is possible to compare what is present there with what is said in uttering S. There can, in principle, be a number of differ ent results as outcome from such a comparison, but those which have con cerned us so far (all restricted to ideational meaning) are: (i)
The propositional content of what is said is identical with a propo sition in CG. (ii) The propositional content of what is said is the contradictory of a proposition in CG. (iii) The propositional content of what is said is not present in CG (and is not directly related to any proposition which is present in CG).
(1) and (2) illustrate (i) in this list. On the basis that the relevance condi tions applicable to an information-conveying exchange are operative, (1) and (2) are redundant and non-relevant in context, by virtue of their ideational meaning. But (1) and (2) are perceived not only as redundant, but also as 'bizarre' in some further sense. On the hypothesis above, this bizarre quality can be seen to result from the fact that in each case the relationship with CG realized in sentence polarity is not that which obtains in context. That is, sen tence negation in (1) marks it syntactically as having the discourse meaning of deleting an element in CG, but that element is not present in CG, and not available for deletion. Positive sentence polarity in (2) realizes the discourse meaning of adding an element to CG, but that element is already present in CG and cannot be added by saying (2). Both (1) and (2) realize a discourse meaning which appears demonstrably false. It is on the basis of the discourse meaning in each case that an addressee will attempt to infer conclusions which preserve, if at all possible, the assump tion that the speaker was abiding by the Co-operative Principle in saying what he did. Given that (1) and (2) are uttered in the course of an information-con veying exchange, the purpose of which could be summarized as progressive emendation of the CG, the addressee has two conclusions he can draw; both have to do with the CG, and are rather unlike the illustrations of conversa tional implicatures given in Grice (1975). In either case the addressee must infer that the speaker's view of what constitutes the CG differs from his own. With a positive sentence his infer ence will be that the speaker believes something to be lacking in CG which he (the addressee) believes to be present, and with a negative sentence he will
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
163
infer that the speaker believes something to be present in CG which he him self had not before thought to be so. From either of these conclusions, and depending on his degree of certainty and estimate of the speaker (and the speaker's sources of information) 9 the addressee will proceed to a judgement as to who is more likely to be right about the fact(s) at issue, himself or the speaker, and will accordingly revise or retain his own view of what is present in CG. 10 The notion that an aspect of discourse meaning (falling under the textual component) is realized in surface grammar by the polarity system, which simultaneously realizes a distinction in ideational meaning, is in harmony with Halliday's general approach. In what follows, I want to explore the ques tion of whether this hypothesis has something to contribute to the analysis of non-deontic MAY in 'predictive' use, relating to future events/states of affairs. Let us begin with MAY by considering cases where the presence/absence in CG of a given possibility or probability is discoverable from context, inde pendently of what is said. I shall adopt a strategy of outlining contexts in impoverished terms, taking only the needed salient features as supplied by them to CG. If we take two locations at opposite extremes with respect to commonly known normal rainfall we can take them as contrasting contexts of utterance. Place A we will say is a tropical rainforest, where it rains on 360 days of the year, and Place let us take as a part of the Sahara desert where it rains, perhaps, on three days in two years on average. To say (5)
It may rain today
in A is odd, that is, apparently unmotivated and redundant; similarly with (6)
It may not rain today
uttered in B. (There is no indication in either A or that the weather on the day in question is going to be atypical.) Let us suppose further that the purpose of the conversation is the same in both cases: to plan what to do in the course of the day in question; and that what can be done depends on whether or not it rains. This topic is therefore, in principle, germane to the progress of the plan-making transaction. If to say (5) in A or (6) in is redundant, we should look for features of meaning in these sentences which are present in the respective CGs. Nondeontic MAY, as here, is generally understood to mean 'possible'; roughly: 'not precluded', 11 but also 'not certain'. If we take the utterance of (5) in A, the CG contains knowledge of a high probability of 'it's raining today', a near
164
EIRIAN . DAVIES
certainty. (5) can be seen as qualifying that degree of near certainty; and it seems intuitively as if the oddness of uttering it in this context results more from the element of 'reservation' which it so conveys than from its being redundant. If so, this oddness has less to do with 'stating the obvious' than with understating it; and more to do with a discrepancy between what is said and what is present in CG than with similarities between them. The analysis I want to propose is as follows: (i) (ii)
(iii)
Part of the meaning of non-deontic MAY is to do with discourse relations and falls under the textual component of the grammar. The meaning feature concerned could be labelled 'contradictory': that is, part of what MAY means in S is 'contradictory of existing CG expectation', The ideational meaning of MAY here is: 'a probability of .5 or more, but of less than 1 (certainty), that the event will (future) occur'.
The expectation supplied from the physical context to CG in A above is that 'it's raining today' is highly probable. The belief that x is highly probable commits the holder to the belief that x is not less than equi-probable; so, in terms of the ideational component alone, (if we accept (iii) as a reasonable analysis of MAY) the use of (5) in A must be seen as expressing a consequ ence of what is held in CG. Though it might be viewed as a 'downtoner', in view of the fact that MAY extends the lower range of envisaged probability to include equiprobability, its ideational meaning still largely overlaps with that supplied by context to CG. In terms of its ideational meaning alone (5) should be perceived in A more as a mild disclaimer conveying some reserva tion in this context than as the bizarre remark which it appears to be. Com pare in this respect (7)
It should rain today
uttered in the same context, which although more clearly redundant (full overlap of probability with CG) seems somewhat less, rather than more, strange. If we take MAY in (5) to realize, in addition to the ideational meaning proposed in (iii), a feature of discourse meaning, 'contradictory', contradic tory of CG expectation, we can analyse as follows: Present in CG in A 'Highly probable that it will rain today'/Probability of
(5) It may rain today Discourse meaning: 'contradictory of CG' Ideational meaning: A probability of
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
nearly 1.
165
.5 or more (but of less than 1). Combination of meanings: Denial of a background CG probability of less than half.
That is, if the ideational meaning is a probability of 'greater than or equal to .5 (but less than 1)' and the discourse meaning is 'contradictory of CG', the combination of these kinds of meaning gives (5) as denying an element in CG which is not present in the CG in Context A, namely that of an estimated probability of less than .5, but greater than 0, 'unlikely'. So to utter (5) in A is to attempt to revise CG by deleting an element which is not present in the CG in Context A: namely that of an estimated of (2) in the context sketched for it above. 12 The distinction claimed between uttering (5) and (7) in A, can be related to the feature of discourse meaning 'contradictory of CG'. I suggest that this is realized by MAY, but not by SHOULD. SHOULD, though it excludes almost the same area of probability as MAY, by virtue of its ideational mean ing, does not realize an explicit contrast with CG. In this way it conveys no sense that what it ideationally excludes is present in CG; and the use of (7) in A does not produce as sharp a clash with context-supplied common knowl edge as that of (5). SHOULD operates in discourse as a device for adding to CG, as in the case of positive indicatives, rather than as a deletion device, as in the case of negative indicatives and MAY. The use of (7) would, on this analysis, be unmarked in cases where there was a lack of any probability expectation in relation to 'its raining' , rather than in cases where there was some expectation of 'probably not' (a probability of less than .5, but greater than 0) as for the unmarked use of (5). The use of (7) in A would be 'marked' as far as expecta tion concerning the probability of 'its raining' is operative in CG; but its use would not convey that the speaker was countering an assumption in CG that rain was unlikely, but would merely suggest that the chances of 'its raining' had not been previously at issue (and were not common knowledge). The proposed analysis can be summarized informally as follows: (i) (ii) (iii)
MAY 'discourse means' that the opposite of its own ideational meaning is present in CG. SHOULD 'discourse means' that its own ideational meaning is lacking from CG. The utterance of a sentence containing MAY acts to delete the opposite of the ideational meaning of MAY from CG.
EIRIAN . DAVIES
166 (iv)
The utterance of a sentence containing SHOULD acts to add the ideational meaning of SHOULD to CG.
I will examine 'own ideational meaning' and the 'opposite' of it in greater detail below. The analysis predicts that to utter (6) in Context will be as bizarre as to utter (5) in A, but that the use of (6) in A, and of (5) in would be unmarked. So far we have restricted discussion to artificial contexts and marked uses of predictive MAY as helping to throw into sharper relief the presence of the claimed discourse element in its meaning. In the nature of things, such instances are unlikely to be found in texts, and it seems a case where appeal to native speaker intuition and constructed examples is justified and useful. Taking the interpretation of MAY so derived, I turn now to some examples of unmarked usage in texts, specifically the racing columns of two 'serious' British daily newspapers (The Daily Telegraph and The Times). (8) (9)
Bill Holden does well with his small string, and his Love Walked In may take the Tonbridge Handicap. 13 However, he shaped like a sprinter on several occasions and, being a son of Mummy's Pet, may not be at his best over seven furlongs on testing ground.14
MAY can be contrasted with SHOULD in racing predictions, as in (10) In the absence of Sarab and Provideo the veteran Vorvados should win the Abernant Stakes. 15 which is a more confident forecast than (8). In horse racing, as in any activity where there is only one winner of a given event, the unweighted statistical chance of any particular individual coming first is a probability of 1 divided by the number of contestants. 16 While I am not suggesting that the ordinary reader is in any way consciously aware of any mathematical calculation of this kind,17 it would I think be fair to say that the common ground shared by writers and readers of racing 'tips' includes the notion that, if there are no specific grounds for thinking otherwise (such as previous performance etc.) the ordinary chances of winning for any particular individual are less than 50 : 50 (in a normal field of more than two entrants). That is, the CG contains an underlying expectation that for any individual there is less than an equal possibility of his winning. On this basis, to assert (8) in the racing pages of the press, can be seen as entirely natural and unmarked in terms of the interpretation given to MAY
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
167
in (5) above. That is, (8) in its second clause, contradicts an underlying assumption of less than even odds that this particular horse will win the race in question. This is why it's worth including (8) in the column. The case is similar for (9). Here the preceding text has discussed three horses which on previous performance and other grounds seem capable of winning the Burton Overy Stakes. This in turn is treated as an indication of likely success in more important races later in the season, the Newmarket classics. The horse refer red to in (9) is 'Rotherfield Greys', but the article has the heading 'Charge Along's chance to put colts in their place'. The CG here, as created through co-text, contains some expectation of Rotherfield Greys' winning and gives grounds for this; he has 'run consistently in top company last season' and came third in a much more important race than the present one then; he was 'a credit to his trainer'. (9) as a whole counters the positive expectation thus previously suggested in co-text, the contrast being marked by an initial adver sative adjunct, however. It gives grounds for predicting the opposite in its first two clauses, and the discourse contradiction itself through MAY NOT in its last. ('Seven furlongs on testing ground' is a description of the course over which the race in question will be run.) In this way MAY(NOT) here can be interpreted along parallel lines to those given for MAY in (5), as contradict ing the expectation in CG. The negative form MAY NOT is taken as realizing an assumption of a probability of more than .5 in CG, and its use acts to delete this, and to replace it with its own ideational meaning (of equal to or less than .5, but greater than 0). The positive probability, having been suggested in cotext, is available for deletion, and the use of MAY NOT in (9) is unmarked. (10), on this analysis, represents unmarked usage if the chances of the veteran Vorvados winning had not previously been of much interest (before the more likely winners Sarab and Provideo were withdrawn). I want now to consider the area of ideational meaning with respect to probability MAY and SHOULD in more detail. MAY is generally taken to realize a semantic feature: 'possible'; but there seem to be two partially conflicting senses in which this term can be commonly used: (i)
(ii)
as in possible world semantics, where what is actual is also, ipso facto, possible: 'possible A'. On this basis, since what is necessary is actual, what is necessary is also 'possible A'. as often used informally, where what is possible is part of what is non-actual: that part which is not impossible: 'possible B'. On this basis, what is necessary is not 'possible B', since what is necessary
168
EIRIAN . DAVIES
is actual and what is actual is other than what is merely 'possible B'. With respect to future events the feature 'actual' cannot apply: but phys ical necessity is not excluded. A sentence such as (11) All men will die can be said to illustrate its expression. Here what is claimed is an 'inevitablity', which contrasts with 'possibility'. We could perhaps say of WILL in (11) that it realizes a probability of 1. On this basis we would then say of WON'T that it realizes a probability of zero. The denial of WILL in these terms can be seen as 'not-inevitable: a probablity of less than ; and the denial of WON'T, 'not inevitable that not: a probablity of greater than zero'. I take the denial of WILL ('not-will') to be realized by MIGHTN'T, and the denial of WON'T ('not-won't') to be realized by MIGHT. This gives MIGHT(N'T) as realizing a discourse meaning of 'contrastive', in parallel with MAY (NOT). The ideational meaning ascribed to MIGHT in this way is a probability of 'greater than zero' ; and that to MIGHTN'T is a probability of 'less than 1'. Hence the ideational meanings of MIGHT and WILL overlap on that of WILL (a probability of 1); and those of MIGHTN'T and W O N T on that of WON'T (a probability of zero). If MIGHT(N'T) is said to realize 'possibility', this is 'possibility A', for the parallel to the 'necessity' term, the absolute WILL(NOT) is included within it. The conjunction of MIGHT and MIGHTN'T on this analysis gives: 'a probability of more than zero and less than . This is the area of 'possibility B', which excludes the absolute terms WILL(NOT). I suggest that MAY(NOT) and SHOULD(N'T) realize ranges of probabilities within this area, based on the relations 'greater than', 'equal to', 'less than', as follows: SCHEMA I: Probabilities greater than 0 and less than 1 MAY: SHOULD: MAY NOT: SHOULD NOT:
'greater than or equal to .5' (gE .5) 'greater than .5' (gt .5) 'less than or equal to .5'. (1E .5) 'less than .5 (1t .5)
Put informally, MAY realizes a probability of 'even odds' or higher (stopping short of certainty/'inevitability'); SHOULD realizes more than 'even odds' (also stopping short of inevitability). The negatives can be seen either as giv ing the same degree of probability for a negative proposition/non-occur rence, or, as above, involving 'less than even odds' for positives.
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
169
I am not suggesting that native speakers of English perform calculations giving decimal point probability ratings when using modal verbs, but that the relational notions concerned are operative in the ideational semantics of those forms. Schema I indicates that the combination of SHOULD and SHOULD NOT, jointly applied to a single predication, results in a self-contradiction: (12) Vorvados SHOULD and SHOULDN'T win whereas the combination of MAY and MAY NOT does not have the same effect (12') Vorvados MAY and MAY NOT win but realizes equiprobability. In terms of Schema I, MAY and SHOULD NOT, MAY NOT and SHOULD are mutually exclusive pairs. We can show this in the following sys temic diagram (adding mnemonics): Diagram 1.0 Ideational semantics of probability modals Probable' (Probability gt .5) SHOULD Not probable' (probability 1E .5) MAY NOT
•Improbable' (Probability 1t .5) SHOULD NOT Not improbable' (Probability gE .5) MAY
If we display the analysis of oppositions of discourse meaning given ear lier for MAY and SHOULD this yields: Diagram 1.1 Textual semantics of probability modals Additive' SHOULD
Deletive' MAY
The most natural presentation of oppositions in linguistic form can be shown as:
170
EIRIAN . DAVIES Diagram 1.2 Surface form of probability modals
Diagram 1.2 also reflects a conflation of diagrams 1.0 and 1.1, as can be seen by mapping them both on to it, to give: Diagram 1.3 Forms and combined ideational and textual meanings of probability modals
It seems, then, that it is a combination of ideational and textual meaning which is realized in linguistic form. The analysis earlier proposed for the ideational meaning of WILL and MIGHT can be set out as follows: SCHEMA II: Probabilities of 1, 0, GT 0 and LT 1 WILL: WONT: MIGHT: MIGHTN'T:
probability of 1 ('inevitable that positive') probability of 0 ('inevitable that negative') probability of Greater than 0 ('not (inevitable that negative)') probability of Less than 1 ('not (inevitable that positive)').
(Note: the following convention is used in abbreviations: UPPER CASE (GT/GE:LT/LE): 'Up to and Including a probabil ity of 1/0'.
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
171
lower case (gt/gE; 1t/IE): 'excluding a probability of 1/0'. 'Equal to' is shown as 'E' in all cases, since it is used only where 'Equiprobability; (a probability of .5) is included in the range given.) In ideational meaning, WILL and MIGHT are not mutually exclusive; and so also for WON'T and MIGHTN'T. The analysis predicts that, jointly applied to the same propositional content within a single sentence, the fol lowing combinations of items will have zero overlap in the probabilities they realize and will produce a self-contradiction: WILL and MIGHTN'T; WON'T and MIGHT (contradictories); WILL and WON'T (contraries). In the remaining pairs there is some overlap between probabilities realized, and the sentence resulting from their conjunction is not self-contradictory: MIGHT and WILL; MIGHTN'T and WON'T; MIGHT and MIGHTN'T. More specifically, MIGHT and WILL overlap on a probability of 1, MIGHTN'T and WON'T on a probability of 0. The area of overlap for MIGHT and MIGHT NOT is the whole range of probabilities less than 1 and greater than 0: that is, it is the area divided up in SCHEMA I between MAY(NOT) and SHOULD (N'T). This analysis can be shown diagrammatically as follows: Diagram 2.0 Ideational semantics Tnevitable-that-positive' (Probability of 1) WILL Not(inevitable-that-positive)' (Probability of Less than 1) MIGHTN'T
'Tnevitable-that-negative' (Probability of 0) WON'T 'Not (inevitable-that-negative)' (Probability of Greater Than 0) MIGHT As with Diagram 1.0, Diagram 2.0 differs from that which would be most natural for displaying distinctions of surface form, namely:
172
EIRIAN . DAVIES
Diagram 2.1 Surface form WILL WONT MIGHT MIGHTN'T We have taken MIGHTN'T as realizing the denial of WILL ('not-will'), and so for MIGHT and W O N T . In terms of ideational meaning this 'denial' has been shown as 'other than': MIGHTN'T realizes the full range of prob abilities other than 1 ; MIGHT the full range of probabilities other than 0. But I want to suggest that an element of discourse meaning also enters into this contrast. If we return to the artificially extreme context of the rainforest on a day following a run of five consecutive dry days, so that the CG contains knowledge of a probability of 1: 'it will rain today', it does not make sense for me to say (13) It might rain today even though ideationally the probability range realized by MIGHT includes the probability realized by WILL. MIGHT can be seen as an item, the dis course meaning of which is to delete a probability of 0 from CG. To use it where a probability of 1 is held in CG makes no 'discourse sense'. An appro priate challenge to the utterance of (13) in the circumstances sketched would be: 'What makes you think it won't?' On this approach MIGHT(N'T) parallels MAY(NOT) in its discourse meaning: both are 'deletive'. There is also a similarity in ideational mean ing in that both include equiprobability (.5) within the probability ranges which they realize. They contrast here with WILL(NOT) and SHOULD(N'T) which both exclude equiprobability. SHOULD(N'T) has been treated as 'additive' (to CG) under discourse meaning. It remains to explore whether this feature also applies to WILL(NOT) as symmetry would demand. In fact it seems reasonable to treat WILL(NOT) as additive. The utter ance of
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
173
(14) Water will boil at 100° Celsius or (15) Asbestos won't burn would be unmarked in a context which was empty of probability expectation with respect to their propositional contents. If they are thought to deny MIGHTN'T (boil), MIGHT (burn) respectively, this is in terms of ideational meaning only. In the same way, the utterance of any positive indicative declarative constitutes a rejection of its negative equivalent. To say (16) It will rain today in the rain forest context sketched for MIGHT above is merely redundant: it does not convey that the speaker assumed a probability of Less Than 1 in CG, and so was countering a (in this case non-existent) misconception. On this basis, the distinction in surface form between the items WILL and MIGHT appears to realize a difference in discourse meaning (as for SHOULD and MAY in Diagrams 1.1, 1.3). If we combine the two ideational analyses we arrive at the following synthesis: SCHEMA III Probability of 1: Less Than 1: less than 1 and greater than .5: less than 1 and greater than or Equal to .5 Equal to or less than .5 and greater than 0 less than .5 and greater than 0: Greater Than 0: 0: One way of displaying this analysis could be as follows:
WILL MIGHTN'T SHOULD MAY MAY NOT SHOULDN'T MIGHT WONT
174
EIRIAN . DAVIES Diagram 3.0 Ideational semantics: probabilities
SHOULD (less than 1 and greater than .5) MAY NOT (greater than 0 and less than or Equal to .5) SHOULDN'T (greater than 0 and less than .5) MAY (less than 1 and greater than or Equal to .5)
This diagram does not, however, indicate that the probability range realized by MIGHT intersects with that of WILL (as also for MIGHTN'T and WONT). The analysis is more satisfactorily represented as follows:
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
175
Diagram 3.0.1 Ideational semantics of probability modals
In this case the network fragment for ideational meaning itself reflects the distinctions between items in surface form without reference to textual meaning. This is because an ideational distinction, 'exclusive/inclusive of equiprobability' co-varies with the discourse distinction 'additive/deletive'. We need to examine the notion of discourse meaning more closely. In discussing Givón's (1978) analysis of negation in English the point was made that it does not hold for other than the first of more than one consecutive (or sequentially close) denials of a single proposition in a limited section of text. Negative indicatives can be used to express agreement in discourse. When this is done, they are often elliptical, but this is not necessarily the case. I take the proposed discourse meaning of probability modals to be exactly the same
176
EIRIAN . DAVIES
with respect to this limitation. Everything suggested above or below on this topic refers to their occurrence in sentences whose utterance constitutes a 'first mention' in this sense. Any probability modal may be used to express agreement, or disagreement with another participant in a conversation if 'first mention' is not involved.18 Where more than one probability modal applies to the same propositional content within one sentence I take dis course meaning as used here to apply only to the first.19 So far, in exploring discourse meaning, we have derived a grouping of MAY(NOT) and MIGHT(N'T) as 'deletive', and proposed that these items be seen as devices for countering a 'current error' in CG, linking them in this respect with the analysis of indicative negative declaratives. The 'error' coun tered by an instance of this last category is given in its corresponding positive. But this is not so for MAY(NOT) and MIGHT(NT) (which I shall refer to for convenience below as the 'M group' of probability modals). I have earlier suggested that MAY is a device for deleting 'Improbable' from CG; and that its use where CG clearly contains 'Probable' is bizarre for this reason. If this is accepted, we can note that 'Improbable' represents the complementary 'probability range' within the area 'greater than 0 and less than : that is, the area in which 'possibility B' applies. We can then propose that, in a parallel way, MAY NOT deletes 'Probable' in CG. If we extend the analysis along these lines to MIGHT(N'T), the range of probabilities con cerned includes 1 and 0, and these limiting terms constitute the complemen tary values. The analysis can be set out as follows: SCHEMA IV M group MAY MAY NOT MIGHT MIGHTN'T
Delete in CG It
.5 ('shouldn't') gt .5 ('should') 0 ('won't') 1 ('will')
This analysis predicts that if MAY is used in a context where the proba bility held in CG is 0, it would be perceived as, to some extent, surprising. For example, to say (17) He may live for another twenty years of a centenarian is to make a more perplexing statement than if MIGHT were used, even though the latter includes within its range the probability value of
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
177
1. To say (17) sounds as if the speaker knows something that the rest of us do not, as if his view of CG is different (that it contains 'shouldn't' here rather than 'won't'). MAY NOT and MIGHTN'T can be analysed similarly in rela tion to a probability of 1 in CG. In a parallel fashion, to use MIGHTN'T where CG holds a probability of 'gt .5' (rather than 1) is also to some degree marked, as to use MIGHT where CG holds 'It .5' (rather than 0). For exam ple to say (18) Shadeed mightn't win the Derby suggests that the speaker believes the general view is that Shadeed is certain to win. MAY NOT in (18), as the form for deleting 'should' rather than 'will' conveys an assumption of less universal confidence here than MIGHN'T. In a context where 'should' is all that is present in CG, (18) is a surprising remark, a marked use. Similarly, to say (19) It might snow tomorrow (on June 1st in London) is more reasonable than to say 'It MAY snow tomor row'. On the present analysis, this is not so because MIGHT is 'weaker/more doubtful' ideationally; its probability range includes 1, which that of MAY does not. Rather, it is because MIGHT acts to delete 'won't' as opposed to 'shouldn't' from CG; and to use it is therefore to indicate an assumption that 'won't' is what CG contains. This is a more reasonable assumption in this case than the 'shouldn't' which would be indicated as believed present in CG by the use of MAY in (19). The analysis predicts that the often noted 'greater doubtfulness' of MIGHT, as compared with MAY, is not a feature of its ideational meaning (the probability range which it realizes itself), but results from the probability range which it is its discourse meaning to delete from CG. If the discourse meaning of the M group can be given for each partly as 'that which it deletes from CG', we should now consider how the discourse meaning of the 'additive' group can be characterized. Here I have suggested that SHOULD(N'T) and WILL(NOT) operate against a neutral background in CG, a lack of any probability-expectation. If so, it would seem that they can be characterized only in terms of the ideational meaning which they themselves each realize. But, as with positive declaratives, the utterance of a sentence containing one of these modals functions to exclude what is other than the ideational meaning realized. The unmarked use of the additive group is said to occur where CG is neu tral. This suggests that where CG contains any probability expectation relat-
178
EIRIAN . DAVIES
ing to the proposition to which they apply their use will be marked. So, the analysis predicts that to use any of this group in (19) would produce a strange effect. It seems self-evident that not only is it strange to say, in London on June 1st, (20) It should snow tomorrow or (21) It will snow tomorrow but that this can be fully accounted for by the clash of the ideational meanings concerned with our knowledge of the world (i.e., the contents of CG). If this is accepted, we can go on to explain the oddness of (22) It shouldn't snow tomorrow and (23) It won't snow tomorrow uttered in the same context, in terms of the redundancy of their ideational meanings. In (22) and (23) we could phrase this as 'strangeness' resulting from attempting to add something to CG which is already present there, and compare it with the earlier analysis of (2)Snow is white. In the first pair, (20) and (21), we could say that the strangeness results from attempting to add to CG a probablility value which is specifically excluded from it by virtue of being the contrary of what it already holds. In both cases the strangeness arises from the fact that CG contains a probability expectation (0) in relation to this proposition. If, on the other hand, someone said, in the same context (24) It should be fine on Friday or (25) It will rain tomorrow neither would usually be strange in this way because, given the unpredictabil ity of English weather, there is unlikely to be any agreed probability expecta tion in relation to 'its being fine on Friday'/'its raining tomorrow' in CG. The same analysis applies here to SHOULDN'T and WON'T. I am arguing, perhaps, on a fine point. But what I want to suggest is that (24) and (25) are unmarked in the context not because they more accurately reflect the CG, by virtue of their ideational meanings, but because they genuinely add to it here as it is their discourse meaning to do. The foregoing discussion can be summarized as follows:
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH (i) (ii) (iii)
(iv)
179
All probability modals discussed here have a discourse meaning of changing CG. They do so in opposite ways: The M group are deletive; the others additive. The discourse meaning of the M group is: (a) that they are deletive and (b) for each modal, that which it deletes (the opposite of its own ideational meaning). The discourse meaning of the remainder is: (a) that they are addi tive and (b) for each modal, that which it adds (its own ideational meaning).
From this analysis we can derive the following network fragment for dis course meaning in this area: Diagram 3.1 Textual meaning
This can be more satisfactorily shown as follows:
180
E I R I A N . D A V I E S
Diagram .1. WILL WONT MIGHT MIGHTN'T
SHOULD SHOULDN'T
MAY MAY NOT
Diagram 3.0.1 gives two types of possibility in terms of what MAY/ MIGHT 'contribute': their ideational meaning whenever used, whether in first mention or not. Diagram 3.1.1 gives two types of possibility in terms of what these two verbs delete from CG when applied to the 'first mention' of a proposition p held there with some probability rating. This type of textual meaning is not operative in 'second mention': so that ideational meaning cannot be said to determine textual meaning in terms of whether it is opera tive or not. Further, ideational meaning in a given instance may be 'consonant' (in cluded in or overlapping) with CG, while the discourse meaning concerned may be at variance (as in the case of MIGHTN'T in (18) if CG probability was less than 1). In ideational meaning, MIGHT realizes a wider range of probability val ues (GT 0) than MAY (gE .5), and could in this respect be said to express more 'doubt'/'uncertainty' in the sense of being less specific. On the other hand, while it allows a lesser probability value (It .5) than MAY, it also allows a higher (1). The greater 'tentativeness' which MIGHT is generally held to communicate in unmarked first mention use can be more fully accounted for by invoking the area of discourse meaning within the semantics of the textual component, which shows it to assume a zero probability in CG as available for deletion. I have argued for the semantic, as opposed to pragmatic, status of the
ON D I F F E R E N T POSSIBILITIES IN T H E SYNTAX O F E N G L I S H
181
'discourse meaning' attributed to the set of modal verbs discussed. If this is so, what is involved is not a matter of semantic presupposition. In the case of the M group, what is taken to be 'assumed' in CG is itself a modality, and its assessment for truth value is philosophically problematical. We have taken a different question: namely that of whether it is true that the 'assumed' prob ability is present in CG. It is on this (pragmatic) factor that the (marked/ unmarked) pragmatic significance of using an M modal is said to depend, by virtue of its having an invariant discourse meaning (the textual semantic fea ture 'deletive of (v)' where 'v' is a probability value). In the narrow area of the grammar discussed, distinctions within textual and ideational meaning yield the same set of surface grammar items. This obscures the claimed degree of independence between the systems of seman tic features realized. But a multi-component approach, such as that funda mental to Systemic Grammar, can both account for such phenomena and encourage further investigation of problems to do with the borderline between pragmatics and the semantics of grammar.
NOTES 1.
Cf., independently, Davies (1979: 141, 156), Palmer (1979: 3-4) on an area of non-deontic modality distinguished from that of epistemic modality by virtue of applying to events as opposed to propositions. This distinction is not uncontroversial. It is not made in Lyons (1977), and is denied in Coates (1983). The area concerned would fall principally under 'modulation, in Halliday (1970b) on syntactic grounds. For the purposes of the present paper, I wish to use it as a restrictive device, to limit the range of the discussion. An integ rated system of contrasts appears to operate within it.
2.
I retain the suggestion made there that the area of modality concerned belongs principally under that area of the grammatical semantics to do with ideational meaning ('the perfor mance plane'). 'Ability' CAN was also included there under this heading, but distinguished from a set of 'prediction' modals of which only WILL and MAY were discussed in any detail. An attempt to integrate SHOULD and MIGHT is made in the present account. (OUGHT TO is taken as equivalent, in terms of ideational meaning, to SHOULD).
3.
As a special case of language events in which the main purpose is to convey information. In other types of event with different purposes, 'statements of the obvious' may be perceived as 'relevant', as in phatic communion.
4.
In this particular instance, the remark could also be thought of as a statement of the obvi ous; but this feature is not necessary to the effect produced. Anything 'unconnected' would do, such as 'John and I are going to Portugal this year' etc. There is, then, a distinction between new information (defined as the statement of propositions not present in the CG) and 'new topic' ('not connected with preceding remarks').
5.
Grice suggests a close connection between the second maxim of quantity ('Do not give more
182
E I R I A N . D A V I E S information than is required') and the maxim of Relation (1975: 52-3); but it is not clear to me why 'too much information' should be treated as different in kind from 'too little infor mation' in this respect.
6.
Other measures can enter into the assessment of 'appropriateness' such as 'politeness' etc.
7.
This second kind of implicature would be equally 'conversational', but based on different premises, namely ones about who knows what. For example, depending on the status of the speaker, and the addressee's independent access to reliable information, the addressee may conclude that the speaker knows (and believes others to know) something which he (the addressee) does not; and he may then add this to what he accepts as CG. Or, he may con clude that the speaker does not know something which he (the addressee) does know with certainty, and knows others to know; and from this he may infer conclusions about the speaker's information, state of mind, etc. Or he may conclude that the speaker thinks that he (the addressee) does not know something which he does in fact know, and which he knows that most others know; and from this he may infer conclusions about the speaker's opinion of his knowledge. And so on; the possible permutations are quite wide, depending partly on a further distinction which can be usefully drawn between a direct addressee, and a 'hearer/reader' who may not include himself, or be included by the speaker, amongst those directly addressed.
8.
In this case, the newspaper reader's reasoning might go as follows: (i) (ii) (iii) (iv) (v)
The purpose of this headline is to convey information; What it says has been part of a nation-wide CG since the last general election; In terms of this known CG the headline writer is stating the obvious and flouting the maxim 'Be relevant'; but, I accept that he is abiding by the cooperative principle; therefore he must be assuming that the contents of CG have altered with respect to the proposition at issue, and that CG now holds its contradictory (otherwise to state that previously held proposition would not qualify as giving 'news', and I accept that this is what he believes himself to be doing).
9.
Other factors are also relevant here, such as what the addressee would like to believe, and so on.
10.
As in fn. 8, he may also revise his view of the speaker as a reliable source of information, or of himself as generally in touch with events. The brief and partial sketch indicated here, represents a very slight incursion into a wide field within pragmatics. One major point is that 'CG' is Common Ground, which raises the important question of 'Common to whom': an 'in-group', a social class, an entire speech community or culture, merely the speakeraddressee pair, and so on. The range of inferences to be drawn is correspondingly varied.
11.
Cf. Perkins (1983).
12.
Substituting 'expectation' as that which is 'corrected'/deleted from CG.
13.
'The Daily Telegraph' 10th April 1985.
14.
'The Times' 25th March 1985.
15.
'Hotspur' in 'The Daily Telegraph' 17th April 1985.
16.
Skyrms(1966).
17.
Where the final size of the field was not yet known it would not in any case be possible to make the calculation.
ON DIFFERENT POSSIBILITIES IN THE SYNTAX OF ENGLISH
183
18.
The question of restrictions in this area, and of differences both in types and degrees of 'op position' between the members of different pairs of probability modals, is of considerable interest. So also is that of allowable combinations of them within, as opposed to across, sen tence boundaries. Links across and within 'turns' are also significant. I discuss some of these issues in a paper currently in preparation.
19.
John may pass his exams easily, and should do so would be equivalent for these purposes to John may, and should, pass his exams easily. In either case, it is only MAY in these exam ples which would be analysed in terms of discourse meaning.
REFERENCES Coates, J. 1983. The semantics of the modal auxiliaries. London: Croom Helm. Davies, E.C. 1979. On the semantics of syntax: mood and condition in Eng lish. London: Croom Helm. . 1985. On types of meaningfulness in discourse. In J.D. Benson and W.S. Greaves, eds., Systemic perspectives on discourse. Vol 1: Selected theoretical papers from the 9th International Systemic Workship. Nor wood, New Jersey: Ablex. 229-247. Givón, T. 1978. Negation in language: pragmatics, function, ontology. In P. Cole, ed., Syntax and semantics 9: pragmatics. New York: Academic Press. 69-112. Grice, P. 1975. Logic and conversation. In P. Cole and J.C. Morgan, eds., Syntax and semantics 3: speech acts. New York: Academic Press. 41-58. Haack, S. 1978. Philosophy of logics. London: Cambridge University Press. Halliday, M.A.K. 1969. Options and functions in the English clause. Brno Studies in English 8:82-88. (Reprinted in Halliday and Martin 1981.) . 1970a. Language structure and language function. In John Lyons, ed., New horizons in linguistics. Harmondsworth: Penguin. 140-165. . 1970b. Functional diversity in language as seen from a consideration of modality and mood in English. Foundations of Language 6:322-361. (Partly reprinted in Kress 1976: 189-213.) . 1973. Explorations in the functions of language. London: Edward Arnold. Halliday, M.A.K. and J.R. Martin, eds. 1981. Readings in systemic linguis tics. London: Batsford. Karttunen, L. and S. Peters. 1979. Conventional implicature. In C-K Oh and D.A. Dineen, eds., Syntax and semantics 11: presupposition. New
184
EIRIAN . DAVIES
York: Academic Press. 1-56. Kress, G., ed. 1976. Halliday: system and function in language. London: Oxford University Press. Lyons, J. 1977. Semantics Vol. 2. London: Cambridge University Press. Palmer, F.R. 1979. Modality and the English modals. London: Longman. Perkins, M.R. 1983. Modal expressions in English. London: Frances Pinter. Skyrms, B.L. 1966. Choice and chance: an introduction to inductive logic. Belmont, California: Dickenson. Stalnaker, R. 1974. Pragmatic presuppositions. In M.K. Munitz and P. Unger, eds. Semantics and philosophy. New York: New York Univer sity Press. 197-213. von Wright, G.H. 1957. Logical studies. London: Routledge and Kegan Paul. Waterlow, S. 1982. Passage and possibility: a study of Aristotle's modal concepts. Oxford: Clarendon Press.
THE ENGLISH PERSONAL PRONOUNS: AN EXERCISE IN LINGUISTIC THEORY
Robin P. Fawcett University of Wales Cardiff, U.K.
1. The Problem The aim of this paper is to explore, in relation to a central and yet rela tively self-contained part of the grammar of English, an issue of theoretical importance in systemic theory: the relation between the levels of FORM and SEMANTICS. 1 It is also an issue in certain other theories, though the different basic assumptions about the nature of language that each theory makes have resulted in their expressing it in different terms. The paper might therefore appear to be addressed primarily to those concerned with the development of systemic theory, and only secondarily to those working in other theoretical frameworks. But I hope that the paper will offer more than this to the non-systemicist. It should also illustrate, in a small way, the value of the systemic approach both in raising questions about features at the semantic level and in capturing relationships between them. For example, Leech 1969: 104 and Leech 1981: 112 both use system networks —indeed, the differences between the two raise interesting ques tions. There is a section heading in Lyons' Introduction to theoretical lin guistics (1968: 413) which reads: '"Having meaning" implies choice'. Sys tem networks (and their derived equivalents in stratificational theory) offer the most comprehensive notation so far developed for modelling paradig matic relationships, and I would like to suggest that they could usefully be in wider use than they are at present. I therefore hope that this paper may encourage those who have not done so to try out this notation. And for those who are already familiar with it, the paper may nonetheless contain one or two surprises.
186
ROBIN P. FAWCETT
Our line of approach will be to examine several alternative ways of modelling the same data (the English personal pronouns), introducing as we go some central concepts in systemic theory. You might ask: 'Why choose such a simple set of data?' The answer is that these relatively uncontroversial data are intended to make it possible for us to focus relatively MORE of our attention on the theoretical issues involved, and relatively LESS on the interpretation and description of the data. Even with these relatively simple data, however, we shall find ourselves in the position familiar to any linguist — and indeed to any scientist — that when one looks closely at even the most apparently straightforward data, one finds more questions arising than one had expected. The theoretical question that is to be examined here is the following: in the framework of an explicit model of language, at what LEVEL do the systemically organised FEATURES which are typically recognized in systemic grammars belong? Do these networks of features belong at the level of FORM (i.e. 'syntax'/'grammar' and lexis') or do they model relations between features that are SEMANTIC? Or, as Hudson's writings some times suggest, do they belong somewhere in between? Or, finally, are there TWO levels of contrasts here: a higher, semantic one and a lower, formal one? (In the alternative metaphor used in the Chomskyan framework, the question would be expressed in terms of 'how deep' the networks of choices between features are.) For further discussions of this issue see Fawcett 1980: 39-46, Fawcett 1983: 111-21, Fawcett 1987, and the Introduction to Halliday and Fawcett 1987. The answer to these questions might be that it depends on the inten tions of the linguist who constructs the system network. But it seems to me that, despite the fact that different systemic linguists may DESCRIBE their networks in rather different terms, the networks themselves are patently intended to capture the same types of insights as each other. I am referring to networks such as those that are often labelled 'transitivity', 'theme', etc., offered in Halliday 1968: 201ff., 1973:40 and 1977: 208ff., Hudson 1971: 71 (reproduced in McCord, 1975: 199) and 1973: 544; Fawcett 1973/81: 155ff., 1980 and forthcoming; and Berry (1975: 189, 190). Despite their differ ences, these are all what Martin 1987 has termed 'first level' networks (as opposed to the higher (or deeper) 'socio-semantic' networks proposed in Halliday 1973: 89 and Turner 1973: 153ff., which Martin terms 'second level' networks). The details of the various first level networks differ in var ious minor ways that are not significant for our present purposes. What is significant is the great difference between the ways in which those responsi-
THE ENGLISH PERSONAL PRONOUNS
187
ble for these networks describe them. Sometimes the features in the net works have no necessary connection with any meaning, while it is claimed of others that they ARE meanings. The strongest syntactic position is taken by Hudson in English complex sentences (1971: 12), where he emphasises that what he has to say concerns 'grammar, to the exclusion of semantics, lexis and phonology'. However, in a later paper (1974: 10) he specifies that the networks 'must be deep enough to make contact with the semantic rep resentation', and in yet another paper (1973: 509) he goes so far as to describe his networks as 'representing syntactic/semantic properties'. An example of the strong semantic position is Halliday, in Text as semantic choice in social contexts (1977: 176). He begins by summarizing his model of language, describing it as having 'three levels or strata: Semantic (semology) Lexicogrammatical (lexology: syntax and morphology and lexis) Phonological (phonology and phonetics)'. He then goes on to describe the four 'functional components' of the seman tic stratum, making it clear that this is the level where the networks of transitivity, theme, etc., are located. However, in other writings (e.g. 1973: 92ff.) he has placed these same networks WITHIN the lexicogrammatical level and reserved the term 'semantics' for the 'second-level' 'socio-semantic' networks mentioned above. (Halliday might however claim that to use the term 'semantics' in these two ways is not as inconsistent as it might appear, on the grounds that BOTH networks have what he terms 'meaning potential' (Halliday 1973: 84ff.). Moreover he might cite in support of the concept of 'meaning at many levels' Firth's notion of 'modes of meaning . . . stated at a series of levels, which taken together form a sort of spectrum' (Firth 1951/57: 220).) Halliday's position remains far from clear. As Butler (1985: 94) points out: 'it is frankly difficult to know what counts as semantic and what as syntactic in [his] later work'. Even Halliday's recent major work on the grammar of English fails to resolve the dilemma. He intro duces the book by describing it (1985: xx) as 'a "functional" grammar . . . based on meaning', and yet on the same page he writes that 'we cannot yet describe the semantic system of a language'—which seems to imply that there is a level of description of rather great importance beyond that which he is describing in that work.
188
ROBIN P. FAWCETT
My own position is that I have argued against the notion that the social semiotic networks constitute an integral level of language, and I have suggested a less direct relationship between them and the tri-stratal model set out above (Fawcett 1975, summarized in Butler 1985: 81-2). However a tri-stratal model does, in my view, approximate to the facts, though in a highly simplified way (Fawcett 1983: 115-8). Enough has been said to show that this issue is still wide open. My intention when I first planned this paper was that, by focussing on an extremely restricted part of the grammar of English, I would be able to bring a small but solid piece of evidence to the discussion. But I must admit from the start that this paper will not result in the kind of logically conclusive evidence that I hoped. I think, however, that a sufficient number of interesting points emerge in the course of the exploration of the topic for the experience to be worth sharing. And the reason for the ultimate failure to clinch the argument is itself a matter of some importance.
2. The Method of Approach Consider the problem, then, of writing an explicit grammar — a generative device — to specify all and only the 'personal pronouns' of Eng lish (henceforth, for brevity, simply 'pronouns'), i.e., I, me, you, he, him, she, her, it, we, us, they, them. Let us follow three principles: 1) 2) 3)
We shall use the concepts and notation of systemic theory. We shall ignore wherever possible any repercussions of this part of the grammar on other parts, or other parts on this part. We shall follow a policy of first trying to account for purely FOR MAL contrasts, and only moving on to consider SEMANTIC contrasts when our inability to construct a satisfactory model forces us to. (The criteria for 'satisfactoriness' will be discussed.)
We shall in fact find that we need to consider three alternative models. As we develop our network to incorporate first the insights of traditional grammar and then a more explicitly semantic approach, we shall encounter problems of various kinds, some of which may come as a surprise even to those familiar with systemic theory. Some of these problems take the form of prices to be paid in additional complexity for what we gain in terms of semantic insightfulness, and in such cases we shall need to consider the
THE ENGLISH PERSONAL PRONOUNS
189
alternatives, and to decide whether or not the price is worth paying. At a number of points you will be invited to test the claims made here by attempting to draw system networks of your own. The basic systemic concepts that we shall need to introduce are the fol lowing: 1) 2) 3) 4) 5) 6)
a system, dependency, a system network, a selection expression, a realization rule, a realization.
We shall also introduce a small number of derived concepts: the six given above are those required in any systemic grammar that is more complex than a single system.
3. Model 1: A Purely Formal Grammar EXPLICITNESS and SIMPLICITY are two criteria for judging alter native models which linguists usually value highly. The concept of explicitness is integral to the concept of model-building, and simplicity is often regarded as equally integral. But we need to be wary of placing TOO high a value on simplicity. Perhaps we might reasonably take the position that so long as simplicity reflects an aspect of PSYCHOLOGICAL REALITY it is right to value it highly. Psychological reality is thus a third criterion. But it must be said that it is a criterion of a rather different sort. Simplicity is the type of criterion that guides the construction of a THEORETICAL model. At a later stage such a model might be subjected to some type of EMPIRI CAL evaluation. The results of such investigations would then be taken into account in revising the model . . . and so on. This is the ideal prog ramme: the reality in linguistics today is that there is much less interaction between these two means of developing models than there should be. And here, reluctantly, we shall not be able to invoke psychological reality. If we now apply the two theoretical criteria of simplicity and explicitness as we construct a system network capable of generating the personal pronouns of English, we shall undoubtedly arrive at that shown in Figure 1. Let us look carefully at how this notation is to be read.
190
ROBIN P. FAWCETT
pronoun-
I me you he him she her it we us they them
Figure 1: Model 1: the simplest system network We need to begin with the concept of SYSTEM. A system is defined as 'a set of features, one and only one of which must be selected if the entry condition to that system is satisfied' (Halliday, 1967: 37). The ENTRY CONDITION in Figure 1 is [pronoun]. (Hudson has proposed the useful convention that features should be enclosed in square brackets when they are referred to in running text.) We note that the features are, in this case, THE ITEMS THEMSELVES. The notation therefore represents a state ment that we may gloss as follows: 'If and only if you select the feature [pronoun], you must then select one of the features [I, me, you, he, him, she, her, it, we, us they, them]'. (Notice that in a system network the verti cal order of the features carries no significance.) Since in this micro-gram mar the FEATURES are identical with the ITEMS, there is no need for realization rules to state how the features are to be realized. This micro-grammar is undoubtedly both explicitly generative and sim ple: it therefore meets the two criteria that we have just introduced. How ever, we will probably agree that the system in Figure 1 fails to satisfy us: we are left with the sense that there is more to be explained about the Eng lish pronouns than has been said so far. We should therefore add as a third criterion EXPLANATORY POWER. (This is not to be confused with the sense of the word 'power' in which a model may be said to be 'too powerful' when it generates outputs that are not wanted. Explanatory power is always desirable.) Let's apply this criterion to Figure 1. It may not appear to explain much, but can we say that Figure 1 has NO explanatory power? What does it say? Two things: it asserts (1) that the items listed are all and
THE ENGLISH PERSONAL PRONOUNS
191
only the pronouns of English, and (2) that one must choose one rather than another of them. This is not as negligible as it may at first appear — a point to which I shall return — but we certainly expect something more. What is lacking in this model? Clearly one lack is any recognition of the traditional subcategorizations of the pronouns as 'first, second and third person' pronouns, and so on. Let's now see what happens if we try to intro duce these into the system network.
4. Model 2: A Grammar Incorporating the Traditional Subcategorizations These traditional subcategorizations of the English personal pronouns are in fact simply a set of systems. There are four of them, and they are set out in Figure 2. It is not always realized that systems in systemic theory and sets of labels of this sort are systems, and that sets of subcategorization state ments of the traditional sort can be essentially the same type of phenome non. What differs is the viewpoint: the concept of systems is associated with the generative approach to language, and the concept of labels with the classification of linguistic units occurring in texts. Thus, if one starts with a bit of TEXT such as the single item me one may reasonably ask: 'How is it to be classified?' A possible answer might be that firstly it is a (or 'has the feature') 'pronoun', and secondly that it is (or 'has the features') 'first per son', 'singular' and, let us say (to use a misleading but nonetheless tradi tional term), 'accusative case'. From the perspective of the TEXT, then, a system network is a CLASSIFICATORY device. But suppose we start with the GRAMMAR, regarding it as a set of rules that specify linguistic behaviour, and ask 'What texts can it generate?' Then the answer — at least, the answer given by a systemic linguist — will be in terms of the rele vant systems (together with the other parts of a systemic grammar that we shall come to shortly). From this second perspective a system network is (part of) a GENERATIVE device. It is sometimes assumed that only generative grammars need to meet the criterion of explicitness, and that grammars designed for analysing texts need not. The truth is that we need an explicit grammar for BOTH tasks, if we are to carry them out efficiently. The difference is that it is rather easier to get away with half-worked-out ideas in analysis than in generation. I would say, then, that a good grammar must be an EXPLICIT grammar, and it must be usable both as a GENERATIVE device in the PREDIC-
192
ROBIN P. FAWCETT
TION of potential texts and as a CLASSIFICATORY device in the ANALYSIS of actual texts. The two emphases are not in conflict in sys temic theory (and should surely not be in any satisfactory linguistic theory.) 2 Figure 2 therefore presents, in the notation of a systemic generative grammar, the traditional subcategorizations of the English pronouns. What, we may ask, are the criteria upon which they are based? They are traditionally said to reflect 'distributional criteria'. What this means in our case is that certain pronouns, when they are Subject (more properly, when they expound the head of the nominal group filling the element of clause structure that we call the Subject), typically CO-OCCUR with certain finite verbs. Thus, the category of PERSON is involved in the co-occurence of I with am and you with are, and the category of NUMBER in I with am and we with are. The category traditionally called CASE is also involved, because we find I at Subject (to slightly over-simplify), but me elsewhere. The usual way of characterizing these PERSON and NUMBER relation ships is through the concept of 'concord', such that the verb is said to 'agree with' the Subject in person and number, and we shall return to this shortly. 'But', one might ask, 'how do we justify the number distinction for pro nouns that are NOT Subject?' How far, for example, are we justified in say ing that us as well as we is 'plural'? The answer is that, once the distribu tional criteria have been set up, linguists use their semantic intuitions to extend the categories to other pronouns. In other words, the criteria are not as formal as they first appear. There is no implication here that there is something dubious about invoking semantic criteria; in the fuller model towards which we are working the central component is the semantics, and semantic criteria have an honourable place. In such a model it becomes possible to see the 'concord' between Subject and verb not as 'cause' and 'effect', but as two 'effects' of a single semantic fact. Let's now examine Figure 2 more closely. Each system, you will notice, has a label, printed in capitals, and this is a common practice in writ ing systemic grammars. However, such labels play no part in the use of the grammar as a generative device, and they are simply there to make it easier to discuss the grammar. In some systemic grammars (e.g. Hudson 1971: 71 and Martin 1983: 50) numbers are used instead. In the second micro-grammar for the pronouns of English that we shall construct we shall use these time-honoured categories. And we shall try to ensure that, as far as possible, it is a network at the level of FORM, not
THE ENGLISH PERSONAL PRONOUNS
pronoun
first second third
pronoun
singular plural
pronoun
masculine feminine neuter
pronoun
nominative accusative
193
Figure 2: The systems considered relevant in traditional grammars SEMANTICS. Form, then, is the level at which ITEMS exist as formal items, and the level of the STRUCTURES that relate those items. Clearly, there is no possibility of realizing the features that we may attach to a pro noun such as [singular] in a structure, in the way that changes in the struc ture of a clause, for example, may indicate a change of MOOD, such as [declarative] or [interrogative]. But we can invoke the formal concept of the ITEM, and we can set ourselves the task of devising a system network in which EACH ITEM IS GENERATED ONLY ONCE. If we succeed in doing this, we could say that we have constructed a system network that is, in one sense at least, genuinely 'formal'. Thus, while we shall want to gen erate a 'nominative' I and an 'accusative' me, we shall not want to generate both a 'nominative' and an 'accusative' it. Nor is there a 'masculine, third person, plural, nominative' they; they is simply 'plural' and 'nominative'. Similarly there should not be a 'singular' and a 'plural' you. By accepting this criterion we can, I think, claim that the network is in some quite solid sense 'formal', even though terms such as 'singular' and 'plural' are evi dently also related to meaning. Perhaps we might claim for a 'traditional feature' system network of this type that it would probably be possible to link it in some quite simple way with a network that expressed truly seman tic choices. It might perhaps be seen as hovering, as it were, somewhere between the purely formal system shown in Figure 1 and some purely semantic system network — perhaps as implied by Hudson's 1973: 509 suggestion that system networks should 'represent syntactic/semantic prop erties', and by Halliday's (1985: xix) description of a 'functional' grammar as 'one that is pushed in the direction of the semantics'.
194
ROBIN P. FAWCETT
We have now reached the point where we must introduce the second basic concept in systemic theory: that of the DEPENDENCY that may exist between two SYSTEMS. Consider the CASE system. We shall wish to say that, if we select the option [first] in the PERSON system, we must then select one of [nominative] or [accusative] in order to generate either I or me. But the CASE system is NOT relevant to [second], where we must gen erate the single form you. The system of CASE is therefore DEPENDENT on a logically prior choice of PERSON. Figure 3 shows the notation for dependency. (Note that the network is far from complete in other respects.)
Figure 3: Dependency in a system network In a precisely parallel way, we shall wish to say that one of the GENDER features [masculine], [feminine] or [neuter] must be selected if we have pre viously selected [third] in the PERSON system. On the other hand, the GENDER system is simply not relevant to the features [first] or [second]. Dependence may therefore be defined as follows: Given two systems, Sys tem A and System B, System is dependent on System A when a feature in System A is (or is part of) the ENTRY CONDITION to System .. We are now in a position to define the third basic concept: that of a SYSTEM NETWORK. A system network is any set of systems that are interdependent in the sense of 'dependent' that we have just defined. (The minimum link between systems in the same network is to share the same entry condition, in a way that we shall meet in section 5.) It may be useful to point out that systemicists sometimes refer to what is, strictly speaking, a SUB-network of a more complex system network as a 'system network' — and sometimes simply as a 'system'. At this point the reader who would like to involve her/himself actively is invited to spend a few minutes with pencil and paper, trying to establish which of the systems in Figure 2 have as their entry condition a feature in another system. Note that we have NOT so far introduced any ruling on the question of whether or not a system may appear more than once in a net work, so that a system may appear as often as it is needed. But we HAVE
THE ENGLISH PERSONAL PRONOUNS
195
ruled that each item must be generated once only, since we are attempting to incorporate that explicitly formal criterion into our network, It is probable that the result of any attempt to capture the dependen cies between systems will produce a network similar to that shown in Figure 4. (Any differences are likely to be due to the characteristic of the notation that the vertical order in a system is not significant.) Figure 4 also shows, on the right, the REALIZATIONS, which we shall come to shortly. We now need the fourth basic notion, which is that of a SELECTION EXPRESSION. This is the cluster of features that are collected, in moving through the network from left to right, and that together specify any ONE of the pronouns. Here are three selection expressions from Figure 4; together with their realizations: [pronoun, first, singular, nominative] .ƒ [pronoun, second] you [pronoun, third, singular, feminine, accusative]... .her It is perhaps surprising that we have found ourselves referring to REALIZATIONS, when we have still not introduced the logically prior notion of REALIZATION RULES. A realization rule is a statement that a given abstract FEATURE will be REALIZED in a given ITEM or STRUCTURE or in some aspect of INTONATION. (We shall here only be concerned with realizations as items, and not with realizations in struc ture, such as those associated with the THEME and MOOD networks, for example, or in intonation.) The work of the realization rules is to make the relatively abstract features of a selection expression 'more real'; hence the term 'realization'.
Figure 4: Model 2: a traditional feature network and its realizations
196
ROBIN P. FAWCETT
In Model 1 there was no need at all for realization rules, because the features in the system were identical with the items that it generated. But we do need realization rules in the present model, if only of a very simple type. As we have seen, each pathway through the network specifies a dif ferent selection expression, in a way that ensures that there is just one REALIZATION for each TERMINAL FEATURE in the network. (This is not always the case with system networks, as we shall see.) In the present case, therefore, the realization rules can therefore take the form of a DOT TED LINE connecting each terminal feature in the network (and so each of the twelve selection expressions that may be chosen) to the appropriate item. To summarize: Model 2 appears to be a reasonably satisfactory genera tive device. It generates all and only the personal pronouns of English; it incorporates the subcategorizations that are traditionally recognized; and its realization rules are extremely simple. It is at the level of form, at least in the sense that it is constructed on the principle that it must generate each item once only, and it is semantic, in the sense that the traditional labels seem to have at least their origins in meanings. However, this network would be severely criticized by most systemicists. We shall take account of these criticisms as we attempt, in the next section, to build a more economical version of the model. (Because it is at the same level as Model 2, we shall consider it as a variant of Model 2 rather than a third alternative, and call it Model 2a.
5. Simplifying the Network in Model 2 Most systemic linguists would wish to modify Model 2 so that each of the two systems of NUMBER and CASE occurs only once in the network. It would be said that a network such as that shown in Figure 4 fails to cap ture certain generalizations. We would, for example, be urged to capture the generalization that both [first] and [third] enter the NUMBER system, whereas [second] does not. Similarly, most systemicists would try to bring together in some way those features that are entry conditions to the CASE system. Two questions that must be asked are: (1) (2)
Is it possible to satisfy these criticisms? If so, is there a price to be paid and, if so, is it worth it?
THE ENGLISH PERSONAL PRONOUNS
197
(Outside the goal of trying to construct the most economical explicit gram mar possible, other considerations come into play, such as usefulness for teaching purposes. See Fawcett 'forthcoming' for a discussion of these). The answer to the first question is 'Yes', as we shall shortly see, and most systemic linguists would also answer 'Yes' to both halves of the second. Yet we shall shortly uncover some problems which should give us pause for thought. It is in fact relatively easy to solve the problem of the CASE system, but to do so we must introduce another of the notational conventions of sys temic theory. We can, if we wish, gather together the five pathways which must lead to the choice between [nominative] and [accusative], and link them together with a left-opening 'or' bracket, as in Figure 5. This is what is termed a DISJUNCTIVE entry condition.
Figure 5: systemic notation convention: a disjunctive entry condition The notation reads: 'If and only if you choose any one of the features [sing ular], [plural], etc., you must then choose [nominative] or [accusative]'. Notice that the feature [plural] appears twice in Figure 5; this indicates that the problem of the NUMBER system remains. We want to know whether it is possible, by rearranging the dependencies between systems, to ensure that the NUMBER system appears only once in the network. If we find that we cannot do this, or that we can only achieve it in a roundabout way, we shall want to ask why. We shall shortly come to the second point at which you are invited to consider the problem for yourself. Before you do so, however, I must make available to you two further conventions: those that are illustrated in Figure 6. So far we have noted the right-opening square bracket found in the SYS TEM itself, and the left-opening square bracket of a DISJUNCTIVE ENTRY CONDITION. Curly brackets, however, mean not 'or' but 'and', so that in Figure 6 the first diagram conveys the meaning of SIMUL-
198
ROBIN P. FAWCETT
TANEITY, thus: 'If and only if you choose a, you must then choose BOTH one of b or AND one of d or e'. 3
Figure 6: Two systemic notational conventions: simultaneity and a compound entry condition. The left-opening curly bracket in Figure 6 is, like the left-opening square bracket in Figure 5, concerned with the ENTRY CONDITION to a system: it is a CONJUNCTIVE entry condition, and it is to be read as fol lows: 'If and only if you choose BOTH 1 AND m, you must then choose one of n or o'. We have now encountered all the conventions that may be used in an attempt to make the network in Figure 3 more economical. Each of the three additional types of bracket shown in Figures 5 and 6 provides a means of expressing not simply a relationship between FEATURES (as a system does) but between FEATURES and SYSTEMS and even, if necessary, between SYSTEMS and SYSTEMS. This is the point at which you are invited to put them to use, in order to construct a variant of Model 2 as shown in Figure 4, in which NO SYS TEM APPEARS MORE THAN ONCE. Let me warn you: it will not be easy, and some ingenuity will be needed to find anything approaching a satisfactory solution to the problem. I shall argue that this fact is, in itself, highly suggestive. Let us now look briefly at some of the solutions that might be prop osed. A tempting 'second-best' would be a network such as that shown in Figure 7, which however introduces the new feature ['not second']. (Note that the labels for systems are omitted, to save space.) The drawback to this solution is that ['not second'] is certainly NOT a traditional feature for a pronoun. Nor does it seem likely such a network would be helpful in mak ing links to some other, explicitly semantic system network: if there is to be any grouping of two of the 'persons' together it seems more likely, on intui tive semantic grounds, that [first] and [second] might be grouped together as 'interactants', as opposed to the 'non-interactant' 'third person'. (This
THE ENGLISH PERSONAL PRONOUNS
199
distinction will in fact be seen to be supported when we come to Model 3.)
Figure 7: A variant of Model 2 (Figure 4) that contains the ad hoc feature 'not second' Another 'second best' solution would be that shown in Figure 8. THIS IS IN FACT Hudson's solution (1971: 60, reproduced in McCord 1975: 200). But you will note that it generates you twice over, and so fails to meet the particular criterion for a FORMAL network that we have set ourselves. We shall however shortly incorporate what is essentially the same pattern in a SEMANTIC system network.
Figure 8: A variant of Model 2 (Figure 4) that generates you twice
200
ROBIN P. FAWCETT
The only possible solution to the problem that meets the criteria is the network shown in Figure 9. The problem is that even those who are fully accustomed to the systemic notation are likely to find this diagram consid erably harder to read than Figure 4. This is because we have been forced to introduce places in the network where a pathway encounters as many as THREE brackets before finding another feature, such that the middle brac ket represents a relationship between one relationship between features and another relationship between features. This undeniably adds a new dimension to the complexity of the model, and so to the difficulty of under standing the relationships it is trying to highlight. We need to ask whether the relationships involved in this part of the grammar really are as complex as this. (This is not the place to discuss whether or not they may be this complex elsewhere in the grammar, but see Halliday, 1968: 206.)
Figure 9: variant of Model 2 (Figure 4) with complex bracketing relationships The position that we have reached is that it seems to be impossible to construct a network which meets the criteria which we have set up and which avoids the complex relationships of Figure 9. If Figure 9 is unacceptably complex, we must ask: Have we made a wrong decision at some point? And, if so, where? Let us consider in turn each of the major guidelines that we have been following. Firstly, systemicists may simply be wrong to try to capture in their net works generalizations such as those concerning the NUMBER and CASE networks, so that Figure 4 is in fact fully acceptable. But agreement that the generalizations SHOULD be captured without repeating systems is so widespread that we shall ignore this for the moment. (I shall return to the point in the concluding section of the paper.) Secondly, the English pro-
THE ENGLISH PERSONAL PRONOUNS
201
nouns may, contrary to our intuitions, really be an area of the grammar that is so complex that it requires the intricate relationships between features used in Figure 9. Thirdly, we may be wrong to regard the traditional sys tems of PERSON, NUMBER, GENDER and CASE that are set out in Figure 2 as a set of FORMAL systems. Could it be that these traditional features are in fact fully fledged SEMANTIC features? If so, we are involved with a new level of language, and we may be wrong to try to make the network 'formal' by generating each pronoun once only. For example, the item you may be semantically ambiguous, one meaning being singular and one plural. There is of course a fifth possibility, which is that the con ventions of systemic theory may be inadequate for describing elegantly what are in fact very simple paradigmatic relationships. This last possibility is particularly interesting, because it draws atten tion to a fact that has not so far as I know been pointed out till now. This is that the model of paradigmatic relations offered by systemic theory is NOT capable of handling ALL types of classificatory relationships. (See the Appendix.) In other words, a systemic grammar is not as powerful as is sometimes suggested, and is consequently a more interesting theory. Thus, the theory incorporates a strong hypothesis as to the particular range of types of paradigmatic relationship that are found in natural languages: that the features between which we choose are related to each other in principle in all and only the ways that we have indicated here. (I say 'in principle', because we must allow for the temporary dislocation of a network in the process of historical change. But even then we would expect the relation ships, if they have been correctly identified, to reassert themselves in the course of time, though often in relation to new features.) My view is that the relationships that the theory highlights ARE likely to be the right ones, and that it is fourth guideline (i.e. the attempt to tie our network to the level of FORM) that has caused the problems. How ever, before we move on to consider a SEMANTIC model, we should remind ourselves of the second of the two questions we asked a few moments ago: If it HAD proved to be possible to draw a fully satisfactory network, would the price to be paid have been worth it? The cost would in fact be that the realization rules would have been considerably more com plex. But since they would have been complex in precisely the same sort of way that the realization rules for our third model will be complex, we shall consider the question in relation to those rules.
202
ROBIN P. FAWCETT
6. Model 3: A Grammar with a Semantic System Network Suppose that we now remove the requirement that each item should be generated once only. To do this is in fact to take the crucial step in estab lishing the network as being at the level of SEMANTICS. The reason is that we have now made it possible for the phenomenon of NEUTRALIZA TION to occur. Neutralization is simply one of the types of what Lamb (1966: 17) calls 'interstratal discrepancy', and it is such indications of the lack of a one-to-one relationship between a meaning and an item or struc ture that alert us to the fact that we are dealing with two distinct strata of language. For example, there are two meanings, which we may gloss as 'not heavy' and 'not dark', both of which are realized in the single item light. Similar cases of neutralization, I want to suggest, occur in the pronoun net work of English. In other words, sometimes A SINGLE ITEM MAY REALIZE MORE THAN ONE SELECTION EXPRESSION. It is because this phenomenon of neutralization occurs in Hudson's network for pronouns (Hudson, 1971: 60, reproduced in McCord, 1975: 200) that we are alerted to the possibility that his network may in reality be a SEMAN TIC network. (It uses the traditional terms, and in it the PERSON and NUMBER systems are entered simultaneously, with [third] and [singular] providing a compound entry condition for the GENDER system: in other words, the relationships are essentially as in the semantic system network shown in Figure 10, even though the labels are different.) Since the system network that we shall construct will be semantic, we shall feel free to replace those of the traditional labels that do not directly express the meaning for which they stand. You will therefore find in Figure 10 the feature [performer] instead of [first], the feature [addressee] instead of [second] and, by extension, the feature [outsider] instead of [third] in what is now labelled the INTERACTANT ROLE system. Sometimes the label is designed to make a specific point, as in my substitution of [male] for [masculine] in what is now the SEX system, where I wish to indicate that in English (as opposed to languages such as French and German) we are con cerned with fully semantic features, as opposed, for example, to cases such as das Mädchen (German for the girl), where the neuter gender is purely formal. There are a number of significant points of detail to be made about the network for Model 3. First, notice that the entry condition is now [thing] instead of [pronoun]. The term 'pronoun' is typically used as a label for a
THE ENGLISH PERSONAL PRONOUNS
Figure 10: Model 3: part of a simplified semantic systemic
203
network
class of items at the level of form: in the semantics we are concerned with meanings relevant to 'things', in a sense of 'thing' that includes persons, only some of which are realized in pronouns. Second, notice the row of dots between [thing] and the two systems in which 'things' must select features. The dots indicate that there are other systems between the two points shown that are omitted here for simplicity. The rows of dots after [quantified], [lexical classification] and [naming] similarly show that there are other systems not shown here. Third, you will see that the overall network allows for the system of QUANTIFICATION (which is large and complex) to be entered in parallel with the INTERACTANT ROLE network, to generate expressions such as all of us, two of you, some of them, two pints of it, etc. (all of which involve pronouns), as well as all of the books, two of your friends, etc., which do not. In what follows we assume, for simplicity, that the feature [unquantified] is always chosen. Fourth, note the crucial point that, when [interactant] is chosen, the realization can ONLY be in a pronoun (or some nominal group such as the ones we have just exemplified with a pronoun at their head), while an 'out sider' has many choices open to her/him/it. Thus there is NOT the close relationship at the semantic level that the traditional approach to the per-
204
ROBIN P. FAWCETT
sonal pronouns of English suggests. The plain fact is that once a 'thing' has been classified as an 'outsider', there are many complex choices open to it which are simply not there when the referent is an 'interactant'. Most of the relevant systems are entered from [lexical classification], and they include the option to show that the referent is or is not some 'specific' entity, and if 'specific', whether or not also 'particularized' (as in books vs. some (=s'm) books vs. the books); whether or not the meaning of the noun at the head of the nominal group is recoverable (as in two recent books vs. two recent ones); and whether the 'thing' is a 'count' or a 'mass' thing (s'm books vs. s' m water). None of these choices, then, are relevant to interactants. And there are a good many others. Fifth, notice that although, as I have said earlier, the vertical order of features in a system is not normally significant in systemic descriptions, it could in fact be put to work to carry the meaning 'Try the features in this sequence'. Thus if we were to put this little network into a computer as part of a grammar of English, it would be more economical to try the features in the sequence shown here, because this is the sequence of the frequency of the three types. In one grammar that I have written (Fawcett in prepara tion) I have actually specified suggested frequencies for each feature in a system. Sixth, you may have noticed that the traditional terms 'singular' and 'plural' do not appear in the network. This is because they are used in the NUMBER system that is dependent on [lexical classification] and [count], which (roughly speaking) puts s onto plural nouns. In the present network the question is phrased in terms of whether the referent is 'one' or 'óneplus'— which says the same thing as 'singular' and 'plural' in different words. Seventh, you may wonder why the 'third person pronoun' network is called the TOKEN CLASSIFICATION network. The answer is that it shows broad similarities to what I term the CULTURAL CLASSIFICA TION network, i.e. the network that specifies the meanings made available to us by our culture's classification of things in terms of the nouns of our language (see further Fawcett 1980: 217- 20). It would be possible to have a slightly simpler network here if we decided to ignore this parallel, because it happens that we use the same pronoun, it, for both 'mass' things, e.g. to refer to some apple that is about to be served as food in purée form, and things that are [count, one, non-person], e.g. an apple on a tree or in a shop. (This is in effect what is done by Hudson (1971:60, reproduced in
THE ENGLISH PERSONAL PRONOUNS
205
McCord 1975: 200) and in Martin 1983: 50.) But I find this less than satis factory, in psychological terms; it seems more plausible to assume a single classification system for 'things', and to allow it to be ambiguous as between 'mass' things and things that are 'count, one, non-person' —just as we shall shortly see that we need to treat you as ambiguous between 'one' and 'oneplus'. (Indeed, even Hudson's rather formal network permits that ambiguity, so why not this one?) Eighth, you will observe that we have introduced a totally new feature, [sex unspecified], into the SEX system. This is intended to account for cases such as they in texts such as when someone wants to borrow more than six books at a time, they must obtain special permission. This is an option that is by no means in the semantics of all native speakers of English. Some use instead he and him and this is reflected in Martin's (1983: 50) characteriza tion of he as [human, unmarked]. Others insist on he or she and him or her, or she or he and her or him, or, in writing s/he and him/her or her/him. Note that this option is only required for singular outsiders, because sex is not specified at all for plural outsiders in English.4 We turn now to a general point. What do the feature labels in such a network actually MEAN? While these replacements for the traditional labels are intended to give a clearer indication of what the selection of any particular feature involves, they still need spelling out further. Let's take as an example the feature [outsider]. Notice that this may well be selected when the referent is a human being who is both physically present and a member of the smallest socially recognisable group that also contains the performer and the addressee. S/he is not an 'outsider' in sociological terms, therefore: only in terms of the interactant roles that the semantics of Eng lish allows. The term 'outsider' is however fully appropriate because, in broad terms, s/he is referred to by selecting from the same semantic fea tures that would be used if s/he were absent. It is of course fully acceptable in our culture to select the feature [outsider] in referring to non-persons (i.e. objects and animals not being treated as 'persons') that are present (though this would not necessarily be the case in all possible cultures), but problems of some sociolinguistic interest arise in the case of persons. When adults or medical practitioners use the semantics for outsiders in referring to children or patients who are both present and conscious, they usually give offence: the person is being treated as if he were an object, and s/he feels that his/her personal status is being diminished. The 'felicity condition' for choosing [outsider], then, might be something like 'the performer con-
206
ROBIN P. FAWCETT
siders the referent not to count as a member of the social group that is set up by virtue of the current communicative action'. Yet in practice we often need to refer to non-interactants who are pre sent. Some conventions have been established that soften the impact of being treated as an outsider, at least to some extent. First, we increase the frequency of the use of the relevant proper name. We thus invoke one of the two uses of the 'vocative' form, that of expressing the appropriate point on the scale of social distance/intimacy, and as a consequence we cut down on our use of the 'third person pronouns', which are markers of outsiders — objects, animals, and absentee persons, etc. Second, such proper names are sometimes followed by here as in John here will show you the way. Third, and most significantly, they are practically always accompanied by a kinesically mediated act of communication such as a gesture or a smile to the 'outsider' — who, in this separate act of communication, is the addressee. This discussion has taken us beyond the specification of the feature [out sider] as we have explored some of the ways in which we overcome a limita tion in the English system. But it has illustrated the point that it is NOT the business of semantics to model ALL POSSIBLE MEANINGS. This brings us to the final question for this section. It is this: What are the criteria that determine what features to recognise in the semantics of a language? For example, in the system that we have just been considering, we might ask: Should we introduce a fourth interactant role, which we might term 'acknowledged overhearer'? In an interactional approach to language, e.g. from the viewpoint of discourse analysis or of social psychol ogy, we might well want to recognize this as a significant category, so should we therefore include it in the semantics? If so, what about 'known but unacknowledged overhearer'? The answer must be that IT DEPENDS ON THE LANGUAGE. We need to make a distinction between 'possible meanings' and 'meanings that are built into the organization of a particular language'. (We shall omit here the question of whether it is possible to delimit the set of meanings that MIGHT be built into the organization of a language, i.e. the question of semantic universals.) The SEMANTICS of a given language, then, contains THOSE MEANINGS THAT ARE BUILT INTO THE ORGANIZA TION OF THE LANGUAGE, in the sense that the items, structures and intonation of the language force us to recognize that such features play a part of some sort in the generation of those items and structures. In other words, every semantic feature must have a reflex at the level of form. In
THE ENGLISH PERSONAL PRONOUNS
207
this approach, then, the level of semantics is not as 'deep' or as 'high' or as 'abstract' as it is sometimes taken to be (though the problems that remain are real enough). When we come to the realization rules for this network in the next section, we shall see what is involved in 'having a reflex at the level of form'. This, as we shall see, is NOT limited to the simple one-to-one type of relationship that we had in Figure 4.5 Thus, the network shown in Figure 10 makes a number of highly specific claims about the semantics of English. For example, it claims that English is not concerned with sex distinctions in relation to the performer and addressee, or to outsiders that are also plural. Other languages, of course, ARE concerned with sexual meanings in these areas, as well as with a number of other distinctions, such as those realised in tu and vous in French, which are not significant in this part of the grammar in English. The semantic system networks of such languages are therefore rather differ ent. The possibility of a comparative analysis of the semantic systems of languages in systemic terms is one possible fruitful future area of applica tion for this model of language — with the proviso just mentioned that the labels on features are not always an adequate guide to the meanings they represent. A feature, then, is specified in THREE different ways. First, 'upwards' through the 'felicity condition' for selecting it to the 'knowledge' that guides the choices (cf. Austin 1962, Searle 1969 and elsewhere, and Fawcett 1984: 165ff.); second, 'sideways', to the other features in the network with which it is in complex paradigmatic relationships of choice, dependency, simul taneity, etc; and third, 'downwards' through the realization rules to items, structures and intonation at the level of form. This brings us to a matter which the discerning reader may already have noticed. If we really are to insist on having a reflex at the level of form, what is the justification for making the [addressee] select between [one] and [one-plus]? The answer depends on a bit of the grammar that lies outside our current brief: the system shown in Figure 11 (which can be slot ted into Figure 10). This apparently minor addition allows us to choose, in appropriate circumstances, not only himself and myself rather than him and me, but also yourself ox yourselves rather than you—thus demonstrating the need to show that a referent that is [addressee] must also be either [one] or [one-plus]. (In the present grammar 'reflexive' and 'intensive' pronouns are seen as essentially the same phenomenon — a position that needs a justifi cation which it cannot receive here.)
208
ROBIN P. FAWCETT
Figure 11: How the INTENSITY system fits into Model 3 (Figure 10) Finally, you may have noticed that the CASE system has disappeared completely. The reason for this is clear within the framework of a semantic approach. The so-called 'case' is determined by whether or not the referent is functioning as the subject of the clause, and the choice as to whether or not to make it the SUBJECT THEME of the clause is a semantic choice related to the 'referent situation' of which the 'referent thing' is an element. We are only slightly simplifying if we say that it is the choice in this system, which is primarily realised in the CLAUSE rather than in the NOMINAL GROUP, which determines that the realization of a selection expression in our network will be, for example, I rather than me. Thus, although these 'situation' features AFFECT the realizations of features in the 'thing' net works, they are not, in themselves, 'thing' options. We shall have to take account of this influence in the REALIZATION RULES for our network in the next section, but it would be misleading to build this system into our network.
7. Model 3: The Realization Rules Let us now turn to the realization rules. The most frequently used notation in systemic theory is to set out the features and their realizations in two parallel columns, the FEATURE column and the REALIZATION column, and I shall use this notation here. Sometimes we must also state the CONDITIONAL FEATURES; that is, features that must also be selected if the realization is to take a particular shape. 6 The realization rules that are needed to complement the semantic sys tem network in Figure 10 are set out in Figure 12. Thus, Figures 10 and 12
THE ENGLISH PERSONAL PRONOUNS
209
together constitute the generative grammar for the pronouns that we are considering. The grammar is used as follows. First, one moves through the system network, collecting the features that make up a selection expression. For example, we might choose [thing, unquantified, performer, one]. We then look up each feature in turn in the left hand column of the realization rules, which is headed FEATURE. In this case we find that only [performer] is listed. The REALIZATION in the right hand column will depend on what other features have been co-selected. In the case of every rule the realiza tion is simply that the head of the nominal group which we are generating (symbolized as h) is EXPOUNDED by the item that is specified. The two CONDITIONAL FEATURES columns show that if [one] is co-selected, and if [subject theme] has also been selected in the appropriate 'situation' network, the realization is that the head of the nominal group will be expounded by the item I. The dashes show that in the unmarked case, when the thing is NOT the subject theme, the realization is me. Finally, FEATURE performer
CONDITIONAL FEATURES one
REALIZATION
subject theme
h
one-plus
subject theme
h < we h < us
addressee
h < you
male
subject theme
h
subject theme
female
h < she h
mass non-person outsider
one-plus
subject theme
h < they h < them
sex unspecified Figure 12: The realization rules f or Model 3 (Figure 10)
210
ROBIN P. FAWCETT
notice the neutralization in two pairs of realization rules. The same simple rule serves both [mass] and [non-person], showing neatly their SEMANTIC difference and their FORMAL equivalence. The rule for [outsider, one plus] and [sex unspecified] is more complex, but works on the same princi ple. This grammar of the semantics and forms of the English personal pro nouns is not simple, but I know of no simpler way to formalize the relation ships between the various meanings that are involved in generating these items: it seems to be the case that this part of the grammar simply has this degree of complexity. The reader may notice that the feature [thing] does not play any part in the realization rules that we give here: it represents the option to process the referent as a 'thing' (rather than as a 'situation') and it results in the use of the particular type of syntactic unit known as the NOMINAL GROUP. (A fairly detailed account of the syntax of this unit is given in Fawcett, 1974-6/80; see also Fawcett 1980: 197ff.) However, all the other features specified in the network have a part to play. (The features [lexical classifica tion] and [naming] are, of course, entry conditions to other networks.) In some cases a selection expression may include a feature that plays no part in the realization of that particular item, and this is normal. The feature [outsider], for example, plays no part in generating he, but it IS needed in generating they. You are now invited to inspect this micro-grammar to ensure that it meets the claims made for it: that it generates all and only the personal pro nouns of English in such a way as to introduce all and only the semantic fea tures that are relevant to these items. An interesting and at first slightly problematical point arises when the features [performer] and [one-plus] are co-selected. A 'performer' of whom there is 'one-plus', taken literally, would be two or more persons speaking in chorus, as in a Greek play. But in fact this compound meaning typically refers to 'the performer plus one or more other persons for whom he con siders he is acting as spokesman'. (In English, though not of course in all languages, 'other persons' may include the addressee or addressees, so that there is no distinction between 'inclusive' and 'exclusive' we in English.) It seems to me that the old problem of royal and editorial we can be explained within the general concept of the performer as spokesman. Compare Sweet 1898: 71, who further suggests that the literary 'plural of modesty' arose from using we 'in the indefinite sense of "myself and the other authorities
THE ENGLISH PERSONAL PRONOUNS
211
on the subject'". We have here, then, a complete generative device for the personal pro nouns of English. The network is wired in, in complex ways that we have minimized here, to other parts of the semantics. The hypothesis — and pending testing of the type advocated by Leech (1970) it is no more — is that the contrasts in meaning that the network offers correspond to the psychological reality stored in the native speaker's mind. If they also corre spond to our own intuitions, then that is all to the good. But only properly conducted psycholinguistic tests can establish whether or not this network is valid.
8. The Price to be Paid I have demonstrated the semantically based micro-grammar that I propose for this area of English, but I have not discussed the price to be paid. The cost, of course, is the additional complexity in the realization rules for Grammar 3 (Figures 10 and 12) as compared with Grammar 1 (Figure 1) and Grammar 2 (Figure 4). Let's consider these grammars in the light of the criteria that we introduced in section 3. Grammar 3 is clearly a less SIMPLE grammar than either of the other two. The three grammars generate the same set of items, so that they are equal in EXPLICITNESS. We must therefore ask ourselves whether the relative complexity of the realization rules in Grammar 3 is compensated for by the greater EXPLANATORY POWER of the grammar as a whole. We rejected Grammar 1 because it failed to give us a sufficient sense that something had been explained. The direction in which we have moved in seeking a greater sense of explanation gives some indication of what it is that an 'explanation' may consist of. We have moved in the direction of 'meaning' because we have been looking for what we might term 'purposive patterning'. In other words, we wish to understand the meanings that are built into a language because it is these that are related to phenomena OUTSIDE language, to the other components of our minds. And man, as the most inquisitive of animals, is inherently ambitious to understand his own mind, by relating one aspect of it to another and ultimately to phenomena in the world outside. If we agree that Grammar 3 is, at the least, a little more insightful than Grammar 2, we must still ask whether the cost in the complexity in the
212
ROBIN P. FAWCETT
realization rules is worth paying. Could we not simply make a few changes to Grammar 2 (Figure 4) in order to incorporate the additional insights of Grammar 3 — while retaining the duplication of the systems of CASE and NUMBER? (We would need to semanticize a number of labels, and to add (1) a NUMBER system to [second], (2) the feature [sex unspecified] to the GENDER system, and (3) a CASE system dependent on it.) The answer is that we could, if we did not mind about our failure to capture the various generalizations concerned, or the fact that the 'CASE' system does not belong at this rank. The result of this would be to present ourselves with a dilemma. We would have in Grammar 2 a more complex network with simpler realization rules, and in Grammar 3 a simpler network with more complex realization rules. There would be nothing to choose between them as generative devices: they might even appear to be notational variants of each other. I would argue, however, that they are NOT simply notational variants. We have now reached the point where we can no longer handle the pro nouns as a set of items that are isolated from the rest of the grammar. Indeed, Grammar 3 has already moved in the direction of acknowledging the influence of other parts of the grammar on what was originally intended to be a system network for the pronouns of English. The pronouns have in fact turned out to be less homogeneous than our initial assumptions led us to expect. As we saw in Grammar 3, the 'third person' pronouns are differ ent from the 'first' and 'second', in that they are open to very many systems of choices that are simply irrelevant to interactants. In other words, we need to add a fourth criterion in evaluating competing grammars: that of COMPLETENESS. Grammar 3 is superior to Grammar 2 in its ability to fit into the rest of the grammar of English. It is on such grounds that I greatly prefer Grammar 3 to Grammar 2, even in a version of the latter that was fully semanticized. The instinct to capture the generalizations about CASE and NUMBER in the network— though not in the 'thing' network for the choices realized in 'case' — and to allow some complexity in the realization rules is, in my view, the right instinct.
9. Conclusions First, let us consider Grammar 1 (Figure 1). What is its status in the light of the position which we have reached? The answer is that this (rather
THE ENGLISH PERSONAL PRONOUNS
213
than Grammar 2 or anything like it) sets out with precision at least a part of the set of contrasts that the language displays at the level of FORM. It is simple, it is explicit, and it explains something about the relations between the members of a set of items. Even if we do not find it particularly illuminating, it is undeniably a set of contrasts that is highly relevant when we consider LANGUAGE IN USE in analysing incoming messages. For example, we need to know what ITEM was intended in deciding whether a speaker has said so he didn't go or so we didn't go. We must therefore allow in our overall model for patterns of contrast at this, the purely formal level of language. But in my view it is NOT the case that, in a GENERATIVE model of language, we should regard the items in such networks as features between which one must select. The choice is made not between ITEMS but between SEMANTIC FEATURES, and the items themselves are merely the reflexes at the level of FORM of certain selection expressions. The question with which we began was: Should 'first level' system net works such as those cited in Section 1 be regarded as being at the level of form or semantics? Plainly my view is that such networks should be regarded as SEMANTIC: that is, that they should be seen as consisting of meanings that are built into the organization of the language, and that demonstrate this by having a reflex in form. The danger in working at the semantic level is that one will go too far, and incorporate into one's model 'pre- linguistic' 'possible concepts' that are NOT built into the organization of the language concerned, and this temptation must be resisted. The way to avoid this danger is to follow the principle: no networks without realiza tion rules. It is likely that you will have disagreed with my judgements at some points in this paper. But if you are in agreement with the overall line of argument and the conclusion to which it leads, this paper may perhaps be accepted as a small but still significant piece of evidence that, unless we are modelling purely formal contrasts (as in Figure 1) the 'first level' networks that systemicists construct are SEMANTIC, in the sense defined in Section 6. If more evidence of the same sort can be presented, as I do for example in Fawcett (1980 and 1987), the inference would be that grammars such as those of Hudson 1971 and 1974 that claim to be grammars at the level of FORM, and so to exclude the semantics, need to be reinterpreted (and, no doubt, ammended in places), in order to exclude contrasts that are NOT semantically motivated (such as 'third person singular' -s on verbs), and to include the types of additional semantic features introduced here. My guess is that rather less adjustment would be needed than the different labels that
214
ROBIN P. FAWCETT
systemicists assign to the components of their models of language would imply: i.e. that such systemicists are in principle working at the same level, but that the labels that we put on the boxes, such as SYNTAX or SEMAN TICS, influence us in relatively small v/ays, when the evidence is less clear than we would wish. An important task facing systemicists is to try to spell out in greater detail the ways in which a feature may be said to have a reflex in form, and Martin 1987 offers a useful contribution to this discussion. As systemicists attend increasingly to the realization components in their grammars, and so come to establish with increasing explicitness the criteria for determining what is and what is not a 'reflex in form5, we may expect the networks that they construct as models of particular parts of the semantics of a given lan guage to become increasingly similar. Let me conclude by summarizing the main points that I have tried to make. (1)
(2)
(3)
If the contrasts to be modelled really are at the level of FORM, we need something like Grammar 1 (Figure 1). But if we wish to include in our network the more abstract features of pronouns such as [one-plus] and [masculine] we shall find difficulties in establishing criteria that will identify the network as being in any sense 'at the level of form' — and, moreover, difficulty in draw ing a satisfactory network. We should therefore accept that the level at which we are working is the level of SEMANTICS, and that our grammar will be more like Grammar 3 (Figures 10 and 12). The danger of developing networks of features which are 'con ceptual' rather than ones that are built into the organization of the language must be recognized: one way to avoid this is to ensure that every feature that is introduced has some reflex or other at the level of FORM. (Note, however, that in any one sys tem there may be a feature whose realization is that there is NO reflex in form.) Having a reflex at the level of form does NOT imply that the net work is at the level of FORM: realization rules (as in Figure 12) are essentially INTER-STRATAL, and they handle conditional realization, neutralization, and other such relationships where there is NOT a one-to-one fit between two levels of language, as well as those cases where there is. Moreover the need for such
THE ENGLISH PERSONAL PRONOUNS
(4)
(5)
(6)
215
rules is evidence that we are here concerned with two distinct levels: semantics and form. Both Grammar 1 and Grammar 3, then, have something to be said for them. In the last analysis Grammar 2 is, as a Yorkshireman might say, 'neither nowt nor summat'. The price paid in Grammar 3 for capturing the generalizations that Grammar 2 in Figure 4 fails to capture is to have the more complex type of realization rule illustrated in Figure 12. The economies that such an approach achieves become increasingly apparent when the systems that we have considered are seen in the context of a rather fuller grammar of the nominal group than we have had space to present here. The original strategy of con centrating on a limited set of data in order to narrow down the argument has therefore failed, at least in this case. Indeed, it may well be that it is only possible to determine such issues in terms of a much more complete grammar. Nonetheless this exer cise in linguistic theory has raised a number of interesting issues in systemic linguistics. Although this paper has taken the form of an argument within the framework of systemic theory, it has also been concerned to demonstrate the particular ability of systemic theory to handle economically the complex paradigmatic relations that exist between the semantic features of a language: the suggestion is that others working at this level will ultimately find that they need to model these relationships — as have, for example, both Leech (1969: 104 and 1974/81: 112) and the stratificationalists (e.g. Lamb 1966; Bennett 1972).
Thus the concept of a 'systemic semantic generative base' for which this paper argues may be relevant not only within the framework of sys temic theory but also to ANY model of language in which the generative base is the semantics.
NOTES 1.
I would like to express my thanks to Professor John Sinclair and the other members of his postgraduate seminar for the valuable discussion with which they responded to an earlier version of this paper, and similarly to members of the Cardiff Linguistics Circle.
216
R O B I N P. F A W C E T T
2.
In the 1960's comparisons were often drawn between Transformational Generative Grammar and Scale and Category Grammar — the precursor of Systemic Grammar — to the advantage of the former: TG was said to be generative (at least in intention) and S and to be 'a mere taxonomy'. But the apparatus of generativeness was in fact already present in the model, in the concept of the system, and since the mid-1960's all systemic grammars have been set within the general framework of generativeness. Moreover sev eral systemic grammars (e.g. those described in Winograd 1972 and Davey 1978) have met the most demanding test of explicitness of all — incorporation in an automaton.
3.
The concept of simultaneity is built deeply into the organization of language (though probably not of other codes), and it is the fact that we can — and must — choose MANY MEANINGS AT THE SAME TIME that gives rise to Halliday's important concept of the FUNCTIONAL COMPONENTS into which the grammar divides itself. In a func tional model of language, each major network, or cluster of networks, is seen as corres ponding to a functional component. (See, e.g. Halliday 1970: 143 and 1973: 104ff., and Fawcett 1980: 25-38.) A functional model of language, then, implies a major hypothesis about the nature of the semantics.
4.
The 'institutionalized authority' they of they've changed the times of the bus, and you of you never know are handled elsewhere in the full grammar.
5.
The range of 'possible meanings' may be referred to as a set of 'concepts'. Thus, it would in principle be possible to construct a network of such 'pre-linguistic' concepts, from which a language user might select when formulating a message. In some cases the CON CEPTUAL meanings would coincide with the SEMANTIC meanings, so that there would be relatively little difficulty in processing the concept, but in other cases the con cept would need to be DESCRIBED in terms of the available meanings. For example, we can understand the concepts that are realized in the Hopi verb when they are explained to us, as Trudgill (1974: 25-6) has reminded us, but these meanings are not built into the organization of English. Hence the value of a distinction between conceptual meanings and a sub-set of them, the semantic (i.e. linguistic) meanings available in any particular language.
6.
An alternative notation would be to use lines and nodes similar to those in the semantic system network, and this is in fact done in the realization portion of a stratificational grammar (cf. Lamb 1966; Lockwood 1972). If we did this, it would save on the repetition of features from the system, but it would greatly increase the complexity of the lines and nodes by which they would be related. Again, the reader is invited to test for himself the claim made here.
APPENDIX It is sometimes claimed that system networks are 'too powerful': that is, that they can generate any selection expression whatsoever, and that they therefore embody no interesting claims about the nature of language. Is this in fact so? System networks are particularly well adapted to generating selection expressions consisting of the sorts of sets of features illustrated below:
THE ENGLISH PERSONAL PRONOUNS NETWORK
217
SELECTION EXPRESSIONS GENERATED c ad be xbf c ad b xbd
c ad b ce bcf xbd
xace xacf xad xb ce xbcf xbde xbdf The crucial question is: Are there any sets of selection expressions that system networks are NOT able to handle simply? The answer is that there are. Consider the apparently simple set xab, xac, xbc. The only system network that will generate just those three sets of features (without repeating a feature) is as follows:
Clearly, this network embodies a number of characteristics that fall outside the normal run of expectations in systemic models of language: we find (1) a system with only one labelled term; (2) two terminal features that are not in systems; and (3) an exceptionally
218
ROBIN P. FAWCETT
heavy use of symbols for relationships, by comparison with the number of features. It is obvious that if the selection expressions that are required to specify the features of lin guistic items were of this type, some model of relationships other than those of systemic theory would be used. It is therefore NOT the case that system networks can handle ALL types of classificatory/choice relationships equally easily. It is in this sense that sys temic theory makes stronger predictions about the nature of relationships between fea tures than might at first be supposed.
REFERENCES Austin, J.L. 1962. How to do things with words. London: Oxford Univer sity Press. Bennett, David C. 1972. Some observations concerning the locative-direc tional distinction. Semiotica 1. Benson, J.D. and W.S. Greaves, eds. (Forthcoming). Systemic functional approaches to discourse: selected papers from the 12th International Systemic Workshop. Norwood, N.J.: Ablex. Bernstein, ., ed. 1973. Class, codes and control: volume II. London: Routledge and Kegan Paul. Berry, Margaret. 1975. An introduction to systemic linguistics: 1 structures and systems. London: Batsford. Butler, C.S. 1985. Systemic linguistics: theory and applications. London: Batsford. Davey, A. 1978. Discourse production: a computer model of some aspects of a speaker. Edinburgh: Edinburgh University Press. Dijk, T.A. van, and J.S. Petöfi, eds. 1977. Grammars and descriptions. Berlin: de Gruyter. Fawcett, R.P. 1973/81. Generating a sentence in systemic functional gram mar. University College London, mimeo. (Also in M.A.K. Halliday, and J.R. Martin, eds., Readings in systemic linguistics. London: Batsford, 1981.) . 1974-6/80. Some proposals for systemic syntax. MALS Journal 1,22,2. (Also published with slight revisions as Cardiff: Polytechnic of Wales, 1980.) . 1975. Some issues concerning levels in systemic models of language. Nottingham Linguistic Circular 4,1:24-37. . 1980. Cognitive linguistics and social interaction: towards an integ rated model of a systemic functional grammar and the other components
THE ENGLISH PERSONAL PRONOUNS
219
of a communicating mind. Heidelberg: Julius Groos and Exeter Univer sity. . 1983. Language as a semiological system: a re-interpretation of Saussure. In J. Morreall, ed., The Ninth LACUS Forum 1982. Colum bia, S.C.: Hornbeam Press. 59-125. . 1984. System networks, codes and knowledge of the universe. In R.P. Fawcett et al., eds., The semiotics of culture and language. Lon don: Frances Pinter. 135-79. . 1987. The semantics of clause and verb for relational processes in English. In M.A.K. Halliday and R.P. Fawcett, eds., New develop ments in systemic linguistics. London: Frances Pinter. . (Forthcoming). What makes a good system network good? — four pairs of concepts for such evaluations. In J.D. Benson and W.S. Greaves, eds., Systemic functional approaches to discourse: selected papers from the 12th International Systemic Workshop. Norwood, N.J.: Ablex. . In preparation. A computer-implementable systemic semantic minigrammar for the English auxiliaries and some related phenomena. Paper given to 25th Jubilee Meeting of the Linguistics Association of Great Britain. Fawcett, R.P., M.A.K. Halliday, S.M. Lamb, A. Makkai, eds. 1984. The semiotics of culture and language. London: Frances Pinter. Firth, J.R., 1951. General linguistics and descriptive grammar. In J.R. Firth, Papers in linguistics 1934-1951. London: Oxford University Press. . 1957. Papers in linguistics 1934-1951. London: Oxford University Press. Halliday, M.A.K. 1968. Notes on transitivity and theme in English, Part III. Journal of Linguistics 4:179-215. . 1970. Language structure and language function. In J. Lyons, ed., New Horizons in Linguistics. Harmondsworth: Penguin. . 1973. Explorations in the functions of language. London: Edward Arnold. . 1977. Text as semantic choice in social contexts. In T.A. Van Dijk and J.S. Petöfi., eds., Grammars and descriptions. Berlin: de Gruyter. . 1985. Introduction to functional grammar. London: Edward Arnold. Halliday, M.A.K. and R.P. Fawcett, eds. 1987. New developments in sys temic linguistics: Volume 1 theory and description. London: Frances
220
ROBIN P. FAWCETT
Pinter. Halliday, M.A.K. and J.R. Martin, eds. 1981. Readings in systemic linguis tics. London: Batsford. Hudson, R.A. 1971. English complex sentences: an introduction to sys temic grammar. Amsterdam: North Holland. . 1973. An item-and-paradigm approach to Beja syntax and morphol ogy. Foundations of Language 9:504-48. (Reprinted in Halliday and Martin 1981.) . 1974. Systemic generative grammar. Linguistics 139:5-42. Lamb, S.M. 1966. Outline of stratificational grammar. Washington: Georgetown University Press. Leech, G.N. 1969. Towards a semantic description of English. London: Longman. . 1970. On the theory and practice of semantic testing. Lingua 24. . 1981. Semantics. 2nd ed. Harmondsworth: Penguin. Lockwood, David G. 1972. Introduction to stratificational linguistics. New York: Harcourt Brace Jovanovich. Lyons, John. 1968. Introduction to theoretical linguistics. London: Cam bridge University Press. , ed. 1970. New horizons in linguistics. Harmondsworth: Penguin. Martin, J.R. 1987. The meaning of features in systemic linguistics. In M.A.K. Halliday and R.P. Fawcett, eds., New developments in systemic linguistics: Volume 1 theory and description. London: Frances Pinter. McCord, M.C. 1975. On the form of a systemic grammar. Journal of Lin guistics 11,2:195-212. Morreall, J., ed. 1983. The Ninth LACUS Forum 1982. Columbia: Hornbeam Press. Searle, J.R. 1969. Speech acts: an essay in the philosophy of language. Lon don: Cambridge University Press. Sweet, Henry. 1898. New English grammar part II — syntax. London: Oxford University Press. Trudgill, Peter. 1974. Sociolinguistics — an introduction. Harmondsworth: Penguin. Turner, Geoffrey. 1973. Social class and children's language of control at age of five and age seven. In B. Bernstein, ed., Class, codes and control: volume II. London: Routledge and Kegan Paul. Winograd, Terry. 1972. Understanding natural language. Edinburgh: Edin burgh University Press.
SEMANTICS FOR A SYSTEMIC GRAMMAR: THE CHOOSER AND INQUIRY FRAMEWORK
Christian Matthiessen University of Southern California/ Information Sciences Institute Marina del Rey, California
1.0 The Task of the Chooser and Inquiry Framework One of the current research areas involving systemic linguistics is text generation (discourse production). Text generation is one way of studying text. It is text study by synthesis rather than by analysis; deconstruction and then reconstruction. 1 The basic question is: given a communicative purpose (goal) in a natural context, how does the system of linguistic processes and resources work to produce a text? Or, how do we get from situation to text? The answer includes several levels (strata) of organization as well as the interaction between these levels: a specification of the goals (purposes) of the intended text, a specification of the relevant parts of the system's field of experience, text planning, grammatical expression, and so on. For an overview of some of the issues involved in text generation, see Mann et al. 1982 and McDonald 1983. For an early systemic text generation system, see Davey 1978.2At the Information Sciences Institute in Southern California, one particular text generation system, called Penman, is being designed and implemented (see Mann 1983c). The grammar of this system is a large sys temic grammar of English, the Nigel grammar. It is the result of a major ongoing research effort into systemic grammar in the context of text gener ation. First begun as a computational grammar in 1980, it is still being expanded and revised. (For introductions to Nigel, see Mann 1984, Mann 1983a, Matthiessen 1983a, and Matthiessen 1984.) The semantics of the Nigel grammar is a chooser and inquiry seman-
222
CHRISTIAN MATTHIESSEN
tics, which is a new development intended to deal with the problem of mak ing purposeful grammatical choices in response to a communicative situa tion. (A presentation of the framework can also be found in Mann 1983b and Mann 1982.) Given a system of options like middle 3vs. effective, trans itive vs. intransitive, or indicative vs. imperative, how is one option in the system to be chosen over another in a purposeful way? The term semantics may suggest other tasks in addition to or instead of what is presented in this paper. In Section 6, I will discuss different approaches to semantics and different tasks for semantics briefly. The chooser-and-inquiry framework does not define a system network of semantic features at present. The framework addresses part of the task Robin Fawcett specifies for the Problem Solver of his model of communica tion: 4 Fawcett (1980; 63) describes the tasks of the Problem Solver as fol lows: The decisions as to which semantic features are to be selected are taken in the light of the relevant knowledge of the universe . . . The problem solver therefore not only has the task of deciding on the best general tactics to help solve the problem, but also of helping in the selection of actual semantic features.5
2.0 The Chooser of a System Asks an Inquiry In a discussion of transitivity in English (1970a), Halliday observes that there are two basic systems in English that 'occur side by side', the transi tive system (transitive vs. intransitive) and the ergative system (effective vs. middle). He characterizes the general semantic difference as follows, using questions: The transitive system asks 'does the action extend beyond the active partic ipant or not?'; the ergative, 'is the action caused by the affected participant or not?' (157-158)
In Halliday 1970a the questions are used as a presentational technique. (For another illustration of the use of questions with systems, see Halliday 1985a: 58.) But in fact they illustrate the approach to the semantic task of making purposeful grammatical choices taken in the chooser and inquiry framework. An inquiry is the question asked, for example 'is the action caused by the affected participant or not?'. It demands the information
SEMANTICS FOR A SYSTEMIC GRAMMAR
223
needed to make a choice. The answers to these questions are called responses. A chooser is the mechanism associated with a grammatical sys tem that asks a question, i.e., presents an inquiry. 6 The chooser starts its work, presenting inquiries, when the system it is associated with has been entered. 2.1 Branching Inquiries: CommandQ I will illustrate the chooser and inquiry framework with examples from the MOOD region and systems such as PRIMARY TENSE that are affected by MOOD choices, since this region is well-known and has been discussed elsewhere by systemic linguists. I will rely on Halliday's interpre tation, and the examples can be read in the light of Halliday 1984b and 1985b. The account encoded in the chooser and inquiry framework can be compared and contrasted with Butler's work (1982), which could also have been encoded in the framework, in the way he and I have done with part of his account of modality in English. As an initial example, I will use the chooser of the MOOD TYPE system. It has two options, indicative vs. imperative. The system is represented in Figure 1.
Figure 1: The MOOD TYPE system The inquiry used to obtain the information needed to choose between indicative and imperative asks whether the speech function is a command or not: Is the speech function of the clause a command or not? — set of responses: command, non-command
Notice that it defines two branches, one for each response. (There can be more than two branches, if the inquiry is an alternative question instead of a simple yes/no question. Moreover, a chooser may contain more than one inquiry.) In Figure 2 a graphic representation is given of this kind of branching inquiry.
224
CHRISTIAN MATTHIESSEN
COMMAND is the speech function of the clause a command or not? NON-Command Figure 2: Branching inquiry in MOOD TYPE chooser We will call this inquiry CommandQ; the Q-tag is used in all branching inquiries. (Instead of simply calling the responses yes and no, we give them labels that indicate the content specific to their inquiry; they can be viewed as semantic features.) 2.2 Identifying Inquiries: MoodID In the inquiry Is the speech function of the clause a command or not?, there is one nominal group with specific reference, the speech function of the clause. It represents the parameter of the inquiry whose referential identities change according to the context. Grammatically, the speech func tion will be expressed by the MOOD of the clause, and we can use the names of these grammatical functions as the names of the parameters: 'is the speech function MOOD a command or not?' We need a means of identifying the referents (values) of these parameters and we use identifying inquiries to accomplish this, one for each parameter that occurs in one of the inquiries. So, for example, MOOD can be identified by means of the following inquiry, called MoodID (all iden tifying inquiries are tagged-ID): What is the identity of the speech function of the clause? Identifying inquiries establish an identifying symbolic relation between a grammatical function like MOOD and a value. The value is defined at the stratum above the inquiry semantic stratum; cf. section 3.1 below. To sum up, there are two types of inquiry, branching inquiries, which are like yes/no questions (or alternative questions), and identifying inquiries, which are like wh- questions. 2.3 The Choose Operator; Choice Conditions I have noted that a branching inquiry like CommandQ is asked so that
SEMANTICS FOR A SYSTEMIC GRAMMAR
225
the chooser can act on the response and choose the appropriate grammati cal feature. This choice is a kind of preselection from the chooser (at the chooser and inquiry stratum) that is activated whenever the response to an inquiry motivates the selection of a grammatical feature (at the grammatical stratum). We call the operator Choose and use it as illustrated in the follow ing example: [if] COMMAND [then] (Choose imperative) The Choose operator takes one operand, the grammatical feature to be selected; and the choice is conditional upon the response to a branching inquiry.7 A response is, then, the choice condition for a grammatical fea ture. One way of representing (one aspect of) the meaning of a grammatical feature is to specify its choice condition(s). 8 We can assume that the response NON-COMMAND is the choice condition for the selection of indicative: [if] NON-COMMAND [then] (Choose indicative) 2.4 Choosers: The MOOD TYPE Chooser I have introduced some of the most important abstractions of the chooser and inquiry framework; we are now in a position to discuss the full chooser of the MOOD TYPE system. First, let's return to CommandQ (the possible responses are given in parenthesis): Is the speech function MOOD a command or not? (COMMAND, NON-COMMAND) The MOOD TYPE chooser contains only this inquiry, since the inquiry fully determines the possible choices in the system. If the response is COMMAND, the feature imperative is chosen, and if the response is NON-COMMAND, the feature indicative is chosen. We can abbreviate the inquiry as (CommandQ MOOD), which makes it easier to represent the chooser graphically in Figure 3. 2.5 Chooser-based Typology of Systems Given the two choice conditions COMMAND and NON-COM MAND, we have a positive condition for choosing imperative and a nega tive one for choosing indicative, i.e., negative in the sense that we choose this latter feature when the response to CommandQ is the negative one.
226
CHRISTIAN MATTHIESSEN
COMMAND (Choose imperative) (CommandQ Mood)
NON-COMMAND (Choose indicative) Figure 3: MOOD TYPE chooser The distinction between positive and negative responses as choice condi tions gives us a basis for a typology of grammatical systems. For example, as characterized here, MOOD TYPE is a privative system (to borrow Trubetzkoy's terminology for phonological oppositions)/We could charac terize it as equipollent, if the MOOD TYPE chooser contained an inquiry or set of inquiries defining positive choice conditions for both features of the system. Space limitations prevent further discussion of the typology here, and I will just list the two types mentioned: 9 1. Privative systems: A positive choice condition is defined only for one of the terms in a two-term system. 2. Equipollent systems: Positive choice conditions are defined for all the terms in the system. 2.6 Descriptive Claims Like all inquiries, CommandQ represents a claim about the semantics of the MOOD TYPE system that we can test and argue about. We may want to elaborate on the inquiry and make the notion of command more detailed and specific. The chooser as a whole also represents a descriptive claim. For exam ple, we may want to argue that there should be two positive choice condi tions rather than just one or that the chooser should contain a disjunction of choice conditions for the choice of a feature (see below).
3.0 Inter-stratal Organization: Above & Below Inquiries Figure 4 gives a diagrammatic summary of the MOOD TYPE system
SEMANTICS FOR A SYSTEMIC GRAMMAR
227
and chooser as a pair in grammar and semantics respectively. Note that only the system is part of a network. The inquiry CommandQ is part of the chooser of the MOOD TYPE system but it is not related to the inquiries of other choosers; see section 4 below. The stratum (level) below the inquiry stratum is the grammatical stratum. There is also a "stratum" above the semantic inquiry stratum, cal led Environment in the diagram. I will discuss it briefly below. 3.1 Above — the Environment: What Inquiries are Asked of How do inquiries get answered; where do responses come from? What justifies the response COMMAND? There is a fairly short answer and a longer one. I will only sketch the short answer here. 10 ENVIRONMENT COMMAND CommandQ
SEMANTICS
group Figure 4: MOOD TYPE: system and its chooser The Nigel component, the grammar with its chooser and inquiry semantics, is embedded in a text generation system. Among other things, the system contains a knowledge base representing the speaker's (i.e., the system's) domain/field of context11 (relevant for the creation of any given text) and a detailed text plan (for any given text). These components are outside ('above') Nigel and are thus part of Nigel's environment. When Nigel presents inquiries, we can think of these inquiries as being presented to Nigel's environment. The knowledge base and the text plan are then the sources for the information used to determine the response the environ-
228
CHRISTIAN MATTHIESSEN
ment should return. For example, the level of social context Halliday iden tifies (see 1984b: 12) in his model of dialogue can be represented in the environment and used to determine the appropriate response to an inquiry like CommandQ. So if the move is one of demanding goods-&-services, the response to CommandQ should be COMMAND. Implemented inquiries. In the text-generation system, the environment is not represented in vernacular English. Rather, formal notations are used. An inquiry operator is said to be implemented if it is represented in terms of a formal notation in the environment rather than just glossed in English. What this means in practice is that the inquiry can be answered automati cally from the environment rather than manually by somebody reading the English question. For example, MoodID is implemented in such a way that it identifies a speech functional operator in a logical form. CommandQ can then be answered after an examination of the type for that operator. 12 3.2 Below — Grammar The operator Choose relates the branch of a branching inquiry (at the inquiry stratum) to a grammatical feature in a system (at the grammatical stratum). It is represented by a vertical link in the diagram above (Figure 4) to show its role as an inter-stratal realization operator. 13 For example, COMMAND is linked to imperative. There may be neutralization between the inquiry stratum and the grammar stratum. In particular, there may be different choice conditions for choosing one grammatical feature, a situation illustrated below for the PRIMARY TENSE chooser in Section 4.1.
4.0 Intra-stratal Organization I have only discussed an isolated inquiry, CommandQ. It is compara ble to an isolated system. A system is just one part of the system network, the paradigmatic organization within the grammatical stratum. I will explore the paradigmatic organization of the inquiry stratum in two steps. First I will take a look at the organization within the chooser of a system. I have only discussed a one-inquiry chooser so far, but choosers often contain more than one inquiry. I will use the PRIMARY TENSE chooser as an example. Then, I will discuss the possibility of inquiry organization beyond
SEMANTICS FOR A SYSTEMIC GRAMMAR
229
the single chooser, the possibility of a network of inquiries. 4.1 Inquiry Organization within a Chooser: PRIMARY TENSE We can start by locating the PRIMARY TENSE system in the system network. The system can only be reached in indicative clauses and then only if the clause is temporal rather than modal since a modal auxiliary does not normally show a primary tense distinction (see Halliday 1970b). The relevant part of the system network is shown in Figure 5.
Figure 5: The PRIMARY
TENSE system in the network
The PRIMARY TENSE system has three options, past vs. present vs. future. Interpreting the system as a three-term one rather than as a two term one has to be justified. The inquiries to be presented also have to be justified. I have done that elsewhere (e.g., Matthiessen 1983b); here I will merely use the chooser as an illustration. The selection of one of the PRIMARY TENSE features expresses the temporal relation between two times, T0 and T . The first one is 'now', the time of speaking. The second one may or may not be the time of the PRO CESS of the clause, depending on SECONDARY TENSE selections (cf. Halliday 1976). If T1 temporally precedes T , past is chosen. If T1 does not precede T 0 , there are two possibilities depending on whether a logical/temporal condition is being expressed or not. If a condi tion is being expressed (resulting in for example a when- or an //-clause), the case in which T1 does not precede T0 leads to the selection of present (e.g., When Henry gets a new job we will all celebrate). If no condition is being expressed, a further distinction of temporal precedence must be made. If T0 precedes T , future is chosen. If no prece dence relation obtains between the two times, present is chosen.
230
CHRISTIAN MATTHIESSEN
How can we represent these cases in the chooser? We need one inquiry, call it PrecedeQ, that asks whether a precedence relation obtains between two times or not: PrecedeQ: Does TIME-X temporally precede TIME-Y or not? (PRECEDE, NON-PRECEDE) PrecedeQ has two parameters, the two times whose temporal relation is under investigation. In the chooser we ask PrecedeQ twice, first for (T , T0) and then (if the response is NON-PRECEDE) for (T 0 , T1 We also need an inquiry to find out whether a logical/temporal condi tion is being expressed, call it ConditionQ: ConditionQ: Does the situation to be expressed constitute a logical or temporal condition on some process, i.e., does it set up logically or temporally the possible world in which this process is instantiated? (CONDITION, NON-CONDITION) For example, when Henry gets a new job is a temporal condition on the process of celebration in we will all celebrate. (The parameter ONUS repre sents all of the referent situation in Fawcett's terms.) The organization of the PRIMARY TENSE chooser is set out in Fig ure 6. (I have turned the tree around simply for reasons of space.)
Figure 6: Chooser of PRIMARY
TENSE
Inquiry presentation conditions. The response to (PrecedeQ T1 T0) is PRECEDE, the feature past is selected. If it is NON-PRECEDE, the
SEMANTICS FOR A SYSTEMIC GRAMMAR
231
chooser asks ConditionQ. We can say that this second inquiry is dependent upon (embedded under) the ÎNON-PRECEDE response from the first inquiry; it is only asked under the condition that NON-PRECEDE is the response. We can think of NON- PRECEDE as the inquiry presentation condition for the second inquiry. It is comparable to an entry condition of a grammatical system consisting of a single feature. When the response to ConditionQ is CONDITION, present is chosen. NON-CONDITION is the inquiry presentation condition for a third inquiry in the chooser. PrecedeQ is used again, this time with the time parameters reversed: (PrecedeQ T0 T1). The positive response to this third inquiry, PRECEDE, leads to the selection of future. The response NON-PRECEDE also leads to a choice, present. How is this justified? Our assumption is that the choice condition for present is that no precedence relation obtains between T0 and T1 and at this point in the chooser that is precisely the information we have obtained. The inquiry presentation condition for the second inquiry is NON-PRE CEDE (i.e. T1 does not precede T ) and the response under which present is chosen is also NON-PRECEDE (i.e. T0 does not precede T1). In other words, the information accumulated at the point at which present is chosen is that no precedence relation obtains between the two times. This illus trates the importance of the inquiry presentation condition of an inquiry. Notice furthermore that the fact that (PrecedeQ T1 T0 ) is the first inquiry about precedence (rather than (PrecedeQ T0 T1)) is significant: the choice reasons for choosing past (the PRECEDE response) are unaffected by the further distinctions made in the chooser. In contrast, the reasons for choosing present and future are not. Consequently, we want the reason for choosing past to be branched off initially in the chooser. There are no complex inquiry presentation conditions in the present version of the inquiry framework. In other words, there can be no conjunc tion or disjunction of branches before an inquiry is presented. Conjunction can be simulated by accumulating responses as the depth increases in the tree of inquiries. Disjunction can be simulated by using the same inquiry in more than one place. The same techniques have sometimes been used in socalled 'displayed system networks' (see Fawcett forthcoming). Neutralization. The PRIMARY TENSE chooser also illustrates how disjunctive choice conditions are handled. Notice that there are two 'paths5 leading to the choice of present. These are two alternative reasons for choosing the feature present. This disjunction of choice conditions is not
232
CHRISTIAN MATTHIESSEN
represented explicitly in the chooser but can only be inferred from the repetition of '(Choose present)'. In principle, this situation is comparable to an upward 'or' in stratificational theory (cf. for example Lamb 1971).14 4.2 Organization beyond the Chooser — Inquiries and the Network: The MOOD Region To explore some aspects of inquiry interaction beyond a single chooser like the PRIMARY TENSE chooser, we can consider the MOOD fragment of the clause system network in Figure 7. The interactions among inquiries I will point to are implicit rather than explicit as in the chooser tree of the PRIMARY TENSE chooser.
Figure 7: MOOD grammar: network First, consider the flow of information involving identifying inquiries as we move from left to right in the network. As soon as the feature clause can be chosen, MoodID is asked to identify MOOD. Consequently, the identity of MOOD has been established by the time the chooser of MOOD TYPE is reached and (CommandQ MOOD) is asked. Second, the branching inquiry of one chooser may set up the context in
SEMANTICS FOR A SYSTEMIC GRAMMAR
233
which the branching inquiry of the chooser of a more delicate system can be asked sensibly. For example, inquiries distinguishing between statements and questions are only asked after it has been established that the speech function is not a command. The system network takes care of this sequence of asking inquiries, since the system INDICATIVE TYPE is dependent on the choice of indicative in the system MOOD TYPE. Similarly, the organi zation of the system network ensures that no inquiries about temporal rela tions are asked if the speech function is a command, since the PRIMARY TENSE system is only entered in indicative clauses (more particularly, tem poral clauses) and not in imperative clauses. 4.3 A Network of Inquiries?: POLARITY To sum up, there is an implicit organization of the inquiries beyond the individual choosers. It follows from the system network organization in the grammar. Is this organization sufficient or do we need an explicit inquiry network comparable to the explicit system network? I think there is reason to believe that an explicit inquiry network would be quite useful. For instance, the system POLARITY is simultaneous with the system MOOD TYPE, which is quite justified in terms of the grammatical network. How ever, the conditions under which negative is chosen are not the same for all speech functions. In particular, questions concerned with the value of the polarity itself, so-called polarity or yes/no questions, constitute a special context for choosing the value of POLARITY: the choice of negative indi cates a biased assumption/expectation on the speaker's part about the value of the polarity; it is an interpersonal comment about assumption/expecta tion. (For example Isn't Henry coming over this afternoon? can mean T thought he was'.) Since there is no way at present of making the polarity inquiries explicitly dependent on the inquiry that establishes whether the speech function is a polarity question or not, that inquiry has to be asked in the POLARITY chooser as well as in the INTERROGATIVE TYPE chooser. This situation could be avoided with an explicit inquiry network. 15 There is another aspect of the use of inquiries which I will only men tion in passing. Systemicists have used daggers and other symbols in system networks to indicate various dependencies between pairs of features that are not captured in the network organization itself. These conventions can be captured at the inquiry stratum, e.g., by asking the same branching inquiry twice in two different choosers and then choosing the appropriate features for a given response.
234
CHRISTIAN MATTHIESSEN
5.0 Factoring of the Framework I have now introduced the whole chooser and inquiry framework and will summarize it and relate it to other work on systemic semantics. To sum marize the framework, I will factor it into five design assumptions: 1. Semantic features like EXTERNALLY CAUSED, STATEMENT, POSITIVE, PRECEDE and so on can be used to specify the semantic distinctions correlating with grammatical distinctions. 2.
Semantic feature labels are abbreviations of explicitly statable choice conditions, specified either in an informal English text or in a more formal notation.
3. Choice conditions can be viewed as the responses to inquiries pre sented to Nigel's environment. An inquiry defines' a minimal deci sion tree with two or more branches. 4. Not only grammatical choices can be associated with the response branches in this tree, but also additional branching inquiries, giving more tree organization. 5. Inquiries come in chooser 'parcels', one parcel for each system. The inquiries of a chooser may be embedded under one another, but they are not related to the inquiries of other choosers in this fash ion. A chooser starts its activity when its system is entered. As a result of the 'chooser parcelling', choices are always local to the sys tem whose chooser an inquiry is a part of. The first design assumption shows how the chooser-and- inquiry framework is related to approaches that make use of semantic features or components, 'componential semantics'. There are differences, some reflected in the sub sequent design assumptions and some having to do with the domain the frameworks are applied to. Componential analysis has typically been used as a tool in lexical semantics.16 The notion of choice condition is similar to Robin Fawcett's procedural felicity conditions (cf. 1980: 253 and 1983). The explicit specification of choice conditions also suggests some relation to the notion of truth condi tion. The latter is of course a narrower notion, being grounded in truth, and is typically not part of a functional approach to semantics. Fawcett notes that linguists 'are discovering that the specification of the meanings built into the code of language requires to be stated in complex terms of felicity conditions, truth conditions (which are really just a sub- category of felicity
SEMANTICS FOR A SYSTEMIC GRAMMAR
235
conditions), etc.; and moreover that these approaches to meaning can always be interpreted as a PROCEDURAL SEMANTICS that relates fea tures of a language to knowledge of the universe' (1980: 253). The notion of choice condition is also related to the use of 'semantic' conditions on the application of 'syntactic rules' in various other approaches to generation; see for example Sowa 1983. The third assumption, the view that choice conditions can be seen as responses to inquiries, determines the minimal organization of semantic features/choice conditions/responses into the paradigms defined by branch ing inquiries. This assumption brings out a similarity with the notion of the systemicization of features. The responses to a branching inquiry are dis junctive just as the output features of a system are disjunctive; in both cases the disjunction represents a minimal alternation. The fourth assumption, that a branching inquiry can be embedded un der the response to another inquiry, determines how inquiries can be organized into a decision tree with more than one branching point and thus more than minimal depth. It determines how inquiries can be related by dependency and vary in delicacy just as systems can. It also determines the current restrictions on inquiry organization (within a chooser), since trees do not have double motherhood (conjunction) or alternative motherhood (disjunction). The fifth assumption, the chooser parcel assumption, defines a limit on how inquiries can be inter-related (at present), since only inquiries within the chooser of one system can be related by embedding. This assumption suggests how a collection of inquiries differs from a collection of systems; the latter collection forms a huge network, the former only fragments cor responding to choosers. The factoring of the framework helps us see how it is composed. It also helps us see how we can create variants of it. For example, we can change the last assumption and let inquiries depend on one another across the boundaries defined by the chooser of a particular system. We can also change the fourth assumption so that we allow disjunction in the embed ding of an inquiry (not presently allowed).
6.0 Semantics and the Consumer There have been different approaches to semantics in systemic linguis tics. They differ, I think, mainly in terms of what questions they took as the
236
CHRISTIAN MATTHIESSEN
starting point in the development of semantics. 6.1 Chooser-inquiry Semantics and the Consumer The present design of the chooser and inquiry framework should be understood in the context of the task it was created to perform. A few years ago, we said to ourselves: we are developing a large computational systemic grammar of English as part of a text-generation system. One key issue for the success of the grammar is that grammatical choice must be made in a purposeful way, in response to a given communicative situation. How can we control grammatical selection in a purposeful way? This was a consumer demand on (systemic) linguistics. In response to this demand, the chooser and inquiry framework was developed. Since we already had a grammar that had been organized to be sensitive to semantic distinctions, we took it as our starting point and created choosers for the grammatical systems. This approach seems similar to the one Robin Fawcett envisions; cf. the remarks on his Problem Solver in the introduction. 6.2 Sociological Semantics An alternative strategy is to take communicative goals as a starting point and write a semantics that is a strategy or set of strategies for pursuing those goals. Turner's work (1973) can be described in that way. For exam ple, if you are a mother and your goal is to control your child so that he or she does not do something undesirable, what are the meanings available to you to pursue this goal? This strategy for approaching semantics, let's call it sociological semantics (cf. Halliday 1973), differs from the chooser and inquiry framework in a number of ways. In particular, sociological seman tics is situation specific and is designed to control grammatical generation from the semantic stratum, e.g., by using preselection. 6.3 Compatibility of Chooser-inquiry Semantics & Sociological Semantics However, it is not incompatible with that framework; rather, the two approach the problem of interfacing context and grammar from different angles: 1. Sociological semantics takes purposes (goals) as a starting point; chooser and inquiry semantics takes the grammatical choices to be made in a purposeful way as a starting point.
SEMANTICS FOR A SYSTEMIC GRAMMAR
237
2. As a result, it can be argued, the semantic unit of a sociological semantics is (a whole) text; the 'domain' of chooser and inquiry semantics is whatever corresponds to a clause (complex) and so on down the grammatical rank scale. 3. Sociological semantics tends to be written for particular contexts; it is situation specific. As a result, the first step in the sociological semantics 'programme' would be a collection of various particular semantices rather than one generalized semantics. In contrast, the work on the chooser and inquiry framework has been carried out under the assumption that inquiries address general distinctions, not situation specific distinctions. The result is one collection of choosers for the entire grammar, not different collections for differ ent tasks. There are obviously a number of ways to interpret these differences. I will not pursue them here, but will leave suggestions for another time. Both sociological semantics and the chooser and inquiry framework are attempts at creating a functional kind of semantics. Both have the task of an inter-level. In Towards a sociological semantics', Halliday (1973: 64) writes that semantics is an interface which relates non-language to lan guage; it 'represents the coding of the "input" to the linguistic system'; 'it is the strategy that is available for entering the language system'. In a similar way, it is the task of the chooser and inquiry framework to relate grammar to expressive demands in the environment. There are other tacks that semantic frameworks have been created to deal with. For example, formal semantics addresses issues having to do with inference and interpretation in a model. The chooser and inquiry framework does not deal with (semantic) inference at all. We should not see this as a failing of the framework. It was developed to deal with a specific task and inference is simply a different task. Inference is a different kind of 'consumer need' and will lead to a different kind of semantics.
7.0 Conclusion The fact that the chooser and inquiry framework was developed in the context of text-generation explains its origins. However, it does not restrict its range of uses. (The reason for this is fairly obvious: text-generation is a
238
CHRISTIAN MATTHIESSEN
natural task, something speakers (and writers) always engage in.) For example, once the chooser and inquiry framework had been developed it was possible to make very definite statements about the choice conditions of grammatical features, As a result, it is also possible to argue from choos ers to systems: arguments having to do with the content of inquiries and their arrangements in choosers are arguments 'from above' about the organization of grammar. In practical work, it is useful to imagine the inquiries for a system that would lead to appropriate choices of its features. If it is not possible to word these inquiries, it is quite likely that the gram mar should and could be re-organized. As an additional example, consider the reliance in one inquiry of a par ticular chooser on the previous identification of one or more of its paramet ers in an earlier chooser. This reliance represents an interdependence in terms of the flow of information. Such interdependences can, I think, be used as strong indications of factoring into meta-functions; they are inter dependences over and above the grammatical interdependences rep resented in the grammatical network.
NOTES 1.
Text generation is usually focused on the system rather than reconstructions of particular instantiations of the system.
2.
The computer has often been a tool in research on text generation. It is not a necessary component, but the computer makes the management of the various parts of the model (such as the grammar) much easier and it also makes it possible to test the whole model of the text generation system or parts of it. At the same time, the computer puts certain demands on the model. For example, all the details have to be made explicit; nothing can be left to the linguist to fill in.
3.
Grammatical features will be marked by underlining.
4.
Fawcett's Problem Solver also deals with tasks other than those handled by the chooser and inquiry framework. For example, it assembles the referent situations to be expressed by the semantics. In general, the tasks of the Problem Solver are decomposed and distri buted across more than one process in our model of text generation; cf. for example Mann 1983c.
5.
What Fawcett refers to as semantic features are grammatical features in Nigel.
6.
Another term we have sometimes used is choice expert. It suggests a little process (proce dure) with a very limited domain of expertise, viz. how to make a purposeful choice in one system by asking relevant questions. The choice experts are oblivious of one another but achieve the right result collectively. I will discuss the issue of what responds to the inquiries presently, but for now we can think of it simply as stratally above semantics. It
SEMANTICS F O R A SYSTEMIC G R A M M A R
239
is important to note in this context that the responses to inquiries are based on presenta tional considerations and not (directly) on an examination of the so-called real world. For instance, the inquiry 'is the action caused by the affected participant or not?' is not a question about the real world but a question about how a given (referent) situation is to be presented. 7.
There is another choice operator, DefaultChoose, which is used in choosers of systems where one feature is to be chosen unless another feature has been preselected by the grammar. Choosers of this kind do not contain any inquiries, since the appropriate choice is a grammar-internal matter. For example, the chooser of the system PRONOUN CASE, nominative vs. oblique, only contains '(DefaultChooser oblique)', which is over ridden when a Subject is preselected to be nominative.
8.
Often the choice condition is a single response to one inquiry, but it is not uncommon for a choice condition to consist of a combination of responses to consecutively presented inquiries; see the discussion of the PRIMARY TENSE chooser below and Mann 1983a.
9.
The distinction between privative and equipollent systems based on chooser differences is often reflected in the feature names of the systems. Privative systems tend to have one term with non- in the name, e.g.. non-cause vs. cause. The distinction is also reflected in realization statements. Privative systems usually only have a statement associated with the positive term. McCord gave graphic recognition to systems of this type in his revision of Hudson's systemic grammar (McCord 1975).
10.
The longer answer to the question above would have to go into detail about exactly how the environment is organized (cf. Matthiessen 1987) and how the inquiries are interpreted by the environment and how the appropriate responses are calculated. It would also have to address how the inquiries we have met in the shape of English questions are rep resented in a more formal inquiry notation.
11.
The term domain is used in computational linguistics to mean something like the systemic notion of field.
12.
Implementation is necessary for automatic text-generation, but it also has other potential theoretical value. When I discuss inquiries here, I use English glosses; CommandQ means 'is the speech function MOOD a command or not?'. These glosses are very helpful when we try to think about the inquiries, but there is a very real problem in using English as its own metalanguage. For example, if 'command' is the inquiry gloss of the grammatical feature imperative, what is the gloss of 'command'? Similarly, what is the gloss of the fea ture plural or the feature past? The problem we face has been discussed by Halliday in his paper on the ineffability of grammatical categories (Halliday 1984a). There are various ways of approaching the problem of ineffability and Halliday dis cusses several. For example, we can explore situation specific semantices as well as one highly generalized semantics. I will return to this topic briefly below. Another possible approach is to use a semiotic system other than English for the glosses. (The two approaches are not mutually exclusive.) This is one way in which an implemented version of the 'environment' in Figure 4 becomes relevant.
13.
The comparable inter-rank operator is Preselect in Nigel.
14.
The correlates of the inter-stratal connection types Lamb identifies are as follows:
240
CHRISTIAN M A T T H I E S S E N
interstratal relation
strat. theory
diversification
downward or
Nigel
neutralization
upward or
repetition of Ch oose-statement
portmanteau realization
upward and
succession of inquiries followed by Choosestatement
composite realization
downward and same inquiry in different choosers leading to different choices
15.
Notice that inquiry dependencies of this kind point to metafunctional organization that is not explicit in the system network since the MOOD TYPE and POLARITY systems are simultaneous. In other words, the use of inquiries can be used as evidence of the metafunctional groupings in a language. In the case of POLARITY, the inquiry depen dency is a piece of evidence in favour of interpreting it as an interpersonal region.
16.
The chooser and inquiry framework is at present only used for what is called grammatical semantics in Halliday 1966;lexical semantics is handled in a different way; cf. Matthiessen 1981.
REFERENCES Butler, C.S. 1982. The directive function of the English modals. University of Nottingham Ph.D. dissertation. Davey, A. 1978. Discourse production. Edinburgh: Edinburgh University Press. Fawcett, R.P. 1980. Cognitive linguistics and social interaction: towards an integrated model of a systemic functional grammar and the other compo nents of a communicating mind. (Exeter Linguistic Studies, 3). Heidel berg: Julius Groos Verlag and Exeter University. . 1983. Language as a semiological system: a re- interpretation of Saussure. In J. Morreali, ed., The Ninth LACUS Forum 1982. Colum bia, S.C.: Hornbeam Press. 59-125. . (Forthcoming). What makes a 'good' system network good? In J.D. Benson and W.S. Greaves, eds., Systemic functional approaches to dis course: selected papers from the 12th International Systemic Workshop. Norwood, N. L: Ablex. Halliday, M.A.K. 1966. Lexis as a linguistic level. In C E . Bazeli, J.C. Cat-
SEMANTICS FOR A SYSTEMIC GRAMMAR
241
ford, M.A.K. Halliday and R.H. Robins, eds., In memory of J.R. Firth. 148-162. London: Longmans. . 1970a. Language structure and language function. In J. Lyons, ed., New horizons in linguistics. Harmondsworth: Penguin Books. 140-165. . 1970b. Functional diversity in language, as seen from a considera tion of modality and mood in English. Foundations of Language 6,3:32751. . 1973. Explorations in the functions of language. London: Edward Arnold. . 1976. The English verbal group. In G. Kress, ed., Halliday: system and function in language. Oxford: Oxford University Press. 136-158 . 1984a. On the ineffability of grammatical categories. In A. Manning, P. Martin and K. McCalla, eds., The Tenth LACUS Forum 1983. Columbia, S.C.: Hornbeam Press. 3-18. (Reprinted herein pp. xxx-yyy.) . 1984b. Language as code and language as behaviour: a systemicfunctional interpretation of the nature and ontogenesis of dialogue. In R.P. Fawcett, M.A.K. Halliday, S.M. Lamb and A. Makkai, eds., The semiotics of culture and language, 1: Language as social semiotic. Lon don: Frances Pinter. 3-35. . 1985a. Spoken and written language. Geelong: Deakin University. . 1985b. An Introduction to functional grammar. London: Edward Arnold. Lamb, S.M. 1971. The crooked path of progress in cognitive linguistics. In R.J. O'Brien, ed., Monograph series on languages and linguistics. Georgetown: Georgetown University Press. 24:99- 123. Mann, W.C. 1982. The anatomy of a systemic choice. Marina del Rey, Cal.: Information Sciences Institute, University of Southern California (ISI/RR-82-104). (Also in Discourse Processes 8,1:53-74.) . 1983a. An introduction to the Nigel text generation grammar. In Nigel: a systemic grammar for text generation. Marina del Rey, Cal.: Information Sciences Institute, University of Southern California (ISI/ RR-83-105). (This paper also appears in J.D. Benson and W.S. Greaves, eds., Systemic perspectives on discourse. Vol 1: Selected theoretical papers from the 9th International Systemic Workshop. Norwood, N.J.: Ablex, 1985. 84-95.) . 1983b. Inquiry semantics: a functional semantics of natural language grammar. In Proceedings of the First Annual Conference, Association for Computational Linguistics, European Chapter. 165-175.
242
CHRISTIAN MATTHIESSEN
. 1983c. An overview of the Penman text generation system. In Pro ceedings of the National Conference on Artificial Intelligence. Ameri can Association for Artificial Intelligence. 261-265. (Also appears as Marina del Rey, Cal.: Information Sciences Institute, University of Southern California (ISI/RR-83-114.) . 1984. A linguistic overview of the Nigel text generation grammar. In A. Manning, P. Martin and K. McCalla, eds., The Tenth LACUS Forum 1983. Columbia, S.C.: Hornbeam Press. 255- 265. Mann, W.C., M. Bates, B. Grosz, D.D McDonald, K. McKeown and W.R. Swartout. 1982. Text generation: the state of the art and the liter ature. American Journal of Computational Linguistics 8,2:62-70. Matthiessen, C. 1981. A grammar and a lexicon for a text production sys tem. In The 19th Annual Meeting of the Association for Computational Linguistics. Sperry Univac. 49-57. . 1983a. Systemic grammar in computation: the Nigel case. In Pro ceedings of the First Annual Conference, Association for Computational Linguistics, European Chapter. 155- 165. . 1983b. Choosing primary tense in English. Studies in Language 7,3:369-430. . 1984. How to make grammatical choices in text generation. In A. Manning, P. Martin and K. McCalla, eds., The Tenth LACUS Forum 1983. Columbia, S. C : Hornbeam Press. 266- 284. . 1987. Notes on the organization of the environment of a text gener ation grammar. In Kempen, G., ed., Natural language generation. Recent Advances in Artificial Intelligence, Psychology, and Linguistics. Dordrecht: Martinus Nijhoff. McCord, M. 1975. On the form of a systemic grammar. Journal of Linguis tics 11:195-212. McDonald, D.D. 1983. Natural language generation as a computational problem: an introduction. In M. Brady and R. Berwick, eds., Computa tional models of discourse. Cambridge: MIT Press. 209-266. Sowa, J.F. 1983. Generating language from conceptual graphs. Computers and Mathematics with Applications 9,1:29-43. Turner, G.J. 1973. Social class and children's language of control at age five and age seven. In B. Bernstein, ed., Class, codes and control 2: applied studies towards a sociology of language. London: Routledge & Kegan Paul. 135-201.
GRAMMATICAL CONSPIRACIES IN TAGALOG: FAMILY, FACE AND FATE — WITH REGARD TO BENJAMIN LEE WHORF
J.R. Martin University of Sydney Sydney, Australia
1.0 Introduction — Some Questions Concerning Linguistic Relativity Writing in 1929, Sapir presented a rather eloquent challenge to the common-sense view that humans use language simply as a means of expres sing thoughts and reflections which are themselves interpretations of reality arrived at quite independently of the particular language a person speaks: Language is a guide to 'social reality'. Though language is not ordinarily thought of as an essential interest to the student of social science, it power fully conditions our thinking about social problems and processes. Human beings do not live in the objective world alone, but are very much at the mercy of the particular language which has become the medium of expres sion for their society. It is quite an illusion to imagine that one adjusts to reality essentially without the use of language and that language is merely an incidental means of solving problems of communication or reflection. The fact of the matter is that the 'real world' is to a large extent uncon sciously built upon the language habits of the group. No two languages are ever sufficiently similar to be considered as representing the same social reality. The worlds in which different societies live are distinct worlds, not merely the same world with different labels attached . . . We see and hear and otherwise experience very largely as we do because the language habits of our community predispose certain choices of interpretation (Sapir 1949: 162).
This was in fact the quote used by Whorf, a student of Sapir's, to introduce his contribution to the Sapir memorial volume, 'The relation of habitual thought and behaviour to language', published in 1941 (Spier 1941). The particular interpretation of linguistic relativity adopted by Whorf will be
244
J.R. MARTIN
outlined below in section 2.0. To begin, however, it is worth noting that a number of interpretations are possible; and it is by no means clear which of these Sapir had in mind. First, there is the question of whether language is a guide to reality or to social reality, Sapir appears to equivocate in the passage quoted above, not making it clear whether language predisposes ways of seeing and hear ing 'the real world', or whether it is the perception of social problems and processes which is prédetermined in some way. But it is important to distin guish the two claims. If language is a guide to reality, then there is a need for linguistics to cooperate with natural science to uncover the connection between how we talk and what we see (cf. the work of Berlin & Kay 1969 on colour terms). If on the other hand language is a guide to social reality, then linguistics must turn to social science for clarification of the relation between how we talk and what we do. Whorf clearly distinguishes between these two hypotheses; and the focus of this paper will be solely on the latter relation — of language to social reality. The second question which needs to be asked is: How does language do it? What is it about the structure of a language that predisposes certain ways of seeing the world or certain ways of acting in it? Sapir, in his work, considers mainly the influence of lexis and morphology. Whorf, making use of his analysis of covert categories, or cryptotypes, goes further, basing his arguments for the most part on a relatively rich, semantically oriented con sideration of the grammatical differences between languages. There are two factors at work here, relating to the influence particular aspects of linguistic structures might have on reality perception, social or otherwise. One has to do with consciousness. The more removed from consciousness a particular category is, the more likely it is to predispose certain choices of interpreta tion without our realising it. Thus grammar is more likely to be influential than morphology, and morphology to have more impact than lexis. Another has to do with generality. The more general a category is, the more scope it has to affect a wide range of interpretations. Again, grammar is concerned with more general meanings than morphology or lexis, and would thus appear to have more power to build the distinct worlds in which different societies might live. Whorf himself placed most emphasis on the unconscious and more general categories he uncovered as cryptotypes. In this he appears to have taken the promising tack. Conscious categories are easier to manipulate than unconscious ones. Indeed, it is by making the invisible visible that
GRAMMATICAL CONSPIRACIES IN TAGALOG
245
humans take the first step towards restructuring their world. Before, for example, feminists can challenge sexism by objecting to the opposition of Mrs. to Miss, or to the wearing of wedding rings by females, it is necessary that the titles and rings be recognised as signs symbolising a particularly dis tasteful form of social discrimination. Similarly, specific categories are much more volatile than general ones. Changes in the lexicon are far more common and rapid than changes in morphology, with grammar lagging behind both of these (of course grammatical changes may be brought about by morphological ones; but a language might well change a morphological category like gender, without a consequent change in grammatical struc ture, as English has done). Lexis represents the flexible side of linguistic structure, constantly adjusting to changes in environmental and social real ity as these appear. Thus while the number of words a language has for snow or rice is not meaningless, neither is it crucial to the discussion of lin guistic relativity. Skiers, or travellers in Asia, very quickly adjust, adding in the space of hours or a few days such lexical items as are needed. Nothing in their language keeps them from doing this. Clearly, if anything deter mines speakers' particular world views, it is not simply the conscious, specific part of their language—its lexical structure. We should be careful, however, in ruling out lexis completely. Poynton (1985) demonstrates convincingly that there are a number of lexical con spiracies in English reflecting sexisim in our culture: for example—the nature and use of diminuitive address forms, the types of adjectives used to charcterise men and women, the distribution of attitudinal lexis, and so on. Nevertheless, because lexis is relatively conscious, specific and volatile, it certainly deserves less weight than it has been given in the past in discus sions of linguistic relativity. The third and final question to be considered has to do with whether a single category is responsible for the predispositions languages condition, or whether several different categories, perhaps even a combination of grammatical, morphological and lexical, are involved. Sapir's examples (e.g. 1949: 91 [1931]; 1949: 158 [1924]; 1949: 443-4 [1916]) tend to focus on single categories or sets of lexical items. Whorf on the other hand based his arguments on sets of categories, which cumulatively oriented speakers in certain directions. Once again Whorf s position looks the more promising. Several congruent categories are likely to have a more powerful effect than single isolated ones. For one thing the congruence between the categories will be even more general and unconscious than the categories themselves.
246
J.R. MARTIN
For another, grammatical 'conspiracies' are harder to find counter-exam ples to than single categories. As far as the question of how language does it goes then, this paper will focus on conspiracies of grammatical categories in Tagalog, making very little use of evidence drawn from lexis or single categories taken on their own.
2.0 Whorf s Interpretation of Relativity 2.1 Whorf s Hypothesis As outlined in section 1, Whorf clearly recognised that the relation of language to reality on the one hand and to social reality on the other were separate issues. And he based his arguments in principle upon conspiracies of grammatical (i.e. covert) and morphological (i.e. overt) patterns, for the most part illustrated from English and Hopi, As Whorf s position has tended to be widely misinterpreted, largely through popularisation (for which Whorf himself is partly to blame), it is worthwhile quoting at length from Whorf here. First, the distinction between reality and social reality: That portion of the whole investigation [i.e. 'a comparison between Hopi and western European languages'] here to be reported may be summed up in two questions: (1) Are our concepts of lime', 'space', and 'matter' given in substantially the same form by experience to all men, or are they in part conditioned by the structure of particular languages? (2) Are there tracable affinities between (a) cultural and behavioural norms and (b) largescale linguistic patterns? (I should be the last to pretend that there was anything so definite as a 'correlation' between culture and language, and especially between ethnological rubrics such as 'agricultural', 'hunting,' etc. and linguistic ones like 'inflected', 'synthetic', or 'isolating' (1956: 1389 [1941]). Next, the significance he attached to covert systems: This illustration [Whorf is discussing Hopi verb morphology] will show how the meaning of a form in a language like Hopi is capable of being more deeply analysed by the cryptotype concept, and how the totality of meaning is a joint product of cryptotypic and phenotypic factors. In many languages the cryptotype concept would be of little use, but there are lan guages like Hopi in which much of the influential material of paradigm production lies in this heavily veiled state, just as there are people whose mental life is much less accessible than that of others. Cryptotypes play a much larger part in Hopi than this rather minor problem of inceptive forms, which however yields a neat illustration. I believe I am the first to
GRAMMATICAL CONSPIRACIES IN TAGALOG
247
point out the existence of this submerged layer of meaning, which in spite of its submergence functions regularly in the general linguistic whole (1956: 110-1 [1937]), The most impressively penetrating distinctions of this kind ['causation, action, result, dynamic or energic quality, directness of experience, etc.'] often are those revealed by analysing to the covert or even cryptotypic levels (1956: 80 [circa 1936]).
Third, the importance of congruent systems (cf. Halliday 1977: 17 [1967]): To sum up the matter, our first question asked in the beginning [as in the first quote of this section] is answered thus: concepts of 'time' and 'matter' are not given in substantially the same form by experience to all men but depend upon the nature of the language or languages through the use of which they have developed. They do not so much depend upon ANY ONE SYSTEM (e.g., tense or noun) within the grammar as upon ways of analysing and reporting experience which have become fixed in the lan guage as integrated 'fashions of speaking' and which cut across the typical grammatical classifications, so that such a 'fashion' may include lexical, morphological, syntactic, and otherwise systemically diverse means coordi nated in a certain frame of consistency (1956: 158 [1941]).
Finally, the nature of the relation between language and social reality: As for our second question [as in the first quote of this section]: There are connections but not correlations or diagnostic correspondences between cultural norms and linguistic patterns. Although it would be impossible to infer the existence of Crier Chiefs from the lack of tenses in Hopi, or vice versa, there is a relation between a language and the rest of the culture of the society which uses it. There are cases where the 'fashions of speaking' are closely integrated with the whole general culture, whether or not this be universally true, and there are connections within this integration, between the kinds of linguistic analyses employed and various behavioural reactions and also the shapes taken by various cultural developments. Thus the importance of Crier Chiefs does have a connection, not with tenselessness itself, but with a system of thought in which categories different from our tenses are natural. These connections are to be found not so much by focusing attention on the typical rubrics of linguistic, ethnog raphic, or sociological description as by examining the culture and the lan guage (always and only when the two have been together historically for a considerable time) as a whole in which concatenations that run across these departmental lines may be expected to exist, and, if they do exist, eventually to be discoverable by study (Whorf 1956: 159 [1941]).
2.2 Whorf s Orientation Whorf s interest in relativity had a great deal to do with his orientation
248
J.R. MARTIN
to the goals of linguistic inquiry. As a student of Sapir's, and in the tradition of American anthropological linguistics (cf. Hymes & Fought 1981 [1975]), Whorf felt that the purpose of linguistic analysis was to make statements of meaning: What needs to be clearly seen by anthropologists, who to a large extent may have gotten the idea that linguistics is merely a highly specialised and tediously technical pigeonhole in a far corner of the anthropological work shop, is that linguistics is essentially the quest for MEANING (Whorf 1956: 73 [1936]). In this Whorf differed sharply from many of his American contemporaries, particularly those associated with mainstream Bloomfieldian structuralism (as represented for example in the Joos 1957 collection). His views are in fact notably Firthian in character: In brief, linguistics accepts speech and language texts as related to the liv ing of, and therefore to the 'meaning' of, life, and applies its theory and practice as far as it is able, to the statement of such 'meaning' in strictly lin guistic terms (Firth 1968: 169 [1957]). Underpinning this interest in the semantics of languages was Whorf s approach to grammatical analysis. Consider for example the following pas sages which illustrate the way in which his analysis of cryptotypes led to the consideration of many aspects of language which linguistics would later come to approach as 'deep grammmar': In English, intransitive verbs form a covert category marked by lack of the passive participle and the passive causative voices; we cannot substitute a verb of this class (e.g. 'go, lie, sit, rise, gleam, sleep, arrive, appear, rejoice') into such sentences as 'It was cooked, It was being cooked, I had it cooked to order' (Whorf 1956: 89 [1945]). The English intransitive verbs as configuratively defined above [i.e. the immediately preceding quote] are a cryptotype. A similar cryptotype com prises the verbs of 'copulative resolution' ('be, become, seem, stay, remain', etc.), which also lack the passive and causative but may be fol lowed by nouns, pronouns and adjectives. Transitives (a cryptotype which includes 'run, walk, return', etc.—indeed most English verbs) possess the passive and causative and may be followed by nouns and pronouns but not by adjectives alone (Whorf 1956: 92 [1945]). Thus, in English, verbs referring to the subject's ego-field experience use the simple present tense for momentary present fact, and not the present progressive. Other verbs employ the present progressive for either momentary or continued present fact, and the simple present (except in special locutions like 'here he comes') for the nomic or customary tense
GRAMMATICAL CONSPIRACIES IN TAGALOG
249
aspect. Foreigners learning English do not know this and hence say 'I am hearing you, he is seeing it' (Whorf 1956: 165 [1940]).
In this connection note the parallels between the latter passage and the fol lowing excerpt from Halliday's most recent discussion of TRANSITIVITY in English: A third distinction between material and mental processes is that of tense . . . In a mental process, the unmarked present tense is the simple present . . . But in a material process the unmarked present tense is the present in present . . . The simple present with a material process is general or habitual . . . The present in present with a mental process is a rather highly conditioned kind of inceptive aspect (Halliday 1985: 109).
2.3 Whorf's Method Whorf s approach to cryptotypes was naturally less explicitly for malised than many contemporary approaches to deep grammar. Neverthe less, the approach was a rich, semantically oriented one. Consider the fol lowing illustration from Tagalog, whose basic clause and verbal structure will be briefly outlined. A basic sentence in Tagalog consists of a Predicate, which may be realised by verbal, adjectival and nominal items, followed by a number of participants and circumstances. One of these complements is selected as Topic, as realised through nominative case markings. Non-Topic comple ments receive objective case markings if participants and oblique case markings if circumstances. This structure is presented abstractly in Figure 1, and illustrated in 1, where Goal, the subversive, has been selected as Topic. The experiential role of the Topic in the clause is signalled by an inflection on the verb — in this case, the infix -in- marks the Topic as a Goal.
(1)
Hinuli ng heneral ang subersibo sa probinsya caught NTM 1 general TM subversive CM province T h e general caught the subversive in the province'.
250
J.R. MARTIN
As well as signalling the experiential role of the Topic through what is generally referred to as its FOCUS system, Tagalog verbs also realise ASPECT, and a number of modulations having to do with causation, abil ity, social or reciprocal activity, and so on. FOCUS, ASPECT and MOD ULATION are realised through a variety of morphological processes: pre fixes, infixes, suffixes, and two types of reduplication—one reduplicating the verb stem's first syllable, another its first two syllables. These patterns are outlined in Figure 2 and illustrated by means of the Goal focus forms for the stem huli 'catch' in (2). There is no simple relation of FOCUS, ASPECT and MODULATION to morphology. The -in- infix in (1) for example is a portmanteau form realising both completed Aspect and Goal focus; and the form itself is conditioned by the affix correspondence class of the stem (for details see Schachter and Otanes 1972, Chapter 5). [prefix] - S - [infix] - Tem - [suffix] R1 R2R2 Figure 2: Basic verb strucutre in Tagalog (R2R2 symbolises reduplication of the first two syllables of the stem) (2)
stem = huli 'catch' (Goal/Topic) hulihin 'to catch' hinuli 'caught' huhuli 'will catch' mahuli 'to be able to catch' huli-hulihin 'to catch a little' 2
Whorf's approach can be illustrated with respect to the covert class or cryptotype which might be glossed along the lines of: mental process verbs of perception (treated as perception verbs in De Guzman 1978: 192, 298 and included under involuntary action verbs by Rafael 1978)—for example kita 'see', dama 'perceive', pansin 'notice', dinig 'hear', masdan 'observe', tanaw 'view', batid 'be aware of, kilala 'to come to know', etc. Verbs belonging to this cryptotype behave differently from other verbs when pre fixed with what would on the surface appear to be straight-forward ability/ involuntary action morphemes (i.e. malmaka). Stems not belonging to this class (or to the class of mental process verbs of cognition which behave similarly), such as huli 'catch', have to be glossed in terms of 'to be able to/
GRAMMATICAL CONSPIRACIES IN TAGALOG
251
to happen to' when prefixed with ma- or maka-. Ma- realises Goal focus as in (3); maka- Actor focus as in (4) (the nalnaka- forms in these clauses sig nal completed aspect). (3)
Nahuli ng heneral ang sub ersibo caught NTM general TM subversive 'The general was able to/happened to catch the subversive'.
(4)
Nakahuli ng subersibo ang heneral caught NYM subversive TM general 'The general was able to/happened to catch a3 subversive'.
With mental process verbs of perception however, the ma- prefix does not carry the expected ability/involuntary action meaning (it is possible that ma- may have lost its ability/involuntary action meaning because of the nature of this class of verbs as will be discussed in section 3.3 below). The parallelism with (3) and (4) breaks down, with the Actor focus form chang ing the meaning of nakita in (5) from simply 'saw' to 'was able to/happened to see' as in (6). (5)
Nakita ng heneral ang sube rsibo saw NTM general TM subversive 'The general saw the subversive'.
(6)
Nakakita ng subersibo ang heneral saw NTM subversive TM general 'The general was able to/happened to see a subversive'.
Whorf would have referred to this morphological peculiarity of these mental process verbs of perception as a reactance. Reactances such as this call covert categories to the attention of the linguist, forcing him to set up cryptotypes alongside the overtly realised categories Whorf refers to as phenotypes (the ability/involuntary action prefix itself is an example of a phenotype). Whorf's point was that a full description of a language could not rest on phenotypes alone, since cryptotypic categories such as that illus trated above are essential to explaining the use of phenotypic forms. Being by their nature hidden, and more semantically revealing than phenotypes, cryptotypes permitted Whorf to make a stronger case for the influence of language on the perception of reality, both social and otherwise, than would have been possible for any other linguist of his generation. There is a caveat however. Whorf notes in a guide prepared for ethnological field workers that covert categories require considerable study
252
J.R. MARTIN
of a language to uncover: III. Cryptotypes. Covert word categories with subtle meaning marked only by reactances. Skip this in the survey except for obvious cases, as determi nation of cryptotypes usually requires deep study of a language (Whorf 1956: 132 [1938]).
Readers will perhaps accept this by way of apology for the dependence on phenotypes in the analysis of Tagalog presented in the next section of the paper (note however the cryptotypical nature of the mental process verbs of perception and reaction discussed in section 3.3 below, and the political ipag- series in 3.2).
3.0 Three Grammatical Conspiracies in Tagalog Whorf s hypothesis concerning the relation between fashions of speak ing and social reality is a very difficult one to 'prove'. Skeptics are not likely to be convinced by the evidence presented in this paper. Nevertheless, in comparing Tagalog and English, from the point of view of both the lan guages themselves and the cultures they realise, a number of striking pat terns emerge. The three most important have to do with cultural values which may be glossed as relevant to family, face and fate. Filipinos differ markedly from most speakers of English (and other northern 'Standard Average European' languages) in their attitudes to aspects of social reality. As far as the family is concerned, Filipinos place a great deal of emphasis on participation, normally in a large extended family group with whom they live, socialise and perhaps even earn their living. In contrast, Westerners place more emphasis on the indivdual, who is able to stand independent and self-supporting, 'on his own two feet'. With regard to face, Filipinos take great care that position is respected and maintained, with as little dis turbance as possible, even where this means that one's true feelings in a given situation cannot be publicly expressed. Westerners on the other hand place a high value on honesty and forthrightness, even when this means stepping on someone's toes—it is better to be what you are than appear to be something you are not. Finally, concerning fate, Filipinos often deny responsibility for what happens to them, whether for better or worse, accepting their fortune somewhat fatalistically. Westerners are rather more concerned with controlling their destiny, try to determine as far as possible what happens in their lives, and develop a keen sense of failure or success measured in terms of taking the future into their own hands. Of course
GRAMMATICAL CONSPIRACIES IN TAGALOG
253
none of these characteristics is without exception; each simply represents a general trend. The question of how to back up these generalities will be considered in section 4.0. First, the grammatical conspiracies associated with each will be reviewed. 3.1 Family —pakikisama The basic theme of the structures reviewed in this section is that of 'to getherness': doing things together or having things in common. In all cases the meaning which is coded grammatically in Tagalog is coded in a less gen eral way in English. It is not that English cannot make the meaning involved. Whorf s hypothesis has nothing to do with being trapped into a particular way of viewing the world. Rather, English has a tendency either not to bother with making the togetherness meaning in contexts where Tagalog is structured to do so, or to make the meaning periphrastically when necessary, through the use of lexical items referring to the meaning encoded in Tagalog structure. This means that in a general and unconscious way, Tagalog predisposes its speakers to a social reality in which participa tion is a key element in a way English does not. a.
social verbs (maki-) 'join in/share in an activity with'
The prefix maki- (sometimes in conjunction with other affixes — see Schachter & Otanes 1972: 7C 5.14) codes the 'participation' of one partici pant with another in some activity. Through this prefix, the idea that Gel and Hector are drinking together, which is left implicit in (7), can be made explicit, as in (8). (7)
Umiinom sina Gel at Hector are drinking TM-pl and 'Gel and Hector are drinking'.
(8)
Nakikiinom si Gel kay Hector are drinking-joint action TM CM 'Gel is drinking with Hector' (or better 'Gel joined Hector for a drink'.)
It might at first appear as if the difference between Tagalog and English here is simply one of an inflecting as opposed to an isolating language: Eng lish codes with the prepostition with what Tagalog codes through the prefix maki. This however does not do justice to the meaning of the maki- form. In (8) for example, there is an implication that Hector has some kind of prior or more direct involvement in the drinking than Gel; this implication
254
J.R. MARTIN
is not present in the English translation. To gloss this implication, English would have to make use of periphrastic forms such as joined in drinking with or shared in drinking with. So more than a typological distinction is at stake. Tagalog is grammaticalising a participational meaning which can only be realised more specifically, through a preposition, or periphrastically, through lexicalisation in English. b.i. reciprocal verbs: 2 or more actors(mag- . . . -an) 'do an activity to one another' (Schachter & Otanes §5.6G) The prefix mag- in conjunction with the suffix -an expresses reciprocal action involving 2 or more actors. Thus (10) contrasts with (9) below: (9)
Hinalikan niya ang singsing kissed NT-pr TM ring 'She kissed the ring'.
(10) Naghalikan sila kissed-reciprocal T-pr-pl They kissed each other'. In (10) Tagalog structures the event referred to as invoving a plural actor engaging in reciprocal action. English grammar on the other hand analyses the experience as one in which a plural actor acts on a goal which happens to refer (reflexively) to the plural actors as individuals, rather than as a group. Where English looks at the participants both as a group and as indi viduals, acting on each other, Tagalog views the participants simply as a group, engaging in reciprocal action. The same pattern of difference is found with nominalisations related to reciprocal verbs. For example, ang kanilang suntukan {suntok 'fight') would have to be glossed as 'their fight with one another'; similarly ang kanilang biruan {biro 'joke') as 'their jok ing with one another'. Alternatively English could lexicalise the meaning through glosses such as 'their fighting session' or 'their mutual joking'. In each case the reciprocal nature of process is coded more overtly in English than Tagalog, and attention is drawn to the interaction of individual partic ipants. b.ii. reciprocal verbs: more than 2 actors {mag- . . . R2R2) 'do an activity to one another' (Schachter & Otanes §5.16.4) Tagalog has a further reciprocal action form which applies to bases not including the suffix -an. This is coded through a mag- prefix followed by reduplication of the first two syllables of the base (R2R2). This form entails
GRAMMATICAL CONSPIRACIES IN TAGALOG
255
that more than two actors are involved: (11) Nagkitakita ang tatlong sube rsibo saw one another TM three subversive 'The three subversives met one another'. The difference between the English and Tagalog perspective is the same as for the mag- . . . -an reciprocal action verbs (note that multiple reciprocal action verbs do not have nominal agnates parallel to suntukan or biruan). Reciprocal {mag- . . . -an) and social (maki-) affixes may both be realised in the same process, in a kind of culmination of joint interaction, as in (12) below. These forms are awkward to translate into English, but might be glossed along the lines of 'joining in mutual action with'. Needless to say, English speakers are not likely to combine reciprocal and social action meanings in a single clause very often, preferring to treat processes as recip rocal or social or neither, where Tagalog naturally combines the together ness meanings. (12) Nakipagbigayan ako sa kanila joined in mutual gift giving T-I CM C-them T joined in the mutual giving (of gifts) with them/I shared in exchanging gifts with them'.
ka- comitative (-mate forms) (Cena 1979: 122)
The prefix ka- codes a relation of 'mateship' between two participants. These can often be glossed by means of the English suffix -mate, as in (13). (13) Kalaro ko siya playmate NT-I T-he/she 'He is my playmate'. 4 When mateship is expressed between members of a single plural partici pant, the forms magka- (2 mates) and magkaka- (more than 2 mates) are used: (14) Magkalaro sina Gel at Hector playmates TM-pl and 'Gel and Hector are playmates'. (15) Magkakalaro sila playmates T-they 'They (2+) are playmates'. The comitative ka- form may be verbalised, with aspect specified as in (16):
256
J.R. MARTIN
(16) Kinalaro ko siya playmate NT-I T-he/she 'He was my playmate'. The ka- comitative prefix has a far wider range of distribution in Tagalog than does -mate in English. English often makes use of a lexical item, where Tagalog uses a ka- form—for example katulong {tulong 'help') 'helper', not help-mate; katalo {talo 'dispute') 'opponent' not *dispute-mate; and so on. Because of this wider distribution, Tagalog can again be seen to more fully grammaticalise participation than does English. d.
ka {sing)- comparison of equality (like-) (Cena 1979: 123)
Tagalog has a superficially similar structure to code similarity in appearance between two participants. In a sense both ka- forms code a meaning of sharing; but here, participants share some feature of their appearance, rather than having an activity in common. (17) Kamukha ko siya like-face NT-I T-he/she 'She looks like me'. Some of these ka- resemblance structures can be glossed through the Eng lish prefix like- (e.g. like-faced, like-eyed, etc.). However, not only is this process somewhat periphrastic in English (many examples sound archaic, odd, or strained in some way), but again, it does not extend to the range of items open to the ka- prefix in Tagalog. For example, in (18), English requires a fully analytic comparison of equality structure to gloss kasinglaki. (18) Kasinglaki ko siya like-big NT-I T-he/she 'He is as tall as I am'. {*We are like-tall.) Sharing, both in terms of activities and appearances, can thus be seen to be more pervasively grammaticalised in Tagalog than in English. e. maka- partisanship (pro-) (Schachter & Otanes §4.10.6) The maka- prefix forms adjectives, typically out of the names of people or places. The basic meaning is that of partisanship, which can often be glossed by the English prefix pro-. (19) -Marcos siya partisan of Marcos T-she/he 'She is pro-Marcos'.
GRAMMATICAL CONSPIRACIES IN TAGALOG
257
There are however cases where the pro- is an ineffective gloss. Maka-ama {ama 'father') is better glossed as 'fond of/close to father'; similarly, makabago {bago 'new') would have to be lexicalised as progressive or perhaps trendy. Again, in terms of distribution, the general meaning of par ticipation, here in the sense of being in favour of someone/thing/place, is more generally coded in Tagalog than in English. f.
sina, nina, kina plural proper name case markers (Schachter & Otanes §3.9)
When two or more partcipants are coordinated in clause structure as in (20), Tagalog has the option of using a single plural case marker to intro duce the nominal group complex as in (21): (20) Dumating si Imelda at si Ferdinand arrived TM and TM 'Imelda and Ferdinand arrived'. (21) Dumating sina Imelda at Ferdinand arrived TM-pl and Tmelda and Ferdinand arrived'. The plural case marker can also be used when only a single participant is involved; in such cases the nominal group would be glossed as that partici pant, plus others: (22) dumating sina Ed arrived TM-pl 'Ed and company arrived'. English can also pluralize proper names with a related implication (e.g. The Martins arrived at 6). But it can do so only with surnames, and the resultant form refers only to a family group. The Tagalog form is not limited in this way; it can be used with a first name as in (22), and the resulting phrase might well refer to Ed's peer group (his b arkada) — there is no restriction to family membership. Thus English requires a coordinate construction to gloss sina Ed in (22), explicitly dividing the group into Ed and others where Tagalog codes the group as a whole. g.
special coordinate construction (Schachter & Otanes §3.10)
Tagalog is also able to avoid coordination where English forces it when pronouns and common or proper nouns function as a single participant structure. Tagalog in fact prefers (24) to (23):
258
J.R. MARTIN
(23) Dumating siya at si Ed arrived T-pr and TM 'She and Ed arrived'. (24) Dumating sila ni Ed arrived T-pr-pl GN 'She/they and Ed arrived'. In (24), a plural pronoun is used, followed by a proper noun in the objec tive/genitive case; the meaning is that she and Ed or they and Ed arrived. The closest English could get to the construction would be the ungrammatical *Ed's them (which in any case would restrict the meaning to Ed and more than one other). Again Tagalog is able to fuse the participants involved into a single group, of which Ed is a member, where English is forced to isolate the individuals involved. h.i. relational verbs: 2 or more objects(pag- . . . -in) 'make things be together' (Schachter & Otanes §5.16.3) Tagalog makes use of the affix pag- . . . -in to derive relational verbs from certain object focus verbs. The objects conjoined by the process to which this affix is attached are viewed as a group, rather than as related entities. In (25), the lettuce and tomatoes are treated as separate entities, one of which has been put with the other. In (26), on the other hand, the lettuce and tomatoes are being treated as a mixture (a salad as opposed to two separate vegetables). (25) Isinama ko ang kamatis sa litsugas put in with NT-I TM tomato CM lettuce T put the tomatoes in with the lettuce'. (26) Pinagsama ko ang kamatis at litsugas put together NT-I TM tomato and lettuce T mixed the tomatoes and lettuce together'. English achieves the same effect in many cases by making use of the adverb together. In some cases, however, an alternative expression must be found: (27) Pinagalapit ko ang silya at mesa put near NT-I TM chair and table T put the chair and table near one another'. Here English is forced to make use of two clause participants {the chair and table and one another) to gloss the Tagalog structure, thereby drawing
GRAMMATICAL CONSPIRACIES IN TAGALOG
259
attention away from the chair and table as a conjoined product of the pro cess. h.ii. relational verbs: more than two objects (mag- R2R2/pag- R2R2 -in) 'make more than two things be together' (Schachter & Otanes §5.16.4) These structures have the same meaning as the forms discussed in h.i. above, but are restricted to combinations involving more than two objects. As well, there is an actor focus form available. (28) Nagdugtong-dugtong ang letter R's niya joining together TM his 'His letter R's (more than 2) are running together'. (29) Pinagtabi-tabi niya ang tatlong silya put beside NT-he/she TM three chair 'He put three chairs beside one another'. Again, in the last example, the Tagalog structure is a more iconic reflection of the combination of chairs than its English gloss. i. implicitness i.i. ellipsis The final aspect of Tagalog's family conspiracy to be considered has to do with implicitness. This general feature can be broken down into two headings: ellipsis and indefinite may phrases (Schachter & Otanes §4.23). In this section ellipsis will be discussed. The most common type of ellipsis in Tagalog is Topic ellipsis. The ellipsed Topic may be recoverable from either the co-text (endophora) or non-verbal context (exophora). For example: endophora (30a) Hindi na tumutunog ang kaniyang stereo not now sounding TM her/his 'Her stereo is already not working'. (30b) Bakit, inano mo? why did do with NT-you-sg 'Why, what did you do to (it)'? (stereo ellipsed) (30c) Wala. Ipinatong ko lang sa mesa nothing put over NT-I only CM table 'Nothing, I just put (it) on the table'. (Flores 1976: 10) (stereo ellipsed)
260
J.R. MARTIN
exophora (31a) May problema ba? existential problem Q Ts anything wrong'? (addressee ellipsed) (31b) Wala naman nothing counter 'Nothing'. (32) Kainin mo nga eat NT-you-sg please 'Please eat (it)' (food ellipsed) Ellipsis of non-topic complements is also common, especially non-topic Actors. The Actor is ellipsed in (33) and the Goal in (34). (33) Kinagalitan ka na naman ba? anger because of T-you-sg again Q 'Did (he) scold you again'? (34) Tatawag na lang ako will call now just T-I T'II call (you)'. Tagalog even has one construction, the immediate imperative (Schachter & Otanes 7C5.30) which does not allow any complements at all: (35) Bili na buy now 'Buy (it/some) (now/right away)'. Ellipsis in basic sentences (paratactic and hypotactic constructions will not be considered here) is somewhat more restricted in English. Subjects are ellipsed exophorically, typically when they refer to the speaker or the addressee. (36) Just got home. (37) Feeling tired? But third person subjects are not easily ellipsed in this way. Moreover, out side of responses to wh questions, Subjects are not ellipsed endophorically in basic sentences. As far as Complements are concerned, again, outside of responses, English does not permit ellipsis of recoverable participants. Transitive clauses require Complements (e.g. I put it on the table. not *I put
GRAMMATICAL CONSPIRACIES IN TAGALOG
261
on the table — cf. (30c) above). Ellipsis in Tagalog is a subject which has not been properly studied, but it is clear that there is greater freedom as to the presence or absence of contextually given elements of clause structure than in English. This has in part to do with the fact that the Focus affixes on Tagalog verbs implicate the presence and role of the Topic participant in a clause, whether the par ticipant is actually mentioned or not. English verbs implicate the presence of Actors in the passive but they cannot explicitly implicate the presence of other roles, and 'passives' are used far less frequently than in Tagalog. Eng lish is further restricted as far as ellipsis is concerned by having to indicate the mood of clauses through the presence or absence, and position of Sub ject (Eat., Did you eat?, You ate. etc.). The degree of implicitness permit ted in the two languages is further affected by the absence of non-human pronouns and of substitution {do, one, and so; see Halliday & Hasan 1976) in Tagalog. Thus where English refers to a participant as it (reference) or as one/some (substitution), Tagalog generally prefers ellipsis (demonstratives can be used where necessary to refer anaphorically to a given participant; and numeratives can be used when needed in place of substitution). Inter ference resulting from these factors is a common feature of Filipino English (e.g. You like?; You want?; Can have.) The net result is that Tagalog is structured to elide participants in clause structure far more frequently than is possible in English. The specification of such a difference in ellipsis potential has been dis cussed by Hasan (1984), comparing English and Urdu. It is not clear at pre sent whether ellipsis in Tagalog has the scope that it does in Urdu, but Tagalog is certainly towards the Urdu end of any comparison of languages as far as ellipsis is concerned. Hasan makes the further point that not only is ellipsis more fully elaborated in the structure of Urdu than English, but that as well, it is exploited in a wider range of contexts in Urdu than in Eng lish culture because of the determinate nature of role systems in Urdu soci ety. Hasan comments on the predominance of the implicit semantic style in Urdu as follows: To say, then, that one is an Urdu speaker is to largely discount the possi bility of being misunderstood. It is to believe that your addressee knows what you are on about; it is to assume that the chances of ambiguity are so low as to be almost negligible (Hasan 1984: 151). It would be premature to apply these observations directly to Filipino cul ture. Nevertheless, it is significant that elliptical constructions rely on
262
J.R. MARTIN
shared knowledge for their interpretation. This knowledge may be present in the verbal or non-verbal situational context, or in an utterance's context of culture. As such, the degree of implicitness permitted in Tagalog text points out yet another aspect of the grammaticalisation of family. Who bet ter to know what you are about than the relations with whom you live, socialise and often work as well? i.ii. indefinite may constructions (Schachter & Otanes §4.23) Indefinite may constructions are another major aspect of Tagalog's implicitness potential. In these constructions clause participants are omitted not because they are recoverable from the context, but because they are nonspecific or can simply be taken for granted. As with Topic ellipsis, the FOCUS affix on the verb implicates the presence and specifies the role of the missing participant. An Actor is omitted in (38), a Goal in (39). (38) May dumating Existential came 'Someone came'. (39) May hinihintay ako Existential waiting T-I T'm waiting for someone'. Depending on the reason for not mentioning the omitted participant, this pattern can be interpreted as contributing to either the family or the face conspiracy. If the participant is left out because it is taken for granted — if for example the listener in (39) knows perfectly well who the speaker is waiting for — then the structure reflects the tendency to implicitness and consequent dependence on shared knowledge. If on the other hand the par ticipant is omitted because the speaker either does not know his identity or is being purposefully vague, then the structure is an aspect of face which will be discussed below. Whatever the case, Tagalog does make use of a structure to avoid men tioning a participant where English is bound to explicitly code its existence at clause rank. English then avoids specifying who it has in mind by using an indefinite pronoun to realise the clause function in question. English has to get even more specific when translating indefinite constructions intro duced with marami 'many'. Note in (40) that an indefinite pronoun cannot be used in the gloss, which turns to the more specific, though still general item, people, to translate the structure.
GRAMMATICAL CONSPIRACIES IN TAGALOG
263
(40) Marami ako -ng susulatan many T-I LK will write to T have a lot of people to write to'. The differences between English and Tagalog as far as the grammaticalisation of the family conspiracy is concerned are summarised in Table 1. J.
pronouns
j.i. inclusive/exclusive distinction (Schachter & Otanes §3.3) As might be expected given the family conspiracy, Tagalog distin guishes systematically between inclusive and exclusive first person plural pronouns. The tayo/natin/atin series contrasts with kamilnamin/amin as in (41) and (42) below (an earlier distiction within the inclusive pronouns between dual and non-dual has now been lost, the dual katalnitalkanita series now being obsolete). (40) Magbeer muna tayo have beer first you and me 'Let's have a beer first'. (41) Nagbeer muna kami had beer first me and others 'We had a beer first'. In English we/us/our/ours are normally ambiguously inclusive or exclusive; only in the first person imperative can an inclusive meaning be unambigu ously made (e.g. Let's go.). j.ii. fused *ko^ ka (kita) Finally Tagalog realises a first person non-Topic Actor followed by sec ond person Topic Goal as the single pronoun kita instead of the expected syntagm of pronouns *ko^ ka. English has no way of realising Subject and Object in a single pronominal form. (43) Makikita kita will meet I/you 'I'll meet you'. 3.2 Face pagkahiya & utang na loob The unifying meaning of the structures contributing to the face conspi racy has to do with appearances: respecting other persons' positions, possi-
264
J.R. MARTIN
Family conspiracy English (individuality): analytic — 'join in/share in doing, do together' b. reciprocal verbs mag- ... -an, mag- R2R2 analytic — 'do to each other' comitative ka-mate (*helpmate) d. comparison of equality ka(sing)like- (*like-walk) e. partisanship makapro- (*pro- new) f. plural case markers sina, nina, kina Martins (*Jims) g. genitive coordinate construction e.g. sila ni Ed parataxis — he and Ed h. relational verbs pag- ... -in, mag- R2R2/ together (*put beside together) pag- R2R2 -in i.i. implicitness: Topic (endophoric & exophoric) Subject (exophoric,l & 2 person only) ellipsisnon-Topic (endophoric & exophoric) no non-human pronouns it, they no substitution do, one, some, so i.ii. implicitness: indefinite may someone, something j . pronouns: i. inclusive/exclusive forms no morphological distinction; some dependence on MOOD ii. fused ko^ ka (kita) Tagalog (togetherness): a. social verbs maki-
Table 1: Summary of grammatical differences related to the family conspiracy
bly at one's own expense. In a sense this is the least dramatic of the three conspiracies considered in this paper — not because Tagalog is less differ ent from English in this respect than in others, but because even more dramatic codings of face are well known in South East Asian languages (e.g. the face-layered lexicon of Javanese or the complex respect morphol ogy of Japanese verbal affixes). The significance of the conspiracy is not however diminished by its predictability. Once again, in most cases, English can be shown to realise the meanings in question more specifically, and with a more limited distribution than Tagalog. a.
respect pronouns (Bautista 1979 §2.23)
As with many languages, Tagalog uses second person plural pronouns to refer to a single addressee as a measure of respect. Third person plural pronouns can be used to show extreme respect. At one time of course Eng lish used you in place of thou with the same effect. This however led to the loss of a second person singular/plural distinction, so that English no longer has the pronominal resources to show respect in this way. Schachter & Otanes (7C3.3) suggest that Tagalog lost its singular/plural distinction for inclusive pronouns in the same way. Whatever the case, retention of the
GRAMMATICAL CONSPIRACIES IN TAGALOG
265
distinction in the second person means that Tagalog has a grammatical resource which English does not for deferring to the position of a superior. b.
address terms (Bautista 1979 §2.22)
Tagalog's system of address terms is more highly differentiated than that of English. Bautista's (1979: 65) analysis of the Tagalog system, though not completely comparable with that of Ervin-Tripp (1972) does neverthe less reflect a greater differentiation of position. Tagalog has for example address terms for older brother, kuya, and older sister, ate, for which Eng lish has no generic names. VOCATION is thus more sensitive to the posi tion of the addressee in Tagalog than English. c.
respect enclitics (Bautista 1979 §2.23)
Tagalog also relies on respect enclitics to show politeness. Ho is very frequently used this way. Po is less common and used to show extreme respect. English has no comparable items whose sole function is to show politeness per se. The respect function of pronouns, terms of address and respect enclitic is illustrated in (44) and (45) below: (44) Maganda -ng umaga ho, Lola. beautiful LK morning grandmother 'Good morning Grandmother'. Ano ang inyo -ng kailangan? what TM C-you-pl LK need 'What can I do for you'? (45) Hi Gel. Saan ka galing where T-you-sg come from 'Hi Gel. Where've you been?' Deference is exemplified in (44), as realised through the respect enclitic ho, the address term Lola for grandmother, and the second person plural pro noun inyo (the choice of the greeting, magandang umaga, is also a marker of respect; but lexicalised clauses and phrases will not be further considered until 5.0 below). Solidarity is illustrated in (45), realised through the address term Gel, the singular second person pronoun ka, and the absence of respect enclitics (and by the greeting Hi as well). d.
polite request maki- (Actor/Topic), paki- (Goal/Topic) (Schachter & Otanes §5.30) Imperatives are generally a sensitive area as far as face is concerned.
266
J.R. MARTIN
Tagalog softens demands for goods and services in several ways. First, there are the polite affixes maki- (for Actor focus) and paki- (for Goal focus). These affixes change the command illustrated in (46) into the polite request in (47). (46) Linisin mo ang kuwarto clean NT-you-sg TM room 'Clean the room'. (47) Pakilinis mo ang kuwarto clean-polite NT-you-sg TM room. 'Please clean the room'. English has no special request forms, and tends to soften its commands by using interrogative instead of imperative mood, often modulated (e.g. Could you clean the room?). Incongruence between SPEECH FUNCTION and MOOD is also used in Tagalog (see Bautista 1979 §3.5). e.
moderative R2R2 (Schachter & Otanes §4.16.2)
This structure has the effect of moderating an action in some respect, with English glosses along the lines of 'do a little/a bit/now and then/here and there'. Schachter and Otanes (340) point out that one of its principal functions is to soften a request; thus the distinction between (48) and (49). (48) Magwalis ka nga ng kuwarto sweep T-you-sg please NTM room 'Please sweep the room'. (49) Magwalis-walis ka nga ng kuwarto sweep-moderative T-you-sg please NTM room 'Please sweep up the room a little'. In English, requests can only be moderated analytically, through the use of more specific meanings such as 'a little/a bit'. f.
politeness request enclitic nga
Like English, Tagalog makes use of a special function word to signal politeness in requests: nga, equivalent in requests to English please. (50) Kumain ka eat T-you-sg 'Eat'.
GRAMMATICAL CONSPIRACIES IN TAGALOG
267
(51) Kumain ka nga eat T-you-sg please 'Please eat'. Reviewing then, as far as demanding goods and services is concerned, Tagalog is able to structure politeness into requests in many cases where English has to rely on indirectness or moderating lexical phrases. g.
reported speech particle daw/raw
The structures considered so far may have made it appear as if the face conspiracy was concerned simply with politeness and status. However, there is more to it than that. Appearances are also crucial: it is not enough to simply defer to superiors—it is important to publicly present oneself in such a way as to maximise social acceptance. This is obviously very impor tant in a culture that places the emphasis on family and participation that Filipino culture does. The remaining structures in the face conspiracy con tribute more to this theme than to politeness per se. One way to increase one's acceptability is not to take personal respon sibility for what you say in case anyone takes offense. Tagalog's reported speech particle can be effectively used in this way, to in a sense disclaim: (52) Subersibo siya subversive T-he/she 'She's a subversive'. (53) Subersibo raw siya subversive they say T-she/he They say she's a subversive'. One commonly noted aspect of maintaining appearances and face is the use of go-betweens to mediate delicate issues. The reported speech particle is often useful to these mediators. Consider for example (54), in which a gobetween extends greetings to the listener on behalf of an admirer who need not be named: (54) Greetings daw sa iyo some says CM C-you-sg 'Someone is saying hi to you'. English has no special disclaiming structure or particle, and draws attention to statements for which the speaker is not taking responsibility through they say, someone says. it is felt, etc.
268 h.
J.R. MARTIN
indefinite may
The indefinite may constructions discussed in connection with the fam ily conspiracy are also relevant here. The vagueness they involve can be used to avoid offending the listener. The speaker in (55) may be perfectly clear in his mind what he is going to do, but it may be inappropriate to make this information public: (55) May gagawin ako Existential will do T-I T have something to do'. Again, English calls attention to the suppressed information, using an analytic structure at clause rank and an indefinite pronoun to fill the neces sary, but semantically empty clause function. Perhaps even more relevant to the face conspiracy than may is its nega tive counterpart wala. This is used by Filipinos as a next to knee-jerk response to any question requesting any information about themselves. Bautista (7C3.6) discussed it in the context of a response to probes about someone's problems, where the existence of a problem may be denied sev eral times before confession. This response pattern, which often appears exasperatingly dishonest to an English speaker, is simply a conversational reflex of the importance of playing one's cards very close to the chest. The less information that is publicly available, the easier face is to maintain. i.
pretense R2R2 . . . -anlmagsa-
Given the significance of appearances, it is perhaps not surprising that Tagalog can grammaticalise pretense, by reduplicating the first two sylla bles of a nominal or verbal stem adding the suffix -an. Thus (56) and (57) where, as the glosses show, English must lexicalise the pretense in order to get the meaning across. (56) NagMarcos-Marcosan siya pretended to Marcos T-she/he 'He pretended to be Marcos'. (57) Nagtulug-tulugan siya pretended to sleep T-he/she 'She pretended to sleep'. In addition De Guzman (1978: 285) notes magsa- verbalisations of pro fessions, nationalities, animals and so on, as in (58). The meaning pretense is again involved.
GRAMMATICAL CONSPIRACIES IN TAGALOG
269
(58) Nagsapari ang subersibo pretended to be a priest TM subversive T h e subversive pretended to be a priest'. j.
mag-,
ipag-,pag- . . . -an 'political' verbs
Tagalog has an intriguing covert class of what might be called 'political' verbs, which are relevant to the face conspiracy. This class has two distinc tive reactances: it takes ipag- (normally a bene factive focus affix) for Goal focus, and it takes pag- . . . -an for direction focus. Some of the verbs in this class are: magbili 'sell out' magmalaki 'boast' magkaloob 'offer' magkatiwala 'entrust' magbawal 'prohibit' magbilin 'give orders magtapat 'confide' maglihim 'keep secret' maglingkod 'serve' magkanulo 'betray' ('deep' Tagalog — malalim na Tagalog; i.e. archaic in Manila) magkaila 'deny' magdiwang 'celebrate' magpaliban 'postpone' magpatuloy 'continue magtanggol 'defend' magbigay-alam 'inform' ('deep' Tagalog) Their distinctive focus affixes are illustrated in (59) through (61). (59) Nagtapat ako ng problema kay Thelma confided T-I NTM problem CM T confided with Thelma about a problem'. (60) Ipinagtapat ko ang problema kay Themla confided NT-I TM problem CM T confided with Thelma about the problem'. (61) Pinagtapatan ko ng problema si Thelma confided NT-I NTM problem TM T confided with Thelma about a problem'.
270
J.R. MARTIN
This series is a classic cryptotype in Whorf s sense, and it is difficult to find a succinct way of expressing the connections between its members. It is prob ably impossible to lexicalise their common theme in a way that nets them all in. However, it is striking that so many of them relate directly to very sensi tive areas of human intercourse, having particularly to do with exchanging information and controlling the behavior of others. Seen in terms of the face conspiracy, the meaning is perhaps easier to grasp. The verbs all refer to activities in which face is very much at stake. There is no corresponding face cryptotype in English, where covert verb classes focusing on appear ance are not found. The face conspiracy is summarised in Table 2, where Tagalog struc tures are again contrasted with their nearest English equivalents. Face conspiracy Tagalog (appearance): English (reality): a. tu/vous — b. elaborated address system less elaborated set of terms of address c. respect enclitics ho,po — d. polite request maki-lpaki- incongruence of MOOD and SPEECH FUNCTION e. moderative request may R2R2 'a little/a bit/now and then/here and there f. polite request marker nga please g. disclaiming enclitic daw/raw (they say, it is said, someone says) h. indefinite may/wala someone, something specified at clause rank i. pretence R2R2 ...-anlmagsa- ('pretended to') j. political mag-, ipag-, pag-...- an -series Table 2: Summary of grammatical differences related to the face conspiracy
3.3 Fate suwerte The basic theme of the fate conspiracy is that of events taking place outside of an individual's control — things simply happen; one cannot really determine one's fate. The structures involved emphasize the acciden tal or uncertain nature of events, equivocate about whether humans make or simply let things happen, and even assign causality for certain events to other than the human participants involved. In Filipino culture, things hap pen, not so much because you make them, but because they fall to you as your lot in life. To begin, consider three systematic equivocations; in each case Tagalog neutralises a distinction having to do with responsiblity for an action taking place that English typically discriminates.
GRAMMATICAL CONSPIRACIES IN TAGALOG
a.
271
causative (indirect action) pa- 'have x do/ask x to do/let x do/make x do' (Schachter & Otanes §5.12)
Tagalog introduces Agents into basic sentences by means of the affix pa-. This affix is however ambiguous as to whether the Agent initiated the actions involved, or simply allowed them to happen. No distinction is made between 'making something happen' and 'letting something happen' as in English. Thus (62) can be glossed in several ways: (62) Nagpahuli ng subersibo ang heneral sa sundalo made catch NTM subversive TM general CM soldier 'The general let/made/asked/had the soldier catch subversive'. b.
ability and involuntary action ma-imaka- 'able to/happen to' (Schach ter & Otanes §5.12)
Similarly, with the ability/involuntary action affixes ma- (Goal focus) and maka- (Actor focus), no distinction is made between being able to do something (thus being in a position to determine whether it happens or not) and happening to do something (i.e. not determining whether it takes place). Schachter & Otanes (330) report some speakers as consistently dis tinguishing ability and involuntary action through vowel length, with ma-/ maka- for ability and ma:/maka: for involuntary action. They note how ever that others, and particularly younger speakers, do not make the dis tinction at all, or at least not consistently. It is not clear whether this means that the distinction is currently disappearing or that one of the forms have simply been in free variation for a considerable period of time. Whatever the case, (63) is typically ambiguous between ability and involuntary action, thus neutralising a distinction between determined and non- determined action which English is inclined to make. (63) Nahuli lang ng heneral ang mga padre able/happened to catch just NTM general TM few 'The general was only able to/just happened to catch the priests'.
possessive magka- 'come to have/get'
Tagalog forms possessive verbal constructions by prefixing magka- to nominals. The resulting structure is ambiguous as to whether the object referred to by the nominal was procured deliberately, or simply fell into the possession of the Topic participant. Thus in (64), Rose's getting money may be the result of diligence or of good fortune:
272
J.R. MARTIN
(64) Nagkakapera si Rose kasi marami siya- ng has money TM because many T-she LK bar-fine noong isang linggo previous one week 'Rose has/got/came to have money because she had a lot of barfines last week'. English tends to distinguish between deliberate or accidental possession (in deed, it was not easy to introduce (64) without biasing the interpretation: cf. Thus in (64) Rose may have procured/secured money . . . vs. Thus in (64) Rose may have come to have/may have received money . . .). d.
accidental result magkang§5.16.5)
R1/magkanda
(Schachter & Otanes
With certain verbs Tagalog is able to add a prefix magkanda or a prefix magkang- plus reduplication of the first syllable of the stem to express a process which is the accidental result of another action. Here responsibility for the effect, usually an undesirable one, is disclaimed: (65) Napuyat si Gel; nagkandasirasira exhausted from not sleeping TM broke as an accidental ang kotse niya result car his 'Gel was rooted; his car ended up smashed as a result'. English can make this meaning only by explicitly lexicalising both the caus ality and the accidental nature of the result. e.
skewing in the enclitic particle system towards possibility
Tagalog has a large system of enclitic particles. One set of these (kasi, tuloy, , na, pa, dinlrin, na naman, pa rin, ulilmuli, man and lang) has to do with signalling logical relationships between sentences having to do with time, cause, expectancy and so on (see Martin 1981 for discussion). The rest have a general orientation to politeness, modality or attitude. Politeness enclitics have previously been discussed (3.1). Of interest here is the skewing in this system away from certainty and towards possibility as far as modality is concerned. Tagalog's disclaiming or reported speech particle daw/raw was described in 3.2 and can be used to diminish the certainty with which a statement is made by assigning responsibility for it to another indi vidual. Beyond this, Tagalog has 4 mood oriented particles, all concerned with uncertainty.
GRAMMATICAL CONSPIRACIES IN TAGALOG
273
ba (polar question particle) (66) Subersibo ka ba? subversive T-you-sg 'Are you a subversive'? kaya (speculation in imperatives, optatives and hortatives; when used in a speculative question, ba is usually omitted) (67) Magbeer kaya tayo T-you and me (inclusive) 'Perhaps we could have a beer'. (68) Subersibo kaya ang pare subversive TM priest 'Do you suppose the priest is a subversive'? yata (uncertainty in statements) (69) Subersibo yata ang pare subversive TM priest 'The priest seems to be a subversive'. sana (volition) (70) Masaya sana ang heneral happy TM general T hope the general is happy'. In addition there is the particle pala, which expresses surprise after receiv ing new or unexpected information or encountering an unforeseen event. (71) Subersibo pala ang pare subversive TM priest 'Don't tell me the priest is a subversive'. Tagalog has only one particle which expresses certainty, nga; and it tends to do so in response to a statement made by someone else. Thus: (72) Maganda si Tetchie, hindi ba? beautiful TM not Q 'Tetchie is beautiful, isn't she'? -Oo nga e yes 'Yes indeed'! The result of this is that the particle system is skewed dramatically towards
274
J.R. MARTIN
possibility and away from certainty. English has a much more balanced sys tem (skewed if anything in the direction of certainty, containing both [cer tain] and [virtually certain] options as it does (see Halliday 1976: 191 [1970]) which has been worked out in detail by Halliday (1970; 1985). f.
mental process perception and cognition cryptotypes (Rafael 1978; De Guzman 1987: 192, 298)
Tagalog has two covert classes of verbs which participate in the fate conspiracy. One of these has in fact been noted by Rafael (1978) because of the relation of its distinctive reactance to the controversy surrounding the interaction of definiteness and focus in Philippine languages (this reactance was described in 2.3 above in connection with Whorf's approach to cryptotypic analysis). The verbs involved refer to mental processes of percep tion (Rafael refers to them as non-volitional verbs): kita 'see' puna 'notice' dinig 'hear' amoy 'smell' tikim 'taste' (in the sense of 'experience') pansin 'notice' masdan 'observe' batid 'beware of' tanaw 'view' As noted in section 2.3, the verbs are prefixed with an apparently 'dead' ability/involuntary action prefix ma- in Goal focus. Thus (73) can be glossed simply as 'we heard the child'; there is no need to make the mean ing 'happen to' or 'able to' explicit. (73) Narinig namin ang bata heard NT-we-exclusive TM child 'We heard the child'. With Actor focus, however, the prefix maka- is used — and a sense of abil ity or involuntary action is conveyed. (74) Nakarinig kami ng bata happened to hear T-we-exclusive NTM child 'We happened to/were able to hear a child'. The explanation for this may lie in the ambiguous nature of mental process
GRAMMATICAL CONSPIRACIES IN TAGALOG
275
verbs of perception as far as the involvement of the Senser (perceiver in this class of mental processes) is concerned. This ambiguity is often reflected in cartoon drawings and has to do with who initiates the act of perceiving: does the Senser actively perceive some phenomenon?; or does the phenom enon simply appear and force the Senser to perceive it? Looked at one way the Senser is active, initiating the process; looked at the other way he is pas sive, a receiver of sensation. The distinctive reactance of perception verbs appears to reflect this ambiguity. When the Senser is in focus, its involvement in the process is felt to be explicitly marked by the maka- prefix. When the Senser is out of focus, playing a 'passive' role in clause structure, its involvement in the pro cess is felt to be unmarked; ma- seems to be a 'dead' prefix in Goal focus. Somewhat ironically, when the Senser is in focus and felt to be explicitly involved, Tagalog uses the ambiguous involuntary action/ability prefix dis cussed above (3.3b) to mark its involvement. The result of this is that in mental process:perception clauses Tagalog treats Sensers as dormant and uninvolved in the passive, and as ambiguously involved (by accident or by being able to perceive) in the active. The phenomenon involved is thus given more prominence than the perceiver of that phenomenon — another reflection of Tagalog's orientation to externally determined events. The closely related cryptotype of mental process:cognition verbs oper ates very similarly. When the Phenomenon is in focus, their ma- prefix lacks an ability/involuntary action interpretation; but to focus on the Senser, maka- and its accompanying ability/involuntary action meaning are required. Verbs in this class include isip 'think', limot 'forget', tanda remember' ,intindi 'understand', tuto 'learn', unawa 'feel out', alam 'know', tuklas 'discover', danas 'experience', balita 'receive news' etc. The cogni tion cryptotype differs from the perception one with respect to the -an reac tance when the phenomenon is in focus (e.g. Nalimutan ko ang pera T for got the money'.). The grammar of English also reflects the ambiguous nature of percep tion verbs as far as the direction of determination is concerned, though in a different way. If responsibility for the act of perceiving is to be uniquely assigned to the perceiver, then a behavioural process is used (e.g. John looked at/listened to the child.). If responsibility is not uniquely determined, then a mental process may be used (e.g. John saw/heard the child.). English is less fatalistic than Tagalog in that it can unambiguously assign responsibilty for the perception to the perceiver by using a behavioural process. The
276
J.R. MARTIN
best Tagalog can do is mark the Senser as involved in Actor focus, with this involvement potentially accidental in nature (for a discussion of the mental and behavioural processes, see Halliday 1985). g.
mental process:reaction cryptotype
The question of directionality comes up again with mental process verbs of reaction. The distinctive reactance of verbs belonging to this cryp totype is that they call explicit attention to the causal involvement of the phenomenon reacted to when the phenomenon is in focus. They do this by means of the causal affixes ika- and ka- . . . -an (these causative focus affixes can also be used with other than mental process reaction verbs— Schachter & Otanes §5.8; but with these verb classes other affixes can be used to focus on the 'object'). The following verbs fall into this class: galit 'anger' gulat 'surprise' inggit 'envy' hiya 'shame' galak 'glad' ('deep' Tagalog) tuwa 'happy' pagod 'tired' inis 'irritate' takot 'fear' hilo 'dizzy' lungkot 'sad' poot 'hate' ('deep' Tagalog) awa 'pity' lito 'confused' When the Senser (reacter in this class of mental process) is in focus, these verbs take a ma- prefix; the phenomenon reacted to appears in oblique case (kay or sa form). This pattern is illustrated in (75). Contrast (75) with (76) where dahil sa kanila is a circumstance of cause, not the phe nomenon feared (cf De Guzman 1978: 65-70). (75) Natatakot ako sa kanila fearing T-I CM C-them T am afraid of them'. (76) Natatakot ako dahil sa kanila fearing T-I because CM C-them T am afraid (of something) because of them'.
GRAMMATICAL CONSPIRACIES IN TAGALOG
277
When the phenomenon is in focus however, the causal involvement of the phenomenon must be made explicit, as in (77) and (78). (77) Kinatatakotan ko sila fearing because of NT-I T-they 'They are making me feel afraid'. (78) Ikinatatakot ko ang ginawa mo fearing because of NT-I TM diding NT-you-sg. 'What you did is frightening me'. By explicitly calling attention to the causal responsibility of the phenome non for the reaction experienced by the Senser whenever the phenomenon is in focus, Tagalog in a sense exports control of the process from the per son reacting to what that person is reacting to. Feelings are thus coded as determined externally, rather than springing from within when the phe nomenon is in focus. Like Tagalog, English can also draw attention to the causal involve ment of the phenomenon. It does this through agency—by making the phe nomenon Subject as in (79), instead of Complement as in (80). (79) The play pleased John. (80) John liked the play. When the phenomenon is Subject, it functions in an agentive role and is interpreted as prompting a reaction in the Senser. English depends on pairs of closely related mental process verbs to create this effect: frighten/fear, surprise/wonder at, puzzle/not understand, delight/enjoy, etc. (see Halliday 1985). This dependence on paired lexical items rather than causative affixes makes focusing on the causal involvement of the phenomenon somewhat less productive in English than in Tagalog. It is hard for example to find 'natural' agentive reaction verbs related to: inggit 'envy' (cf. make x jealous) galak 'glad'(cf. gladdened — now archaic; but cf. saddened and lungkot 'sad') tuwa 'happy'(cf. make x happy) poot 'hate' (cf. piss x off in North America; it shits me in Australia) awa 'pity' (cf. make x feel sorry for) Note that the make in the periphrastic glosses on several of these verbs is directly related to Tagalog's pa- causative affix (3.3a above), not its ika-
278
J.R. MARTIN
and ka-...-an focus forms. Thus while English can focus on the causal involvement of the phenomenon in a mental process of reaction, it is often much less natural for it to export responsibility for the reaction than in Tagalog. h,
voice: avoiding Actor focus
As noted by Bloomfield (1917:§93-94), Tagalog has a strong tendency to avoid focusing on the Actor if another clause constituent is available. Actor focus is thus limited to intransitive clauses, and to clauses in which other focus candidates are nonspecific as far as definiteness and number are concerned. This has the effect of skewing the voice system dramatically towards the passive. Actors are rarely in focus in transitive clauses where they play an agentive role; and when they are in focus in such clauses, the Goal is so low in specificity that it functions more as a Range (a kind of cog nate subject, if you will—see Halliday (1985)—an extension of the process rather than a fully fledged participant which is acted upon by the Actor. In ergative terms, Tagalog prefers middle voice, with the Medium in focus (i.e. the intransitive Actor and the transitive Goal), avoiding focus on the Agent (i.e. the transitive Actor) in most cases. Thus structures such as those in (81) and (82) predominate: (81) Dumating siya kagabi came T-he/she previous evening 'He came last night'. (82) Kinuha niya ang libro took NT-he/she TM book 'She took the book'. But much more rarely: (83) Kumuha siya ng libro took T-he/she NTM book 'He took some book (or other; it doesn't matter which one)'. This has the effect of drawing attention away from the participant who is responsible for something being done to another participant, and onto the participant who undergoes or simply does an action (not to someone or something else). Tagalog's focus system, in other words, plays down agency, orienting speakers not to the instigator of an event, but to its affected participant. Thus voice mirrors the fatalism discussed with respect to other structures in this section.
GRAMMATICAL CONSPIRACIES IN TAGALOG
279
English's voice system makes greater use of the passive than that of other western Indo-European languages (see Mathesius 1964: 62-63 [1928]), but it is nowhere near as skewed towards passive as Tagalog's. A large part of the reason for this is that in English, selection of Theme (equivalent to Topic selection in Tagalog for the purpose of this discussion) is relatively independent of the definiteness of participants in clause struc ture. In Tagalog on the other hand, selection of Topic and definiteness are closely related: definite Goals have a strong tendency to be selected as Topic and non-Topic Goals are normally interpreted as indefinite (cf. the glosses for (73) and (74) above). Thus where English would say John took the book., Tagalog would prefer The book was taken by John. In general then, although English often uses passives in many contexts where closely related languages would not, it is nowhere near as oriented to middle voice as Tagalog. In English informal counts show only 1 clause in 10 selecting passive5, whereas in Tagalog passive predominates in many text types (Bloomfield 1917). i.
aspect
Tagalog verbs inflect for aspect rather than tense. The basic system is set out in Figure 2. Actions are viewed as begun or not; if begun, as com pleted or not; and if completed, completed recently or simply over with (Schachter & Otanes §5.22).
Figure 2: Aspect in Tagalog This means that events are not tied to the time at which they are spo ken about the way they are in a tense system. Tense is a kind of verbal deixis, placing events before, during or after the act of speaking about them. Aspect is not deictic in this sense; it divides events into those which have taken place (realis) and those which have not (irrealis), then commments on whether they are currently going on or not, and if not, how
280
J.R. MARTIN
recently they have stopped. The way in which Tagalog treats unfulfilled intentions clearly illustrates the difference between the two systems. Unful filled intentions are irrealis—the events they encode never took place. This is true regardless of whether the intention itself occurred in the past, before the act of speaking, or not. Thus, whereas English's tense system (would have bought) situates the event in (84) in the past with respect to the act of speaking, Tagalog selects contemplated aspect, coding the action as not begun. The adverbial kahapon 'yesterday' serves to set the clause in time with respect to the past; the verb itself does not place the event in time rel evant to the act of speaking. (84) Bibilhin will buy wala negative T would money'.
ko sana ang libro kahapon, pero NT-I optative TM book yesterday but ako- ng pera existential NT-I LK money have bought the book yesterday, but I didn't have any
Compare (84) with (85), a clause encoding a future event. Tagalog uses the same verb form in (84) and (85). (85) Bibilhin ko ang libro bukas will buy NT-I TM book tomorrow 'I'll buy the book tomorrow'. Or, to take another example, Tagalog uses the same realis, not completed form for events which English tenses structure in terms of the past or pre sent: (86) Umuulan noong dumating ako was raining when arrived T-I 'It was raining when I arrived'. (87) Ayokong umalis; umuulan I-not-want LK leave is raining T don't want to leave; it's raining'. Thus while English's tense system is speaker oriented, situating the speaker with respect to the events he is talking about (i.e. did the events take place before, during or after the act of speaking), Tagalog's aspect system is event oriented, concerned with the 'reality' of the event involved (i.e. has the event taken place or not, and if so, is it finished), not with the speaker's relation to the event. Something of the flavour of this distinction can be
G R A M M A T I C A L CONSPIRACIES IN T A G A L O G
281
captured by comparing English's simple and certain of its compound tense forms. The compound tenses are more aspectual in nature: I went I go I'II go
vs. vs. vs.
I was going I am going I will be going.
The simple forms focus on the speaker's relation to the action, whereas the compound forms draw attention in part away from this relation and onto the on-going nature of the act itself. The significance of this as far as fatalism is concerned is that tense sys tems orient the speaker towards the past, present or future, thus encourag ing him to adopt a perspective in which he views himself as controlling or being controlled by whatever has happened or will happen. Aspect systems do not orient speakers in this way, so do not encourage an orientation to time in which speakers control their destiny. This is a fairly abstract line of argument, and certainly could not be taken as convincing on its own (cf. Whorf 1956, who in several chapters articulates a more detailed argument concerning the relation between Hopi's treatment of time and their world view). However, taken in conjunction with the rest of the fate conspiracy, it would appear to contribute to a fatalistic as opposed to a deterministic perspective on events and individuals' control over what happens to them. The fate conspiracy is summarised in Table 3, with Tagalog structures aligned to the relevant English structures. Fate conspiracy Tagalog (fatalistic) : a. causative pa- (indirect action) b. ability/involuntary action ma-lmakac possessive magkad. accidental result magkang- R1/magkanda e. particles skewed to possibility f. perception and cognition verbs —inactive Senser in passive; equivocally involved via ma-lmaka- (cf. b above) in active g. reaction verbs — causal focus on phenomenon (ika-lka-...-an) h. voice — skewed to passive i. aspect Table 3: Summary
English (deterministic) : had x do/made x do/let x do/asked x to do able to/happen to come to have/get 'happen as an accidental result' balanced modality system verbal vs. behavioural (see vs. look at); unequivocal in behavioural
2 way please/like pairs; less productive passive less common tense
of grammatical differences related to the f ate conspiracy
282
J.R. MARTIN
4.0 The Relation of Grammatical Conspiracies to Social Reality Although Whorf took great pains to promote the idea that there was a connection between cultural and behavioural norms and large scale linguis tic patterns, he was never very explicit about the nature of this connection. He does suggest the linguistic patterns do not specifically determine the shape of a culture in that one cannot diagnose or infer the existence of a cultural norm on the basis of a pattern or a pattern on the basis of a norm. And his emphasis on the habitual and unconscious nature of the connection suggest that he did not view global linguistic patterns as absolutely pre determining the behaviour of individuals, nor as preventing individuals from avoiding norms by consciously stepping back and observing their lan guage and their culture (a process in which Whorf himself was constantly engaged). The way in which grammmatical conspiracies and social reality were in fact related was never however made clear. The precise nature of this relation is certainly beyond the scope of this paper. But there are three significant points which can be made. First, it is important to echo Whorf's perspective on the habitual and unconcious nature of the relation. There is no reason why this paper could not in fact be written by a Filipino, any more than the comments made on English lan guage and culture herein are invalidated because the author is a native speaker of English. It is always possible to step back from a language and culture and through a conscious act of semiotic reconstruction transcend the connections Whorf was on about. In addition, there will always be individu als who transcend the connection more often than others, and every indi vidual is likely to contradict the relation between conspiracies and culture from time to time. The point here is that Whorf's hypothesis is about typical behaviour, not about humans as 'helpless prisoners of the categorisation schemes implied' by language (Sampson 1980: 102). Second, language and culture are distinct levels of abstraction. Lan guage realises a culture, but it does so indirectly. It does not make the meanings which constitute a culture (otherwise there would be no reason to distinguish language and culture): rather, it encodes them. As such the rela tion between the two levels can perhaps best be understood in terms of phonaesthesia. Just as there is a tendency in many languages for phonolog ical patterns to transcend arbitrariness and reflect lexicogrammatical mean ings (for example the [ŋ] 'oral' series in Tagalog—examples from Panganiban 1973; starred items are technical and not used by many Manila speakers:
GRAMMATICAL CONSPIRACIES IN TAGALOG
283
ngiti 'smile' ngipin 'teeth' *ngisbi 'expression of lips when sobbing' ngisi 'grin' *ngisngis 'continued opening of mouth to grin or show teeth when laughing' angil 'growl with teeth showing' ngalangala 'palate' *ngalot 'crush by chewing' nguya 'mastication' *ngasab 'loud smacking of lips while eating' *ngawa 'up and down movement of jaws' *ngidngid 'gum of teeth' ngiwi 'crooked mouthed' *ngubngob 'completely toothless with lips sunken' etc.), Languages vary in terms of the degree to which phonaethesia is found, just as Whorf suggests that languages may vary in the extent to which grammat ical conspiracies are present. And phonology never codes phonoaesthtically more than a small portion of the meanings the words and structures of that language make. Phonaethesia is perhaps best interpreted as a kind of leak age of the lexicogrammmatical meanings into phonology. The way in which grammatical conspiracies code a culture probably represents a leakage of a similar degree. There is another aspect of phonaethesia which is especially interesting, and this brings up a third point about the nature of the relation between language and culture. This is that phonaesthesia is symbolic — phonaesthetic series symbolise lexicogrammatical patterns. It is not just that the phoneme /ŋ/ happens to be found in lots of words having to do with 'orality'; but somehow, Tagalog speakers associate the acoustic and articulatory substance of the velar nasal with oral activity (another example of this kind of symbolic relation between grammar and phonology would be that func tion words tend to be short and lexical items longer, reflecting the specific ity of the meanings they encode). The ng series thus acts as a metaphor for eating, chewing, grinning, the parts of the mouth and so on. As such it pow erfully reinforces the semantic association that would in any case be felt among the lexical items connected with orality.
284
J.R. MARTIN
The relation between grammatical conspiracies and culture is also of this kind; it is not an accident that the particular series reviewed in section 3.0 are the ones connected to family, face and fate. Rather, these systems symbolise the cultural meanings considered. The structures reviewed are conventionally connected to the family, face and fate — another language might use a different set of structures to realise the same cultural values (just as cats meow in English but ngiyaw or ingaw in Tagalog; the relation is conventional but not arbitrary — the substance of all three words is onomatopoeic); but the relation is not arbitrary—the structures involved are predictable vehicles for the more abstract meanings a culture involves. The grammatical metaphors related to the family conspiracy have to do with joint action, sharing (both qualities and understanding) and grouping. Each of these is relatable to the general theme of togetherness through par ticipation. Both social verbs (3.1a), reciprocal verbs (3.1b) and comitative ka- (3.1c) symbolise joint action. Ka- (comparison of equality (3.1d) sym bolises shared qualities, while the implicitness (3.1h) involved in elliptical and indefinite structures acts as a metaphor for shared understanding. And Tagalog's plural case markers (3.1f), inclusive and fused pronouns (3.1j), and special coordinate construction (3.1g) symbolise an orientation to groups as a collection of individuals. The metaphors related to face have to do with distance, mitigation and disclaiming. Each of these is related to a general theme of oblique or indi rect engagement with others. Distance is symbolised in the use of plural or 3rd person pronominal forms with singular references (3.2c). Moving from the singular to the plural or from the first person to the third has the effect of removing the speaker slightly from the context in which he is speaking, in deference to the person he addresses. Mitigation is most apparent in directives, where requests for goods and services are softened by treating the action involved as a joint one (via maki-: 3.2d) or one which simply has to be done now and again (moderative verbs; 3.2e). And both the reported speech particle daw/raw (3.2g) and indefinite may/wala (3.2h) structures have the effect of distancing the speaker from what he is saying. Fate is powerfully symbolised through agency, with powerlessness and irresponsibility the unifying theme. The equivocal causative (pa-; 3.3a), abilitative (ma-, maka-; 3.3b) and possession (magka- ;3.3c) structures are all ambivalent about who is actually responsible for initiating an act. And the peculiar reactances of both perception (3.3f) and reaction (3.3g) mental process verbs, along with the middle orientation of voice in general (3.3h)
GRAMMATICAL CONSPIRACIES IN TAGALOG
285
all draw attention away from the agency associated with a human partici pant in the clause, and focus on the typically non-human participant involved in the process. Modality (3.3c) also functions as a metaphor for powerlessness, with the system strongly skewed towards possibility. These metaphors are summarised in Table 4. family
face
fate
joint action
social and reciprocal verbs; comitative kasharing: quality kacomparison of equality sharing: understanding implicitness grouping plural names; special coordinate con struction; relational verbs; inclusive and fused pronouns distance plural and 3rd person address with sin gular referent mitigation joint and moderative requests disclaiming reported speech particle and indefinitestructures agency causative, abilitative and possession equivocation; perception and reaction reactances; voice certainty particle system skewed to possibility Table 4: Summary of grammatical metaphors symbolising family, face and fate values
Grammatical conspiracies can thus be seen not simply to realise or code family, face and fate in Tagalog, but to act as metaphors for the par ticular interpretation of family, face and fate reflected in Filipino culture. The point is, in Halliday's terms, that language serves 'both as a vehicle and as a metaphor, both maintaining and symbolising the social system.' (1977: 25). Filipinos do use their language as a vehicle—as a means of directly cod ing cultural values. They write articles about being a good neighbour (fam ily), discuss shame and debts of gratitude (face) and often explicitly credit fate or God when something goes very right or very wrong. Tagalog, like any language, can be used to directly code cultural values in a purposeful conscious way. But beyond this, in an unconscious and invisible way, Tagalog structures continually symbolise and reinforce these values even when the Filipinos are talking in a literal sense about something else. It is
286
J.R. MARTIN
this indirect expression of culture that fascinated Whorf. What makes it so powerful, alongside its unconscious habitual practices, is its symbolic nature. Tagalog functions as a metaphor for the culture it sustains at the same time as it serves the myriad of more specific and apparently unrelated functions it must serve if the Filipinos are to use it on a day to day basis to live. Moreover, it is not hard to see that the family, face and fate con spiracies are themselves congruent. Taking the family conspiracy as the point of departure, there are a number of ways in which the social organisa tion being symbolised leads to a preoccupation with face. Since privacy is at a minimum, many occasions arise in which appearances are more effective than fact. One is constantly surrounded by relatives of different genera tions, and their status has to be acknowledged. The extended family includes many members whose blood ties are weaker than those of others, from which various sorts of indebtedness and responsibilities arise. And so on. Similarly fate follows on from family and face. This seems to works as follows. The extended family unit with complex face relationships intro duces a very large number of variables into the life of the Filipinos. Obvi ously an individual has control over only a few of these, so mastering one's fate is a difficult task. On the other hand the family unit introduces consid erable flexibility into one's future. One can't plan, but there are always a number of possiblities at hand; and each of these leads on to a number more. There is no need to complete the circle at this stage. Suffice it to suggest here that the grammmatical conspiracies reviewed in this paper har monise with each other at a cultural level, and that this harmony contrasts with the equally congruent western set of individualism, forthrightness and mastery of one's destiny.
5.0 Related Linguistic Evidence Because of their unconscious nature and symbolic power, grammatical conspiracies clearly constitute the most important piece of evidence rele vant to Whorf's interpretation of the relation between language and social reality. However, once established, these conspiracies form a foundation to which further evidence may be added. Three promising additional sources
GRAMMATICAL CONSPIRACIES IN TAGALOG
287
of evidence in Tagalog have to do with (1) words and lexicalised phrases which somehow resonate in the culture; (2) words with an unusually large range of derivationally related forms in the language; and (3) certain types of 'play' with language. All languages have a number of words and phrases which strike a chord in the culture as it were. They are perhaps easiest to recognise when first learning a language. When first used, they bring a knowing smile to the native speaker, and often prompt a compliment on the progress the lan guage learner is making: Now you are really talking Filipino, or Now you are really starting to talk like us. Expressions related to the family conspi racy in this way include: sige na
'c'mon — join in'; usually in the context of joining in some group activity May asawa ka ba? 'Are you married?'; usually asked within a few sec onds of meeting a Filipino — followed by May anak ba? 'Do you have children?'; followed by Bakit? 'Why?' and a discussion of contraception if none Sino ang kasama mo? 'Who was/is/will be your companion?'; whenever going out or coming home; followed by pity and some consternation if you have been or are going out alone barkada 'peer group'; who you should have been going out with in the first place Or, to take another example, fate resonates in the following words or phrases: bahala na mamaya mayamaya titingnan wala
ko
'it's up in the air — undecided' 'later'; with reference to future plans 'soon' 'I'll see'; if pressed to consent to a future event 'nothing'; when asked what is wrong, why something is the way it is, when looking for something
To take some English examples — drawing on Australian English and cul ture — consider the following expressions centering around the theme of 'irreverence': chucking a Uey
'making a u-turn in a road'; normally delighted in, especially if illegal
288 bludger wanker she'll be right mate take the piss out of stir
J.R. MARTIN
'someone exploiting the system through minimal effort 'someone who mentally masturbates — to show off — or even for a living'; (e.g. academics) 'don't worry'; usually in situations of dire stress 'criticise'; a person, especially wankers 'tease'; generally through a blunt verbal attack
These resonating words and phrases probably do not stand as conclusive evidence as far as the existence of the underlying cultural values are con cerned (they are certainly seriously over- rated in the social anthropological literature); and it is not clear how lists of such items can be empirically established. Nevertheless, native speakers do have clear intuitions about their significance, as lexicalised indices of a culture, and as such they are a valuable source of evidence when taken in conjunction with that proved by grammatical conspiracies. A second source of evidence relates to the morphological scatter of certain lexical items. The clearest example of this in Tagolog is the word sama 'companion', related to the family conspiracy. This term has 40 sepa rate entries in its list of morphological derivatives and idiomatic uses in Pangangiban's Diksyunaryo-Tesauro Filipino-Ingles (1973). This scatter is larger than that of any other item in the dictionary, and many times larger than that of the average item, even the average item of core vocabulary. Again, collecting evidence of this kind is problematic. Pangangiban's criteria for including a separate derivational or idiomatic entry are unclear. But it may well be that certain culturally salient items have a far greater morphological scatter in a language than items of other kinds. Not unrelated to this scatter is McFarland's 1984 report of sama as among the 10 most frequent lexical items in his study of some 5,000 sen tences, exceeded only by general material, verbal and mental processes: gawa 'do', sabi 'say', dima 'know', ibig 'love', isip 'think', gusto 'like', and yari 'happen'. Finally, the predilection of Filipinos for English acronyms needs to be noted. These are used not simply to refer to institutions and organisations as in English (UNESCO, LSA, etc.) but in a large variety of more social contexts as well. So we find S.I.R. 'smooth interpersonal relationship', A.I.D.S. 'acute income deficiency syndrome' and so on. This kind of lan guage play reflects the family conspiracy iconically symbolising through abbreviation and actively realising through the assumption of shared under-
GRAMMATICAL CONSPIRACIES IN TAGALOG
289
standings the solidarity of those participating in this type of code-switching language game. Probably the most troublesome kind of evidence as far as Whorf s hypothesis is concerned is lexical. There are two reasons for this. First of all speakers are far more conscious of words than structures, and are thus in a stronger position to adapt them to their own particular ends. Subcultures in any society continually operate in this way, evolving, often quite dynami cally, alternative lexicalisations which distinguish them from the rest of soc iety and may even oppose them to it. Slang is an example of this process; the language of homosexuals in Manila, which is not intelligible to outsid ers, is an extreme case, and is the source of slang terms in the community at large (see Halliday on anti- languages for further discussion; 1976 [1978]). The second reason for the weakness of lexical evidence is its extreme volatility. Lexis is far more flexible than grammar, adapting more quickly to specific changes in the social environment of a community. Bor rowing from Spanish and English in Tagalog clearly illustrates this adapta bility. Lexis is simply so unstable that it is hard to demonstrate just how it could determine or somehow condition the view of social reality assumed by some community (however, cf. Poynton 1985). One problem with lexical evidence can be illustrated with respect to sexism and lexical items referring to humans in the Philippines. Urban Manila society is one of the most sexist in the world, clearly dividing women into two classes: god's police (sisters, mothers, daughters and other female relatives) and 'others' (waitresses, dancers, massage parlour attendants, fashion models, 'bold actresses' and so on). There are so many impoverished women in the city that life as a sex object is the only viable alternative (it needs to be kept in mind in this regard that the vast majority of these women serve the local market, not the tourist trade). Yet Tagalog makes fewer sex differences in its terms referring to people than English: sanggol 'baby' anak 'child of' bata 'young person' kaibigan 'friend' kapatid 'sibling' pinsan 'cousin' There is no single word for boy or girl, son or daughter, or brother and sis ter. (In addition Tagalog pronouns do not distinguish sex, which shows the
290
J.R. MARTIN
danger of relying on a single piece of grammatical evidence rather than a conspiracy, and the purely symbolic value of feminists' criticism of the use of he in English to refer neutrally to human participants in discourse.) Lex ical evidence may be more important as far as the relation of language to reality is concerned (Tagalog does have some 40 words for rice). But as far as social reality is concerned, it would appear that great care must be taken when arguing from lexis in favour of, or against, Whorf s claims.
6.0 Related Cultural Evidence It is admittedly somewhat unseemly in a paper making use of Whorf s interpretation of the relation between language and social reality to turn to non-linguistic evidence related to the discussion of family, face and fate. This problem will be taken up again in section 7.0 below. However, given that the nature of the relation of language to culture is far from understood, it may be instructive to look at what scholars approaching the problem from other points of view have to say about the Filipino way of life. Bautista (1979) is extremely helpful in this respect. In connection with her sociolinguistic analysis of Filipino radio dramas, she makes a number of observa tions which appear to support the analysis of family, face and fate used here. The following quotations are suggestive. family: Lynch and Hollsteiner's Understanding the Philippines (1967) is instruc tive. In that coursebook, they view 'social acceptance' as a main goal in Philippine life, with smooth interpersonal relations as intermediate value. The instrumental values for achieving smooth interpersonal relations — pakikisama 'going along with the group', euphemism, and go-between — come to life vividly in the dialogues prepared to 'show how ordinary con versations can reflect cultural values.' (Bautista 1979: 141)
face: It would be good to approach the matter from the other end this time. Without being predisposed to any particular values and themes, we can begin with speech acts and then try to see if the speech acts point to certain values or themes in the culture. For example, the patterns for compliments and probes seem to point to modesty and reticence as seemly on the part of the addressee. Why are modesty and reticence valued? To what theme in the culture can modesty and reticence be related? (Bautista 1979: 142)
GRAMMATICAL CONSPIRACIES IN TAGALOG
291
fate: Samson, Piquero, and Banzon (1976) have come to the conclusion that Philippine radio dramas (together with comicbooks and movies) are pur» veyers of a philosophy of poverty with the following features: (1) Poverty is God's will. (2) Life is like that, that's fate . . . Nofuente (1978) has derived the following values from his analysis: (1) It is fate that decides all things (Bautista 1979: 142). Another useful source of evidence is found in the cultural note included by Ramos and Goulet (1981) in their Intermediate Tagalog developing cultural awareness through language. In this book, an attempt is made to teach the grammar and culture of Tagalog hand in hand, with a short dialogue forming the basis of each chapter. Consider the following quotations: family: Filipinos are usually not inhibited about initiating conversations because talking to a stranger is generally not considered intrusive. If thrown together for almost any reason, someone will break the ice. A common conversation opener is Taga-saan ka?, Where are you from? (38) Friendships are very important, and a friend is usually considered to be a part of the family. Note the use of anak, 'child', by Meding's mother when addressing Lisa. It is also common to call the older relatives of one's friend by their kinship title. Thus, Lisa may call Meding's mother Inay, 'mother'. The use of kinship names is one way of extending the family. (74) Filipinos find it hard to give small parties because they do not want to leave other people out. When it is necessary to limit a party to a few friends, it is kept a secret. (434) Peer pressure to go along with the group, makisama, is strong. A does not believe Jaime's alleged excuse for not going with the group. (462) face: Another common greeting in Tagalog is Saan ka gating?, Where have you been?', which is also equivalent to the English 'Hello'. Note the vague response Diyan lang, There only'. Hayaan mo or ayaan mo, lit., 'Let it be', as a response to the invitation to stop by, is vague and can mean that the person will indeed stop by or is too polite to say he/she cannot. (5) Kathy uses the polite po not only because Rudy is older but also because he is a new acquaintance. Note that she also calls him Mang Rudy. Mang/ mama, an honorific used with males. (The female honorific is Ale/Aling.) A young woman who is not interested in a closer relationship with a man uses po or ho as a cue for him to lay off. (82)
292
J.R. MARTIN Girls have to be careful not to offend men who ask them for a dance. If a girl refuses to dance with a man, she has to have an excuse or a pretext like Hindi ako marunong ng sayaw na iyon, 'I don't know that kind of dance', or Napapahinga ho ako sandali,'T'mresting'. In addition she has to sit out that particular dance or not dance with any other man until she has danced with the man she has refused earlier. (82) In general one does not acknowledge a compliment with 'thank-you'. Instead one downgrades what is being complimented and then returns the compliment. (283) In general, situations involving reprimands are unpleasant for both parties. If they cannot be avoided, one goes about reprimanding another indirectly and hesitantly. Usually, it is justified by laying the blame on a higher authority or external factor. In general, the lower the status of the person being reprimanded, the more tactfully the reprimand is given. One is careful not to 'pull rank' and to hurt the feelings of a social inferior. It is important not to prolong the unpleasantness caused by a reprimand. After a reprimand, one tries to restore the previous friendly relationship by such devices as: 1. giving excuses for why the reprimand had to be made, 2. changing the subject, 3. asking about the person's family, or 4. indicating in some other way that the unpleasant situation is over and should be forgotten. (308)
fate: A begins to realise that life is not that easy in the United States. In spite of A's qualifications, is saying that luck might be a factor in A's landing a job. (274) A is being fatalistic and believes that certain events are inevitable. believes prayers help. (344) believes that the hour of one's death is predetermined. (348) Many Filipinos firmly believe that God does not give a person a 'cross' he cannot bear, and that one's misfortune won't last. Bahala na does not imply passivity but an optimistic acceptance of life's ups and downs. (351) Filipinos believe that one's fortune, whether good or bad, is transitory; it is like a revolving wheel. (353)
A third source of evidence is anecdotal (the traveller's tale). Most Eng lish visitors to the Philippines are able to entertain their friends and col-
GRAMMATICAL CONSPIRACIES IN TAGALOG
293
leagues with a myriad of stories, most of which illustrate cultural clashes relevant to the general themes of family, face and fate. Visitors soon learn the difficulties involved in doing things on their own instead of going along with the group. The art of bribery without appearing to bribe and of bar gaining without losing face must be noted. Filipinos' approach to planning, particularly making dates or appointments, soon distinguishes itself from that of westerners. However, as Bautista cautions (1979: 143-144), although the cultural commentary provided by sociological analyses, cultural notes in language teaching manuals, and anecdotes ring true, they are all at some remove from the linguistically based discourse analysis of the text they are explicitly or implicitly commenting on. The commentary is encouraging; but it has no real empirical validity — particularly if the symbolic relation ship between grammatical conspiracies in Tagalog grammar and Filipino attitudes to family, face and fate is as important as Whorf suggests.
7.0 Language and Social Context Bernstein, in an insightful discussion of the relation of his work to Whorf's, comments as follows: Although Whorf insists that the 'influence of language upon habitual thought and behaviour does not depend so much on any one system . . . [as quoted in section 2.1]', these fashions of speaking, the frames of consis tency are not related to an institutional order, nor are they seen as emerg ing from the structure of social relations. On the contrary they are seen as determiners of social relations through their role in shaping the culture . . . The view to be taken here is different in that it will be argued that a number of fashions of speaking, frames of consistency, are possible in any given language and that these fashions of speaking, linguistic forms, or codes, are themselves a function of the form social relations take. Accord ing to this view, the form of the social relation or, more generally, the social structure generates distinct linguistic forms or codes and these codes essentially transmit the culture and so constrain behaviour . . . It [this thesis] shares with Whorf the controlling influence on experi ence ascribed to 'frames of consistency' involved in fashions of speaking. It differs and perhaps relativises Whorf by asserting that, in the context of a common language in the sense of a general code, there will arise distinct linguistic forms, fashions of speaking, which induce in their speakers dif ferent ways of relating to objects and persons. It leaves open the question of whether there are features of a common culture which all members of a
294
J.R. MARTIN society share which are determined by the specific nature of the general code or language at its syntactic and morphological levels. It is, finally, more distinctly sociological in its emphasis on the system of social rela tions. (1973: 143-4 [1965])
The issue of whether language determines the shape of culture or vice versa is an intriguing one; the relation is presumably symbiotic, rather than direc tional — but this point will not be pursued here. But Bernstein's work is especially relevant in terms of the emphasis he places on analysing, rather than simply commenting on, the system of social relations which constitute a culture and which must be related to language if Whorf s hypothesis is to be carefully investigated. Whorf himself was not equipped to provide a sociological analysis of this kind; indeed, he visited the Hopi on their reser vation only once for a brief period during his research. Thus his analysis of Hopi culture was necessarily somewhat circumscribed by his lack of experi ence living in it. The analysis of the Filipino system of social relations pre sented in this paper is if anything even more primitive than that provided by Whorf for the Hopi. Attention is drawn to this gap to make the point that in the long term Whorf s hypothesis will have to be investigated on an interdisiplinary basis, with linguistically oriented sociologists such as Bernstein working in cooperation with sociologically oriented linguists. There is a limit to how far a linguist can go on his own. There are however schools of linguistics that can go further than others. Halliday for example has suggested as a universal feature of lan guage that grammars are organised paradigmatically into three main metafunctional components: the ideational, the interpersonal and the tex tual (e.g. Halliday 1978). These components reflect the ways in which lan guage is used and can thus be directly related to the traditional Firthian categories of field (realised mainly through experiential meaning), mode (realised largely through textual meaning) and tenor (realised for the most part through interpersonal meaning). These three register categories, field, mode and tenor, can be characterised in general terms as follows: field refers to what is going on, where this is interpreted institutionally (in terms of what people are doing with their lives: e.g. linguistics, cooking, tennis, music, accounting, etc.); mode means the channel of communication, where this is interpreted in terms of spatial and temporal distance— between speaker and addressee and between language and what it is talking about (e.g. face to face, telephone, television, video, film, book, stream of consciousness writing, etc.); tenor is relevant to the social relations between
GRAMMATICAL CONSPIRACIES IN TAGALOG
295
the interlocutors where these are analysed in terms of status, frequency of contact and affect (e.g. power or solidarity, love or hate, frequent or occa sional contact). Some kind of register analysis is an essential step in unravelling the nature of the relation between language and social reality. The reason for this is that register skews the probablities inherent in a language according to the use to which language is put in one situation or another. Unless this skewing is taken into account, texts collected from particular situation types may show people behaving differently than would generally be the case for their language and culture. As far as field is concerned, Halliday has shown (1971 [1973]) that Golding, in The Inheritors, systematically skews transitiv ity selections towards the middle voice when writing from the point of view of the less 'civilised' of the two ancestors of modern man he pits against each other in his novel. Texts written from this point of view will thus resemble those found much more generally in Filipino culture as far as agency and fate are concerned. To take another example from mode, there are contexts, TV commentary on a football match for example, in which context dependent language oriented towards joint action will predomi nate. Again, English texts in this mode, reflecting probablities skewed in terms of shared knowledge and cooperation, would resemble Tagalog texts taken from a far wider range of situation types. Similarly with tenor, there are power differentiated situations, an application for a job interview for example, in which deference, variously realised, would reflect a concern with face typifying Filipino culture. The point is that these examples of Filipino orientation to family and face in English culture reflect English being used for particular discourse tasks. The general system bends, and adapts to a particular need. The difference is that in Tagalog, the orienta tion to family, face and fate discussed above is inherent in the system — either structured into it or included by way of inherent probablities as far as the selection of particular meaning is concerned; in English, on the other hand, similar orientations are contextually specific, reacting to particular register selections. Distinguishing inherent from noninherent probablities presents methodological problems which cannot be resolved here. But the distinction between inherent and register specific predispositions to mean ing must be maintained if the relation between language and culture is to be investigated on a contrastive basis (the question of how Bernstein's notion of code and the notion of genre are similarly involved is important but will not be pursued here).
296
J.R. MARTIN
In short, then, in the long run any thorough investigation of Whorf s interpretation of the relation between language and social reality must incorporate a model of the social context in which language functions. As far as linguistics is concerned this means developing models of language use — of register, genre and code — that can be related systematically to the grammar and lexis of a language and to the work of linguistically oriented sociologists on the structure of social relations. Needless to say, the discus sion of family, face and fate in Tagalog has by no means been thorough in this sense. But hopefully, enough evidence has been presented to show something of the promise of inquiry extending along these lines.
8.0 Conclusion Is language a guide to social reality? Taking conspiracies of grammati cal categories as evidence, and the difference between English and Filipino cultures as a test case, then the answer to the question would appear to be yes. Tagalog grammar does differ from English grammar and many of these differences are not arbitrary. They fall into at least three general groups, each relatable to an identifiable theme in Filipino culture — a family con spiracy oriented to participation, a face conspiracy concerned with appear ances, and a fate conspiracy preoccupied with the external determination of events. Moreover, each of these relations is symbolic; each set of general categories in the grammar functions as a metaphor for one of these cultural themes. As such, these conspiracies function subliminally, unconsciously predisposing Filipinos to ways of saying, meaning and behaving which taken as a whole constitute their culture. The exact nature of this predisposition is not yet fully understood. Models of register, genre and code have not been sufficiently articulated to distinguish inherent from contextually conditioned orientations to meaning. And a linguistically oriented model of Filipino social structure has not yet been developed. One further problem which needs to be taken into account is the phylogenesis of such predispositions. It is this problem that Whorf alludes to when commenting that language and culture will be seen to be connected only when the two 'have been together historically for a consid erable time' (1956: 159 [1941]). The Philippine languages show considera ble lexical evidence of exposure to at least four imperialising languages and cultures: Chinese, Indonesian (Moslem culture), Spanish and American.
GRAMMATICAL CONSPIRACIES IN TAGALOG
297
The extent to which grammar and culture have also been affected is hard to ascertain. As far as the family, face and fate conspiracies reviewed in this paper are concerned, this contact does not appear to have affected matters greatly. The interpretations of family, face and fate given here are usually referred to in terms of 'traditional' or 'real' Filipino values, representative of the culture (Bautista 1979: 143). And there is considerable grammatical homogeneity among Philippine languages, regardless of contact — so much so that simultaneous grammars have been conceived covering a number of different languages (Constantino 1965), and plans for a common lingua franca amalgamating several languages have been under way for almost two decades. The test case of course involves examining an aboriginal linguistic community in the Filipines, which would have experienced little contact with outsiders, and seeing if the connections between grammar and culture discussed here hold true. If they do, this would mean that the influence of foreign cultures on the Filippines, especiallly that of Spanish and American colonists, has been rather superficial. There is more to a culture than cocacola (or coffee), jeans (or dresses), disco music (or guitars) and a few hundred common lexical items — so the superficiality of foreign influence as far as grammar and culture are concerned is hardly surprising. Whatever the result of these investigations, the systematic relation between language and social reality is not one that can be dismissed lightly. Yes, as Sampson suggests, men do have the ability 'to break conceptual fet ters which other men have forged' (1980: 102). Men can stand back from their language and culture, bring them to consciousness, and interpolate whatever changes they see fit. But beyond this, as Halliday points out, 'Soc iety, language and the mind are indissoluble: society creates mind, mind creates society, and language stands as mediator and metaphor for both these processes' (1977: 3). Until both the mediating and metaphorical power of language in relation to culture is better understood, we will indeed remain 'prisoners of the categorisation scheme implied by our language' (Sampson 1980: 102). The luxury of escape is something that only semioticians can afford: and in escaping they incur a debt which is in no measure repaid by means of romantic dismissals of the connections between lan guage, its structure and use, and the social reality in which we live.
298
J.R. MARTIN NOTES
1.
Abbreviations used in glosses for Tagalog examples: TM — Topic case marker NTM — non-Topic case marker CM — circumstantial marker T — Topic form of pronoun — nominative case NT — non-Topic form of pronoun — objective/genitive case — circumstantial form of pronoun — oblique case LK — linker.
2.
e.g. speaking to your cat: Hulihulihin mo ang mga daga catch a little you TM some rats 'Catch a few rats (why don't you)'.
3.
See Martin 1983 for a discussion of the change in definiteness in 4.0.
4.
These structures are glossed to note the parallels between Subject and Topic as unmarked Themes in the two languages, thus 'He is my playmate' for (13), not 'I am his playmate' (cf. Kalaro niya ako).
5.
Halliday, personal communication; approximation only.
REFERENCES Bautista, M.L.S. 1979. Patterns of speaking in Filipino radio dramas: a sociolinguistic analysis. (Monograph Series, 13). Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies. Berlin, B. and P. Kay. 1969. Basic color terms. Berkeley: University of California Press. Bernstein, Basil. 1973. Class, codes and control. Vol 1: Theoretical studies towards a sociology of language. St. Albans, Herts.: Paladin. Bloomfield, L. 1917. Tagalog texts with grammatical analysis. Part 1: Texts and translation. Part 2: Grammatical analysis. (University of Illinois Studies in Language and Literature, 2 and 3). Urbana: University of Illinois. Cena, R.M. 1979. Tagalog counterexamples to the accessibility hierarchy. Studies in Philippine Linguistics 3:119- 24. Constantino, E.A. 1965. The sentence patterns of twenty-six Philippine languages. Lingua 15:71-124. De Guzman, Videa P. 1978. Syntactic derivation of Tagalog verbs.
GRAMMATICAL CONSPIRACIES IN TAGALOG
299
(Oceanic Linguistics Special Publication, 16). Honolulu: University of Hawaii Press. Firth, J.R. 1968. Selected papers of J.R. Firth, 1952-1959. F.R. Palmer, ed. London: Longman (Longman Linguistics Library). Flores, F.G. 1976. Tagalog for missionaries—phase II: enrichment. Manila: Interchurch Language School and Phillippine Christian University. Halliday, M.A.K. 1970. Functional diversity in language, as seen from a consideration of modality and mood in English. Foundations of Lan guage 6,3:322-61. (Reprinted in G. Kress, ed., Halliday: system and function in language. London: Oxford University Press, 1976. 189-213). . 1971. Linguistic function and literary style: an inquiry into the lan guage of William Golding's The Inheritors. In S. Chatman, ed., Literary style: a symposium. New York: Oxford. 362-400. (Reprinted in M.A.K. Halliday, Explorations in the functions of language. London: Edward Arnold, 1973. 103-138.) . 1976. Anti-languages. American Anthropologist 78,3. (Reprinted in Halliday 1978: 164-182.) . 1977. Aims and perspectives in linguistics. (Occasional Papers, 1). Brisbane: Applied Linguistics Association of Australia. . 1978. Language as social semiotic: the social interpretation of lan guage and meaning. London: Edward Arnold. . 1985.An introduction to functional grammar. London: Edward Arnold. Halliday, M.A.K. and R. Hasan. 1976. Cohesion in English. London: Longman. Hasan, R. 1984. Ways of saying: ways of meaning. In R. Fawcett, M.A.K. Halliday, S.M. Lamb and A. Makkai, eds., The semiotics of culture and language. Vol 1: Language as social semiotic. London: Frances Pinter. 105-162. Hymes, D. and J. Fought. 1981. American structuralism. (Januarum Linguarum, Series Maior, 102). The Hague: Mouton. Joos, M., ed. 1957. Readings in linguistics I. Chicago: University of Chicago Press. Kress, G., ed. 1976. Halliday: system and function in language. London: Oxford University Press. Lynch, F. and M. Hollsteiner. 1967. Understanding the Philippines: a study of cultural themes. Quezon City: Ateneo de Manila Institute of Philip pine Culture.
300
J.R. MARTIN
Martin, J.R. 1981. Conjunction and continuity in Tagalog. In M.A.K. Halliday and J.R. Martin, eds., Readings in systemic linguistics. London: Batsford. 310-36. . 1983. Participant identification in English, Tagalog and Kate. Australian Journal of Linguistics 3,1:45-74. Mathesius, V. 1964. On linguistic characterology with illustrations from English. In J. Vachek, ed., A Prague school reader in linguistics. Bloomington: University of Indiana Press. 59-67. (Actes du Premier Congres International de Linguistes à La Haye, 1928. 56-63) McFarland, Curtis, D. 1984. Computers and Tagalog grammar. In A. Gon zales, ed., Language planning, implementation and evaluation: essays in honour of Bonifacio P. Sibayan on his sixty-seventh birthday. Manila: Linguistic Society of the Phillipines. 230-246. Nofuente, V.L. 1978. Pananaw sa buhay na hatid ng drama sa radyo at telebisyon 'Views of life in radio and television dramas'. Manuscript. Department of Psychology, University of the Philippines. Poynton, C. 1985. Making the difference: the language of gender. Geelong, V i c : Deakin University Press. Rafael, T.C. 1978. Topic in Tagalog revisited. Studies in Philippine Lin guistics 2:36-48 Ramos, T.V. and R.M. Goulet. 1981. Intermediate Tagalog: developing cultural awareness through language. Honolulu: University of Hawaii Press. Samson, L., M.R. Piquero, and C R . Banzon. 1976. Mga mukha ng kahirapan T h e faces of poverty'. Sagisag 2,3:7-9 Sampson, G. 1980. Schools of linguistics: competition and evolution. Lon don: Hutchinson. Sapir, E. 1949. Selected writings in language, culture and personality. D. Mandelbaum, ed. Berkeley: University of California Press. Schachter, P. and F. Otanes. 1972. Tagalog reference grammar. Berkeley: University of California Press. Spier, L., ed. 1941. Language, culture and personality: essays in memory of Edward Sapir. Menasha, Wisconsin: Sapir Memorial Publication Fund. Whorf, B.L. 1956. Language, thought and reality: selected papers of Ben jamin Lee Whorf. J B. Carroll, ed. Cambridge, Mass.: M.I.T. Press.
GENERIC SITUATION AND REGISTER: A FUNCTIONAL VIEW OF COMMUNICATION.1
Michael Gregory Glendon College, York University Toronto, Canada
1 Any language event engages the knowledge of both user and receiver: knowledge of the world and knowledge of a language. These are not, of course, two distinct knowledges. Our knowledge of the world is something which we largely receive, store, and process in terms of language; and the languages we know themselves answer to the worlds in which they are used. It is in terms of this dialectical socio-cognitive view of language that, together with associates at York University, Toronto, I have been revising and developing the framework for describing intra-lingual variety which I first proposed in 1967 and expanded in 1978 (cf. Gregory 1967, Gregory and Carroll 1978). A socio-cognitive perspective is of necessity a functional perspective: knowledge for the social scientist is knowledge as function, as social fact. Mao Zedong put the matter well: Start from perceptual knowledge and actively develop it into rational knowledge . . . Practice, knowledge again, practice, and again knowledge. This form repeats itself in endless cycles . . . Such is the whole of the dialectical-materialist theory of knowledge, and such is the dialecticalmaterialist theory of the unity of knowing and doing (Mao 1975: 1, 308).
Our language behaviour is surely a matter of both knowing and doing, of a stored, always ready to be activated knowledge of the code and of its rele vance to our lives, and a matter of what J. R. Firth called the noises we make with our faces in order to live. It has been an increasing awareness of
302
MICHAEL GREGORY
this dynamic, dialectical relationship between code and behaviour, behaviour and code, an awareness fostered by confronting texts (records of language events) that has led to the development of that kind of systemicfunctional linguistics called COMMUNICATION LINGUISTICS (cf. Gre gory 1985b, 1982- and forthcoming, and Asp 1985 for the theoretical princi ples of communication linguistics). Rather than point by point revising the framework I presented in the 'sixties and 'seventies, I am going to make explicit how I presently regard the phenomenon of diatypic language variety, synchronically considered. This matter of synchrony and diachrony should not, however, be taken as a strict dichotomous distinction. What I will be concentrating on is 'variety according to use' as opposed to 'variety according to the user'. But 'use' involves the 'user' and the user has a human history. The user has had and exists in a temporal dimension; he shares with other users a TEMPORAL PROVENANCE. He has learnt his language and uses it in certain areas of the world in which he lives; he shares with others a GEOGRAPHICAL PROVENANCE. The user spends most, if not all, of his life within a par ticular socio-economic political system; he has and shares a SOCIAL PROVENANCE. The user is an individual as well as a member of tem poral, geographical, and social provenances; so he has an INDIVIDUAL PROVENANCE. He or she is, then, a participant in a COMMUNICAT ING COMMUNITY CONTEXT. The communicating community context is the macro-environment both of the user's knowledge and of the language code (or codes in the case of multi-lingual societies) he controls; it is the environment that is reflected in language in terms of TEMPORAL DIALECT, GEOGRAPHICAL DIALECT, and SOCIAL DIALECT. The user's own unique response to this environment is reflected in his indi vidual dialect, or IDIOLECT. The communicating community context as a knowledge environment is a source of message potential: messages which may, or may not, be transmitted by a language code, which is why, descrip tively and theoretically, we need the distinct, although related, parallel categories: temporal provenance — temporal dialect; geographical prove nance — geographical dialect; social provenance — social dialect; indi vidual provenance—idiolect. A user's knowledge is, of course, only accessi ble to us as knowledge in so far as it is communicable, which is one reason why there is deliberate repetition in the categorical term 'communicating community context'. Moreover, our human, social communication takes place, and is
GENERIC SITUATION AND REGISTER
303
describable, in terms of communication in situations which, although unique and non-repeatable, have features in common with other unique, instantial situations. As Sir Alan Gardiner put it some fifty-six years ago: In the traffic of daily life, situations are constantly arising so closely similar that we do not hesitate to speak of them as the 'same situation'. Every lan guage has its own fixed ways of coping with certain recurring situations (1932: 6).
I suggest that we communicatively cope with these 'certain recurring situa tions' because we have internalized a generalized knowledge of them. Any INSTANTIAL SITUATION we respond to in terms of this knowledge which is a knowledge of what I term GENERIC SITUATION, the microenvironment of our knowing and doing. This knowledge includes a knowl edge of the meaning potential of the media of communication open to us. In the case of language as a kind of doing, the meaning potential associated with a generic situation is called REGISTER. I suggest, too, that any particular communicating community context and generic situation are in a model — instance relationship; in the same way, any configuration of temporal, geographical, social and individual dialects is describable in terms of its available registers. Furthermore, I am suggesting that we cannot conduct linguistics, the systematic study of language, in a way which accounts for both language homogeneity and language variety, without involving ourselves with the study of language users' and receivers' knowledge as it is relevant to the description of the discourses they construct between them. Such study, fol lowing a suggestion of Sydney Lamb (1971), I call GNOSTOLOGY. I am not saying that the linguist has to account for the whole of human knowl edge, but that he has to be prepared to account, in a principled way, for the knowledge that is at risk in the texts of the language he is describing. (In conversation I have heard Kenneth Pike say that in coming to grips with texts in a new language one needs not only a phonetics manual, a grammar, and a lexicon, but also an encyclopaedia.) The knowledge at risk in a dis course is circumscribed by the parameters of time, place, and social struc ture I have mentioned; what was known in the sixteenth century is different from what is known now; what is known in Beijing is different from what is known in Mallorca; what was to be known within feudalism differs from what is to be known within capitalism or socialism. What one individual knows differs from what another individual knows. This is important for the linguist concerned with text because what we 'can mean' and 'can say' as
304
MICHAEL GREGORY
forms of 'doing' (cf. Halliday 1973: 48-71, 1978: 39ff.) is in terms of what we 'can know'. Knowledge is both activated in and gained through instances of the generic situations in which we are communicatively involved in terms of a relationship to a human experience or experiences, in terms of an inter active relationship to our fellow human(s), and in terms of a relationship to a medium of communication. This configuration of EXPERIENCE RELATIONSHIP, INTERACTION RELATIONSHIP, and MEDIUM RELATIONSHIP, provides a complex COMMUNICATIVE FUNCTION for the generic situation, and they are the dimensions along which it is describable. In the case of linguistic communication, a knowledge of a linguistic code is involved, a knowledge of appropriate dialects and registers. Such a code is describable in terms of a realizatory code cycle: 'meaning', the stratum of SEMOLOGY, encoded as 'wording', the stratum of MORPHOSYNTAX, encoded as 'sounding', the stratum of PHONOLOGY, or 'signing', the stratum of GRAPHOLOGY. From the receiver's point of view, 'soundings' or 'signings' are decoded as 'wordings' and as 'meanings' in terms of available knowledge. What I have said so far is summarized in Figure 1.
Figure 1: User's/receiver's gnostology
GENERIC SITUATION AND REGISTER
305
2
Following M. A.K. Halliday, the code is seen to be not only tri-stratally organizable but also tri-functionally (cf. Halliday 1973, 1978, 1979). The functional organization of language is taken to be both extrinsically and intrinsically justifiable. Extrinsically, language has to answer to the demands of both ideational meaning (experiences, real or imagined and their logical relationships to each other) and interpersonal meaning (role assignment and adoption, intention and attitude). As we are message giving animals whose behaviour takes place in space and time, whose media of living are serial, our communicative activities also have organizational meaning: points of message departure and highlight points. In other words, we have to have something to say, somebody to say it to, and we have to say one thing before another. So we have the three -FUNCTIONS of lan guage: the IDEATIONAL, the INTERPERSONAL, and the TEXTUAL (which is our term for language's message organizational function). Extrin sically, these functions of language correspond to the dimensions of the micro-environments of generic situation: the dimension of experience relationship corresponds to the ideational function; that of interaction relationship to the interpersonal function; that of medium relationship to the textual function. Intrinsically, the three functions are justifiable because the message carrying units of language can, in their structural realization, be seen as the configuration of choices from three different kinds of options: in technical terms, as the selection of features from system networks assignable to the three different functions. I will illustrate in terms of English. Let us con sider, in terms of simultaneous configurations of meaningful structures, three English propositions realized by independent clauses: 1. John washed the dishes. 2. The dishes were washed by John. 3. Did John wash the dishes? (' indicates the tonic syllable). What do they have in common? All three have selected a semological pro cess type Action: Affective (WASH) with the participant roles Agent (JOHN) and Patient (DISHES). All three have selected the same semolog ical Time of Event (BEFORE NOW) realized by syntactic Past Tense. So they share a common ideational meaning.
306
MICHAEL GREGORY
The first two have selected the same semological Speech Function: Comment: Statement (as opposed to Comment: Exclamation, or Directive: Question, or Directive: Command) realized syntactically by Indicative Declarative clauses; that is by clauses which have a Predicator (washed, were washed) and in which the Subject (John, the dishes) precedes the Finite Predicator element. They differ from the third example, Did John wash the dishes?, which has selected the semological Speech Function Directive: Question realized by an Indicative Interrogative clause (one in which the Finite part of the Predicator (did) precedes the Subject (John). That is, they differ interpersonally. The user has selected the role of 'giver of information' in John washed the dishes, and The dishes were washed by John, and 'demander of information' in Did John wash the dishes', the first two necessitate no response from the interlocutor but the third one does, and so the user has assigned a different role to the receiver. All three differ from each other in their organization, their 'textuality'. In John washed the dishes the semological Focus (message starting point) is on John as participant and on the speech function Comment: Statement realized morphosyntactically by Subject as Unmarked Theme in a Declara tive clause. The semological Prominence (message highlight) is on dishes as Patient realized by the tonic syllable (New) in the information system of the phonology. In The dishes were washed by John the semological Focus is on dishes as participant and on the speech function Comment: Statement realized by Subject as Unmarked Theme in a Declarative clause. The Prominence is now on John as Agent realized syntactically by optional Adjunct and phonological New. The semological Focus in Did John wash the dishes? is on the speech function Directive: Question realized by the Finite element of the Predicator as Unmarked Theme in an Interrogative clause of the Polar (yes/no) sub-class. The Prominence is on dishes as it was in the first example. The structural realization of all three sentences can be seen then as a simultaneous configuration. I have discussed the extrinsic and intrinsic tenability of a multi-func tional view of language because it is central both to an understanding of the sources of 'varieties according to use' and to their principled prediction and description.
GENERIC SITUATION AND REGISTER
307
3
I have said that it is in a situation as an instance of a generic situation that, as users and receivers of a language, we have a communicative func tion. It is only in terms of a relationship to an experience or experiences, in terms of a real or imagined audience, and in terms of a particular medium that we have, if you like, anything to 'say'. Corresponding to the function the instantial and generic situation provides us with is our knowledge (im plicit or explicit, more or less conscious) of register, the meaning potential of the language we are using or receiving appropriate to the situation. The experience relationship predicts the range of options in the idea tional component of the language code: what process types (action, mental, or relational) that are at risk; the nature of the participants, animate or inanimate, known or unknown etc., the circumstances of time, place, and manner of the event; and the perspective we have on the experience as something which is durative, habitual, completive, inceptive, and so on. These semological options correspond to the particular kind of experience we are relating to and are realized in the morphosyntax of English, for example, by the selections made in the transitivity and adjunctivization sys tems of the clause, the modification, head, and qualification systems of the nominal group, and the tense and aspect systems of the verbal group. The particular range of such options we have in a given register constitute what I called in my earlier work the potential FIELD OF DISCOURSE avail able; the actual selections taken in a particular language event constitute the field of that particular discourse. Let us compare and contrast two short portions of discourse in these respects. The full discourses are given in Appendix I. The first is from a recipe for Spaghetti Bolognese: Melt the butter in the olive oil in a heavy saucepan on top of the stove. Finely chop the onion, garlic cloves, salt pork or bacon and sauté in the oil and melted butter until soft but not brown.
The second is from a political position pamphlet: Let's be clear: I personally favour the initiative and ardently support disar mament negotiations to reduce the risk of war. But I don't think endorsing a specific nuclear freeze proposal is appropriate for California Common Cause. We should limit our involvement in defense and weaponry to mat ters of process, such as exposing the weapons industry's influence on the political process.
308
MICHAEL GREGORY
In the recipe text we note the incidence of Action Affective processes (melt, chop, sauté), and the participant roles of Patient that are inherent with such processes realized by butter, onion, garlic cloves, salt pork, bacon; and we note too that such participants are inanimate and concrete, as are the participant referents in Adjuncts realizing the circumstantial roles of Place: olive oil, heavy saucepan, stove. There are no Mental processes and only one Relational predication (until soft but not brown) which is attributive to the inanimate concrete entities as a result of action processes. The verbal groups are tenseless and without aspect. The cooking experi ence is timeless! In the political text, the process type is predominantly Mental. There are the Mental: Reaction processes realized by favour, support, think (reac tive rather than cognitive in this text), endorse, limit . . . involvement', and the Mental: Verbalization: expose. The clause Lef s be clear realizes an Attributive Relation but the attribute in question is a mental one. With the exception of the user, and the organization he is addressing, the participant realizations are abstract: initiative, disarmament negotiations, risk of war, specific nuclear freeze proposal, defense, matters of process, weapons indus try's influence on the political process. We note too the density of nominalizations of predominantly mental processes: initiative, disarmament, negotia tions, risk, proposal, involvement, influence. With the exception of should limit, the verbal groups are present tense realizing Now and After Now semological choices. Experientially the interpersonal modal should also realizes Now and After Now time. So the two texts are different experientially; they have different fields of discourse; they have selected from different meaning potentials in this respect. It is part of our living and linguistic knowledge that the experience of cooking involves actions affecting concrete entities and that the experi ence of political position in a U.S. organization such as California Common Cause in the late eighties of the twentieth century involves mental reaction and abstraction. As what Halliday has called 'on-going social activities', they are very different and this is reflected in the ideational potential of their respective registers (cf. Halliday 1978: 221-227). The interactive relationship in the situation as an instance of a generic situation predicts the selection of options in the interpersonal component of the language code. This involves the choice of Speech Functions (Com ments and Directives of different kinds) realized in English by the syntactic system of Mood with options between Indicative and Imperative, and
GENERIC SITUATION AND REGISTER
309
within Indicative between Declarative and Interrogative, and by the choice of phonological 'tune' or intonation contour. It also involves the matter of user's Turn, whether the speaker is Initiating or Responding or Continuing which is related in the syntax to choices between Major or Minor, Full or Elliptical clauses. There are options which indicate Social Distance between speakers, reflected, for instance, in the use of address terms, ranging from My Lord, or Sir, to Mary, darling, and duckie. Interlocutor Assessment, a checking by the user of his receiver's stance, is reflected in syntactic tags such as isn't he? and items such as eh?, huh?, right?, and what? with charac teristic phonological features. The systems referred to so far are concerned with the interpersonal function as the negotiation of roles between user and receiver. There is another set of interpersonal systems which allow the user to indicate to the receiver his assessment of, and attitude to, the events and relations he is verbalizing. The range of syntactic modalities in the English verbal group: may, might, can, could, shall, should, will, would, must, ought etc. allow the user to mediate as regards the event he is realizing in terms of different kinds of assessment of likelihood, capacity, and hypothesis. There are sentence rank elements in the syntax realized by items such as probably, possibly, perhaps, and maybe which fulfill a similar semological role as well as comment items such as fortunately, mistakenly, happily, and sentence rank final Attitudinals realized by clauses such as which is a pity, which was a good thing, which allow the user to intervene, personally as it were, in his message exchanging. The particular range of interpersonal options at risk in a given register constitute the TENOR OF DISCOURSE available: the actual selection taken in a particular language event constitutes the tenor of that particular discourse. If we consider the Speech Functions selected throughout the spaghetti recipe I quoted an excerpt from, we find that of the twenty-seven proposi tions realized by independent clauses in the Method part of the text, twenty-six have selected Directive: Command realized by the congruent syntactic mood selection, Jussive: Imperative; that is, by a clause with a neutral form Predicator and no Subject: melt the butter . . . chop the onion . . . add half the wine . . . turn the heat up, and so on. The one Comment: Statement, the sauce will now be ready, occurs as a transition between the Method and Service parts of the recipe. This is a pattern maintained in all the weekly recipes I have read over the last six months in the English ver sion of the China Daily. It reflects an important aspect of the interactive
310
MICHAEL GREGORY
relationship in the generic situation which predicts the tenor of the register. In the recipe situation (as also in barbecue or bicycle assembly situations), a relationship of 'instruction' is accepted from user to receiver and by user and receiver, and this allows the Jussive Imperative, which otherwise is rarely used by English speakers in realizing commands. Unless there is such a marked feature in the generic situation, Commands in English tend to be realized by Interrogative clauses with second person Subject, a Modal Aux iliary and the item please, as in Would you close the window, please?. But we would find it inappropriate in the recipe register if we met Would you please melt the butter? . . . Would you please chop the onion? Such selec tions might be made in some kinds of co-operative cooking situations but they are not recipe situations. In other respects the tenor of this recipe is characterized by the absence of mediation and attitudinal features. The one Modal Auxiliary verb, will, in the sauce will now be ready, is used predictively with no fea ture of intention. The political pamphlet exhibits a very different tenor, a very different selection from the interpersonal resources of the language. In the portion I quoted, all but one of the Speech Functions are Comment: Statement realized by Indicative Declarative: I personally favour . . . (I) ardently sup port . . . I don't think . . . we should limit. The one exception is the Direc tive: Command realized by Optative Imperative: Lef s be clear, which is a rhetorical command, demanding no response. One notes, too, the attitudinals, personally and ardently, and the modal, should, realizing obligatory responsibility in We should limit our involvement, which selection moves the comment function toward directive. None of these features are surpris ing when the 'persuasive' feature in the interactive relationship of the generic situation that produces political pamphlets is taken into account. Neither discourse, nor the registers they are instances of, can be pre dicted to have the initiation-response turn pattern, involving elliptical and minor clauses, of the following dialogic excerpt from a written narrative of a joke: 'How much for that bird in the window?' 'Seven hundred dollars.' 'Seven hundred dollars for a bird!' 'Sure' said the shopkeeper. 'That bird speaks seven languages.' 'Seven languages, eh. That's just the thing. I'll buy it.'
The medium relationship in the situation predicts the selection of options in
GENERIC SITUATION AND REGISTER
311
the textual component of the language code and MODE OF DISCOURSE. All media of communication have their organizing resources, resources which allow the communicator to structure the 'message', that is, to relate its parts to each other, and indicate importances. In language, this involves the relationships Halliday and Hasan (1976, 1985) have discussed under the heading COHESION. Reference, ellipsis, substitution, conjunction, and lexical relations, are the major resources which create TEXTURE in Eng lish discourse. Semologicai FOCUS and PROMINENCE, realized in the clause by way of Theme and Rheme and in the phonological tone group by Given and New, are the crucial resources which permit the user to start message units and indicate their highlights. When language is the medium of communication, there is the choice of the sub-mediums, speaking or writing. In 'Aspects of Varieties Differentia tion' (1967) and Language and Situation (Gregory and Carroll 1978) I pre sented more delicate distinctions of medium relationship: SPEAKING SPONTANEOUSLY (MONOLOGUING OR CONVERSING), SPEAK ING NON-SPONTANEOUSLY (RECITING OR SPEAKING WHAT IS WRITTEN), WRITING TO BE SPOKEN AS IF NOT WRITTEN, WRITING TO BE SPOKEN, WRITING TO BE READ. Halliday in a recent article 'Spoken and written modes of meaning' (forthcoming) has eloquently indicated the distinctiveness of the modes: About twenty five years ago . . . I began observing natural spontaneous discourse in English and from the start I was struck by a curious fact. Not only were people unconscious of what they themselves were saying; they would often deny not just that they had said something I had observed them to say, but also that they ever could say it. For example, Ï noticed the utterance It'll've been going to'v e been being tested everyday for the past fortnight soon . . . Other things I noted regularly included present in pre sent participial non-finites like being cooking in I never heard you come in — it must have been with being cooking', marked thematic elements with reprise pronoun, as in that poor child I couldn't get him our. of my mind; and relatives reaching into dependent clauses such as that's the noise which when you say it to a horse the horse goes faster. These are all systematic fea tures that people are unaware that they incorporate in their speech, and often deny have said even when they are pointed out; or at least reject as unsystematic—after 'I didn't say it', the next line of defence is 'well, it was a mistake'. But of course it was not a mistake: it was a regular product of the system of spoken English.
Halliday notes in the same paper that written language tends to have a much higher lexical density than spoken, and at the same time a much sim-
312
MICHAEL GREGORY
pler sentence structure. 'Spoken and written language do differ', he points out, 'in their preferred patterns of lexicogrammatical organization. Neither is more organized than the other but they are organized in different ways'. So far I have exemplified the difference between registers, ideationally and interpersonally, from snippets of text which were written. But the pre vious snippets are very different in their selections from the textual function of English from this transcription of a dialogue at a pumpkin stall during a Canadian 'fall fair': A: B: A: B: :
Going to buy one? Don't know. Perhaps. Better hurry, they're packing up. Oh, all right. (to ): That one over there, please, the big one.
The following ellipses can be recognized: 1. 2. 4. 5. 6.
Are you from Going to buy one?; I and 3. if I'm going to buy one from Don't know; I m going to buy one from Perhaps; You had from better hurry; I'll hurry from Oh, all right.
Only three of these six ( # 3 , 4, 6) are elliptical in Halliday and Hasan's cohesion sense of the term in which the ellipsis is 'endophorically' retrieva ble (i.e. from within the text). The other ellipses ( # 1 , 2, 5), and the repeated nominal substitution item one are 'exophorically' retrieved by way of the instantial situation and the participants' knowledge of it and so are COHERENT if not COHESIVE in the discourse (cf. Malcolm 1985b and forthcoming). These ellipses also mean that the syntactic Themes, the start ing points of the independent clauses, are in three instances {Going . . . Dont know . . . Better hurry) differently realized from what would be pre dictable in other modes which are not determined by a medium relationship which is that of spontaneous conversation in a face-to-face situation. It could also be argued that some of the gnostological, if not the semological, prominences are not syntactically realized but retrieved from the ellipses; that is to say, if the propositions were realized by full indepen dent clauses, the phonological New would fall on what in this text are the ellipted elements. However, none of this means that this discourse is in any way ill-formed as an instantiation of message exchange by way of language; it functioned as communication with an economic adequacy. Its mode of
figure 2
GENERIC SITUATION AND REGISTER
313
314
MICHAEL GREGORY
discourse was appropriate and predictable in terms of the medium relation ship of its generic situation, particularly when the experience and interac tion relationships are also taken into account. Its registerial selections, in all respects, reflect the communicative function of its situation as an instance of a generic situation. Figure 2 summarizes the relationships that I have been discussing amongst instantial situation, generic situation, register, and the tri-functionally arranged semological and syntactic systems, and a brief indication is given of phonological systems which are directly relevant in semological realization. All systems refer to the codal cycle for English. I would hypothesize that any language (be it Catalan, French, Spanish, or Chinese) has systems which could be regarded as analogous or parallel. All languages have to cope with experience, interaction, and medium organization. They do so in their own way. Appendix 2 gives a simplified display of the major English systems referred to in the analyses.
4
It is appropriate for me now to comment briefly on what important changes I have made from my previous writings on these matters (cf. Enkvist, Spencer and Gregory 1964, Gregory 1967, Gregory and Carroll 1978, Gregory 1980). I no longer make the distinction between personal and functional tenor which used to be seen as answering linguistically to dis tinct personal and functional addressee relationships in the generic situa tion. Influenced by J. R. Martin (1980/81 and private communication) and by the task of reconciling prediction and description, it now seems to me that such a distinction, if it needed to be made at all, was at an inapprop riately generalized degree of delicacy; and that it gave too limited a view of communicative function by confining it to the interactive relationship. Working from the direction of text variety, I used to distinguish functional tenors (expository, didactic, persuasive, and so on) from personal tenors (formal, neutral, informal). However, when one works on the interpersonal systems of the code it appears that systems of Speech Function, Turn, Social Distance, Interlocutor Assessment, Mediation, and Attitude are best seen as simultaneous entries in the interpersonal network of options avail able at the semological rank of proposition, and so there is no strong reason for giving a primary importance to those parts of the network that involve
GENERIC SITUATION AND REGISTER
315
the matter of formality. (It would be like saying language has four metafunctions: the ideational, the interpersonal, the textual, and the polite!) Furthermore, the theory is stronger if the different dimensions of the generic situation and the different functional components of the code are directly related; they can be seen as mutually defining until such a time as a system, or part of a system network, is found to be totally incongruent. Such has not been the case. Influences between one dimension and another, one functional component and another, can best be dealt with when the general 'geography' of each has been made explicit. This means that there is no place for a functional tenor, with multi-functional realiza tion, as a dimension of variety on a par with field, mode, and personal tenor, all of which had a corresponding functional realization: ideational, textual, and interpersonal, respectively. (See Gregory 1980 where I argue the opposite position with more vigour than acuity because Î was, in fact, only approaching the matter from the point of view of instantial text vari ety.) I now see the function of the generic situation as being constituted by the interrelationship of experience, interaction, and medium; and, as was pointed out earlier, it is this complex COMMUNICATIVE FUNCTION which, when language is the channel of communication, corresponds to REGISTER. I am still therefore in agreement with Halliday (1975: 26) that register can be seen as 'a configuration of the semantic resources that the member of a culture typically associates with a situation type. It is the meaning potential that is deployed in a given social context. It thus repre sents the semantic options and configurations of semantic options that are at risk under particular environmental conditions'. I postulate generic situa tion as well as register in order to organize what can be generalized about 'particular environmental conditions' whatever channel of communication may be selected by people to exchange their messages, their knowledge as a function of living, as social fact. Writing about language, Halliday (1974) has said 'there is not a great deal one can predict about the language that will be used if one only knows the field . . . or only the mode or the tenor. But if we know all three we can predict quite a lot'. I suggest that a fully developed socio-cognitive linguis tics has to recognize that there is not a great deal one can predict about the knowledge that is at risk between people in a culture if one knows only the experience they are relating to, or only the interactive relationship they have, or only their medium of communication; but if we know all of these we can predict quite a lot. It is as we move to do so that we might begin to
316
MICHAEL GREGORY
justify Claude Levi-Strauss' frequently voiced assertion that linguistics is not only the exactest of the social sciences but also the central one for the understanding of human cultures and the human personality within a cul ture.
5 Prediction, of course, entails description and description, prediction; and neither, in my view of scientific endeavour, is justifiable if it is con ducted without a concern for utility. It would be remiss of me, if I failed to assert the need for the continued description of the language, and to assert that such description should be carried on with theoretical consistency as regards intra-lingual variety, and with an eye on usefulness. No theory without praxis; no praxis without theory. This paper has been largely theoretical, but let me now recall the eight major uses of the distinction and description of language variety which I proposed in 1967 and developed in Gregory and Carroll 1978. The order in which I present them is not a mat ter of priority; priorities are for individuals to decide. The first use is in the language teacher's choice of materials. In this respect I wish to emphasize that hospitality to the range of registers within a language, and a constant concern for appropriateness as regards the lear ners' needs are of the essence if the materials are to maintain relevance and the language is to be learned as a living language. The second use, in the making of rhetorics, is related to the first. There can be nothing but health in prescriptions for how to use language well and effectively if they are based on sound and thorough descriptions of the appropriate linguistic and situational conditions of the type of discourse in question. Such a prescription is what I mean by a 'rhetoric'. This, in turn, is linked to the third use: in the judgement of correctness. or as I prefer, appropriateness of usage. Sadly, there seems to be a general human compulsion to pass judgement on other people's use of language. Many of the traditional criteria for correctness are based upon such untena ble criteria as ideals of etymological purity, and grammatical unchangeability, upon confusions between logic and language. They all frequently involve social judgements masquerading as linguistic judgements. Register theory and description provides a framework for the rational assessment of degrees of appropriateness of usage in terms of instantial and generic situa-
GENERIC SITUATION AND REGISTER
317
tion. This is not to take a permissive view of language, a view that anything will do; it is to take a productive view; it aims to extend people's awareness and appreciation of the developing range of varying language habits sum med up by the word 'Catalan', for example. The fourth use: in the study of style, especially in the establishment of relevant norms for comparison and contrast, is, I think, self-evident and does not need expansion in this paper. I would like, however, to emphasize the fifth use: in the establishment of translation equivalence. Thanks to the work of scholars such as Eugene Nida (1964),'Charles Taber (Nida and Taber 1969), and J . Catford (1965), there has been a growing awareness that translation is not just a matter of item to item equivalence, or, indeed, of sentence to sentence equivalence; rather it is a matter of discourse to discourse equivalence and this involves dialect and register considerations. The establishment of regis ter to register equivalence can be seen as a major factor in the process of translation: 'the problem of establishing such equivalence, a crucial test of the limits of translatability' (Gregory 1980: 466). This is one of the factors that reinforce the importance of the sixth use of register description: in the collection and assessment of a representative corpus for descriptive statements about a language. Many grammars of the major world languages have, until recently, been based on and have adequately described only a portion of what the language is, usually the reasonably formal and written, with a bias towards the literary. What are needed are grammars and lexicons reflective of the many registers, the many fields, modes, and tenors of a language. The framework discussed in this paper can be of use in ensuring the comprehensive nature of a collected corpus intended to be genuinely representative of a language, and of use in the selection of register specific corpora; and for different purposes we need both kinds of corpus. Work on the basis of such corpora can facilitate the seventh use: in making more sophisticated the institutional labelling of items in dictionaries. Definitions and citations for dictionary entries can be linked to more deli cate statements of their registerial provenance as a guide to their appropri ate usage. The eighth use, in statements of meaning in regard to specific texts, in other words, in a semologically sensitive discourse analysis, is too large and complex a matter for elaboration here. Suffice it to say that, together with my colleagues Elissa Asp, Janet Dill, Karen Malcolm, David Watt, and Lynne Young, I have been developing from register theory, the concept of
318
MICHAEL GREGORY
discourse PHASE in order to facilitate description of the dynamic process of discourse as the linguistic exchange of message. Phase is used to charac terize those stretches of discourse in which there is a significant measure of consistency and congruity in what is being selected from the three metafunctional resources of the language. What are termed TRANSI TIONS in and out of phases are indicated by a marked change in choice from one or more functional components or a backwards' and 'forwards' functional relevance. Because of its tri-functional basis, phase can be thought of as a very delicate statement of register realization. Phase and transition are used to capture the dynamic instantiation of micro- registerial choices in a particular discourse (cf. Asp 1983, Dill 1985, Gregory and Mal colm 1981, Gregory 1982b, 1984, 1985a, 1985b, forthcoming; Malcolm 1983, 1985a, 1985b, forthcoming; Watt 1985; Young 1984). It is important, in this matter of language variety according to use, to recognize the unique ness of any particular discourse at the same time as we see it in relation to generalized potentials of language behaviour and code at varying degrees of delicacy. Only so can we have a fruitful and insightful renewal of connec tion between phenomena and abstraction. Such a constant renewal of con nection is necessary for a theory to be descriptively revealing and of use. Such a theory I have attempted to sketch for you. It is concerned with the instantial situations of our 'living'; it is concerned also with our 'know ing', our building up of gnostologies which include knowledge of context and knowledge of communicative codes such as languages; it is concerned with a language as a kind of 'doing', a reaiizatory code cycle of meanings (the semology of a language), of wordings of meanings (its morphosyntax), and of their soundings and signings (its phonology and graphology). These soundings or signings in instantial situations lead to the physical manifesta tion of discourses which we perceive as we receive from, and give to each other, messages. Living, knowing, doing, meaning, wording, sounding, signing, and perceiving: all are relevant when we address the question of a language's homogeneity and heterogeneity, the question of variety within a language. And this is so whether that language be English, Chinese, or Catalan.
NOTE This is a revised version of a paper prepared for the II International Congress of the Cata lan Language, Social Linguistics section, Palma de Mallorca, May 1986.
GENERIC SITUATION AND REGISTER
319
REFERENCES Asp, Elissa. 1983. The dynamics of discourse: Dylan Thomas' Memories of Christmas. Honours thesis, Glendon College of York University, Toronto. . 1984. Meaning and metaphor: a revised theory for stylistics exemplified. Paper given to the 11th International Systemics Workshop, University of Stirling, Scotland. . 1985. The communicative functions of metaphor: a cognitive, cultur al, and lińguistic catalysis. M.A. thesis, York University, Toronto. Catford, J.C. 1965. A linguistic theory of translation: an essay in applied linguistics. London: Oxford University Press. Dill, Janet. 1985. The space-time novel as a message event: Pursewarden's Suicide. M.A. thesis. York University, Toronto. Enkvist, Nils Erik, John Spencer and Michael Gregory. 1964. Linguistics and style. London: Oxford University Press. Gardiner, A.H. 1932. The theory of speech and language. Oxford: Claren don Press. Gregory, Michael. 1967. Aspects of varieties differentiation. Journal of Linguistics 3:177-198, . 1974. A theory for stylistics exemplified: Donne's Holy Sonnet XL Language and Style 8,2:108- 118. . 1978. Marvell's To His Coy Mistress: the poem as a linguistic and social event. Poetics 7,4:351- 362. . 1980. Perspectives on translation from the Firthian tradition. Meta 25,4:455-466. . 1982a. The nature and use of metafunctions in systemic theory: cur rent concerns. The Eighth Lacus Forum 1981. Columbia, S.C.: Hornbeam Press. . 1982b. Hamlet's voice: aspects of text formation and cohesion in a soliloquy. Forum 198 Linguisticum 7,2:107-122. . 1984. Propositional and predicational analysis in discourse descrip tion. The Tenth Lacus Forum 1983. Columbia, S.C.: Hornbeam Press. . 1985a. Discourse as the instantiation of message exchange. The Eleventh Lacus Forum 1984. Columbia, S.C.: Hornbeam Press. . 1985b. Towards communication linguistics: a framework. In J. Ben son and W. Greaves, eds., Systemic perspectives on discourse. Vol 1: Selected theoretical papers from the 9th International Systemic Work shop. Norwood, N.J.: Ablex.
320
MICHAEL GREGORY
___ (Forthcoming). Phasal analysis within communication linguistics: two contrastive discourses. In J. Copeland, ed., Proceedings of the Sec ond Rice University Symposium on Linguistics and Semantics: text semantics and discourse semantics. Chicago: University of Chicago Press. . 1982-. Notes on communication linguistics. With the assistance of K. Malcolm, E. Asp, D. Watt, and J. Dill. Mimeograph. Six chapters com pleted. A.L.R.W.G. Glendon College of York University, Toronto. Gregory, Michael and Susanne Carroll. 1978. Language and situation: lan guage varieties and their social contexts. London: Routledge and Kegan Paul. Gregory, Michael and Karen Malcolm. 1981. Generic situation and dis course phase: an approach to the analysis of children's talk. Mimeo. A.L.R.W.G. Glendon College of York University, Toronto. Halliday, M.A.K. 1973. Explorations in the functions of language. London: Arnold. . 1974. Language and social man. (Schools Council Programme in Linguistics and English Teaching Papers, 11,3). London: Longman. . 1975. Language as social semiotic: towards a general sociolinguistic Theory. The First Lacus Forum 1974. Columbia, S.C.: Hornbeam Press. . 1978. Language as social semiotic: the social interpretation of mean ing. London: Arnold. . 1979. Modes of meaning and modes of expression: types of gram matical structure and their determination by different semantic func tions. In D.J. Allerton, ed., Function and context in linguistic analysis. London: Cambridge University Press. . (Forthcoming). Spoken and written modes of meaning. In R. Horowitz and S. Samuels, eds., Comprehending oral and written lan guage. Academic Press. Halliday, M.A.K. and R. Hasan. 1976. Cohesion in English. London: Longman. . 1985. Language, context and text: a social semiotic perspective. Deakin University Press (Language and Learning Series). Lamb, Sydney M. 1971. The crooked path of progress in cognitive linguis tics. Monograph Series in Languages and Linguistics. 24:94-123. Malcolm, Karen. 1983. The paragraph: what use is it? some traditional and contemporary proposals. The Ninth Lacus Forum 1982. Columbia, S.C.: Hornbeam Press. . 1984. Different approaches to the description of casual conversa-
GENERIC SITUATION AND REGISTER
321
tion. The Tenth Lacus Forum 1983. Columbia, S.C.: Hornbeam Press. ___ 1985a. Communication linguistics: a sample analysis. In J. Benson and W. Greaves, eds., Systemic perspectives on discourse. Vol 2: Selected applied papers from the 9th International Systemic Workshop. Norwood, N.J.: Ablex. . 1985b. The dynamics of casual conversation from the perspective of communication linguistics. Ph.D. Dissertation. York University, Toronto. ___ (Forthcoming). Casual conversation: a message focussed register. In J. Benson and W. Greaves, eds., Systemic functional approaches to dis course: selected papers from the 12th International Systemic Workshop. Norwood, N.J.: Ablex ___ (Forthcoming). Situational and gnostological exophora. The Eleventh Lacus Forum 1984. Columbia, S.C.: Hornbeam Press. Mao Zedong. 1975. On practice. Selected Works. Eng. ed. Vol. 1. Beijing: Foreign Languages Press. Martin, J.R. 1980/81. Register and metafunction. Department of Linguis tics, University of Sydney. Revision of paper given to the Seventh Inter national Systemics Workshop, Sheffield. Nida, E.A. 1964. Towards a science of translating with special reference to principles and procedures involved in Bible translation. Leiden. Nida, E.A. and C.R. Taber. 1969. The theory and practice of translation. Leiden. Watt, David. 1985. A communication linguistics perspective on the analysis of interlanguage via focus and prominence. M.A. thesis. York Univer sity, Toronto. Young, Lynne. 1984. Discourse analysis within the framework of communi cation linguistics. Paper given to the 18th Annual TESOL Conference, Houston.
APPENDIX I TEXTS SPAGHETTI BOLOGNESE INGREDIENTS 1 lb. lean ground beef 1/4 lb. salt pork or green cured bacon 1 4 oz. tin tomato purée
322
MICHAEL GREGORY
1 8 oz. tin Italian style tomatoes 1 tsp. oregano 1 tsp. basil 1/4 tsp. nutmeg 1 bay leaf 1 tbs. olive oil 1 tbs. butter 1 cup dry red wine pinch of salt black pepper chopped fresh parsley 1 lb. spaghetti METHOD Melt the butter in the olive oil in a heavy saucepan on top of the stove. Finely chop the onion, garlic cloves, salt pork or bacon and sauté in the oil and melted butter until soft but not brown. Add half the wine and turn the heat up until the wine evaporates and glazes the friture. Then add the beef and cook until there is no sign of blood. Add the tomatoes and tomato pune and bring to the boil. Then turn the heat down and allow the mixture to simmer. Sprinkle the oregano, basil, nutmeg, salt and pepper on top, place the bay leaf in the mixture and gently stir, adding the remainder of the wine. Cover the pan and adjust the heat so it simmers gently. In a large pan bring two litres of water to a fast boil and add the spaghetti slowly so the water continues to boil. Cook in the open pan for 10 minutes (al dente), then drain the spaghetti and return to the pan, adding the butter. Gently turn the pasta over. The sauce will now be ready. SERVICE Arrange the spaghetti in a large dish. Cover with the meat sauce and sprinkle with Parmesan and chopped parsley. Serve with a light dry Italian wine such as Valpolicello.
I don't believe that endorsing the Nuclear Freeze Initiative is the right step for California Common Cause. Tempting as it may be, we shouldn't embrace every popular issue that comes along. When we do so we use precious, limited resources where other players with superior resources are already doing an adequate job. Rather, I think we will be stronger and more effective if we stick to those issues of governmental structure and process, broadly defined, that have formed the core of our agenda for years. Open government, campaign finance reform, and fighting the influence of special interests and big money, these are our kinds of issues. Let's be clear: I personally favor the initiative and ardently support disarmament negotiations to reduce the risk of war. But I don't think endorsing a specific nuclear freeze proposal is appropriate for California Common Cause. We should limit our involvement in defense and weaponry to matters of process, such as exposing the
GENERIC SITUATION AND REGISTER
323
weapons industry's influence on the political process. Therefore, I urge you to vote against a CCC endorsement of the nuclear freeze initiative.
There was this young man who wanted to buy a present for an Aunt who had every thing. He looked and looked until at last he found a bird in a pet store. He asked the shop-keeper, 'How much for that bird in the window?' '700 dollars.' '700 dollars for a bird!' 'Sure' said the shopkeeper, 'That bird speaks seven languages.' 'Seven languages eh. That's just the thing. I'll buy it.' So the young man bought the bird and sent it off to his Aunt. About a week later he 'phoned her and asked, 'So how do you like the bird?' And she said, 'Oh it was delicious!' 'You ate it! That was a 700 dollar bird! It spoke seven languages!' And she said, 'Well for heaven's sake! Why didn't it say something?'
1: 2: 3: 4: 5:
Going to buy one? Don't know. Perhaps. Better hurry, they're packing up. Oh, all right. That one over there, please, the big one.
A P P E N D I X II Major semological and syntactic systems referred to in the discussion of English dis courses: simplified networks. Order: 1) Process type 2) Transitivity 3) Speech function etc. 4) Mood 5) Theme 6) Information system.
324
MICHAEL GREGORY
SYSTEM: PROCESS TYPE LOCI: METAFUNCTION: IDEATIONAL STRATUM: SEMOLOGICAL RANK: PROPOSITON/PREDICATION
SYSTEM: TRANSITIVITY LOCI: METAFUNCTION: IDEATIONAL STRATUM: MORPHOSYNTAX RANK: CLAUSE
(Related in realization to the system of predicational roles and process types in the semology.)
GENERIC SITUATION AND REGISTER 325
326
MICHAEL GREGORY
SYSTEM: SPEECH FUNCTION AND RELATED SYSTEMS LOCI: METAFUNCTION INTERPERSONAL STRATUM: SEMOLOGY RANK: POSITION
Realizations in the morphosyntax are complex and in volve status and mood sys tems in the clause, and mo dality in the verbal group.
SYSTEM: MOOD LOCI: METAFUNCTION: INTERPERSONAL STRATUM: MORPHOSYNTAX RANK: CLAUSE
GENERIC SITUATION AND REGISTER 327
(realization oí focus and in marked cases, focus and prominence in semological propositions)
SYSTEM: THEME LOCI : METAFUNCTION : TEXTUAL STRATUM: MORPHOSYNTAX RANK: CLAUSE
328 MICHAEL GREGORY
GENERIC SITUATION AND REGISTER
SYSTEM: INFORMATION LOCI: METAFUNCTION: TEXTUAL STRATUM: PHONOLOGY RANK: TONE GROUP (tonic syllable: bearer of primary stress at significant point in intonation pitch contour.) (New: information considered by speaker not to be retrievable from dis course or situation Given: considered retrievable
329
WHAT DID MILTON SAY BELIAL SAID AND WHY DON'T THE CRITICS BELIEVE HIM?
Terry Threadgold University of Sydney Sydney, Australia 1
The word is not a thing, but rather the eternally mobile, eternally changing medium of dialogical intercourse. It never coincides with a single con sciousness or a single voice. The life of the word is in its transferral from one mouth to another, one context to another, one social collective to another, one generation to another. In the process the word does not forget where it has been and can never wholly free itself from the domin ion of the contexts of which it has been part. (Bakhtin 1973: 167)
There are many questions I want to ask, and attempt to answer in this paper. In order to begin somewhere, I can perhaps start with the one that in some sense prompted all the others. The question I want to address through Michael Halliday's notion of language as social semiotic is an old one in critical circles. It is this: how can it be that a text in which an omnis cient narrator's voice says very clearly, 'this voice, Belial's voice, which you will now hear, is the voice of Satan, the voice of evil, of a liar', can be read as saying something quite 'other', if not opposite. And indeed we have nearly three hundred years of critical debate to prove that this can be done. I quote just two examples here, Shelley, who said in 1821, 'It is a mistake to suppose he [i.e. Belial] could ever have been intended for the popular personification of evil', and Blake, 1793, 'The reason Milton wrote in fet ters when he wrote of angels and God, and at liberty when of devils and hell, is because he was a true poet of the devil's party without knowing it' (see Appendix 2). Let me point out, however, that what I have told you the narrator's voice says 'very clearly' is already a paraphrase, a citation, a re-contextualization of the voice (but not the voice, the written word) of an 'other', a re-
332
TERRY THREADGOLD
semanticisation, a meaning mapped onto my voice, my words, as a projec tion (Halliday 1985). That process of projection of the words, the meanings of others includes the voices/words of Shelley and Blake: which is why the apparently monologic is always dialogic, and therefore why Bakhtin (1981) was wrong in regarding poetry as monologic (see Thompson 1984 for the argument that he actually saw poetry too as heteroglossic). It is also why Bakhtin's word/voice is here re-contextualised dialogically in support of an argument he would not have espoused (that poetry, even classical poetry, is dialogic) and why the primary focus of this paper will be the attempt to eluci date (vision, light, clarity, shades of the seventeenth century and Foucault's episteme, other 'voices', discourses, contexts mapped onto my discourse) the question of dialogism. Dialogism involves precisely the kinds of issues that are raised by the systemic-functional stylistic and semiotic analysis of the Miltonic textual problem referred to above. The 'text' that initiated this dialogism is from Book II of Milton's Paradise Lost, 11. 43-228. The semiotic analysis will be discussed in detail below. I mark text above with quotation marks because what I am here looking at is inevitably and irretrievably marked as a metonymy (a part of a debate which is a part of a whole epic), but is also a whole which has its distinct parts, two speeches and the frames with which the narrative voice introduces (and dismisses), projects and constructs, the speakers, and their speaking/written voices. This is where the dialogism begins. For it is the multiplicity of ways in which the 'word', the lexico-grammatical level of language, and the patterns at that level, can be re-co- textualised and con-textualised to produce new or altered meanings or polysemy that I take Bakhtin (1973: 167) to be talk ing about as dialogism: and these are not issues which our semantic and semiotic theories can yet cope with or explain. This paper has involved try ing to establish, for this textual problem, what dialogism might mean, what are its linguistic and semiotic parameters in this context, and what are the implications for linguistic and semiotic theories of this semantics and semio tics of dialogism. One cannot however talk about dialogism in Bakhtin's (1973) sense without also talking about genre and heteroglossia (Bakhtin 1981), some or all of which Kristeva 1980 and others have renamed intertextuality: and to talk of genre in a systematic-functional context and in the context of this Milton text invokes a further series of questions in the light of current debate within and without systemic circles. What is genre (Weber 1980;
WHAT DID MILTON SAY BELIAL SAID
333
Rosmarin 1985; Thompson 1984), what is the relationship between genre and literature (Hasan 1979; Martin forthcoming), is genre enough to account for what happens in texts (Martin forthcoming, 1986; Bakhtin 1973; Thompson 1984), what is the relationship between genre and foreg rounding (Martin forthcoming; Hasan 1985), and finally, what is foreg rounding, and what is foregvound-edi-able for whom? There is not space here to discuss all of these issues theoretically, although they need to be so discussed. The answers will have to emerge from my textual practice. However, the problems raised by Bakhtin's (1973, 1981) discussion of dialogism, genre and heteroglossia implicate all of the questions involved in the detailed systemic functional analysis of the literary text as artefact/object/product and as a dynamic process through which - textual and con-textual meanings are realised as coding orienta tions (Bernstein 1971, 1982), genres, heteroglossia (Bakhtin 1973, 1981) and intertextual semantic frames (Thibault 1986) or codes (Barthes 1974; Kristeva 1980). The attempt to deal with these issues sheds light on the constraints exerted by generic norms (and the nature of those 'norms'), on the making of meanings, and thus on the question of what meanings can and cannot be made at a given historical conjuncture. It focuses on the material, linguistic/ semantic nature of texts and contexts as system and process. It enables the exploration of the making and re-making of genres, the semantic system and the social semiotic in and through text as dynamic intertextual process; and it suggests ways of analysing the re- structuring and re-constitution of genres and intertextual semantic frames like them through the text/context dialectic in and across intertextual relationships and time. It will be my argument that the consistent semantic patternings and lexico-grammatical structures that are accessible to analysis in the Milton text are generically predictable; that this is a text which is constrained by not one but many intertextual generic systems, and that this is the source of much of the text's polysemy and ambiguity. This is how I will begin to define heteroglossia: the tendency for the text to speak in many voices, or, to avoid the mimetic metaphor, the tendency for the one lexico-grammati cal wording to have mapped onto it many different and often contradictory discourses. Out of this complexity of generic types/systems (which only ever exist as processes, semantic 'chunks'/frames in other texts, and are thus transmit ted from text to text (or discourse to discourse), not from system to text as
334
TERRY THREADGOLD
our theories would have us believe) a new generic type, the Biblical epic, is constructed in this Milton text as token/process. This then becomes the intertextual semantic frame or type for the production/performance of new tokens — other similar epics or the reproduction of the 'Miltonic style' and so on. The point about types is that they are never exactly replicable (Eco 1976). Every time a type is re-made as a textual/discursive token, it is sub ject to not only re-co-textualisation at the lowest lexico-grammatical or syntagmatic level, but also to --textualisation at the global or highest semiotic levels. And at both levels - and multi-variate structures (Halliday 1985) seem to operate. That is, there are co-variate processes which seem to operate across texts as sequences (for example, what happens when different generic frames follow one another — narrative frame, speech, narrative frame, speech, narrative frame), for which Halliday and Hasan's (1976) account of cohesion can serve as a model; and there are multi-var iate, simultaneous and recursive processes, involving embeddings of, for example, one generic frame within another, or the simultaneous mapping of several generic frames onto the same lexico-grammatical structure. These kinds of processes produce an extremely volatile and unstable semantic/semiotic configuration and produce many 'new' meanings, in the sense of 'not previously coded by any system' (Eco 1976): but these proces ses are not, I believe, 'unpredictable' (Martin forthcoming) in terms of the kind of view of - and con- textualisation and generic type/token relations that I have outlined above. I take this to be something like what Halliday (1982) is suggesting when he speaks of choices at the lowest lexico-grammat ical levels being 'partially freed' from the 'control of the semantics'. Rather than using this metaphor of freedom and control, I would prefer to say with Thibault (1986) that this is actually where the syntagm-paradigm distinction breaks down, and the two operate simultaneously but independently of one another. That is, it is not either/or, but both/and. The semantics seems to me to be still in place, but its values must change/slip/slide, as soon as what I am calling co-textualisation begins to operate, and this will always happen as soon as the systemic (which is only ever a myth of our own construction) (Silverstein 1979) is realised as text and process. This is of course what the Halliday/Hasan (1976) account of cohesion was always saying, voicing the partial independence of the syntagmatic from the paradigmatic. Thibault's related discussion (1984b) of activity at the lowest lexicogzrammatical levels 'bypassing' the socio-semantics raises I think a different
WHAT DID MILTON SAY BELIAL SAID
335
set of problems. Is 'genre* ail there is? Is it enough to theorise the 'higher order social semiotic' (Halliday 1978) as genre and register? Martin's own work, in which he now includes ideology as the third of three connotative semiotics, where ideology is everything not covered by genre and register (Martin 1986) suggests that it is not. Leaving aside the problems of separat ing ideology from genre or the other levels, register and text (genre is an ideological frame of extraordinary power), it is clear that much more hap pens in texts than genre, even allowing mixed genres, will account for: and this brings me back to heteroglossia and is why Thibault sees a need to explain what happens in lexico-grammar as not only independent of but also 'by-passing' the socio-semantics, and in some sense being re-con-textualised at a more global level directly from the higher order social semiotic of discursive formations (Halliday 1978; Foucault 1972) and from the heterogiossia of the many 'voices' (Bernstein 1982) of the centres and peripheries of the culture. Such global re-contextualisations would involve the mapping onto the same lexico-grammar, and the reconstituting at that level (co-textualisation), of the intertextual semantic frames of epistemic (Foucault 1972, 1973) and cryptotypical 'fashions of speaking', or habitual modes of communication (Whorf 1956; Hasan 1984) and of social heterog iossia. These, because they tend to pervade innumerable, otherwise dis tinct, 'ways of speaking' at many different social and contextual levels are apparently 'trans-generic'. It would take more than another paper to attempt to sort these out. Perhaps, in the end, we will be able to do no better than Eco (1976) and call them all codes or sememes and represent the ineffable conjunctions of these as the encyclopaedia of his model Q. Suffice it to say that all of them need to be thought of in relation to speaking subjects with particular coding orientations (Bernstein 1971) who are constituted in and through language/ discourse, and socially/historically positioned at the intersection of a specific conjunction of discourses. Thus it is, that without resorting to biog raphical or historical criticism, and without denying the 'autonomy' of the artistic text with respect to the immediate context of utterance (Hasan 1985), we cannot afford to neglect the position of the writing subject and the immediate and more global context in which the text is produced if we want to construct probabilistic or predictive models of the factors which constrain and contextualise choices at the lexico-grammatical level. There always is a writer and an immediate context of utterance, verbal art or not, and these factors are never immaterial to the semantics of the text.
336
TERRY THREADGOLD
Thus I would suggest that much more in 'artistic' texts is 'predictable' than we usually allow or can see, given the ideologies of autonomy and indi vidual creativity which constrain our perceptions of art, if only we work with a heteroglossic and not normative/monologic notion of what consti tutes the semiotic construct of context, and if we are prepared to theorise that construct not in narrow linguistic terms but in terms of a broadly based social semiotic theory. Jay Lemke's notion of context as dynamic open sys tem is helpful here (1985a,b). The Milton text is characterised by multiple generic embedding involv ing the simultaneous mapping onto single lexico- grammatical wordings of several, and often conflicting, generic frames. Generically predictable parody, irony, politeness and indirectness produces an incongruent (Halliday 1985) semantics. Conflicting rhetorical and heteroglossic discourses produce nominalisation, grammatical metaphor, and multiple levels of overcoding and coding conflict. There is a textual polemic between the frames and speeches in the text in which the norms of dialogue (as genre) are reconstructed and recontextualised in many ways. First, this is a parody of a debate — a debate about God's will is generically impossible; the debate is a performance of the intertextual semantic frame of sophistry/sin/ wrong reason and involves the semiotics of deceit; the simple exchange of information (Halliday 1985) becomes an exercise in the re-structuring, resemanticising, and co-textualising, of the voice/voices of the other, so that the meanings of the 'other' are mapped onto the wordings of the speaking T' and involve the projection of the voice/meanings of the 'other' as word ings in indirect speech. In the process whole semantic fields are restruc tured (in generically predictable ways in terms of rhetorical purpose) so as to produce new, generically and systemically, uncoded meanings. All of this produces a volatile and unstable textual semantics. This is instantiated in the patterns of the lexico- grammar and realised in the text, co-text, con-text, contra-text (Martin 1986) dialectic in ways which allow us to witness/participate in the re-structuring of and construction of the seman tic system, generic frames and the social semiotic. At the same time the dialogic interaction between text and reader and between text and other historical and ideological frames (semiotic systems) reproduces the textual semantics of deceit at the level of reading practices and allows us to witness/ participate in the further restructuring and reconstruction of generic frames. The reader, now reading mimetically rather than rhetorically and generically (Colie 1972), identifies with the 'voice' of the deceiver and is
WHAT DID MILTON SAY BELIAL SAID
337
deceived. Satan becomes the hero, and the Biblical and classical epic genres are reconstituted as a new secular epic in a humanist and individualistic world. But the humanist and individualistic discourses, the conflicting codes of social heteroglossia, were always there, 'lurking between the lines, the future of semiotic systems' (Martin 1985a), simply invisible in a world where Hell was still a reality and the Bible truth not fiction. There is a certain inevitable semantic slippage in the above between the account of a seventeenth century production of the text by a Milton and the twentieth century reproduction of the seventeenth century production by a female semiotician. In the end or in the beginning the production of a text is always a reproduction and a new dialogic encounter. But the making 'visible' of the nature of this encounter questions another of our reflectionist and visually based ideologies. Our prevailing notions of foregrounding (Halliday 1982; Martin forthcoming; Hasan 1985) have to be rethought. As they stand, whether foregrounding is defined as consistent and recurrent patterns which are unpredictable in terms of the linguistic system (Halliday 1982), or as patterns or even single instances which contrast with the norms of the text as system (and are not therefore necessarily defined by frequency (Hasan 1985; Mukarovsky 1964b)), our notions of foregrounding are grounded absolutely in the text. This, of course, serves to maintain the visual epistemology — a view of the text as product, and of the reader as passive observer. If instead we perform the text as process, we have to ask foregrounded/-able for whom and why? And if we allow the kind of view of text as dialogic process outlined above, then we will have to admit that consistent semantic/lexico-grammatical patterns are neither unpredictable nor peculiar to the texts of verbal art and that the question of foreground and background, visible and invisible, is relative, not absolute, and relevant only and always in terms of questions like: which reader? which writer? what reading or writing practice? what coding orien tation and what specific social semiotic contextual construct? It is very clear that a humanist, individualist, mimetic reading practice does not foreground the same patterns as a rhetorical, generic reading prac tice: and very clear to me, as female twentieth century semiotician, that once I had begun to read the text in both these ways, and in several others, I could no longer see any of the patterns my systemic-functional analysis produced as being fore- or back-grounded. They were rather a set of pat terns, all of which in some consistent and contextually predictable way con tributed to the semantics of the text, but remained a potential in search of
338
TERRY THREADGOLD
a foregrounder until singled out for prominence in some particular 'con cretisation' (Vodička 1984) or reading of that potential. Readings seem to involve 'foregrounding' against the total potential of the text itself: that is, some patterns are functional, others not at all, in certain readings. Colie's (1972) excellent account of a number of quite incompatible readings of Donne's Anniversary poems provides a good illustration of the way this works. (See also Threadgold 1986b). But what are we to say of foregrounding against 'the system' as opposed to foregrounding against the text? This creates further difficulties. We are back to the question of what is 'the system', to Silverstein's (1979: 205) very pertinent comments about the 'analytic fiction of a well-defined system', and even to the question of which system or whose system. It is certainly true, for example, that neither Quirk and Greenbaum's (1973) University Grammar of English nor Michael Halliday's (1985) Introduction to Functional Grammar contains anything that will enable us to predict the degree of nominalisation and lexical density or the nature of the semantic fields (collocational sets) realised in the Milton text which is the subject of this paper. In this sense, these consistent patterns in Milton's text would have to be seen as foregrounded against those systems: but I have already argued that other 'systems', specifically a number of literary generic codes, which I have above referred to as intertextual semantic frames, are predic tive of the first two patterns (as mode categories) and of the third (as field and tenor categories). Foregrounding then is an area, and a theoretical concept, which we have been reading transparently for too long. There is a need for much more work on the subject. I shall have to be satisfied here with pointing to some of the problems.
Dialogue as Genre, Dialogism, the Prague School and the History of Literary Transmission In order to unpack some of the density of the above arguments I want to go back and look at some of the questions I have raised from different perspectives. Dialogism as I have defined it above may seem a long way from our generic notions of dialogue (in folk-linguistic terms 'everyday con versation') and this may seem a far-cry from Prague school linguistic and lit erary theory and even further from the history of literary transmission.
WHAT DID MILTON SAY BELIAL SAID
339
However, as Umberto Eco, arguing for the sememe as a model of the text, once said: 'one could say that there are more things in a text than are dreamt of in our text theories. But there are also fewer things than are dreamt of (1979: 38), and, in a similar vein, Michael Halliday (1985) hav ing described the extraordinary complexity of transitivity patterns in Eng lish, says: Tt is true that, from one point of view, all these types of process are different. . . . At the same time, looked at from another point of view they are alike'. I am thus in good company when I say that the apparently disparate phenomena I have listed above are indeed very different, but that looked at from another point of view there is a real sense in which they are all alike. Halliday's account of the meaning of the clause as interaction (1985: Ch.4) outlines the genre of dialogue in the sense that it describes the most fundamental types of speech role which are recognisable in all the more specific and complex kinds of dialogue we may ultimately encounter (see also Halliday 1984). But, and this is crucial, we need to remember that the basic interactive grammar of this genre is written into the grammar of the clause in the mood system and the other meanings of the interpersonal function (Halliday 1985). This means that the clause, and therefore all text as discourse, is dialogic in nature; and this in two senses. First, the potential for 'dialogue' as exchange of meanings/goods and services is built in. Sec ond, the clause in its basic grammar carries the lexical items, the words 'which do not forget where they have been'. Halliday and Hasan's account of cohesion (1976), but particularly of lexical patterns and collocation, pro vides fundamental insights into this process at clause level (even if these processes are too 'intuitive' to satisfy linguistics (Martin forthcoming)). And, finally, the genre allows for/predicts the semantics of refusal and con tradiction, what I will later call the semantic reversals of argumentation and debate as kinds of dialogue. The Milton text as a dialogue between speakers in a fictional debate, a dialogue between the narrator's frames and those speeches (which have a typical sayer : said structure), and a dialogue between narrator and reader/ s, speakers and reader/s involves all possible choices encoded in the generic type as outlined above. However there is also more than this. At the point in history when Milton writes, debate as generic-type has already been con-textualised and co-textualised in countless other texts. It has already been classified, hierarchised, taxonomised (subject to framing, classification: Bernstein 1982) into a number of recognisable, contextually
340
TERRY THREADGOLD
Role in Exchange
Commodity Exchanged
Expected Initiation
Discretionary Response
Alternative
give
goods and services
offer
acceptance
rejection
demand
goods and services
command
undertaking
refusal
give
information
statement
acknowledgement contradiction
demand
information
question
answer
disclaimer
Speech Functions and Responses (Halliday 1985)
specific kinds of debate. We need mention only a few here. There is the lit erary/classical type of debate among the gods (relevant because this debate is embedded in an epic), the political type (relevant because of the context of the English revolution), and the religious types of cosmological and theological debate (relevant because this is a debate between rebel angels in Hell). This means that the generic realisation of the basic type of dialogue/debate we started with is overcoded, globally contextualised in at least these terms. Moreover, there are certain 'trans-generic' 'ways of meaning, of saying'/discourses, such as 'humanism', 'empiricism' and 'sec tarian heresy' — which are likely to be an aspect of the realisation of any of these genres. They are, if you like, the common ways of saying (and there fore of knowing and believing) which must have pervaded everyday talk as well as they certainly do pervade the non-canonised popular literature and non-literary texts of the period (the political ballads and broadsheets, the romances, the scientific texts and so on) (Wright 1935) and, as here, the works of 'high' literature. What one should make clear at this point is that many of these genres and discourses are quite contradictory and incompati ble with one another (and here the 'dialogue'/'debate' potentially begins anew at another level). I would like to suggest that while the 'choice' of genres by a writer (al though never unconstrained by coding orientation and subject positioning) may well have been a conscious and deliberate strategy at this time in his tory—and there is good evidence that the choice of 'mixed genres' by writ ers like Milton was precisely this (Colie 1972)—the other kinds of 'dis courses', what I have called 'trans-generic', have much less to do with 'choice'. They are, rather, precisely those discourses which have to do with
WHAT DID MILTON SAY BELIAL SAID
341
the construction of subjectivity, of identity, and they operate at a much less conscious and far more pervasive level than generic discourses (Henriques et al. 1984; Steedman et al. 1985). They are the discourses which 'speak' the subject, and explain the sense of 'the subject constructed in and through language' (Halliday 1978). This is, of course, not to deny the possibility of making those discourses 'visible' and thus disrupting them, which is pre cisely what the books referred to here (Steedman, Henriques) are doing. However it does indicate a need to think differently about these kinds of generic and trans-generic effects as processes in texts and in the construc tion of subjectivity. It should at least be clear by now that the multiple overcoding of the generic-type of dialogue in this text must produce a very considerable 'deautomatisation' of the genre, to use the Prague School term (Garvin 1964). That is, the dialogic can now be seen to involve all of those complex mean ing-making processes which I referred to initially as intertextuality (what Kristeva re-defined as the interaction between texts as semiotic systems (1980)) and heteroglossia, which I take to refer to the coding in texts of social diversity and contradiction in the form of conflicting generic and other discourses. But one aspect of this needs still further discussion and is the reason why I distinguished above between 'generic' and 'other' transgeneric discourses (although the distinction is never more than contingent and constantly shifting in the history of the construction of the subject or the social). It is an aspect which has to do with genre and de- automatisa tion. Every one of the genres (as opposite to the 'other' discourses) I refer red to above as constitutive of the debate in this text involves the semiotics of conscious manipulation, sophistry, deceit and politeness. If, as speaker/ writer, one is going to be/do these things, one has to be ambiguous, obscure, metaphorical. One has to say one thing and mean another: or, to use Eco's terms, one has to flout all of Grice's conventional maxims (Eco 1976: 89). And this of course involves a further kind of de-automatisation of the genre of dialogue with which we began. All of this has implications for theories of language which concern themselves with questions of predictability. It suggests what we can and what we cannot predict and even different ways of thinking about what pre dictability is. There seems to me to be a certain angst in linguistic circles about the question of 'new', uncoded, unsystematic, and therefore unpre dictable meanings—meanings which 'escape' (and the metaphor is quite common) the 'systems' linguistics has constructed. It is an old anxiety, as
342
TERRY THREADGOLD
old as the history of Western thought about language and meaning. In this Milton text, without any attempt at being exhaustive, or rather of claiming the kind of closure which would have said everything, one can say that cer tain things are predictable and others are not (from one point of view). From another point of view, one can say that everything is predictable, but that would, paradoxically, involve putting into question the existence of a pre-established process of semiosis which would in some sense precede the production and interpretation of new meanings, 'ground' them. It would involve seeing new meanings not as something we make sense of because we know the code, but rather as the means by which we construct the code (Eco 1976): and this of course involves a considerable dislocation of the current Western linguistic discourses of literal and figurative, denotative and connotative, and system and process, to name but a few. In this Milton text, the basic dialogue-genre and the overcoding of that genre by others allows one to predict a number of characteristics of the text. Moreover, given the kind of de- automatisation that I have described above, one can even begin to predict the areas of the lexico-grammar which are likely to be most 'at risk' in this sort of configuration: that is, the areas where - and con-textualisation processes are most likely to produce new meanings. What one cannot predict is what those new meanings will be (al though one can of course explain them after the event). But if we envisage semiosis as a process, a dynamic open system and not a closed system, then it is even quite predictable that this state of affairs will be so. Otherwise we would be condemned to forever saying the same thing. Let me give some precise examples. What I have already said 'about the genre of dialogue in this text makes the interpersonal semantics, the logical tactic and semantic structures of projection, the collocational lexical patterns, the patterns of token/value relationships in identifying clauses, and the nominalisations as grammatical metaphor seem the likely sites for processes of c- and con-textualisation of the lexico-grammar to begin to operate independently of the socio-semantics to generate new meanings. Patternings in these areas are already predicted by generic norms. For example, the code of deliberative oratory is based on lexicalisation, nominalisation, identifying processes and the assignment of values to tokens. The rhetorical purpose of that code (tenor) is to persuade/manipu late and to argue, and to do it politely; this is overcoded by a theological debate genre in which the (same) speakers are sophists, deceivers, and a cosmological debate which deals with the unspeakable (God); this must
WHAT DID MILTON SAY BELIAL SAID
343
produce the semantics of indirectness, or a high level of incongruent coding (Halliday 1985). Because certain kinds of argumentation (Ramist hypothetico-deductive logic and the topics of deliberative oratory) (Ong 1958; Caplan 1954) are already predicted generically, and these depend already on a high degree of lexical density, nominalisation, and parataxis, we will be likely to find the incongruence in the nominalisations in this text. Moreover, the individual motivations of speaking voices in the debate (pur pose/tenor again) will predict a skewing of the resources of the basic dialogic-genre — the interpersonal choices, the collocations (semantic fields), and the modes of projection, in particular directions. Thus we will expect the mapping of one debate voice onto the other and the re-co-textualisation that results (one voice arguing with the other and contradicting/ refusing the meanings of the other): and at a certain point we will even be able to predict the direction of the argument. Thus: if Moloch, then war with God, then God is incompetent, able to be defeated, unforgiving, unforgetful: if Belial, then not war, then God is a tyrant, omnipotent, for giving, forgetful. It is interesting that a number of these areas are also those which seem to have the greatest systemic potential for the generation of new meanings (de-automatisation). Thus Halliday (1985) sees metaphor as interpretable in terms of identifying relational processes, and Eco (1976) would see metaphor as a tool for restructuring the semantic universe. Vološinov (1930) suggested that reported speech and its various transforma tions (here projection) are the prototype of metasemiotic discourse (that is, communication about another communication at another level) and thus dialogic in the sense of foregrounding multiple contextualisations and the production of new meanings. And Martin has pointed to collocation, con versational structure and reference as 'dynamic' areas difficult of 'synoptic' description (forthcoming, 1983). I would like to stress that the degree of predictability for which I have been arguing in the case of the Milton example in no way lessens the enor mous complexity of reading, making sense of these texts. Which is to delocate another of our ideologies, that the typical, the predictable is cliché (Hasan 1979) and not to be found in verbal art. There is every indication that contemporary readers of Milton were as skilled in making meanings with the genres as he was (Colie 1972), that indeed to read and write gener ically were contemporary practices. What remains enormously complex is the effect of the realisation of a genre or of mixed genres and the - textualisation and con-textualisation of this and non-generic discourses at the
344
TERRY THREADGOLD
lexico-grammatical level. In so far as all texts participate in these processes, a monologic text is not possible, and reading 'literature' as 'generic' in no way reduces the complexity of the reading process. Nor is it 'reductive' of 'literature' as verbal art. This was the question Mukařovsky addressed a long time ago and is also why dialogism in a text like Milton's is the same as dialogism through a text like Milton's with history. This is the sense in which Mukařovsky (1977) argued that all monologic texts are potentially dialogic, defining the specific character of the semantic structure of dialogic discourse as follows: several or at least two contextures interpenetrate and alternate in dialogic discourse . . . . Because the contextures which interpenetrate in this way in a dialogue are different, often even contradictory, sharp semantic reversals occur on the boundaries of individual . . . replies . . . these [semantic rever sals] have their linguistic correlate in lexical oppositions of evaluative character, and in ambiguity and paradox.
He thus defines dialogue as a special kind of semantic structure oriented towards a maximum of semantic reversals (p.109) which may take one of three basic forms: dialogue as 'a certain external form of utterance' (ques tion/answer: address/reply), dialogic speech as 'a special kind of linguistic (semantic) structure' marked by semantic reversals but not necessarily by the address/reply structure and dialogic quality which involves the potential for the alternation of two or more semantic contextures and may be man ifested in monologue or dialogue. It will be my contention that all of these aspects of dialogism are involved in the Milton text which exhibits a constant process of re-contextualisation and re-structuring of semantic fields and voices. As Berger and Luckmann 1966 put it: The dialogic process is concerned with the creating of alternative, partially differing realities which are defined by the dialectical oppositions among the constitutive 'voices'
in the dialogue. What is interesting about the Milton text is the fact that the epic/ generic convention of monologic, omniscient narrator, the generic norms which guarantee the authority of the narrator, attempt to suppress this interplay of 'voices', and thus to efface the socio-semiotic processes by which the voices are produced and in and through which they are articu lated.
WHAT DID MILTON SAY BELIAL SAID
345
Thus one effect of the exploration of dialogism in this text is to de naturalise the fictional process by reconstituting the play of 'voices' in Bakhtin's (1981), Mukařovsky's (1977) and Berger and Luckmann's (1966) sense. 'Dialogism' involves the recognition that as Mukařovsky put it: monologic and dialogic qualities comprise the basic polarity of linguistic activity, a polarity which reaches a temporary and always renewed equilib rium in every utterance, whether formally monologic or dialogic.
The analysis of 'dialogism' is thus a way of questioning the ideological practices (coding orientations) and generic and intertextual codes by which the authority of the monologic narrator is maintained in a text like this of Milton's. What I would like to suggest is that the interaction of all these elements in the text as dynamic social process produces a dialogic quality (Mukařovsky 1977) which is 'linguistic in nature and that thus the very lin guistic aspect of the text seems to be oscillating between monologue and dialogue' (Mukařovsky 1977: 106). This dialogic quality, described as a characteristic semantics, a set of specific choices and patterns of choice in the text under the constraints of dominant coding orientations, generic norms, and intertextual codes, is characterised by a polysemy, polyphony and heteroglossia (as conflict/contradiction) which provides the potential for change. As the material text interacts at different times with different coding orientations, and intertextual codes, this dialogic semantics provides the material for the restructuring of the semantic system and even the generic and register norms. In the work of the Prague school, and notably that of Vodička (1984), this semogenic and again essentially dialogic process — what is involved is always discourse about discourse, and a type of discourse in which multiple levels of contextualisation appear to be foregrounded — is called transduc tion. This term is used to describe the chains of transmission involved in the processing and reprocessing of literary texts across time and thus according to differing intertextual practices, discursive formations and coding orienta tions involving different generic norms and register disjunctions. Vodička (1984) describes this as an active reprocessing of the 'message' (with all its characteristic polyphony/dialogism) over which its source has lost control. He argues that the judgements of critics (as metadiscourse) provide a description of a reader's 'concretisation' of this process (the term is his), and as such are representative of a particular context of reception which compels us to take account of 'the confrontation of the works' formal lin-
346
TERRY THREADGOLD
guistic properties with contemporary literary codes', and — I would argue — the wider generic and trans-generic systems I have referred to above. In the case of this Miltonic text we have in fact almost 250 years of crit ical comment to explore in these terms. I have included in Appendix 2 a selection of critical texts which illustrate this dialogic process of transduc tion in Vodička's sense and which provide concrete evidence of the way socio-semiotic norms (genres/registers) are actively constituted and trans formed in discourse as intertextual and metasemiotic, semogenic process. Briefly, Milton says Satan is a liar — he says it very clearly as omnis cient narrator — but within 40 years of his writing, the dialogism of trans duction is arguing that Milton got it wrong and Satan is really a hero. It is the extraordinarily complex dialogic, heteroglossic and intertextual ramifi cations of that transformation which an account of dialogism and a sys temic-functional analysis of the texts allows us to explore. The Text Milton, Paradise Lost, Book II, 11.43-228. Frame (Moloch) He ceased, and next him Moloch, sceptered king, Stood up, the strongest and the fiercest Spirit That fought in heav'n, now fiercer by despair. His trust was with th'Eternal to be deemed Equal in strength, and rather than be less Cared not to be at all; with that care lost Went all his fear: of God, or hell, or worse He recked not, and these words thereafter spake: Speech (Moloch) 'My sentence is for open war. Of wiles, More unexpert, I boast not: them let those Contrive who need, or when they need, not now. For while they sit contriving, shall the rest, Millions that stand in arms and longing wait The signal to ascend, sit lingering here, Heav'n's fugitives, and for their dwelling-place Accept this dark opprobrious den of shame, The prison of his tyranny who reigns By our delay? No, let us rather choose, Armed with hell flames and fury, all at once O'er heav'n's high tow'rs to force resistless way, Turning our tortures into horrid arms Against the Torturer; when to meet the noise
WHAT DID MILTON SAY BELIAL SAID Of his almighty engine he shall hear Infernal thunder, and for lightning see Black fire and horror shot with equal rage Among his angels, and his throne itself Mixed with Tartarean sulphur and strange fire, His own invented torments. But perhaps The way seems difficult and steep to scale With upright wing against a higher foe? Let such bethink them, if the sleepy drench Of that forgetful lake benumb not still, That in our proper motion we ascend Up to our native seat; descent and fall To us is adverse. Who but felt of late, When the fierce foe hung on our broken rear Insulting, and pursued us through the deep, With what compulsion and laborious flight We sunk thus low? Th' ascent is easy then; Th' event is feared? Should we again provoke Our stronger, some worse way his wrath may find To our destruction, if there be in hell Fear to be worse destroyed: what can be worse Than to dwell here, driv'n out from bliss, condemned In this abhorred deep to utter woe; Where pain of unextinguishable fire Must exercise us without hope of end The vassals of his anger, when the scourge Inexorably, and the torturing hour Calls us to penance? More destroyed than thus We should be quite abolished and expire. What fear we then? What doubt we to incense His utmost ire? Which to the highth enraged Will either quite consume us, and reduce To nothing this essential, happier far Than miserable to have eternal being; Or if our substance be indeed divine, And cannot cease to be, we are at worst On this side nothing; and by proof we feel Our power sufficient to disturb his heav'n, And with perpetual inroads to alarm, Though inaccessible, his fatal throne; Which if not victory is yet revenge." Frame (Moloch Belial) He ended frowning, and his look denounced Desperate revenge, and battle dangerous
347
348
TERRY THREADGOLD To less than gods. On th' other side up rose Belial, in act more graceful and humane; A fairer person lost not heav'n; he seemed For dignity composed and high exploit: But all was false and hollow, though his tongue Dropped manna, and could make the worse appear The better reason, to perplex and dash Maturest counsels: for his thoughts were low; To vice industrious, but to nobler deeds Timorous and slothful: yet he pleased the ear, And with persuasive accent thus began: Speech (Belial) T should be much for open war, peers, As not behind in hate, if what was urged Main reason to persuade immediate war Did not dissuade me most, and seem to cast Ominous conjecture on the whole success: When he who most excels in fact of arms In what he counsels and in what excels Mistrustful, grounds his courage on despair And utter dissolution, as the scope Of all his aim, after some dire revenge. First, what revenge? The tow'rs of heav'n are filled With arméd watch, that render all access Impregnable; oft on the bordering deep Encamp their legions, or with obscure wing Scout far and wide into the realm of Night, Scorning surprise. Or could we break our way By force, and at our heels all hell should rise With blackest insurrection, to confound Heav'n's purest light, yet our great Enemy All incorruptible would on his throne Sit unpolluted, and th'ethereal mold Incapable of stain would soon expel Her mischief, and purge off the baser fire, Victorious. Thus repulsed, our final hope Is flat despair, we must exasperate Th' almighty Victor to spend all his rage, And that must end us, that must be our cure, To be no more. Sad cure, for who would lose Though full of pain, this intellectual being, Those thoughts that wander through eternity To perish rather, swallowed up and lost In the wide womb of uncreated Night
WHAT DID MILTON SAY BELIAL SAID Devoid of sense and motion? And who knows, Let this be good, whether our angry Foe Can give it, or will ever? How he can Is doubtful; that he never will is sure. Will he, so wise, let loose at once his ire, Belike through impotence, or unaware, To give his enemies their wish, and end Them in his anger whom his anger saves To punish endless? Wherefore cease we then? Say they who counsel war; we are decreed, Reserved, and destined to eternal woe; Whatever doing, what can we suffer more, What can we suffer worse? Is this then worst, Thus sitting, thus consulting, thus in arms? What when we fled amain, pursued and strook With Heav'n's afflicting thunder, and besought The deep to shelter us? This hell then seemed A refuge from those wounds. Or when we lay Chained on the burning lake? That sure was worse. What if the breath that kindled those grim fires Awaked should blow them into sevenfold rage And plunge us in the flame? Or from above Should intermitted vengeance arm again His red right hand to plague us? What if all Her stores were opened and this firmament Of hell should spout her cataracts of fire. Impendent horrors, threat'ning hideous fall One day upon our heads; while we perhaps Designing or exhorting glorious war, Caught in a fiery tempest shall be hurled Each on his rock transfixed, the sport and prey Of racking whirlwinds or for ever sunk Under yon boiling ocean, wrapped in chains; There to converse with everlasting groans, Unrespited, unpitied, unreprieved, Age of hopeless end? This would be worse. War therefore, open or concealed, alike My voice dissuades; for what can force or guile With him, or who deceive his mind, whose eye Views all things at one view? He from heav'n's highth All these our motions vain, sees and derides; Not more almighty to resist our might Than wise to frustrate all our plots and wiles. Shall we then live thus vile, the race of heav'n Thus trampled, thus expelled to suffer here
349
350
TERRY THREADGOLD Chains and these torments? Better these than worse, By my advice; since fate inevitable Subdues us, and omnipotent decree, The Victor's will. To suffer, as to do, Our strength is equal, nor the law unjust That so ordains: this was at first resolved, If we were wise, against so great a foe Contending, and so doubtful what might fall. Î laugh when those who at the spear are bold And vent'rous, if that fail them, shrink and fear What yet they know must follow, to endure Exile, or ignominy, or bonds, or pain, The sentence of their Conqueror. This is now Our doom; which if we can sustain and bear, Our supreme Foe in time may much remit His anger, and perhaps, thus far removed, Not mind us not offending, satisfied With what is punished; whence these raging fires Will slacken, if his breath stir not their flames. Our purer essence then will overcome Their noxious vapor, or inured not feel, Or changed at length, and to the place conformed In temper and in nature, will receive Familiar the fierce heat, and void of pain; This horror will grow mild, this darkness light, Besides what hope the never-ending flight Of future days may bring, what chance, what change Worth waiting, since our present lot appears For happy though but ill, for ill not worst, If we procure not to ourselve more woe'. Frame (Belial) Thus Belial with words clothed in reason's garb, Counseled ignoble ease, and peaceful sloth, Not peace; and after him thus Mammon spake. (headings are mine)
Paradise Lost, Book II, lines 43-228 contains two speeches which are part of a debate in Hell. The speeches enter into a textual polemic with one another. They are projected by the narrator as direct speech, as the speeches of two rebel angels, Moloch and Belial. The speeches as projection are framed by the projecting context in which the omniscient narrator addresses the
WHAT DID MILTON SAY BELIAL SAID
351
reader and declares the speeches and the speakers to be tokens of values which are the opposite of what they seem. The text as a whole, frames and speeches, is situated within an Epic poem by a Puritan poet about the Fall of man. This gives some indication of the complexity of subject positions, cod ing orientations, generic choices, and semiotic systems encoded in the text as intertextual process. I have attempted below to outline some of the potentially contradictory social activity types which these generic choices encode both at the level of normative socio-semiotic potential (genres as types) and across field, tenor and mode as heteroglossia and conflicting dis courses (register as token). For example, the lines are part of a religious epic as a newly coded genre constructed out of existing generic text types. This choice determines the field as both 'heroic' and 'Biblical' and thus constrains the contradiction heroic and liar in the construction of Satan. It also predicts the encoding in this debate context of conflicting meanings from the discourses of theology and politics. Field, tenor and mode choices are constrained for example by Biblical discourses, the discourses of Puritanism, sectarianism, religious debate and the contemporary politics of revolution. The frames, realising the discourse of omniscient narrator, articulate the coding orientation of middle class élitist intellectual and thus the theological discourse of author ity. This discourse attempts to control polysemy, contain contradiction and silence discordant voices. But Belial's speech as debate/polemic encodes not only 'quotations' from Moloch, but also 'quotations' from the frames, and realises the same intertextual semantics as the frame — mapping the many voices of social heteroglossia and the discursive formation onto its 'single wording' as polysemy, paradox and ambiguity — and implicating the voice of the narrator, the voice of authority, in the very voice of rebellion, anti-authority, anti-theology and anti-language which that dominant voice declares to be a lie. The narrative voice does this by equating eloquence with sophistry, inactivity with sloth, and asserting Belial's speech to be a token for values which are the opposite of what they seem to be. The dialogism between this narrative position and Belial's 'quotation' of it must ultimately threaten the control of the narrator's frame in the process of the 'transduction' of the text. The equation of opposite lexical values and the reversal of token/value identifications involved in this dialogism and polysemy is totally constrained by the generic norms of deliberative oratory. (See Appendix 1.) The
352
TERRY THREADGOLD
dichotomising antinomies of the place logic (topics) of deliberative oratory govern the choice of semantic fields in the ideational function, patterns of lexical cohesion in the textual function, and tenor choices in the interper sonal function. The axiological and polarity values which enable the discur sive strategy of the reversal of token and value identifications produce the semantic reversals and re-structuring of semantic fields within the text which are the very stuff of polysemy and dialogism and which ultimately make possible the textual transformation of generic (register) norms them selves. Halliday has argued that dialogue structures arise in the exchange of symbols. The dialogue structures he has outlined in the semantics of mood in the interpersonal function and in the logical semantics of projection, reported and reporting contexts, constitute a normative codification of the genre (register) of dialogue types and of the various transformations of reported/reporting speech contexts as the prototypes of metasemiotic dis course (1985: Ch.7). But as I pointed out above, the dialogic process that occurs in and through co-textual, contra-textual and intertextual relation ships in a text like this Miltonic one involves a crucial de-automatisation of the semantic potential of these generic norms of dialogue. This de-automisation does more than simply foreground the polemic between antithetical or contradictory positions among subjects in discourse: 'What it fore grounds is the nature of the dialogic process itself, the interactive process in which such antithetical positions are defined and redefined' (Thibault 1984b). This is realised through the semantic reversals and recontextualisations described by Mukařovsky, what I call (in the analysis below) the re structuring of semantic fields and the metasemiotic process, which actually contructs the higher order social semiotic in discourse. This happens as the socio-semiotic norms of dialogue are transgressed (Eco 1976) by the semantics of 'quotation' and dialogism at the lowest levels of lexico-grammatical realisation. The process which de-automatises the norm dialogue structure by mapping multiple voices onto single word ings and the higher order processes (mediated by intertextual semantic frames) which produce a discourse characterised by multiple levels of contextualisation and semantic conflict are the principle means for effectively by-passing the semantic determinants of the generic norms of dialogue and producing a new range of meaning possibilities at the lexico-grammatical level. The implication of this is that the contradictory meaning potential or
WHAT DID MILTON SAY BELIAL SAID
353
semiosis of the higher order social semiotic as a dynamic open system is never entirely absent from texts precisely because of the nature of the lexico-grammatical level which simultaneously realises the potential of the normative socio-semantics, the partially independent and processual semantics of co-textualisation and the independent processes of more global contextualisations: as in the explicitly coded case of deliberative oratory (Appendix 1) every meaningful act may be a token of many differ ent types, according to the way it is coded or overcoded (Eco 1976). Thus any meaningful act, or lexico-grammatical structure, may be contextualised simultaneously in many different and contradictory domains (see analysis of Belial's speech below). What this means is that the global patternings of meanings, which are immanent in the organisation of the higher order social semiotic/discursive formation at all levels, and which I described earlier as characterised by both normative and heteroglossic tendencies, culture with conflict, can never be reduced to a unique or monologic meaning in some specific local context of realisation. The dialogic process, as a process of de-automatisation, promotes instability in the system, and seems to ensure that heteroglossia in the form of differences and conflicts in a continual state of dynamic interplay actually constitutes the structure of texts as dynamic processes. But some principle of regulation is always maintained, the process is never totally uncon strained, and it seems that the level of the normative coding of genre is one level at which some degree of replicability and stability is maintained. However, the re-structuring which occurs at the lexico-grammatical level through dialogic conflict (for example, the re-structuring of semantic fields in the polemic of quotation between Moloch's and Belial's speeches) may also effect the system in that it provides the potential for higher level changes in which generic norms themselves are subject to structural trans formation (for example, the restructuring of the omniscient narrator norm through the process of intertextuality and higher order contextualisation and re-contextualisation that is evident in the 'transduction' or transmission of this literary text, and in the critical judgements that are part of this ongo ing dialogism). (See Appendix 2.) Thus it is that in dialogic discourse, the antithetic polemic between rival 'voices' and the conflict between intertextual semantic frames, and between generic norms and heteroglossia, have the potential to re-define the relations among 'voices', to re-define the interpretation of previous
354
TERRY THREADGOLD
'voices' and even to change the rules according to which discourse is to be interpreted (Thibault 1984a). This is the kind of truly metasemiotic situation which is involved in the negotiation of meanings and the meaning-making practices involved in dialogic discourse at every level of its semantic organisation.
The Analysis Representing, within the space of a paper such as this, the systemicfunctional analysis of a text as complex and as long as this one is extremely difficult. I have therefore chosen to look only at certain aspects of the analysis which are relevant to the arguments made earlier in the paper. What fol lows includes an outline of the generic issues, an analysis of the frames to the speeches which focuses on deliberative overcoding and the question of heteroglossia as intertextual frames mapped onto the lexico-grammar of the text, and a much more detailed analysis of Belial's speech, which looks closely at the areas of the lexico-grammar of the text which involve the realisation of semogenesis (Halliday 1985) and semantic reversals (Mukařovsky 1977) and contribute to the dialogism of this text. In Appen dix 1/2 I have included my own version of the taxonomic choices of the deliberative code (adapted from Caplan 1954) and some examples of criti cal 'readings' of the poem which in various ways interact with this textual semantics/lexico-grammar. 1. Genre Since this paper is part of the proceedings of a systemics workshop, it is necessary to say that I do not take the same view of genre as either Mar tin (forthcoming) or Hasan (Halliday and Hasan 1985). Halliday has argued (1978) that the register of a text constitutes those choices actually made across field, tenor and mode, in a particular textual instantiation. I want to use this here to characterise register as token and distinguish it from genres as types (mediated through intertextual semantic frames). Thus Figure 1 attempts merely to characterise some of the generic types which are embedded within one another, follow one another in sequence, or are simultaneously mapped onto the same lexico-grammar within the text. Figure 2 looks at the Register of this text, very schemati-
WHAT DID MILTON SAY BELIAL SAID
355
cally, as sets of conflicting generic discourses mapped onto the lexico-grammatical choices across field, tenor and mode. It should be noted that heteroglossic discourses of a non-generic kind are not included in the account in Figures 1 and 2. Because I am working here with intertextual semantic frames, or genres, of a literary or aesthetic (in Mukařovsky's 1964b sense) kind, all of which include the basic dialogue-genre (Halliday 1985) described above, it is very clear that genre constrains choices in field, tenor and mode. As a text-type each of the literary or aesthetic genres involved specifies what can be talked about (field), how it should be talked about (mode) and who should talk about it to whom and with what purpose (tenor). I do not there fore see how either theme (Hasan) or purpose (Martin) can be singled out as the overriding generic constraint. Indeed, I would like to substitute 'cod ing orientation' in Bernstein's (1971) sense for what Martin (1985c, 1986) calls goal-oriented social activity and see this 'orientation to meaning' as ultimately constraining the choice of genres itself. In the case of these genres, we would have to say that theme is in fact constrained by the genre, and that while purpose constrains some choices (those in tenor), it doesn't constrain field or mode which are already generically specified. Hence my suggested register/token, genre/type dialectic in which the type is only ever instantiated as a token, mediated by an intertextual semantic frame, specifying field, tenor and mode choices, and becoming in its turn the type for new tokens. Figure 1: Genres (Intertextual semantic frames as social practices controlling choice in field, tenor and mode) Biblical epic: (This is a new genre, constructed out of other genres, and other texts — epic and Biblical narrative genres are mapped onto one another) e.g. Field Tenor Mode
The Fall of Man, the activity of heroes. 'To justify the ways of God to man'. To tell what happened. To know. heroic, poetry, written. 3rd person narrative This includes (embeds, contains) The Debate Between the Rebel Angels in Hell. which is made up of:
356
TERRY THREADGOLD
(a) The Narrative frames which project the debate and (i) Maintain the epic/Biblical generic choices (ii) Adopt the topics and logic of deliberative Oratory and of Ramism and thus themselves enter the debate genre. tenor: To instruct the reader that the speakers lie field: What they say mode: 3rd person narrative projecting the speech of charac ter (b) The debate: embedded within the epic and contained by the frames. The genres involved are theological and cosmological debate, classical and political debates, deliberative oratory and Ramist logic as genres of argumentation. These involve: (i) The topics and logic of deliberative oratory and Ramism. Deceit/ sophistry, (ii) The polemic between the speeches, (iii) The polemic between the speeches and the narrative frames. ((i) and (iii) involve politeness and irony (indirectness), and parody (includ ing the 'quotation' and re-co-textualisation of the meaning of the other speaker as this speaker's wording) — all contributing to polysemy, vague ness and ambiguity). field: tenor: mode:
Constrained by topics of deliberative oratory, To argue the expediency of war or peace, Written representation of speech: 1st person and 3rd person narrative: quotation/projection.
Narrative Power and Reader Positioning (as generically constrained tenor relationships): Epic omniscient (absent) narrator instructs the reader, controls polysemy, according to generic aesthetic norms of reading and writing practices. Debate:
(as fictional mimetic representation of action). (i) Omniscient narrator addresses reader — to persuade the reader of his narrative control (I say these men lie) (ii) The speakers address the fictional audience in Hell—to per suade them to act or not against God. (iii) The second speaker (Belial) addresses the first (Moloch), and the fictional audience, quotes the omniscient narrator (implicating this voice in the debate), and quotes Moloch.
WHAT DID MILTON SAY BELIAL SAID
357
This involves a textual mapping of several voices onto the speaking voice, the reconstructing of semantic fields and speaker positions, and the production of polysemy and ambiguity. Figure 2: Register as sets of conflicting generic discourses (1) Field (a) Ideational: Subject-Matter (topic):
People: (Characters)
The Biblical Account of the Fall of Man Classical Models — councils of leaders (Iliad, 1. 54ff., Tasso etc.) (Steadman 1959) Revolution and political debate (Hill 1977) better vs. worse — deliberative oratory (Appen dix 1) Rebel Angels allegorical types orators/sophists heroes: Achilles-type — injured merit, military valor Agammemnon-type — dux, takes command, diverts war Odysseus-type — wiles, ruses, deceptions to gain success in war Revolutionaries God and the Angels omniscient narrator
Action-types:
material — attack/not attack God (as part of debate) Relational — identifying/attributive (as objective telling) (x is brave: courage is despair) mental — thoughts (projection) (as part of debate) verbal — quotes (projection) — (persuasion)
Transitivity (ergative): God as medium vs. Angels as Medium (constrained by speaker position).
358
TERRY THREADGOLD
(b) Logical (1) Speech as quotation of the other Structure'. (2) Hypothetico-Deductive logic — (Ramus hypothetical and connex Syllogisms) (3) Nominal group, participial phrase parataxis — (topical logics and figures — Rhetoric) (2) Tenor
narrator/God. speakers/God Omniscient narrator/reader speakers/narrator — polemic speakers/fictional audience — persuasion speakers/readers — persuasion Status, affect, contact, emphasis Interpersonal rhetoric — irony, politeness, deceit
Speech roles: Pronouns:
poet as pedagogue, speakers, poet as informer, ques tioner, persuader, quoter, interpreter I/you. he/they. etc. we/them etc.
(3) Mode written, poetry, heroic epic style (high Rhetoric Virgilean/Latinate models) (1) Grammatical metaphor (nominalisation/participial phrases) — incongruent (2) Representation of speech in writing — coding conflict between topical logic and Rhetoric of deliberative oratory (nominalisation, parataxis inductive) and the Ramist logic (hypotaxis, parataxis deductive).
2. The Frames, Deliberative Oratory and Intertextuality (other than generic) The rhetorical genre of deliberative oratory constrains choices in field, tenor and mode in a number of interesting ways. As set out in Appendix 1 it would appear to be mainly relevant to field and it certainly constrains choices in this area: it specifically constrains lexical choices, and particularly nominalisations, since the topics are seen as a taxonomy of 'things' to be debated: but this already involves mode and an axiology (tenor: what is good versus what is bad) is built in. It is realised always in the form of a positive/negative polarity either through lexical items with opposite affect
WHAT DID MILTON SAY BELIAL SAID
359
values (courage + /despair- : patience + /sloth - ), or through the overcoding of transitivity patterns in terms of these values ('He said/did/is x' = ± wisdom, i.e. ± good), or through the semantics of identification, involving a con stant ambiguity of token/value relationships which is due to the polarities and oppositions built into the code. These make it possible for any value to be identified with its opposite (his courage is despair). What we are dealing with here is the semiotics of deceit, sophistry, political rhetoric — the semiotics of the lie — where everything is by defin ition the opposite of what it seems —- the de-automatisation of the genre of dialogue. This has the function of assigning positions to subjects as readers, writers, narrators and fictional characters — at least as long as the delibera tive genre is part of reading as well as writing practices — and thus con structs tenor relationships as well as constraining field and mode choices. In Figure 3 I have tried to represent what happens when the lexicogrammar of the frames of Moloch's and Belial's speeches is co-textualised within a debate sequence within an epic and con- textualised (overcoded) by the deliberative code (genre) and a number of heteroglossic transgeneric discourses. The possibilities for ambiguity, polysemy and multiple readings are now considerably increased and the dialogism considerably extended. Figure 3: Narrative frames to Moloch's and Belial's speeches Genre and heterogloss ia poet as teacher: didactic poet as speaker: debate poet as teller: description
(1) Moloch
11.43-50 Opening Frame Narrative: Moloch stood up to speak Description: (facts) He is strongest and fiercest spirit His trust was to be equal with God He is not
Value (Axiology: good/bad polarity: pos./neg.) Deliberative Heteroglossia overcoding (trans-genericconflicting discourses)
Topics: Courage/despair Courage - wisdom
Theology: Manichean Heresy
360
TERRY THREADGOLD
Therefore he does not care He does not fear He cares not to be
despair courage/despair despair
Position of poet as speaker: He is fiercer by despair (in embedded clause, backgrounded)
11.106-8 Closing Frame
heroic: courage in defeat Theology: (1) suicide (2) debate about annihilation mortalism
despair = reality courage = appearance appearance reality .'. courage = despair (Token) (value) Topics: Courage/despair Expedience/Inexpedience
Description: (facts, comment) He looks fierce - courage He threatens war Courage He promises revenge power/success
not victory
- power/- success
battle is dangerous Position of poet as speaker: His courage (revenge) is dispair (desperate) What he says is expedient (battle) is inexpedient (dangerous)
Inexpedience
Theology: (1) revenge must lead to damnation (2) battle cannot be successful
courage = despair
better = worse (token) (value) appearance reality
(2) Belial 11.108-118 Opening Frame Narrative: Belial stood up to reply to Moloch reply to Moloch Topics: Honour Description: (facts, comment) & its sub-topics esp. wisdom/courage/ temperance He was more graceful and + temperance humane He seemed for dignity honour/courage
heroic code: (1) type of hero
WHAT DID MILTON SAY BELIAL SAID composed and high exploit
361
wisdom = -honour/-courage/ - wisdom appearance ≠ reality
His tongue dropped manna + honour + wisdom + justice
He pleased the ear With persuasive accent
+ temperance
And could make the worse appear the better reason to perplex and dash maturest counsels All was false and hollow His thoughts were low To vice industrious to nobler deeds timorous and slothful
expedience
(2) orator who uses wiles/deceit to achieve heroic ends (3) Inaction in defeat = lack of courage political code: (Revolution) deceitful orator (dangerous) Scholastic Code: disputation. The type of the Sophist eloquence = deceit
-wisdom -justice -temperance -courage -honour
Position of poet as speaker: Theology: His eloquence is (a) persuasive (b) dangerous
He is a liar {false, vice) He is afraid to act {to deeds timorous, slothful) He is a sophist (false eloquence - manna)
appearance reality
post-lapsarian knowledge is discursive: wrong logic/ sophistry = sin
(token) (value)
Theology: he is not free to act vs. Bourgeois individualism! humanism, freedom, self-discipline, control 'man must be free to prevent God from
362
TERRY THREADGOLD being responsible for sins' (Hill 1977: 261)
11. 116-228 Closing Frame Narrative: that is what Belial counselled Description: (facts, comment) His words were cloth'd in reason ≠ reason reason's Garb He counselled ignoble ease patient acceptance and peaceful sloth not peace (Courage) = inactivity (-courage) Position of poet as speaker: He says peace/inactivity is Courage It is not. It is fear, sloth Appearance reality (token) (value)
≠
as above
Protestant work ethic: sloth is sin heroic inactivity in defeat = patience, courage = fear, disgrace
3. Belial's Speech My analysis of the speech is eclectic and does not actively follow any methodology I know of in systemic-functional textual analyses. What I have tried to do is to point to those areas of consistent, generically constrained patterning, in this text, which are processual (as social action/discourse) and involve that kind of - and con-textually produced semantic slippage that we call the production of new meanings: which effectively is the construc tion of the social semiotic in discourse. These processes occur at the lexicogrammatical level of co-variate and multi-variate relationships and produce a multiply ambiguous (dialogic/polyphonic) semantics which is the potential source of further new meanings once it is re-contextualised at more global levels of further co-variate and multi-variate relationships. The graphic representation of the analysis in parallel columns is an attempt to indicate the simultaneous, interactive, consistent, ambiguous and often contradictory nature of the processes involved at the lexico-grammatical level. In at least two columns I have grouped together patterns that would not normally be seen as in any sense related, or at least not grouped together. Thus my linking of anaphoric reference, textual cohesion and pro jection as metasemiotic discourse, which is how they all function in this text, constantly mapping the words of the other in the other contexts where
WHAT DID MILTON SAY BELIAL SAID
363
the word has been', onto Belial's words as a projection by the narrator, involves a recognition of all three processes as quotation' or citation of the voice of the other in Bakhtin's (1981) sense, as forms of dialogism, Mukařovsky's (1977) dialogic quality. Likewise the column I have called semantic fields, where I look at the restructuring of semantic fields from speech to speech under the demands of argumentation, involves patterns of lexical cohesion and collocation (thus mixing the syntagmatic and the paradigmatic) and dislocates the lexical/grammatical opposition. This text is marked by a high degree of generically constrained nominalisation and grammatical metaphor (incongruence) and as a result many of the ideologi cally crucial transitivity and ergativity patterns are coded only implicitly in nominal lexical items. It goes without saying that both columns also blend field, tenor and mode categories to some extent. I have also divided the analysis into sections on the basis of what Í have called 'thematic schema', a deliberate blend of textual (mode), generic (schema), field (topic) and logical categories (the periodic sentence). There is a reason for this in the semantics and structure of the text. First, each element of the text that I have included under 'thematic schema', seems to be functioning as a 'paragraph theme' (in Halliday's 1985 sense) and clearly marks the stages of development of the text. Each sec tion of the analysis then includes that section of the text which seems to be constrained by the 'theme' in this column. Thus 11.119-128 are constrained by the clausal theme, T should be much for open war', and so on. I use 'constrain' in the sense that the theme here seems fairly obviously to control choices in field (lexis) and cohesive ties within the paragraph unit of the text: but it also, in important ways, seems to control everything else as well. That is, for example, it usually includes a 'point of view' (the interpersonal choices in the clause — tenor) and thus suggests (1) which earlier parts of the text will be relevant to argue against (metasemiotic discourse/cohesion/ reference, quotation) (2) what semantic fields will have to be restructured to make the argument work (3) what the tenor relationships will be and (4) what will need to be backgrounded as the argument of the 'other' (mode). What the theme does not predict is the Conflicting Codes column. What this suggests is that such control as the thematic schema exerts on the text is very much dependent on co-textualisation at the lexico-grammatical level and on meaning-making as a form of social practice: and indeed the choice of the theme itself is partially controlled at that level and only partially at the generic level. That is, each 'theme' in Belial's speech is a
364
TERRY THREADGOLD
topic of deliberative oratory, but it is also a response to Moloch's argument, as refusal, contradiction, and so on. This then is where the lexico-grammatical level begins to operate independently of any pre-established code or semantics and indeed to construct the code, the semiotics, the social semiotic, in the text as social practice at the points where the 'semantic reversals' of dialogism come into play. I do not want to call what happens in this column 'generic schema' (after Hasan (Halliday and Hasan 1985) or Martin 1985a) precisely because it does not seem to me to be entirely generically constrained and because in this text, one would need at least the first four columns of my analysis to begin to account for the generically constrained 'schemata' in the text: for there is certainly more than one. It is in every sense a dialogue and Halliday's (1985) dialogue 'genre' is part of what is going on here; but Ramist logic (Ong 1958) as a genre of argumentation with a specific logical struc ture (in Halliday's 1985 sense) is also in play, and so is the place logic and the topics and axiology of deliberative oratory (Steadman 1959). (A whole paper could be written about the interplay, in the text, between hypotaxis and parataxis, lexical density, nominalisation and participial phrases as an issue involving the conflict between Ramist and deliberative codes and the speech, writing nexus involved in the embedding of a debate within a writ ten epic.) Moreover, although these 'genres' account for mode and logical experiential choices, and the deliberative code is the primary determinant of field, the genres of theological and cosmological debate and Biblical nar rative also contribute at least to field choices, and a whole series of classical genres and aesthetic/literary codes constrain lexical choices, and aspects of mode which I have here left entirely out of account (e.g. diction/decorum; parallelism, variation and so on). Part of the dialogism is then this generic dialogism. Finally, I would like to suggest that the thematic structure of this text may have a lot to do with wider epistemological issues of what I called earlier the 'trans-generic' kind. Because each section of the thematic schema is constrained in part by a place logic, each tends to be partially independent of the others, with its own pattern of coherence, lexical cohesion and so on. There are of course cohesive patterns across these section boundaries, but the sense of a sequence of juxtaposed, partially independent units remains very strong and is supported by the analysis. What I want to suggest is that this may very well be related to what Land (1974) has called the concept of form in eighteenth century semantic theory. Milton's text is of course half a century
WHAT DID MILTON SAY BELIAL SAID
365
earlier, but the semantic theory is not unrelated, then, or later, to contem porary scientific theories (Slaughter 1982), or the way contemporary cul ture views the world: and the taxonomic arrangement (dispositio) and hypothetico-deductive logic of this text, are one of the less obvious ways in which its semantic structure constructs a different view of the world, a dif ferent reality. A Note on Transitivity The majority of main clause patterns in Belial's speech involve: (a) Relational processes: The tow'rs of heav'n are filled with armed watch How he can is doubtful (b) Material processes: we besought the deep to shelter us The breath that kindled those grim fires should blow them into sevenfold rage. (c) Verbal and mental processes and speech functional metaphors (perfor matives) are rare in main clauses: (1) Who knows whether our angry foe . . . (2) What was urged . . . dissuades me (3) We are decreed, destined . . . to eternal woe) and are mostly coded (as the words, speech acts, thoughts etc. of Moloch (2) or God (3) within nominais or attributes as grammatical metaphor (see tenor and mode). (a), (b) and (c) are constrained by the debate genre (as strategies of argumentation, rhetorical tenor) and the rhetorical/logical generic norms — parallelism/parataxis, nominalisation, participial phrases. All the grammat ical complexity is in these areas, and accompanied by extraordinary lexical density and grammatical metaphor. This means that the semantics of lexical items as cohesive chains (mode) or collocations (realising topic within field) and realised in the text as recurrent or contradictory semantic fields is cen tral to the question of transitivity. Two major ergative fields are involved — God as medium vs. Angels as medium — and these are constantly negotiated and re-negotiated at every level of the textual semantics — as the two major ideologies which have to be foregrounded or backgrounded in arguments about war against God or acceptance of the fate he decrees. See Figure 4, Semantic Fields, Tenor, Mode, for examples.
366
TERRY THREADGOLD
Figure 4. Analysis of Belial's Speech
WHAT DID MILTON SAY BELIAL SAID
367
368
Figure 4. Continued
TERRY THREADGOLD
WHAT DID MILTON SAY BELIAL SAID
369
370
Figure 4. Continued
TERRY THREADGOLD
WHAT DID MILTON SAY BELIAL SAID
371
372
Figure 4. Continued
TERRY THREADGOLD
WHAT DID MILTON SAY BELIAL SAID
373
374
Figure 4. Continued
TERRY THREADGOLD
WHAT DID MILTON SAY BELIAL SAID
375
376
TERRY THREADGOLD
Deliberative Code and Conflicting Codes These columns represent the same principles of overcoding, involving - and con-textualisation and generic and trans-generic intertextual semantic frames as in Figure 3 above. Note however that Sections 3 and 5 of the analysis of Belial's Speech, where the debating voice engages in theological debate in order to refute Moloch's Conclusion (3) and draws its own impossible but absolutely logi cal (syllogistic, logic of necessity) conclusion (5) are the two sections of the speech marked internally by the greatest degree of metasemiotic discourse and contradiction respectively: and these processes involve a constant re construction of the social semiotic. The remainder of the Analysis is self-explanatory. The columns should be read vertically and horizontally where indicated by the arrows.
Two Readings of the Analysis — Dialogism at Work 1. The Frames and Belial's Speech The frames to the two speeches in the debate pose a number of fas cinating semantic and textual/contextual problems. They are the places in the text where Milton's voice as omniscient narrator declares Satan to be a liar. This framing voice of narrative authority, then, ought to control the reading of the speeches to which it has assigned the value 'lies/deceit'. In the critical controversy that has continued since 1692 (approx.) no-one seems to doubt that Milton said Belial was a liar (and I will take just this one example), they simply refuse to believe that he is or assume that Milton somehow got it wrong or lost control over his characters. (See critical judgements, Appendix 1). One of the things that needs to be explored then is whether there is any explanation to be found in the material lexico- grammatical characteristics of the frames as texts for the way their reception by readers is later 'con cretised' (Vodička) and how within the total framework of a critical social semiotic theory as outlined above we can explain what is going on here. I refer you to the analysis above for the detailed analysis of the Moloch and Belial Frames. The narrative frame is very clear about Moloch. What appears to be courage here is to be read as despair. What Moloch argues for
WHAT DID MILTON SAY BELIAL SAID
377
(war—battle, annihilation) is (a) impossible, doubtful (God is omnipotent) and (b) must lead to damnation. Note: These arguments are generically constrained (Biblical, theologi cal). But the deliberative code as generic constraint and the problematic of eloquence (as cultural, intertextual code) produce: (1) (2)
The constantly shifting values of token and value in the seman tics of identification. The conflicting axiological and polarity values realised in the lexico-grammar and the deliberative overcoding.
(3)
The semantics of intellectual doubt coded as the semiotics of deceit and pervading every functional level of the text as generic constraint (Satan is a liar/eloquence is sophistry) — realised, for example, in the semantic oppositions appearance/reality (delu sion), being/not being (ambiguity and polysemy) and truth/false hood (deceit) — and coded in lexis, transitivity, tenor relations and mode choices.
(4)
The semantics of indirectness coded in: a. The deliberative stance of politeness and irony towards opponents — and the poet is here positioned as a speaker in the deliberative debate. (See Analysis, Belial's speech, tenor/mode). b. The difficulty of representing ineffable states like being/not being, doubt, belief, desire as facts in an apparently objective argument — which produces abstract nominalisations as grammatical metaphor, and dizzying identifying clauses and polarity shifts:
e.g. (1) (2)
And rather than be less Cared not to be at all with all that care lost went all his fear
and these ensure that this frame to Moloch's speech is always potentially polysemous and ambiguous. My second example involves a kind of mapping of meanings onto wordings which engenders even greater semantic instability. In the context of Moloch' speech alone the narratorial control is main tained and the polysemy contained. However, as soon as Belial's speech re-
378
TERRY THREADGOLD
co-textualises this frame by adopting its axiology/polarity values and remak ing its meanings as the starting point for its own argument: e.g.
(1) I should be much for open war If what was argued . . . did not dissuade me most.
(i.e. he adopts the position of the frame: (a) revenge without victory = lack of courage = despair (b) annihilation = dangerous = doubtful: coding Moloch's argument as fact and agent of dissuasion) — the frame's potential polysemy, polyphony, heteroglossia can no longer be contained. The voice of narrative authority is subsumed by, mapped onto the voice of the very speaker the authoritative voice has declared to be a liar. Thus the semantics of dialogism, of re-co-textualisation in the voice of the other, as the voice of the other, potentially deconstructs the generic norm of narrative control in the material structure of the text itself at the lexico-grammatical level. This will not be a problem, not perceived as a possible meaning, until these conflicts which produce structural changes at these low levels bring about a number of adjustments at other levels in the system — thus it is not perceived as a meaning while generic norms inscribe monologism, and narrative control over characters, and while cultural codes (intertextual semantic frames) maintain the essential truth of Satan as liar. It will become perceptible and a problem as genres and codes are con structed and re-negotiated in the ongoing intertextual process of 'transduc tion'. This dialogism has then wide implications and is also at work in the frame to Belial's speech. As with the Moloch frame, the position of narrative control here is both absolutely clear and potentially polysemous, polyphonic and heteroglossic. The real polyphony is implicit in the ambiguous status of Belial's free dom to act (as intertextual semantic frames involving conflicting codes). The arguments of Belial's speech, anticipated (and thus pre- co-textualised) in the frame, seem to accord with the voices of authority — God, the Bible, the narrative voice of the Moloch frame. All of these voices, mapped, in turn, onto the lexico-grammar of the Biblical frame (and thus onto the narrator's voice) and then onto the lexico-grammar of the speech (and thus onto Belial's voice) encode the meanings that peace, in the sense of maintaining the status quo, is the only possible course of action. For a
WHAT DID MILTON SAY BELIAL SAID
379
Rebel Angel, in Hell, cast out by God from heaven, action of any kind is generically impossible. And yet the voice of narrative authority (constructing inactivity as sloth according to the intertextual semantic frames of the deliberative code and the Protestant work ethic) simultaneously codes the transitivity patterns that realise these voices—the patterns that say 'action is impossible' as (a) Belial's voice and (b) a lie. The narrative voice then in re-co-/con-textualising Belial's arguments in this contradictory way potentially questions all the voices of authority which are mapped onto these arguments. The voices of authority in the new co-text/context, become implicated in, are mapped onto, the voice of rebellion, political dissent, individualism and generic impossibility. Not only generic narrative control, but also the generic control of the great cultural narratives of religion and the Bible is implicated in the extraordinary contradictions of this dialogism.
2. The Semiotics of Deceit and Literary Transmission The deliberative generic norms provide the codes for the representa tion of Delusion and Deceit as token-value reversals in the semantics of identification (Halliday: 1985) and interact with the Biblical generic norms and theological codes (God is omniscient, the rebel angels are liars, there is no alternative to Hell) to contextualise the whole debate sequence as not only sophistry, false logic and lies, but as ultimately meaningless. If generic norms constrain the status quo, then debate about better/worse courses of action is the worst kind of intellectual delusion. It is talk for the sake of talk and can have no relevance in this situation. The ultimate ambiguity and contradiction then is that the debate is contextualised as a parody of a debate. Because of the problematic of eloquence/sophistry, that is, the fact that rhetoric/oratory may be used to good or evil ends, the appearance of any one of the topics of deliberative oratory may actually represent its opposite (see the discussion of Figure 3 above). Thus, for example, what seems to be courage may be despair: what seems to be patience may be sloth. This results in a high degree of potential instability in the identifying semantics of token and value in the Miltonic frame and the speeches.
380
TERRY THREADGOLD
(1)
To begin with, every transitivity pattern has a value as a token of some topic within this generic framework:
e.g.
he is the fiercest spirit - courage
(.*. courage = despair) he does not care - despair and, as in this case, the juxtaposition of opposite values produces a new identification as a representation of Moloch's sophistry/deceit: courage = despair. (2)
This process continues in the polemic between the speeches, where for example: Moloch : Belial :
but
revenge = possible annihilation = better revenge = victory/success annihilation = worse .'. revenge = worse inaction = better : narrator: inaction = sloth = worse ... (presumably) revenge = better
Does this mean that the poet's voice, the voice that undertakes 'to jus tify the ways of God to men', is actually advocating revenge against God? And does this mimetic reading suggest an answer to the question of why the critics read Milton as 'losing control' of his characters?. The extraordinary degree of 'semantic slippage' here would seem to suggest the basis for such readings. It is, then, this constant renegotiation of meanings, involving a restructuring of semantic fields, from one context to another within the frame/speeches polemic, which constitutes the most vertiginous aspect of the semiotics of deceit in the text. In effect, the semantics of both speeches and frames realised in choices at every level in field, tenor and mode, is constrained by the semiotics of deceit as articulated in the possibilities of deliberative oratory as genre. The state of 'appearing and not-being', or of 'being and not-appear ing', or the other possible oppositions between appearance and reality as exemplified in the Greimassian (1983) model of the elementary signifying structure of deceit:
WHAT DID MILTON SAY BELIAL SAID
381
true
false (Maddox 1984: 25) find their expression at every level of the narrative and discourse organisa tion of these passages. This pervasive polysemy, and shifting of discursive position, is, of course one reason for the arguments that have gone on around the texts, about the meaning of the speeches (see Steadman: 1959). Critical judge ments, as concretisations of the interaction between text and context are firmly based in the lexico-grammar and semantic polysemy of the text itself. I want to look again then briefly at the question of narrative power and reader positioning in these texts in order to suggest, as my final point, how the question of dialogism and the semantics of deceit is implicated in the ongoing dialectic between text and context that Vodička called 'transduc tion', a dialectic which involves the construction and reconstruction of generic norms themselves. While the debate is 'contained' by the narrative frame as the voice of an omniscient narrator with narrative control (epic genre) and addresses only a fictional audience in a fictional Hell, the reader is positioned so as to read the speeches as the narrator instructs, that is, as lies, and according to generic norms which specify: (i)
These speaking voices are 'positions', almost allegorical types, representing relatives of the 7 deadly sins, 'despair' and 'ignoble sloth' (an overcoding not hitherto taken into account in this paper). (ii) Their eloquence is sophistry, a representation of 'wrong reason' (and not to be identified with 'real speech'). (iii) These voices are the voices of Satan (and God, the Angels, Christ, The Word (Bible) all call Satan a liar) and are not to be identified with 'human voices', or the voices of 'real people'.
382
TERRY THREADGOLD
However, what we witness in the critical debate over these speeches, in the 'concrétisation' (Vodička) in critical opinion of the interaction between material text and contextualisation, is the re-making of these generic norms. What we see in this interaction, in this ongoing semanticisation which takes the form of a retrospective re- evaluation and a radical re-reading of the same material text, is the development of mimetic reading and writing practices, the alignment Òf literature with truth, sincerity, psychological reality, and the separation of the 'literary' from the generic—that is, we wit ness the construction of new generic norms, involving new kinds of narra tive control and new reader positions, and implicating different generic and discursive disjunctions. In the case of the critical controversy over this debate, the speakers become men, speaking to the audience of the poem as 'real people', as indi viduals, as 'characters' in the modern sense. The narrative voice becomes another voice among many and has to struggle to be heard (Bakhtin's dialogism (1981)). This has much to do with the changing generic/discursive disjunctions which characterise the end of the seventeenth century and after and are already implicit in Milton's text. As theology, poetry, science and history become disjoined, separate discourses (and the process is underway in Mil ton's text) so the voice of the narrator in the Biblical epic ceases to be the voice of the Logos. Secularised literature involves a new kind of knowing subject and a new kind of democracy, relation among voices. What results is the performance in critical readings of the classic or invariant paradigm of deceit identified by the Greimas school. And we witness, in the interaction between critical opinion (reader) and text, the modal structure of manipula tion which involves the 'transformation of the semiotic competence of the addressee' that is characteristic of the semiotics of deceit and evidence of the ongoing dialogism between textual process and contextualisation. As mimetic reading practices become the norm, and older generic norms (e.g. that this is a parody of a debate) cease to be possible con straints on meaning, the addressee (reader/critic) has access only to the delusive, persuasive knowledge made available by the manipulator (the fic tional narrative/speaking voices of "characters"). The exchange of meanings follows a circular system in which the man ipulator (deceiver) is "operant" in the cognitive/semantic world of nonbeing and appearing, or Delusion, generated verbally, while the addressee
WHAT DID MILTON SAY BELIAL SAID
383
is 'responsive' in the pragmatic world of being and appearing, or material truth, seen as depicted within the discourse: a process in which the dichotomies semantic/pragmatic, text/reality are themselves more than marginally functional. Thus we have first the epic, monologic genre in which a narrator or characters maintain and control the axiological bias and the monological equilibrium which orients reception of aesthetically controlled denoted and connoted values, which determines the relationship true vs. false at the level of manifest textual organisation, where the point of view of the nar rator which corresponds to that of a God or a community inhibits any dialogism between the historical and discursive aspects of the enunciation. This is then replaced by a dialogic, carnivalesque structure, involving the simultaneous manifestation of the monologic and its transgressions (the dominant world-view and the heteroglossia) which were always implicit in the textual structure but controlled, repressed by generic norms. We can perhaps let one of the critics, realising Vodička's notion of 'transduction' in the mode of a mimetic, historical critical tradition, have the last word. C.S. Lewis (1949) sums this up as follows: Milton was relying on two predispositions in the minds of his readers, which, in that age, would have guarded them from our later misun derstanding. Men still believed that there was such a person as Satan, and that he was a liar. The poet did not foresee that his work would one day meet the disarming simplicity of critics who take for gospel things said by the father of falsehood in public speeches to his troops (Thorpe 1951: 95).
This, I take it, is what Vodička meant by 'an active reprocessing of the mes sage, over which its source has lost control'. It represents the essence of that contextual and semantic slippage which constitutes dialogism at the global level. It also supports the argument that genres function to maintain stability in the system. As the mimetic genre develops as a reading practice, and 'generic' reading declines, the generic structures of the text (Colie: 1972) become invisible and readings find harmony, cohesive wholes and again monologism in the text as object.
Conclusion I would therefore argue that it is the many-voicedness of this text as dialogism — the semiotics of deceit realised at every functional level in the
384
TERRY THREADGOLD
text, the points at which the generic norms and the intertextual codes enter into a dialogism of conflict and contradiction at the lowest levels of lexicogrammatical realisation through the realisational forms of quotation and semantic reversal, the points at which the 'single wordings' of the lexicogrammar, mapping and carrying simultaneously the conflicting voices of authority and anti-authority, the voices of poets/debaters/God and Satan and mapping and carrying the conflicting intertextual codes from the many other centres and peripheries of the culture — that answers the question which is the title of this paper: 'What did Milton say Belial said and why don't the critics believe him?' What Milton said is debatable and the dialogism and the dialogue continues . . .
NOTE 1.
This paper was first delivered at the 12th International Systemic Workshop, University of Michigan, Ann Arbor, August 18-24, 1985.
APPENDIX 1 Deliberative Oratory Adapted from Cicero, Ad Herennium(Cap\an 1954) Rhetorical code/genre: controlling choices in field (topics), lexical choices and lexi cal cohesion, constraining mode choices and offering specific subject positions for or against war (tenor). Topics for the debate of war vs. peace as courses of action. If one argues for the bet ter, one argues in terms of the topics below. If ones argues for the worse, one argues in terms of the negation of these topics or their opposites.
WHAT DID MILTON SAY BELIAL SAID
385
386
TERRY THREADGOLD
APPENDIX 2 Some Critical Judgements (from Thorpe 1951) John Dennis (1692)
p. 345
and the Devil is properly his hero, because he gets the better. Pope (1723)
p. 349
since in his speeches (where clearness above all is necessary) there is fre quently such transposition and forced construction, that the very sense is not to be discovered without a second or third reading: and in this certainly he ought to be no example. ('Usual style' — is suitable for 'marvellous' subjects, like Heaven, Hell, Chaos . . . 'Imaging and picturesque parts' — but not for 'the lower sort of narrations'.) William Blake (1793)
p. 353
The reason Milton wrote in fetters when he wrote of Angels and God and at liberty when of Devils of Hell, is because he was a true poet of the Devil's party without knowing it. (Bibliographical criticism — personality of the author). Shelley (1821)
p. 358
Nothing can exceed the energy and magnificence of the character of Satan as expressed in Paradise Lost. It is a mistake to suppose he would ever have been intended for the popular personification of evil . . . . Milton's rebel as a moral being is as far superior to his God, as one who perseveres in some purpose which he has conceived to be excellent in spite of adver sity and torture, is to one who in the cold security of undoubted triumph inflicts the most horrible revenge upon his enemy . . . . He mingled as it were the elements of human nature as colours upon a single pallet and arranged them in the composition of his great picture according to the laws of epic truth . . . to excite the sympathy of succeeding generations of man kind. (Romantic view — Milton as rebellion/Satan, biographical criticism) Macaulay (1825)
p. 360 (personality of poet, - comparison with Dante)
Poetry which relates to the beings of another world ought to be at once mysterious and picturesque. That of Milton is so . . . . His fiends are won derful creations . . . . Their characters are, like their forms, marked by a certain dim resemblance to those of man, but exaggerated to gigantic
WHAT DID MILTON SAY BELIAL SAID
387
dimensions, and veiled in mysterious gloom. . . . Yet it would be difficult to name two writers whose works have been more completely . . . coloured by their personal feelings. The character of Milton was particularly distin guished by loftiness of spirit. Landor (1846)
p. 368
It is Adam who acts and suffers most. . . . This constitutes him the main character; although Eve is more interesting, Satan is the more energetic, and on whom the greater force of poetry is displayed. The Creator and his Angels are quite secondary. . . . It [Paradise Lost] is . . . the noblest speci men in the world of eloquence, harmony and genius. (mimetic criticism, ahistorical.) Matthew Arnold (1888)
p. 373ff.
Milton is of all our gifted men the best lesson, the most salutary influence . . . . Now, no race needs the influences mentioned, the influences of refin ing and elevation, more than ours; and in poetry and art our grand source for them is Milton . . . the great style . . . . But it triumphs in Milton, one of our own race, tongue, faith and morals. Milton has made the great style no longer an exotic here; he has made it an inmate among us, a haven, a power . . . . The English race overspreads the world, and at the same time the ideal of an excellence the most high and the most rare abides a posses sion with it forever. (Moral value of literature — literature — teaches us about life — Nationalism). Hanford (1919)
p. 164
obviously his belief is anything but naive. And as the actual experience of Adam and Eve, not to mention the desperate plot of evil men in Hell to overthrow the reign of righteousness and law, they are richer in human truth than anything in English Imaginative Literature outside of Shakes peare . . . a high seriousness . . . . Its practical results are Paradise Lost . . . in which the total Renaissance is summed up and revealed as one, through a harmony of its great ideals of beauty, righteousness and truth. (humanist reading—mimetic, historical — man the knowing subject, ideals of unity, coherence, truth).
REFERENCES Apple, M.W., ed. 1982. Cultural and economic reproduction in education: essays on class ideology and the state. London and Boston: Routledge and Kegan Paul.
388
TERRY THREADGOLD
Atkins, J.W.H. 1951. English literary criticism: seventeenth and eighteenth centuries. London: Methuen. Bakhtin, M.M. 1973. Problems of Dostoevky's poetics. R.W. Rostel, trans. Ann Arbor: Ardis. . 1981. The dialogic imagination: four essays. M. Holquist, ed. Austin: University of Texas Press. Barthes, R. 1974. S/Z, trans, by R. Miller. New York: Hill and Wang. Benson, J. and W.S. Greaves, eds. 1985. Systemic perspectives on dis course. Vol 1: Selected theoretical papers from the 9th International Systemic Workshop. Norwood, New Jersey: Ablex. Berger, P.L. and T. Luckmann. 1966. The social construction of reality: a treatise in the sociology of knowledge. London: Allen Lane, Penguin Press. Bernstein, B. 1971. Class, codes and control 1: theoretical studies towards a sociology of language. (Primary socialization, language and educa tion). London: Routledge and Kegan Paul. . 1982. Codes, modalities and the process of cultural reproduction: a model. In M.W. Apple, ed. [see above]. Caplan, H. ed. and trans. 1954. Cicero ad C. Herrennium. Loeb Classical Library. Cambridge, Mass. and London: Camb. U.P. Colie, R.L. 1972. 'All in Peeces': Problems of interpretation in Donne's Anniversary Poems. In P.A. Fiore, ed. Just so much honor. Univ. Park and London: The Pennsylvania State U.P. Easthope, A. 1983. Poetry as discourse. London and New York: Methuen. Eco, U. 1976. A theory of semiotics. Bloomington: Indiana U.P. . 1979. The role of the reader: explorations in the semiotics of texts. London: Hutchinson. Foucault, M. 1972. The archaeology of knowledge. Trans. Sheridan Smith, A.M. London: Tavistock. . 1972. The order of things. New York: Vintage Press. Fowler, A. 1982. Kinds of Literature: an introduction to the theory of genres and modes. Oxford: Clarendon Press. Fowler, R. 1981. Literature as social discourse: the practice of linguistic criticism. Bloomington: Indiana U.P. Garvin, P.L. ed. 1964. A Prague School reader on esthetics, literary struc ture and style. Georgetown: Georgetown Univ. Press. Greimas, A.J. 1983. On anger: A lexical semantic study. Monographs, Working Papers and Prepublications of the Toronto Semiotic Circle 4.
WHAT DID MILTON SAY BELIAL SAID
389
Halliday, M.A.K. 1978. Language as social semiotic. London: Arnold. . 1981. Text semantics and clause grammar: some patterns of realisa tion. In J.E. Copeland, and P.W. Davis, eds. The Seventh Lacus Forum, Columbia: Hornbeam Press. . 1982. The de-automatization of grammar: from Priestley's 'An inspector calls'. J. Anderson, ed. Language form and language varia tion: Papers dedicated to Angus Mcintosh. Amsterdam: John Benja mins, pp.129-159. . 1984. Language as code and language as behaviour: a systemic-func tional interpretation of the nature and ontogenesis of dialogue. In R.P. Fawcett, M.A.K. Halliday, S.M. Lamb, & A. Makkai, eds. The semio tics of culture and language, Vol. I. London: Frances Pinter. . 1985. An introduction to functional grammar. London: Arnold. Halliday, M.A.K., & Hasan, R. 1976. Cohesion in English. London: Longman. . 1985. Language, context and text: aspects of language in a social semiotic perspective. Geelong, V i c : Deakin U.P. Hasan, R. 1973. Code, register and social dialect. In B. Bernstein, ed. Class, codes and control, Vol. 2. London: Routledge and Kegan Paul. . 1979. Language in the study of literature. In M.A.K. Halliday, ed., Working conference on language in education: report to participants. Sydney: Univ. of Sydney Extension Program. . 1984. Ways of saying: ways of meaning. In The semiotics of culture and language, Vol. I. R.P. Fawcett, M.A.K. Halliday, S.M. Lamb, & A. Makkai, eds. London: Frances Pinter. . 1985. Linguistics, language and verbal art. Geelong, V i c : Deakin LLP., 1985. .1986. The ontogenesis of ideology. In Threadgold et al., eds. Semio tics — Ideology — Language. Sydney: The Pathfinder Press. Havránek, B. 1964. The functional differentiation of the standard lan guage. In P.L. Garvin, ed. pp. 3-16. Hill, 1972. Milton and the The English Revolution. London: Faber and Faber. Kristeva, J. 1980. Desire in language: A semiotic approach to literature and art. New York: Columbia U.P. Land, S.K. 1974. From signs to propositions: The concept of form in eighteenth century semantic theory. London: Longman. Lemke, J.L. 1983. Thematic analysis: Systems, structures and strategies.
390
TERRY THREADGOLD
RS/SI 3, 159-187. . 1985a. Textual politics: heteroglossia, discourse analysis and social dynamics. Unpublished paper given at The International Summer Insti tute for Structuralist and Semiotic Studies. Bloomington: University of Indiana. . 1985b. Ideology, intertextuality and the notion of register. In J.D. Benson and W.S. Grèaves, eds. Systemic perspectives on discourse. Vol. I: Selected theoretical papers from the 9th International Systemic Workshop. Norwood, New Jersey: Ablex. Maddox, D. 1984. Semiotics of deceit: the pathelin era. London and Toronto: Associated University presses. Martin, J.R. 1983. Conjunction: the logic of English text in J.S. Petöfi & E. Süzer, Micro and Macro Connexicity of texts. Hamburg: Helmut Buske (Papers in Text Linguistics 45) pp. 1-72. Martin, J.R, 1985a. Texts: Two aspects of human semiosis. In J.D. Benson and W.S. Greaves, eds. Systemic perspectives on discourse. Vol. I: Selected theoretical papers from the 9th International Systemic Work shop. Norwood, New Jersey: Ablex. . 1985b. The language of madness: method or disorder. Sydney: Lin guistics Department Unpublished manuscript. . 1985c. Factual writing: exploring and challenging social reality. Geelong, V i c : Deakin U.P. . 1986. Politicalising ecology: The politics of baby seals and kan garoos. In T. Threadgold et al., eds. Semiotics — Ideology — Language. Sydney: The Pathfinder Press. . Forthcoming. Lexical cohesion, field and genre: parcelling experi ence and discourse goals. Proceedings of the Second Rice Symposium in Linguistics and Semiotics: Text Semantics and Discourse Semantics. Medvedev, P.N. The formal method in literary scholarship. Baltimore, London: Johns Hopkins U.P. Mukařovsky, J. 1964a. Standard language and poetic language. In P.L. Garvin ed. 17-30. _. 1964b. The esthetics of language. In P.L. Garvin, ed. 31-69. . 1977. The word and verbal art: selected essays. J. Burbank, & P. Steiner, Transl. & ed. Foreword by R. Wellek, Yale: Yale U.P. Ong. W. 1958. Ramus' method and the decay of dialogue. New York: Octagon. Quirk, R., & S. Greenbaum. 1973. A university grammar of English. Lon-
WHAT DID MILTON SAY BELIAL SAID 391 don: Longman. Rosmarin, A. 1985. The power of genre. Minneapolis: Univ. of Minnesota Press. Silverstein, M. 1979. Language structure and linguistic ideology. In P.R. Clyne, W.F. Hauks, & C.L. Hofbauer, eds. The elements: a parasession on linguistic units and levels. Chicago: Chicago Linguistic Society. Slaughter, M.M. 1982. Universal languages and scientific taxonomy in the seventeenth century. Cambridge: Cambridge U.P. Steadman, J.M. 1959. Milton's epic characters. Chapel Hill: University of North Carolina Press. Steedman, , Urwin, & V. Walkerdine. eds. 1985. Language, gender and childhood. London: R.K.P. Thibault, P.J. 1984a. Narrative structure and narrative function in Vladimir Nabakov's Ada: A contribution to the development of a critical, neomaterialist social theory of texts as the products of social meaning making practices. Linguistics Department Sydney: Unpublished Ph.D. thesis. . 1984b. Narrative discourse as a multi-level system of communica tion: some theoretical proposals concerning Bakhtin's dialogic principle. Studies in Twentieth Century Literature, Fall 9, 89-118. . 1986. Thematic system analysis and the construction of knowledge and belief systems in discourse. Sydney: English Department Unpub lished manuscript. Thompson, C , ed. 1984. Studies in Twentieth Century Literature: special issue on Mikhail Bakhtin, 9(1). Thompson, C. 1984. Bakhtin's 'theory' of genre. Studies in Twentieth Cen tury Literature, 9(1), 29-40. Thorpe, J. 1951. Milton criticism: Selections from four centuries. London: R.K.P. Threadgold, T., E. Gross, G. Kress, & M.A.K. Halliday. eds. 1986. Semiotics — ideology — language. Sydney: The Pathfinder Press. Threadgold, T. 1986a. Semiotics — ideology — language. In: Threadgold et al., eds. semiotics — ideology — language. Sydney: The Pathfinder Press. . 1986b. Subjectivity, ideology and the feminine in John Donne's poetry. In: Threadgold et al,, eds. Semiotics — ideology — language. Sydney: The Pathfinder press. . 1986c. 'It's great music ■— But can you dance to it?' Review of ISISSS '85. Semiotica, 62(1/2).
392
TERRY THREADGOLD
Ventola, E.M. 1979. The structure of casual conversation. Journal of Prag matics, III. Vodička, F. 1984. Response to verbal art. In Semiotics of art, L. Matejka, & I R. Titunik, eds. Cambridge, Mass. and London MIT Press. Vološinov, V.N. 1930. Marxism and the Philosophy of Language Trans, by L. Matejka & I R. Titunik, New York + London: Seminar Press. Weber, S., ed. 1980. The Strasbourg Colloquium: genre. A selection of papers. GLYPH: Textual Studies 7. Baltimore and London: The Johns Hopkins U.P. Whorf, B.L. 1956. Language, thought and reality, J.R. Carroll, ed. Cam bridge: MIT Press. Wright, L.B. 1935. Middle class culture in Elizabethan England. Chapel Hill: Univ. of North Carolina Press.
THE INTERACTION OF LANGUAGE AND MUSIC AS SEMIOTIC SYSTEMS: THE EXAMPLE OF A FOLK BALLAD
Erich Steiner Universität des Saarlandes Saarbrücken, West Germany
Introduction The purpose of this paper is the examination of two semiotic systems: language and music. The question asked is what kinds of meanings are realized in what way, and more specifically, how the two systems interact in weaving the texture of linguistic text and musical text in a folk ballad. When embarking on this kind of investigation, we can rely on a fairly comprehen sive knowledge about language up to and including the clause, as well as a growing amount of knowledge about text and discourse above the clause (cf. Halliday and Hasan 1980; Beaugrande and Dressier 1981; Journal of Pragmatics 6,5-6, special issue on Story, 1982; Text 2,1-3, special issue on text processing, 1982). With regard to music, the situation is similar as far as the 'syntax' of music is concerned: there is a respectable tradition of describing systems of rules of tonal harmony in interrelation with systems of rules for polyphonic or homophonic setting of voices or melodies. There are even theories of musical structure which refer to the arrangement of phrases and sets of phrases into larger 'texts', such as the theory of classical sonata form or that of the baroque fugue. What is conspicuously absent, however, is an account of the 'semantics' of music, the meanings realized by the formal patterns of tonal harmony and melodies (Cooke 1959; Caudwell 1973: 267ff.). This lack of knowledge, due in large part to the formalistic orientation of theories of art in our century, makes the present task difficult and interesting. Given the purpose of this paper, and given the present state of knowl-
394
ERICH STEINER
edge, as far as I am aware of it, it was thought best to answer the questions raised, by analysing a particular instance of a message, where both semiotic systems are involved, which is why the folk ballad presented in this text was chosen (section 1). With regard to this ballad, we shall ask what is the meaning conveyed by it and how it is realized in both language and music. On a more general level, this is the question of what is different between language and music, on the one hand, and what is common to both of them on the other. It is becoming more and more widely accepted that meaning cannot be discussed without reference to a situation, or the mental image of a situation (cf. Malinowski 1935; Firth 1950; Parret and Bouveresse 1981; Schank and Abelson 1977). Into this 'situation', the human agent enters as the acting subject. In our context, it is of particular importance that his/her activities can take the form of linguistic activities. Obviously, then, the 'ac tivities' and 'situations' mirrored or portrayed in the linguistic and musical texts have to be described within a reasonably consistent framework, because only then can we ask the question of how 'meaning' is realized and portrayed in our texts. In doing so, our focus will be on 'conceptual' or 'ide ational' meaning, i.e. 'the story told', but we cannot afford to neglect those meanings that are more directly a consequence of the situation in which the text is realized, the 'textual' and 'interpersonal' meanings. I have drawn upon the following theories and models in the analysis presented in this paper: the framework of the theory of ACTIVITY is derived from the writings of a certain school of Russian psychologists and physiologists, which is attracting more and more interest in the west (cf. Vygotsky 1962; Vygotsky 1978; Luria 1973; Leont'ev 1975; Leont'ev 1978). I have given the reasons for this particular choice elsewhere (Steiner 1983). The framework for the linguistic analysis is the general framework of Sys temic Linguistics (Halliday 1967-68; Halliday 1985; Berry 1975 and 1977; Monaghan 1979; Steiner 1983), particularly the version presented by Fawcett 1980, 1983 and (1987)). Influences from linguists outside the Systemic framework will be acknowledged where appropriate. The musical analysis was done with a general background knowledge of tonal harmony and mus ical composition, not traceable to any particular school of thought within the theory of music. I owe many of the thoughts on the meaning of music expressed in this paper to Cooke's The language of music (1959). As usual, when applied to a particular task, the theories and models mentioned had to be further developed and adapted. It is hoped that the developments and adaptations will prove useful in other contexts, too.
THE INTERACTION OF LANGUAGE AND MUSIC
395
Apart from the purpose stated above, this paper has a more general aim; to bring together a set of ideas that have not been brought together before in an analysis, ideas about the structure of ACTIVITY in relation to the structure of semiotic systems. In writing this paper, I have learned more about the problems of bringing together ACTIVITY, SITUATION and LANGUAGE, and I hope very much to have presented my ideas clearly enough to enable the reader to form a critical opinion of them.
1.0 The ACTIVITY and the SITUATION Let me start with a brief general outline of the theory of ACTIVITY to be used here. This theory will then be applied in the description of the story told in the text of our ballad. The theory operates with three ranks of structure within the overall structure of ACTIVITY—the ranks of ACTIVITY, ACTION and OPER ATION. ACTIVITIES are the most general form of behaviour, answering to people's (cultural and/or physiological) needs. They are aroused by those needs and have objectives which will satisfy those needs. The ACTIVITY ceases with the achieving of the objective, the satisfaction of the need. Looking more closely at the whole phenomenon of ACTIVITY, we find that an ACTIVITY can be realized by one or more ACTIONS, and conversely, that one ACTION can be the realization of quite different ACTIVITIES. An ACTION is defined by its conscious purpose or goal, designed to (partly) satisfy the need, achieve the objective of the ACTIV ITY which it realizes. ACTIONS, in this framework, are composed of 4 phases (elements): MOTIVATION, where a course of ACTION is invoked in relation to the need of the general ACTIVITY; PLANNING, where a programme for the ACTION is developed, alternative hypotheses are evaluated and related to their probable out comes; EXECUTION, where the ACTION is actually carried out; and EVALUATION, where the outcome is checked against the original plan, and the EXECUTION checked. In the case of failure, the whole ACTION may be re-entered. Each of these phases of ACTION is filled by units of activity (used
396
ERICH STEINER
here in the general sense), again from one rank below: OPERATIONS. These are units of activity defined by a task which helps towards the achievement of the purpose/goal of the ACTION of which the OPERA TION is a part in any particular SITUATION. When we describe a particu lar individual in the course of an ACTIVITY, the OPERATIONS filling MOTIVATION, PLANNING and EVALUATION will take place inside the organism in the majority of cases, whereas OPERATIONS filling EXECUTION will be external in the case of bodily ACTIONS. OPERA TIONS are thus the lowest unit in this framework, and do not have a struc ture on this level of the theory. They do, however, have a structure in terms of the sensory-motor components of behaviour, as analysed in detail in neurophysiology (cf. Luria 1973; Pribram 1971). In describing ACTIVITY, we do not go 'straight down the line' in every analysis: elements on the ranks of ACTIVITY and ACTION can be filled by units from the same rank or even ranks above in the case of complex or otherwise difficult ACTIVITIES. This is the outline of the theory as far as we need it in the present con text. Let us then, before embarking upon an analysis of a certain ACTIV ITY, namely the one narrated in our folk ballad, have a look at the text we are going to analyse. Here are the lyrics of our folk ballad. The numbering in the left hand margin refers to the clause in the text, for reasons which become clear in the textual analysis itself. One line of the text, as printed here, corresponds to one melodic and rhythmical unit as analysed in section 3 of this paper. Verse 1 1 2,3 4,5 6,7
Old Mr. Johnson had troubles of his own He had a big yellow cat that wouldn't leave his home He tried and he tried to give that cat away And he gave it to a man who was going far away Refrain
8 9,10,11 12,13
And the cat came back the very next day The cat came back, you know, he thought he was a gonner Yeah, the cat came back, he just couldn't stay away Verse 2
14
So, he gave it to a boy with a five-dollar-note
THE INTERACTION OF LANGUAGE AND MUSIC
15,16,17 18,19 20,21
397
Said: 'Take it down to the river, take it out in a boat' He tied a weight around its neck, must have weighed a pound Now they're searching the river for the poor boy that's drowned Refrain Verse 3
22,23 24 25,26 27,28
The man around the corner said 'I'll shoot that cat on sight' And he loaded up his shotgun with nails and dynamite He waited and he waited for that cat to come around But ninety-seven pieces of that man was all we found Refrain Verse 4
29 30 31,32,33 34
The A-Bomb fell, just the other day The H-Bomb fell in the very same way England was gone, and France was gone, and even the good old USA The whole world was demolished on that day Refrain American Traditional
Concentrating on the 'story' told in the text, we can see that it derives its characteristic features from the conflict of two ACTIVITIES, which drives the story to its climax: need 1
activity 1
objective 1
Untroubled Remove state of mind. | disturbance If disturbed (cat) (by presence of cat) →
Removal of disturbance; Satisfaction of NEED 1
AGENT: 'Mr. Johnson'
need 2
activity 2
objective 2
'Pleased' state Remove of mind. If disturbance disturbed, (by absence from home) →
Removal of disturbance; Satisfaction of NEED 2
AGENT: 'The cat'
398
ERICH STEINER
Already on this general level, we can see the source of conflict, which does not lie in the NEEDS, ACTIVITIES or OBJECTIVES themselves, but in the fact that what disturbs the satisfaction of NEED 1, in this particu lar SITUATION, is exactly the state of affairs that makes for the SATIS FACTION of NEED 2: The cat's being in the vicinity of Mr. Johnson, or at least his borne. Both ACTIVITIES provide the entry condition for ACTIONS which are followed through in the course of the song. From ACTIVITY 1: GOAL/PURPOSE Cat be far away
ACTION 1 change possession of cat MOTIVATION 'having troubles' = 'cat'
PLANNING not in text to be inferred
EXECUTION give cat to man
(••
EVALUATION not in text to be inferred
We are not told anything about the actual OPERATIONS carried out, because these do not matter, in the opinion of the author and in the context of the story as told here. There is a structural property of ACTIONS to be seen in ACTION 1 which is certainly worth pointing out in this context: embedding. 'Interpreting' an ACTION means among other things expecting the EXECUTION to provide a sufficient condition for the state of affairs described under GOAL/PURPOSE, the GOAL/PURPOSE on the other hand providing the reason for the EXECUTION (cf. Lowe 1983 for a dis cussion of the terms 'condition' and 'reason'). This, however, is not the case here; so, either we have to consider ACTION 1 as badly planned, or else we have to look out for embedded ACTIONS which could provide the missing links in a chain of cause-effect, to make the EXECUTION a suffi cient condition of PURPOSE/GOAL. Looking at VERSE 1 of our text, we can see an embedded ACTION of the kind we are looking for: The linguistic reflection of EXECUTION of ACTION 1 (lines 6,7) contains an embedded clause. This embedded clause tells us that the 'receiver' of 'give cat to man' is himself an argument in an embedded proposition in the following way: (Johnson x CAUSE (mana POSSESS catb)y ) and (mana CAUSE (mana DISTANCE FROM X)z )
THE INTERACTION OF LANGUAGE AND MUSIC
399
where the whole proposition after the and sign can be embedded after 'man' in the first line. a
In other words, 'Mr. Johnson' performs ACTION 1 by involving in EXECUTION another person who will function as an AGENT in yet another ACTION, and these two ACTIONS together will provide the sufficient condition for GOAL/PURPOSE of ACTION 1. What is still lacking as an element of ACTION 1 is the EVALUATION. Again, there is none reflected in the text, but it is to be inferred from what we are told about ACTION 2: only if we infer a 'failed' GOAL/ PURPOSE in the EVALUATION of ACTION 1, do we get the necessary entry condition for ACTION 3 later on. Let us, therefore, have a look at ACTION 2: From ACTIVITY 2: ACTION 2 Come (back) to Johnson MOTIVATION PLANNING can't stay not in text; to away' 1 be inferred
GOAL/PURPOSE (cat) be close to Johnson EXECUTION cat comes back
EVALUATION not in text probably 'successful'
The cat's ACTION is not narrated in any great detail, the only important effect being that it is successful, which leads us, in the 'Situation of Utterance' (Davies 1979: 57ff.) to infer a 'successful' EVALUATION by the cat as well. Note that there is a big difference between the EVALUATION of the listener! reader of the story, which is an Evaluation of ACTION 2 by a person different from the AGENT of ACTION 2 and in a different SITUATION ('Situation of Utterance' of the text), and EVALUATION of ACTION 2 by its own AGENT, which is not reported in the story, and which the reader/listener only can infer to have actually taken place in the 'Situation of Performance'. In section 2.3 of this paper on PROCESS we shall deal with the linguistic clues for the 'successful'-EVALUATIONS just mentioned. After ACTION 2 has taken place, then, the old conflict between ACTIVITIES 1 and 2, deriving from the particular constellations of NEEDS and OBJECTIVES in this SITUATION, is in full force again, permitting re-entry to the cycle of ACTION and counter- ACTION. Therefore:
400
ERICH STEINER From ACTIVITY 1: ACTION 1 kill cat
MOTIVATION not in text; to be inferred
PLANNING to be inferred 'create subACTIONS' 3.1.-3.3
GOAL/PURPOSE Cat be dead EXECUTION filled by 3.1.-3.3
EVALUATION indirect, see lines 20-21 of text: 'failed'
From EXECUTION 3: A C T I O N 3.1. give cat to boy MOTIVATION not in text to be inferred
PLANNING not in text; to be inferred
GOAL/PURPOSE boy possess cat ( b o y = A G E N T in some further A C T I O N x = c a u s e cat to be in water) EXECUTION Johnson gives cat to boy
EVALUATION not in text; to be inferred: 'successful'
From EXECUTION 3: A C T I O N 3.2. Cause boy to take cat to water (and do A C T I O N x ? ) MOTIVATION not in text to be inferred
OPERATION, Interpersonal: Ideational: Į Textual: Realization:
PLANNING not in text; to be inferred
GOAL/PURPOSE boy take cat to water (and do A C T I O N x) EXECUTION tell boy to do P U R P O S E / G O A L 3.2 A N D
EVALUATION not in text; to be inferred: 'successful' pay boy
from E X E C U T I O N 3.2. Enter linguistic system command, imperative jussive, no politeness formula... locational, third party agent, change, +accompanyment... Theme not realized, unmarked intonation... T a k e it down to the river'
We are also told the second OPERATION: T a k e it out in a boat.' There are good reasons for drawing the boundaries between ACTION and OPERATION in cases like these between 'direct' and 'reported' speech, because what is consciously planned is the content of an utterance, so that this stage should be analysed on the rank of ACTION, whereas the particu lar wording and the intonation structure of the utterance will not normally
THE INTERACTION OF LANGUAGE AND MUSIC
401
have to be consciously planned and monitored, so that the actual wording will be analysed on the rank of OPERATION. Obviously, all this will not apply in certain SITUATIONS, as for example in learning a foreign lan guage, or in the case of a patient recovering from aphasia. Writing a paper, such as this one, or preparing a speech for a public occasion will also at many points require dealing with the wording of utterances at ACTION rank. From EXECUTION 3: ACTION 3.3 Add weight to cat MOTIVATION not in text; to be inferred
PLANNING not in text; to be inferred
GOAL/PURPOSE Cat be so heavy as to go under EXECUTION tie weight around cat's neck
EVALUATION not in text; to be inferred: 'successful'
In ACTIONS 3.1-3.3 we find a type of embedding of ACTIONS which is different from the embedding in ACTION 1: in ACTION 1, further ACTIONS are embedded as expected ACTIONS of AGENTS reflected as 'arguments' participant roles' under GOAL/PURPOSE of ACTION 1, their agent not being identical with the AGENT of ACTION 1. In ACTION 3, further ACTIONS are embedded as realizations of elements of the ACTION carried out by the AGENT of the ACTION in which they are embedded. Note, however, that due to the linguistic realization in our case, there is an ambiguity of AGENT in 3.3 (cf. section 2.2). Furthermore, we cannot discuss the question of the actual temporal sequence of ACTIONS and elements, but we should take note of one question connected with embedding: the PLANNING-stages of 3.1-3.3 could be already realized within the PLANNING-stage of the superordinate ACTION 3, rather than in their unmarked place in the sequence of embedded ACTIONS. This, however, is entirely an empirical question for any single ACTION under description. We do have to make sure, though, that our model provides for this possibility, which the present model indeed does. Also, the PURPOSE/ GOAL of ACTIONS .1.- 3.3 will have been created as part of PLAN NING of ACTION 3, because it is this very fact that makes them 'superor dinate' to ACTION 3. Back to our story now, where ACTION 3 is again followed by ACTION 2 (refrain), leading to another re-entry:
402
ERICH STEINER From ACTIVITY 1: ACTION 4 kill cat
PURPOSE/GOAL cat be dead
There is, of course, the question of whether we still have a re-entry from ACTIVITY 1 here. The AGENT has changed to 'the man around the corner' and the listener is free to infer that either there is a new ACTIVITY 3 with a new AGENT, which would have to have many features in common with ACTIVITY 1, or that 'Mr. Johnson' is using 'the man around the corner' as an INSTRUMENT in the course of ACTIVITY 1. This ambiguity is part of our story, which continues with the elements of ACTION 4: MOTIVATION not in text; to be inferred
PLANNING to OPERATION
EXECUTION to ACTION 4.1 and 4.2
EVALUATION
0
OPERATION (from PLANNING 4); enter linguistic system Interpersonal: statement, indicative, modalized... ideational: action, agent-centred, +affected, overt affected, textual: theme 1st. inherent role... realization: 'I'll shoot that cat on sight.'
From EXECUTION 4: A C T I O N 4.1 load gun MOTIVATION not in text; to be inferred
PLANNING not in text; to be inferred
PURPOSE/GOAL gun be loaded EXECUTION load up with nails and dynamite
EVALUATION not in text; to be inferred: 'successful'
From EXECUTION 4: A C T I O N 4.2 wait for cat MOTIVATION not in text; to be inferred
PLANNING not in text; to be inferred
PURPOSE/GOAL be close to cat EXECUTION wait for cat
EVALUATION not in text; to be inferred: ± successful
THE INTERACTION OF LANGUAGE AND MUSIC
403
The final sentence of verse 3 leads us to the conclusion that the EVALUATION of ACTION 4 can be nothing but negative; in fact, it can not have been performed by the AGENT of ACTION 4. The following repeat of ACTION 2 leads on to the end of the story (verse 4), Here we run into some interesting difficulties: firstly, we do not find an AGENT men tioned, which means that we may either infer 'Mr Johnson' to be the ulti mate AGENT, or else decide that we simply do not know the AGENT. If we choose the latter, we run into the further complication that we do not find any NEED or OBJECTIVE mentioned. In that case, we have to inter pret it as an ACTIVITY and as ACTIONS with all but the EXECUTION unknown, or else not as an ACTIVITY, but as an event, or a series of events happening without agency, much like natural disasters. Both interpretations are, as far as I can see, valid, with the second one being weaker in the sense that less structure has to be inferred. If we choose the event-interpretation, verse 4 is nothing but a scenario, a SITUATION, which seems to put an end to our endless re-entry of ACTIONS. To everybody's surprise, however, Action 2 does re-occur, in spite of the SITUATION created in verse 4 and with all possible AGENTS of further ACTIONS against our cat being eliminated. That makes our cat a symbol of power and perseverance, and, at the same time, makes human efforts a symbol of futility. So far, we have been looking at the dynamic aspect of what is going on in our story, at ACTIVITIES, ACTIONS, and OPERATIONS. We have been operating with units that are states and processes (NEED, ACTION etc.) or mental images of states and processes (GOAL/PURPOSE). If we look at the ACTIVITIES going on, as if they were frozen for a moment, we get what is known as SITUATION, that category so essential for the mean ing of any linguistic utterance, and, indeed, human behaviour in general. The SITUATION, the background to ACTIVITIES, can be seen as a con cept, a matrix, with the following 'slots' to be filled in every given instance: AGENT- ONGOING PROCESS-MATERIALS/STRUCTURES acted upon-INSTRUMENTS- PRODUCT. 1 Thus, ACTIVITY and SITUA TION are nothing but two aspects of the same reality, looked upon from different angles. How are the slots of SITUATION filled in our story? It will have to be tested in the textual analysis (section 2) if and how the categories mentioned in section 1 are reflected in the language used to tell the story (see Steiner forthcoming for a brief introduction of the con cept of SITUATION used here). It should by now be obvious that while
404
ERICH STEINER
Verse 1
instruments
Process
Materials
Johnson (in states
give away
cat, man
man
go away
man
the cat
come back
the cat
the cat
think
Johnson
(try to) kill
cat
give command
cause state 5
cat, boy Action = boy throw cat in water cat, weight
they
search
river, boy
man around the corner
(try to) kill project Actionx
cat
gun
see refrain
Action.
language
gun self man around
nails, dynamite
we
load up wait find
ment. image of Actionx = man around the corner shoot cat gun loaded not in text we have man man around...
not given
fall
bombs
not given
demolish
world
cat be with man a
= state 3 a
Refrain
2
3
4
Product
Agent
man a be away cat be with J. = states 1,2 cat be a gonner = ment. image of state 4 boy, weight, river, $5, language language
see refrain
cat with boy to be inferred
weight be around cat's neck = state 5 not given
bombs
bombs are down world is gone world is demolished
section 2 of this paper could stand on its own as a sort of 'exercise in linguis tic description', it derives its justification from an interest in the function which the text serves, and only through this in the formal properties of the text.
THE INTERACTION OF LANGUAGE AND MUSIC
405
2.0 The Text After analysing the ACTIVITIES portrayed in our folk ballad, let us concentrate on the question of what the structure of the text is like and howit achieves its function of placing the ACTIVITIES described in section 1 before our 'mental eyes'. We shall be looking at certain elements of struc ture in the text, particularly at certain features of the clauses making it up, in order to show what each contributes to the weaving of the text. 2.1 Subject Theme The Subject Theme of a clause extends over the S-element in clause structure, which, filled by an NG or a clause, is analysable in terms of the Participant Roles realized in it. The lexical items expounding S-elements in the clauses of a text are analysable in terms of lexical cohesion. Looking at the analysis in Figure 1, we can establish the percentages of certain Participant Roles in the whole text (the Appendix contains the pro cedure which is used here for assigning semantic structures to clauses, i.e. types of process and participant roles): Participant Role
No.
AGENT-ATTRIBUANT AGENT 3RD PARTY AGENT SENDER PROCESSOR SIMPLE-ATTRIBUANT AFFECTED AFFECTED-ATTRIBUANT
3 9 6 2 1 8 2 3
Counting the numbers in the table above does not tell us very much. Instead, it is helpful to look at the clustering of certain Participant Roles in certain sections of the text, and at the relation of these sections to the ACTIVITIES reflected in them: The first verse has a lot of variation as far as Participant Roles under Subject Theme are concerned, with THIRD PARTY AGENT (2), AGENT-ATTRIBUANT (2), SIMPLE ATTRIBUANT (2) and AGENT (1) all being found. If we remind our selves that the two AGENT-ATTRIBUANT roles stem from embedded clauses and if we remember that the first two lines represent the MOTIVA-
406
ERICH STEINER
Difficult decision between AGENT ATTRIBUANT and AGENT. Does 'leave' take an AFFECTED or a LOCATIVE as 2nd role? Difficult decision. Figure 1: Subject theme
TION stage of ACTION (3RDPARTY)-AGENCY of The refrain is represented in that of AGENT (4 out of
1, we can see the dominant pattern of the acting AGENT (Mr. Johnson) emerging. lines 8-13, and there the dominant pattern is 6) with lines 9,10 containing some 'aside'-
THE INTERACTION OF LANGUAGE AND MUSIC
407
thoughts, possibly meant to be taken from the PLANNING stage. Verse 2 (14-21) has the clearly dominating pattern of 3RD PARTY AGENT which, at first sight, seems like a straightforward realization of ACTION 3 with its threefold sub-ACTIONS: Mr. Johnson tries to achieve his GOAL by using somebody else. A straightforward deduction from the structure of clauses to the structure of ACTIVITIES is, however, decep tive: Of the 3RD PARTY AGENT roles (lines 14, 16, 17, 18), only line 14 represents the '3rd party-agent' of the ACTION, Mr. Johnson. Lines 16,17 represent the direct AGENT in terms of ACTIVITY, 'the boy'. Line 18, again, represents Mr. Johnson as 3RD PARTY AGENT in terms of the semantics of the English language; yet, in ACTION 3.3. which is narrated in this line, Johnson is the direct AGENT (cf. section 1). The SENDER of line 15 is a straightforward realization of the EXECUTION of 3.2. Lines 20 and 21 follow a different pattern and, indeed, do not represent ACTION 3. In verse 3, AGENT is the dominant pattern under Subject Theme, with line 22 (SENDER) containing a straightforward realization of the OPERATION entered from PLANNING under ACTION 4. Lines 27,28 represent a different pattern from the rest of the verse, again not belonging directly to ACTION 4, As in the case of verse 2, the AGENTS of clauses do not have to be the AGENTS of the ACTIONS represented in the same lines (.26). Verse 4 has a somewhat conspicuous pattern as compared with the remainder of the text, featuring predominantly SIMPLEATTRIBUANT, AFFECTED-ATTRIBUANT and AFFECTED. There is not a single AGENT-roie to be found. And, indeed, there is no ACTION openly described in these passages, as we have seen in section 1. As the analysis under PROCESS will show, we do not even find AGENTCENTRED verbs in this passage, except in line 34, which in this very con text could almost receive a SIMPLE-ATTRIBUANT interpretation, so the syntactic structure does not even hint at hidden AGENTS. Ail of this partly explains the special 'feeling' of verse 4, as well as the ambiguity in terms of the analysis of ACTIVITY. Looking back on Subject Theme as a whole in the case of this text, we notice a clear lack of SENDER- and PROCESSOR-roles (3 out of 34), which is a straightforward, though by no means necessary consequence of the fact that PLANNING and OPERATION stages are, for most of the time, not narrated in the story, nor are the EVALUATION stages. The role of lexical cohesion will have to be discussed under Thematic Progres sion'.
408
ERICH STEINER
2.2 Thematic Progression I shall not discuss the concept of Thematic Progression itself here (Danes 1970), as I have done this elsewhere (Steiner 1983). Let us only look at the types of Thematic Progression that we find in the text, and try to find out how they contribute to the meaning of the text.
* gaps which are wider than 1 line are especially marked ** ambiguous Figure 2: Thematic
progression
THE INTERACTION OF LANGUAGE AND MUSIC
409
An immediately obvious phenomenon indicating types of Thematic prog ression is lexical cohesion between fully lexical items in Subject Theme position (cf. Figures 1 and 2). In the refrain (lines 8-13), lexical cohesion establishes a series of continuous Themes by repetition, the definite article in the first instance indicating that the Subject Theme of line 8 is linked, as a Linear Progression in this case, to a theme that has occurred before. The second obvious instance of lexical cohesion between fully lexical items is to be found in the last verse (lines 29-34). Firstly, we have to note the repetition in lines 29/30, as well as the hyponymy-relation of the two lexical items to a third item 'weapon'. This constitutes what is known as a Derived Theme. Next, the Themes in 31-33 are 'instances of, or 'tokens' of a concept ('type') 'country', which in turn stands in a 'part-whole'-relationship with the items expounding the NG in 34, so that we have a case of Generalized Theme in lines 31-34 (cf. Steiner 1983: 18 for this term). Turn ing to pronominal chains, we immediately see the important role of anaphoric reference in this text. The pattern is as follows. Verse 1 begins with a fully lexical item as Subject Theme, obviously a New Theme. All the following lines, except lines 3 and 7, continue this Theme by anaphoric reference. The two cases of Linear Progression occur in embedded clauses and are expounded by relative pronouns. The refrain again displays the basic pattern of a Continuous Theme. We have to note, though, the interesting case of possible ambiguity in anaphoric reference in line 10, between a reference to 'Mr. Johnson' and 'the cat'. This can only be resolved with recourse to the analysis of ACTIV ITY. Verse 2, again, starts with a New Theme (lines 14-21), which, how ever, is only New with respect to the refrain preceding it. The item 'he' establishes anaphoric reference to the Subject Theme of line 1, which again cannot be established by linguistic criteria alone, but only by also referring to the concepts of ACTIVITY and SITUATION, in the light of which this interpretation is plausible. The pattern of Thematic Progression in verse 2 is different from that in verse 1, displaying two conspicuous gaps in lines 16 and 18. These can be explained by the fact that ACTION 3 contains embedded OPERATIONS (utterances). The AGENT of the higher level ACTION 3 is maintained throughout the verse in terms of the analysis of ACTIVITY; the linguistic analysis, however, bears evidence of a shifting agency, because linguistic OPERATIONS entering into a constituent structure in terms of the analysis of ACTIVITY, have to be reflected (narrated) in linear order in the medium of language. Note, by the way, that the reference in line 18 is
410
ERICH STEINER
ambiguous between referring to the Subject Theme in line 17 or in line 15. As it happens, this ambiguity cannot be resolved at all within the present framework, which means that the story is ambiguous at this point for our analysis as well as for the listener/reader. Lines 20 and 21 display the same New-Linear pattern which we shall encounter in lines 27/28. The function in the text is the same in both instances: to induce an EVALUATION in the hearer or listener. Line 22 introduces a new Theme with respect to the preceding text. In this new Theme, the AGENT of ACTION 4 is encoded. Lines 22-26 are closely linked by Thematic Progression, and indeed, it is one ACTION that is presented here. Lines 27/28 with their sequences of New-linear have a parallel function to lines 20/21. We have already noted the conspicuous pattern of Thematic Progression in lines 29-34: part of the feeling of a 'climax' in verse 4 derives precisely from this type of Thematic Progression. Let us finally in this section discuss one methodological question. With regard to Subject Theme, what is the meaning of 'same' and 'dif ferent'? There seem to be three ways of giving an answer. We might only count 'sameness' or 'differentness' of lexical items. This, however, leads to two kinds of false conclusions. On the one hand, we would establish lexical cohesion for Themes that are not in fact referentially identical in the ACTION, i.e., we might get cohesion in our analysis where there actually is none, especially in the case of pronominal chains. On the other hand, we would fail to establish cohesion for Themes, just because they are expounded by different lexical items. We could follow a second line of approach in using cohesion in dis course in addition to the narrow procedure just outlined (cf. Halliday and Hasan 1976). This would give us the means to take into account pronouns and deixis, at least in anaphoric and cataphoric use. In a third type of approach to 'sameness' and 'differentness', we could try to take into account reference to the elements of the SITUATION (cf. section 1) or to the mental representation of such SITUATIONS (cf. for useful discussions of such an approach: Johnson-Laird 1980, Brewer and Lichtenstein 1982). This, admittedly, brings in a concept which is as yet dif ficult to control, but yields a very rich description. What is more, one does not really seem to be able to consistently avoid it in establishing exophoric, and, to a considerable extent, even anaphoric and cataphoric reference. For an analysis such as the one we are engaged in here, the third line of approach should be followed, as the first and second lines would seem to impoverish the analysis.
THE INTERACTION OF LANGUAGE AND MUSIC
411
2.3 Process Looking at the analysis of types of 'Process' expressed by the Verbal Groups in our text, we can see a pattern which again makes good sense in relation to our analysis of ACTIVITY (cf. Figure 3).
* Difficult decision between ACTION and LOCATIONAL ** The preposition/adverb is here considered to be part of the meaning of the process (phrasal/ prepositional verbs). Tests: questioning of constituents, clefting Figure 3: Process
412
ERICH STEINER
Each of the verses as well as the refrain are characterized by a domi nant type of process, with the possible exception of the first verse: AGENT ONLY in the refrain, LOCATIONAL (3RD PARTY AGENT) in verse 2 (see Willis and Jeffries 1982 for a useful discussion of the role of caseframes in determining 'type of process', LOCATIONAL in particular), AGENT-CENTRED in verse 3, and CLASSIFICATORY in the final verse. The structure of the first verse is less easy to characterize, which, however, is to be expected, as it includes an account of NEED 1, realized as MOTIVATION of ACTION 1. However, we can see the pattern of dominant AGENT-CENTRED and ASSOCIATIVE (3RD PARTY AGENT) processes. Remember also that line 7 reflects an embedded ACTION. The final verse, too, has a conspicuous pattern: CLASSIFICATION dominates, LOCATIONAL (AFFECTED-ATTRIBUANT) coming next. Both of these types of process are not a straightforward realization of any ACTION, which again explains the difficulty of interpreting this verse. Even more so as the pattern over the preceding verses provides a kind of climax with respect to directness of ACTION towards a GOAL, and AGENT on MATERIAL in terms of the SITUATION: dominance of direct AGENT in verse 1, the MATERIAL (cat) in itself, however, remaining unchanged; dominance of AGENT ONLY in the refrain; domi nance of 3RD PARTY AGENT in verse 2; dominance of AGENTCENTRED in verse 3, this time, however, aimed at changing the MATE RIAL itself very drastically, rather than merely causing a change of loca tion. In this context, the final verse provides a definite anti-climax with not a single AGENT expressed in the actual clauses. Note that this anti-climax in one aspect of 'inter-verse relationships' is in contrast with the 'verseinternal' climax in verse 4 brought out by our analysis of Thematic Progres sion. The impression of climax from 'movement' to 'caused movement' to 'direct action' is reinforced by the dominant sense relationship of 'direc tional opposites' (Lyons 1977, vol. 1: 281ff.) in the first half of our text. If we look at all the verbs in lines 2-17, we notice that nearly all of them seem to be - hyponyms of a superordinate term like 'movement + direction'. This pattern is practically extinct in the second half of the text, where we only occasionally get related verbs like 'shoot' and 'load' (23,24; necessary condition) or 'search' and 'find' (20-28; purpose?). Looking back at what has been said in 2.3, one should again point to the fact that semantic categories like LOCATIONAL, or AGENT do not
THE INTERACTION OF LANGUAGE AND MUSIC
Figure 4: Rheme
413
414
ERICH STEINER
necessarily mirror the analysis of ACTIVITY in a straightforward way. The two analyses, however, have to be related to each other, if they are to make sense, i.e., the story told cannot be said to exist without reference to the texts through which it gets realized and the structure of the text does not make sense without recourse to the structure of the ACTIVITY and SITU ATION which it places before our eyes. 2.4 Rheme The analysis of the semantic constituents over which Rheme extends is given in Figure 4. Remember that 'Process', although belonging under Rheme, has been analysed separately. We do not have to interpret the analysis of Participant Roles very extensively in this context, as most of the results in this area are fully pre dictable from what has been said about Subject Themes, Thematic Progres sion and Process; if we know the first Participant Role and the type of pro cess of a particular clause, we can predict the second and third roles, except in the case of covert roles. Quite naturally, therefore, our analysis of Rheme mirrors the analysis of ACTIVITY in the same way as the analysis of Subject Theme, Thematic Progression and Process.
THE INTERACTION OF LANGUAGE AND MUSIC
415
Let us therefore turn to the non-predictable constituents under Rheme, the CIRCUMSTANCES. The preceding figure contains the options in CIRCUMSTANCE which we have found necessary in the tex tual analysis so far (many of these options can be found in Quirk et al. 1972). This is not the place to give detailed realizational rules. CIR CUMSTANCES are syntactically realized in Adverbial Groups, Clauses, Prepositional Groups or Nominal Groups. Not all CIRCUMSTANCES named in the network above have the whole range of syntactic realizations. Where the realization is a clause, certain classes of conjunctions ('Binders' and 'Linkers') will have to be stated; where the realization comes in the form of a Prepositional Group, certain classes of prepositions characterize certain CIRCUMSTANCES. Remember that within the semantic-syntactic model used in this paper, obligatory LOCATION, CLASSIFIER and ASSOCIATED are not regarded as CIRCUMSTANCES, but as inherent Participant Roles. So: (1) (2) (3)
Rudi went home Isidore treated her teddy bear badly Mr. Johnson gave the cat to President Reagan
do not have a CIRCUMSTANCE in their semantic structure. This is why the category 'local, directive' does not have a sample realization as CIR CUMSTANCE. Remember also that in cases where an Adverb or a Preposition is part of a verb phrase, rather than introducing or constituting a clause element of its own, it is not regarded here as either CIRCUMSTANCE or Participant Role, but simply as part of the verb (cf. 'come back', 'stay away'). The usual tests (clefting, front shifting, substitution) are applied to make this decision. Using the network above in the analysis of our Rhemes, we find the CIRCUMSTANCES almost exclusively concentrated in the second half of our text (lines 20-34). These are: TIME (8,23,29,34), MEANS (24), PUR POSE (20), MANNER (30), OTHER (14). Part of the explanation for this lies in the semantic structure of the whole text which displays the already stated predominance of the one-role processes towards the end, thus 'mak ing room' in the syntactic-semantic structure of clauses for CIR CUMSTANCES. In line 34, the Adjunct even directly helps to make the lack of the covert AGENT less obvious. Some of the Adjuncts may simply
416
ERICH STEINER
be due to the needs of rhyming (lines 29, 30, 34). This may not be very good poetic craftsmanship, but in writing in the rhyme form used here (AA, BB, CC . . .) it is sometimes difficult to avoid. The classification of the Adjunct 'with a five dollar note' in line 14 in terms of the network above is difficult, which is why we have labelled it as OTHER. If, however, we use SITUA TIONAL categories (cf. section 1), we see that this constituent does repre sent the INSTRUMENT in terms of our analysis of ACTIVITY. We can see in this case again that syntactic-semantic categories do not necessarily mirror SITUATIONAL categories directly. Looking at lexical cohesion under Rheme, we see that it functions more strongly verse-internally than between verses, which means that the ACTIONS established in the analysis of ACTIVITY are in their linguistic reflection bound together by internal lexical cohesion much more strongly than by lexical cohesion between the ACTIONS within one ACTIVITY. If we take into account the heads of elements of clause structure ('heads' of Nominal Groups, the 'apex' of Adverbial/Adjectival Groups, the 'heads' of the Nominal Group within the Prepositional Group; verbs will generally be dealt with under 'Process'), the following picture emerges (cf. also Figure 5 under E ). In verse 1, lines 2 and 4 are linked by the repetition of the item 'cat', and line 6 is linked to these by anaphoric reference. From line 6, another chain of anaphoric reference extends to lines 10, 14, 15, and 17. The repet ition of 'cat' is continued in lines 22 and 25. This pattern seems to contradict my statement about the predominance of ACTION-internal cohesion, but note that these + the item 'cat' under Subject Theme are not the only examples of inter-ACTION cohesion in the whole text. The explanation in terms of the analysis of ACTIVITY is easy. As in a text largely built on a transitive rather than an ergative basis the MATERIALS ('cat') in terms of the SITUATION are bound to be represented largely in syntactic-semantic AFFECTEDs or CIRCUMSTANCES, and as in this particular story the MATERIAL is the only relatively stable constituent of the SITUATION (other than AGENT, which will, however, in this text not appear in Rhemes very often), the lexical cohesion between lexical items realizing MATERIAL are bound to display the only significant inter-verse cohesion under RHEME. In the refrain, we find hardly any cohesion under Rheme, again not surprisingly, as we have already shown that the refrain is domi nated by AGENT-ONLY structures. For the same reason, lexical cohesion under Subject Theme is very strong in the refrain.
THE INTERACTION OF LANGUAGE AND MUSIC
417
Verse 2, again, displays internal cohesion by repetition (lines 14 and 20; lines 16 and 20) by a type-token relationship ('weight' and 'pound' in lines 18 and 19). The items 'river' (16,20) and 'boat' (17) share at least one significant feature, which is 'water'. Lexical cohesion between verse 2 and other parts of the text is only to be found in the pronominal chain of the item 'it' (lines 14,16,17,18), referring to 'the cat'. Strong internal cohesion is also a feature of verse 3, with repetition of the item 'cat' (lines 22,23,25) and with the presence of the co-hyponyms 'shotgun' and 'dynamite' (line 24; superordinate term 'weapon'). Both of these can be conceptually linked to 'shoot' (22) as INSTRUMENT to ACTION. As before, the internal cohesion is stronger than the external one. The items in verse 4 are internally linked by repetition (lines 29,34 by 'day'; lines 31, 32, 33 by 'gone'). If we remember the amount of lexical cohesion stated for this verse under 'Thematic Progression', we see the very strong pattern of cohesive ties in this verse. There is no cohesion between this verse and the rest of the text. This is as far as we shall go in our analysis of the elements over which Rheme extends in our text. Section 2.5 will indicate how the results of 2.4 and the preceding sections can be integrated in looking at the text as a whole. 2.5 Integration of the Textual Analysis Let us remind ourselves at this point that our textual analysis does not exactly follow the text as it is sung in two respects. Firstly, the analysis is done by clauses rather than by the lines of each stanza. This is an inevitable feature of the technique of analysis employed here, which is based on semantic features that are defined for the clause. Secondly, when the text is sung, the refrain (lines 8-13) is repeated after every verse, which does not, however, invalidate the statements made here, as the only thing added in this way is, from the textual point of view, an extra amount of cohesion. From the angle of the analysis of ACTIVITY, each recital of the refrain provides, via EVALUATION, the necessary entry condition for a new ACTION. Figure 5 brings together the results of the textual analysis done so far. In interpreting Figure 5, we shall deal with the most important features only. Figure 5 represents a type of span-analysis (cf. Grimes 1975: 91ff.) with a set of variables slightly different from the ones investigated by Grimes and with an almost complete restriction to linguistic criteria for
418
ERICH STEINER
identifying each occurrence of the variable in the text. Let us briefly discuss the variables in turn (see key to Figure 5 first).
Figure 5: Span analysis Spatial Setting (Set.spat ): 1 = context independent 3 = no direction
2 = context dependent 4 = +direction
THE INTERACTION OF LANGUAGE AND MUSIC
419
Event. (E.1proe-( '): 1= 3= 5= 7=
locational associative action, agent only mental process
2 4 6 8
= = = =
classificatory action, agent-centred verbalization action, affected only
Identification (I): 1 = agent-attributant 3 = 3rd party agent 5 = sender 7 = affected-attribuant 8 = 1st person (speaker) 10 = 3rd party entity
2 = simple attributant 4 = agent 6 = affected 11 = processor 9 = 2nd person (addressee lexical cohesion and Thematic Progress indicated by connecting lines
Event2 (E2(Rheme)): 1 = Associated 3 = affected 5 = classifier 7 = other role 9 = circumstantials lexical cohesion indicated by Temporal Setting (Set. 1 = context independent 3 = no direction
2 4 6 8
= = = =
location attribuant message phenomenon
connecting lines
): 2 — context dependent 4 = + direction Key to Figure 5
E 1 process(event/process): this variable is solely identified by the analysis by verb meaning in terms of TRANSITIVITY (cf. the Appendix for the procedure for determining this type of meaning). As a result, we get, in this case, 8 different types of process distributed over the text. The number of different types would be greater if we included more delicate features in the analysis, which, however, does not seem to be necessary in the present context. What we see in this column is that we do get dominant patterns for the verses and for ACTIONS reflected in them, but that at the same time the semantic type of process is not necessarily a direct or 'congruent' reflection of the corresponding cognitive category. ACTION 1 'change possession of cat', for example, is linguistically reflected in LOCATIONAL, ACTION and ASSOCIATIVE Processes. This general result will be confirmed in all
420
ERICH STEINER
the columns of Figure 5. E Rheme (event/Rheme): This variable is identified by classes of ele ments under Rheme. The same general remarks apply as under EI The additional information we get here derives from the analysis of lexical cohe sion. The E -column shows very nicely the interplay between interACTION and intra-ACTION cohesion. I (Identification/key participant): applying our analysis of Transitivity, we get the distribution of key participants (cf. 2.1) left of the comma. On the right of the comma, we have marked the communicative roles of these participants. In addition, for certain types of lexical cohesion and Thematic Progression on the right of the comma, we have marked the communicative roles of the participants. In addition, certain types of lexical cohesion and Thematic Progression are entered, again lending support to our analysis of ACTIVITY. SetTemp (Temporal Setting): we have not discussed the question of temporal setting of (parts of) the story so far. In Figure 5, the following dis tinctions are made:
Let us have a brief look at the features and their main realizational structures. The 'point of reference' can be 'implicit', i.e., not mentioned in the clause at all. In that case, there will be no realization in the particular clause under description at the moment. 'Context independent' points of reference are typically realized by mentioning a date giving a unique point or span of time. Examples are 'the 1st of June 1983' or 'since 1950'. Also typical are some Adverbs denoting absolute spans and quantities of time, like 'always', 'never', 'every day'. 'Context dependent' points of reference are typically realized by 'time CIR CUMSTANCES' denoting a span or quantity of time relative to the SITU ATION in which the communication or the event portrayed is taking place. Examples are 'later', 'now', 'recently', etc. In principle, the whole range of units that can fill the element of (Time)-Adjunct in clause structure is avail-
T H E I N T E R A C T I O N O F L A N G U A G E A N D MUSIC
421
able. The criterion for the + direction decision is whether the ACTION or SITUATION portrayed in the text at a certain point is located parallel in time to a reference point (-direction), or either before or after this refer ence point (+direction). The point of reference can be either the situation of the utterance itself, or a point of reference established in the text in a second SITUATION introduced in text (cf. Davies 1979 for the distinction between 'situation of utterance' and 'situation of performance'). Typical realizations are the tenses, certain Adverbs ('now', 'later on', 'soon' . . . ), certain conjunctions ('when', 'while', 'after', 'before' . . . ) and certain pre positions ('during', 'after', 'before'. . . ). Looking at Figure 5, we can see that the temporal setting of our story is relatively simple. The general pattern is, that points of reference are introduced at the beginning of each verse with the following structure being rather linear. Note that our analysis does not show a new point of reference for verse 2. Whereas the reader/listener will probably infer a new point of reference at this point, there is no realization of this in the linguistic mate rial of the text, and accordingly, it is not shown by our analysis. Note finally that 'time' and 'tense' have to be kept strictly separate (cf. for useful treat ments of 'tense' and 'time': Davies 1979; Polanyi 1982; Bybee and Slobin 1982). Set (Spatial Setting): the choices in this area are, at least as a first approximation, similar to the choices in 'time reference'.
We may need an additional 'no reference* option here for 'generic' situations. The same applies to the TEMPORAL SETTING-network.
A context dependent point of reference is given typically by deictic ele ments ('here', 'there' . . .), and, more generally, by LOCAL-CIR CUMSTANCES whose referent cannot be determined in isolation from the text. Examples are: 'behind me', 'in the house I've told you about', 'where I was born, everything is different', etc.
422
ERICH STEINER
A context independent point of reference is most obviously established by a proper name ('In London', 'above the Alps', e t c ) , but also by giving absolute units of measurement, like 'ten yards', '6 miles', etc. All indefinite Nominal Groups are context independent, but their point of reference can not be located in a unique way. Realizations of the '+direction' choice and the dependent system are: LOCATIONAL verbs ('come', 'go' , . .), ACTION verbs of the structure 'Main Verb +C m (Prepositional and Phrasal Verbs and some others), LOCAL-CIRCUMSTANCES, certain post-modifiers containing Preposi tional Groups, and some others. Realizations of the '-direction' option include stative verbs, 'herethere' deixis, Prepositional Groups with the Preposition expressing position rather than direction, and some others. The options in 'SPATIAL SETTING' express the location of the events expressed in the clause in relative or absolute terms. As Figure 5 shows, the spatial setting in our text is relatively simple and only of limited importance for the understanding of this text. This brings our textual analysis to an end. Many of the semantic-syn tactic devices with the help of which our text narrates ACTIVITIES 1 and 2 (cf. section 1) should by now have become more obvious to us. Note that our text narrates ACTIVITIES 1 and 2, but does not realize them, as under stood in the present framework. Parts of the ACTIVITIES are realized by OPERATIONS narrated in our text as direct speech, yet these are two dif ferent kinds of relationship. We have seen how conceptual categories of the theory of ACTIVITY (AGENT, INSTRUMENT etc.) may be directly mirrored in linguistic structures, i.e. in semantic elements bearing the same labels as categories of the theory of ACTIVITY, or, alternatively they may be cast into linguistic structures of an entirely different nature (cf. Halliday 1985, ch.10 on the question of 'Congruence'). Within our descriptive framework, linguistic-semantic structures and conceptual structures are related, but distinct. Take our analysis of ACTION 3. There is no linguistic indication of the superordinate ACTION in our text, so ACTION 3 'kill cat' is a product of inferences we have drawn. I suggest that it is our knowledge of possible structural configura tions, particularly about necessary/sufficient conditions for certain states of affairs, which leads us to draw such inferences. For example, neither of the EXECUTIONS 3.1.-3.3. nor a conjunction of them permits a non-trivial
THE INTERACTION OF LANGUAGE AND MUSIC
423
conclusion with respect to the rest of the story. So, what happens is that the hearer infers an additional ACTION ('boy throw cat into water'), which will provide the essential link. If we want to explain 'understanding', there is no possible explanation based on linguistic structures alone. This is an illustration of the fact that the structures we use in under standing the world as a whole, and, more specifically, ACTIVITIES, and the semantic structures of languages, are, while certainly interdependent in many ways, not isomorphic, not just a one-to-one reflection of each other. Let us, then, turn to the musical accompaniment in order to see what its contribution to our story is.
3.0 The Contribution of Music Turning to the relationship between language and music, we are enter ing territory that has been far less explored than any of the areas dealt with in this paper so far. Contributions known to me include David Cooke's The language of music (1959), Terry Winograd's 'Linguistics and the computer analysis of tonal harmony' (1968), and a number of shorter contributions on issues connected with the interrelation of language and music in one way or the other (cf. Fonagy and Magdics 1972; Nauck-BOrner 1982; McLean 1982; Thogmartin 1982; The Journal of Music Theory). The following is the musical score of our folk ballad. The words in this particular version are German, but this does not affect the structure of the musical accompani ment as far as we need it as a point of reference for the following discus sion. Our question now is: what meanings does music contribute to the meanings already described in sections 1 and 2 of this paper, and how is this achieved? The approach taken here is that of examining the building blocks of music, as far as they have a direct relevance to our purpose, and of describing what their functions and meanings are within music. The results will then have to be seen in the context of what we have already found out about the structure of our story on the ACTIVITY and textual level respec tively. On a more general level the question could be framed like this: what do language and music have in common, what separates them? For the time being, we shall use 'music' to refer to melodies and rhythms without words or in abstraction from them. We start at the obvious common basis of music
424
ERICH STEINER
THE INTERACTION OF LANGUAGE AND MUSIC
425
and language: sounds and their arrangements into patterns. 3.1 The Sound-basis of Language — Intonation and Phonology The information about the sound-basis of language, as far as we need it in our present context, is presented here in tabular form, largely following the presentation in Hailiday 1970. Table 1 has in the first column the phenomena we have to take into account here. The second column names the basic units we need in order to
Phenomenon
Units
Definition
Other relevant facts
Rhythm
Syllable Foot, Beat
Number of Syllables per beat or number of beats per foot, + place of salient syllable(s)
A beat may be silent (^), i.e., there may be no syllable on it.
Tempo
Beat, Unit of Time
Number of beats per unit of time (minute; second)
The tempo is roughly constant, each foot being roughly equal in length of time.
Tonic Prominence
Tone group, Tonic prominence is the place Possible structures: tonic ment of salient vs. non-salient Pre-tonic/tonic; or information in the tone group. Pre-tonic/tonic tonic Realization: stress, length pitch range
Melody
Tone group, Pitch
Contour of pitch movement over tone group
Structures: falling; rising; 1 combinations. Jumping; gliding; combinations.
Tone (primary)
Tone group, tonic, pitch
pitch movement over tonic
Structures: falling/high rising, or falling-rising (pointed)/low rising/fallingrising (rounded)/risingfalling (rounded)/falling plus low rising/risingfalling (rounded) plus low rising
Sound/ phoneme or: Phonematic
Articulatory/ Unit of sound described as auditory combination of features feature, unit of sound (C,V, sylla ble, word...)
Unit vs. Prosody
LANGUAGE: Table 1
Sample features: vibrating parts of vocal tract; kind and place of obstruction; pitch of sound; length of sound...
426
ERICH STEINER
Phenomenon
Meaning
Rhythm
Does not seem to be part of the grammar of English. Can signify emotions. Rhythm provides the necessary metrical background for the production of speech. It is consciously exploited in poetry.
Tempo
Same as under 'Rhythm'
Tonic Prominence
Distribution of 'given' and mew' information over the tone group
Melody/Tone
Realize options in M O O D and M O D A L I T Y
Sound/Phoneme
Distinguish Meanings. Build up morphemes, the minimal units of referential meaning, and separate them ('Grenzsignale')
LANGUAGE:
Table 2
account for the phenomenon given in the first column, while the third col umn gives a definition of the phenomena in terms of the basic units intro duced in column 2. Column 4 mentions other factors that are relevant for understanding the phenomenon under discussion. Table 2, again, lists on the left phenomena we are discussing. In the remainder of the table, the functions of these phenomena within the system of language are given. We should bear in mind that, while all the phenomena listed have to work together to constitute 'meaning' in language, the characteristic phenomena, i.e. those unique to language and to man, seem tó be the phonemes, or in alternative framework, 'phonematic units' (Palmer 1970; Steiner 1983: 105ff.) 3.2 The Sound Basis of Music—Intonation and Tonal Harmony We are now turning to the question of the meaning of music, starting by identifying the essential elements of which music as a process is com posed, thus emphasizing what is similar to language. These elements are given in table 3 in a parallel arrangement to that of table 1. So far we are still on safe territory in the sense that hardly anyone would dispute the statements made. All the terms used in Table 3 are reducible to the language of physics, and the definitions given can be fully operationalized in these terms. When we turn to the possible functions or meanings of these phenomena we enter less tightly defined areas.
THE INTERACTION OF LANGUAGE AND MUSIC
Phenomenon
Units
Definition
427
Other relevant facts
Rhythm
Bar, Beat
No. of beats per bar, + place of salient beat(s), or No. of notes/syllables per beat, + place of salient note/syllable
The definitions given define 'basic rhythm' and 'rhythm of the melody' respectively
Tempo
Beat. Unit of time
Beats per minute ( = Maelzel's Metronome)
Tempo is constant, each bar taking up equal time Exceptions especially marked (ritard., stringendo etc.)
Tonic Prominence
Phrase (4 or 8 bars), Tonic
Musical Phrase consists of tonic ( = salient part), + non-tonic.
Tonic is realized by stress, pitch, length, or any combination of these. There can be 'secondary' and 'tertiary' tonics.
Melody
Phrase, pitch
Melody is the pitch movement over one or more phrases. Pitch is relative, intervals are absolute
Melodies are composed of falling, rising, and level stretches; they are recognized as individual melodies.
Key
Sounds/Notes Intervals
A key is a system of sounds in a defined relationship of intervals, from which one is chosen at each point of the melody.
Each key is named after its absolute starting note The key defines boundaries for the possible structures of melodies.
Sound
Cycles unit of time
A sound is defined by the frequency of cycles per second.
Additional criteria: duration, vibrating material. 'Place of articulation' and 'kind of obstruction' are not normally used to differentiate.
MUSIC: Table 3
Tempo and rhythm provide a background, the canvas relative to which the other factors are foregrounded. Against this background, melodies are set. These in turn consist of phrases establishing 'given-new' patterns within the flowing melody, very similar to the 'given-new' pattern in the clause, as explained in Table 3. So far, our statements seem rather obvious, closely linked to general properties of human processing strategies within percep tion and, to that extent, few people would probably want to dispute them. For the same reasons, these statements are of only limited interest. Let us,
428
ERICH STEINER
therefore, look at what probably most people would regard as the candi date for the core meaning of music: melodies, tonal keys and their struc tures. In doing so, we are embarking on a discussion of the differences between language and music, which we can also identify by looking back at Tables 1-3: language utilizes Phonemes as building blocks for the construc tion of units of sounds to be classified as Morphemes, i.e. minimal units of lexical meaning. Music does not have Phonemes in this sense. On the other hand, music utilizes tonal keys as systems of sounds in certain defined relationships to construct its structures, the melodies. Language does not display any systematic use of tonal keys, at least not to any significant extent. In talking about the 'systems' of tonal keys and the 'structures' of melodies, we are now entering the area of 'tonal harmony'. There exists in our culture a tradition of the 'grammar of tonal har mony' which is no less ancient than the tradition of 'grammar' in language. This grammar of tonal harmony is taught to every student at a music department or conservatory. It defines all and only the acceptable melodies of the major/minor harmonic system. Recursive rules can be used to ensure 'syntactical' creativity. Using a Chomskyan-type of grammar, one would also want to make use of Transformational Rules, not so much for reasons of generative power, but for reasons of economy. And, as with language, one needs different grammars for different 'musical languages' (key sys tems), such as the major/minor system, 12-tone-harmony, harmonic sys tems used in different styles of Jazz, the tonal keys of ancient Greece (aeolic, doric etc.), the Persian key systems, and others. And again, as in the case of language, grammar can be described and also taught without taking account of meaning, the desirability of which approach, however, can be disputed on various grounds. For the present task, it is fortunate that one of the few formalized grammars of tonal harmony has been written within the systemic framework (Winograd 1968). His grammar generates structures of units on 5 ranks: Composition, tonality, chord group, chord, and note. The range of tonal harmony defined by his grammar is basically the same as the one I am referring to here: our classical major/minor har monic system with its more delicate intricacies left out. Winograd's grammar of tonal harmony has a status within an overall description of music which is comparable to the status of Halliday's 'Categories of the theory of grammar' (1961) relative to an overall descrip tion of language, i.e. it describes the possible arrangements of formal items.
THE INTERACTION OF LANGUAGE AND MUSIC
429
In the present context, however, we want to find out about the level of 'meaning', so we are looking for semantic networks, describing (part of) the meaning potential of music. We have already outlined some of the choices in 'textual meaning' in music. Let us now find out about the meanings of keys, tonal harmony, and, above all, melodies. The reader must be warned,
Feature
Realization
neutral + finality anguish 4- finality longing, + pleasure + finality stoic acceptance, tragedy joy pathos, neutral devilish inimical forces neutral, - finality anguish, - finality longing, pleasure, - finality mournfulness violent longing or aspiration + finality
Tonic Minor Second Major Second Minor Third Major Third Normal Fourth Sharp Fourth Dominant Minor Sixth Major Sixth Minor Seventh Major Seventh
Realizational rules (not complete)
430
ERICH STEINER
however, that I am approaching this question somewhat hesitantly, because while most people agree that music has a meaning, far fewer people will agree what the meaning of a piece of music is in any particular instance. In what follows, we are dealing with very intuitive terms, which, how ever, have had a long tradition in the teaching of musical composition. What follows is, of course, only valid for major/minor harmonics. In the following, I am giving in a system network notation the choices in (emo tional) meaning underlying sequences of sounds, such as are generated by Winograd's grammar. I am formalizing the insights gained by D. Cooke in his book The language of music (1959) in a consolidated network. Every selection expression from the network giving choices in meaning for intervals defines one note in a melody, relative to the note preceding it, rather than defining one chord in a composition. Viewed from the present perspective, chords will have to be chosen dependent on the notes of the melody. Relating this statement to Winograd's approach, we see that every note of the melody preselects a set of alternative paths through his five net works. At least two open questions have to be made explicit in this connec tion: firstly, the REALIZATION-column given after the network refers to intervals. Intervals, however, can be arrived at from two different direc tions: either between the notes of the melody, i.e. between elements of structure within one melodic line, or between any note of the melody and between the basic note of its accompanying chord. In the melody under description here (see above), the fourth note, for example, (g in bar 1) is in a 'major third' interval to the b preceding it. At the same time, it is the minor third of the basic tone of its accompanying chord e-minor. In this way, chords define the context, without which melodies have ambiguous meanings. The question of exactly how this interplay should be modelled within a Systemic framework is beyond the scope of this paper. Generally speaking, melodies preselect chords, or chords preselect melodies according to the rules of tonal harmony. The direction of this preselection depends on the working technique of the composer. Secondly, in our system network we choose the meaning of the Mor phemes of music. It remains to be seen how these form larger structures. These larger structures should not simply be regarded as the sum of the (meanings of) elements making them up, but rather as a higher unit of meaning characterized by the 'Morphemes' that are salient/foregrounded in
THE INTERACTION OF LANGUAGE AND MUSIC
431
Before going on to 3.3, it should be made clear that the statements made in 3.2 about the meaning of intervals should be taken as what they are: informed speculations and hypotheses rather than well established facts about the meaning of music. (Cf. Bernstein 1967: 15 for critical com ments on such an approach.) These are, however, speculations which are founded above all in what we know about the practice of composing and performing music. 3.3 Interaction Having established the common basis of language and music in sound and the meanings associated with it, and having isolated areas where lan guage and music are clearly different, by virtue of their exploiting one area each where the other semiotic system is relatively poor — Phonemes in the case of language and intervals in melodies in the case of music — we shall now have a final look at how these structural resources interact in our bal lad. In what follows, I shall assume a non-free mode of performance, that is to say, the tempo will be stable and there will not be many ritar dandi and stringendi (Cf. the musical score of the ballad at the beginning of 3). In its basic rhythm, that of a 4/4 bar, the music imposes a fixed pattern of 4 beats per bar, with beats 1 and 3 being salient. Note, that this does not fix the number of syllables per bar; it establishes only where the salient syllable will be in the unmarked case (the placement of stressed syllables/notes on non-salient beats is a well known feature of many types of music). The number of syllables per bar is only limited by comprehensibility, and can be quite different. If we compare, say, line 8 of our text with lines 9,10,11, we see that both are sung over 2 bars with line 8 containing 10 syllables, and lines 9,10,11 containing 13 syllables. The function of the basic rhythm is the establishment of a stable background against which any patterns in melody, tonic prominence, vol ume etc. will be clearly perceptible. The tempo is assumed to be stable. It has to be chosen with the aim of ensuring maximum intelligibility and of maximum coherence with the other aspects of meaning of the song. With regard to intelligibility, those bars which have to carry the greatest number of syllables have to be taken as a measure of the upper limit of the tempo. Above the bar, the next higher musical unit is that of the 'phrase', which corresponds in form and function to the tone group in language. And, as in language, where tonic prominence can be distributed over struc tures of ideational meaning in various ways, so in music parts of melodies
432
ERICH STEINER
can be phrased in different ways. From the experience of performing the song under consideration here a great number of times, to me the obvious way of phrasing it seems to be that of having one phrase (tone group) extending over two bars, with beat 3 of the second bar constituting the tonic and beats 1 and 2 of bar 2 leading up to it. In other words, we get a simple tone group in linguistic terms (pre-tonic/tonic), where the tonic is marked by stress or volume (stress can be realized by intensifying the tone without necessarily raising the volume). So, we get a fixed pattern of 'given-new', which, however, may be broken in lines where the text suggests it, as in lines 1, 22, or 29, where new7 AGENTS are introduced and the realization of two phrases/tone groups in one line seems to be adequate. This way of phrasing leads to a range of conspicuously foregrounded lexical items in the text (cf. Figure 6). 1 2 4 6
Old Mr.Johnson home away away
own
8 10 13
the cat gonner stay away
day
14 17 19 20
note boat pound drowned
22 24 25 27
The man around the corner dynamite around found
sight
29 30 33 34
The A-bomb way USA day
day
Figure 6: Lexical items of the 'tonic' of the musical phrase
THE INTERACTION OF LANGUAGE AND MUSIC
433
There are at least two alternatives to the way of phrasing suggested here. In the first, we could have one tone group per bar with prominence following the basic rhythmical pattern of the bar. This would mean tearing the ballad to pieces, making phrases too short and the pattern tedious. This way of performing is quite frequently to be found at a stage when one is still learning to perform a particular piece, and one still has technical difficulties with it. In the second alternative, we could have one tone group over four bars with prominence (tonic) in bar 4. This, however, is to be rejected because of the large amount of information which listeners would have to process within one phrase. Let us next turn to the phenomenon of 'tone' which is, to my knowl edge, not known under this name in music. There seem to be clear signs of its existence, though: when, for example, two jazz musicians perform a 'question and answer sequence' in improvisation, they will exploit, at least very often, exactly the phenomenon of 'tone', i.e. the melodic movement around the tonic of a phrase. The same seems to apply to many, though not all, the 'question and answer' sequences in double concerts of classical music. I think, therefore, that there are good reasons for trying to take it into account in our folk ballad. There is no pitch movement on the tonic syllable of our melody (beat 3 of bar 2 of each phrase). If we only looked at the tonic syllable or beat, this would always give us tone 1, signifying neutral statements. As, how ever, in musical phrasing the tonic is normally 'prepared', i.e. 'played towards' (realized in pitch/volume/stress); as, in other words, the 'tonic' will be a group of beats rather than just one beat, we have to observe not only beat 3 of bar 2, but rather the whole movement in bar 2. In this case we get: Phrase A Phrase
tone 4 = falling - rising tone 5 = rising - falling
In the system of English intonation, tone 4 carries the meaning of, roughly, 'statement, expressing reservation or contrast or personal opinion offered for consideration'; tone 5 that of 'statement, asserting or expressing some other form of commitment' (Halliday 1970: 26). The only tone group breaking the constantly alternating pattern of tones 4 and 5 is the final one in the refrain, which is 'falling' (tone 1), signifying an untempered, final statement. This can be easily interpreted as it is always this line which, lead ing up to an EVALUATION on the part of the hearer as well as leading one to infer an EVALUATION on the part of the acting AGENT, termi-
434
ERICH STEINER
nates one ACTION. It is, of course, a somewhat problematic step to take over the meanings of, in this case, English intonation and claim that music will be interpreted in the same way as far as this area of meaning is concerned.. It seems to be fairly clear that the meanings of music in this area are not universally the same, as anybody will witness who is trying to share a musical experience with somebody belonging to some distant culture. On the other hand, the meanings carried by intonation in music are not bound to cultural bound aries in the exactly same way as those of a language. There will probably be little difference in interpretation among people from, say, Western Euro pean countries and cultures. It is in the context of this reservation that my interpretation of 'tone' in music should be understood. In our ballad, then, music provides a wave like pattern of 'reservation-commitment', constantly alternating (in a basic context of 'statement'), and being only broken by the 'firm' tone 1 phrase which makes the return of the cat the only stable and safe thing in the whole ballad. Let us turn to the melody then, bearing in mind the network for the meanings of intervals. Phrase A carries an overall falling melody, which is 'stepping' rather than 'gliding'. With the reservations made above, I suggest that in this case again the meanings from the English intonation system are carried over into music, at least for a British listener. These would be 'cer tainty; polarity known'. Phrase with its basically rising melody signifies 'unknown polarity', i.e. uncertainty. Again, we get a wave-like pattern of constant alternation, this time between 'certainty' and 'uncertainty'. On the background of this general pattern, let us look at the 'Morphemes' of the melodies, the intervals. Leaving the 'up-beat' out of account for the moment, phrase A consists of 1 major third, 2 minor thirds and 3 major seconds. Phrase consists of 1 minor third, 2 major seconds and 1 major second. It would be wrong just to add up the meanings of these 'Morphemes', just as it would be wrong in language to define the meaning of a clause as the sum of the meanings of its Morphemes. Instead, we look at the salient points in these melodies and at the relations into which they enter. Using numbers to signify 'distance from basic note of key (e)', we then get the following structures for the melodies: Phrase A: Phrase
:
5-4-3-2 1-2-3-2
The sum of these intervals then gives us for phrase A the meanings of
THE INTERACTION OF LANGUAGE AND MUSIC
435
'pleasurable longing + anguish' (2 major + 1 minor second). This combina tion of intervals constitutes a descending minor scale, which, as one basic melodic element, is usually taken to express 'grief, sorrow' (Cooke 1959: 133). Yet, note how the melody is harmonized: instead of, say, an e - 7 - e -B7606' - e accompaniment, what we get is e - D - - 76 . (Capital letters signify 'major', small ones 'minor'.) This leads us through major keys while still being basically minor. So, our 'g' in bar 1, which would be a 'tragic' minor third according to the first harmonization, becomes the 'neutral' dominant of C-major. The other notes change their meaning similarly in this context, so that the melody as a whole receives a flair of optimism. The final F # leads us from a 'neutrally-pessimistic' assertion to another phrase of message, in minor again. Line B, in isolation, signifies desperate emotions judging by the indi vidual intervals and overall key (cf. network for intervals). Yet, again, the harmonization provides a context for the intervals of the melody which pro duces a somewhat different range of meanings in the end:
The seconds and thirds of the minor (e) scale are harmonized as major third and dominant of major keys, while the final leads us back to a darker minor. Note a typical instance of 'plurifunctionality' here: while the second bar of phrase constitutes an open ending as a realization of melodic structure (ending in the dominant), its 'tone' signals assertion, i.e. rather a closed ending. 'Plurifunctionality' is as typical of music as it is of language. The rhythm of our melody is dominated by punctuated notes. In a rather fast tempo, this rhythm is normally chosen as a sign of liveness, very often in dances. The syncopation at various points of the melody (cf. bar 1), in the actual song occurring more frequently than this score suggests, gives the melody an additional element of lightheartedness. tion, let us look back at what we have said in 3.3. There is one general point that seems worth making because it relates to what is often assumed to be the general character of music as semiotic system. We have seen that the sound-basis of language and music is the same to a large extent, but that there is an important difference in the way 'Phonemes' are utilized in lan guage and melodies and tonal harmony are utilized in music. Might this not
436
ERICH STEINER
be an indication that these two systems emphasize different types of mean ing? Systemic Linguistics has always emphasized that language should not be seen as reflecting cognitive meaning only, in its 'intellectual function', but that language, because it has evolved to serve a whole variety of func tions, also has a variety of meanings. This has always seemed to me to be an important insight on which to base a study and a science of language. If we make a comparison between language and music, however, I think that in both of these systems there is one prominent area of meaning, relative to the other. Seen in this context, language does emphasize cognitive meaning, whereas music does emphasize emotional meaning. Each of these types of meaning does not have an isolated existence, though, and this is why, in spite of the respective emphases, emotional meaning is always present in language, and cognitive meaning is always present in music. This particular feature of the two systems under investigation is nowhere more obvious as in what we have been looking at here: a story, narrated in its perhaps more adequate form, the form of words set to music.
4.0 Conclusion We have now succeeded in tearing to pieces, in fragmentating some thing which in actual life exists only as a whole. We have been looking at actions as a hierarchy of units of activity; yet, when observed, these actions would be nothing but a continuous flow of activity. We have been looking at a text as having constituents, which we would hardly ever think of as such in the natural business of processing; yet, when perceived, this text, too, is nothing but a continous whole. We have finally separated all this from its music, that part which makes it all so lively, intense and interesting to see and hear. If a speculation is permitted, one might say that we have in a way repeated the history of these component parts as they often appear today in their manifestations on a larger scale, as activity, thought, and art, dis sociated from each other. This is when we get mindless and joy-less work, as on the assembly line; this is when we get thought dissociated from prac tice and feeling, as in the solipsistic theories of science or in the ivory-tower ideologies of some scientists; this is when we get art that seems in danger of losing 'content', joy and purpose, as in 'art for art's sake'. However, as it is to be hoped that in the end this schizophrenia in our
THE INTERACTION OF LANGUAGE AND MUSIC
437
lives will finally be resolved in a re-union of the component parts, each part more fully developed than it was before the separation thus making the new unity more enjoyable, so the analysis undertaken here derives its justifica tion only from the hope that it can make our appreciation of the ballad, when we finally hear it again, more thorough and more vivid than it was before the analysis. If it cannot achieve this, it might as well not have been carried out.
ACKNOWLEDGEMENTS I am very grateful to the following colleagues for their comments and criticisms; Johannes Burgard, Roger Charlton, Robin Fawcett, Martina Horn, Martin Montgomery and particularly my colleagues in the Linguistic Colloquium of the English Department of the University in SaarbrOcken. All of the remaining mistakes and weaknesses are, however, entirely my responsi bility.
NOTES 1.
This will often describe a 'generic' rather than an 'individual' state of affairs; cf. von Wright 1972: 39ff.
438
ERICH STEINER APPENDIX
A method for assigning semantic structures to clauses
THE INTERACTION OF LANGUAGE AND MUSIC
439
REFERENCES Beaugrande, R. de and W. Dressier. 1981. Einführung in die Textlinguis tik. Tübingen: Niemeyer. Bernstein, Leonard. 1967. The joy of music. New York: The New Ameri can Library. Berry, M. 1975. An introduction to systemic linguistics: 1 structures and systems. London: Batsford. . 1977. An introduction to systemic linguistics: 2 levels and links. Lon don: Batsford. Brewer, W.F. and E.H. Lichtenstein. 1982. Stories are to entertain: a structural-affect theory of stories. Journal of Pragmatics 6,5-6:473-486. Bybee, J.L. and D.I. Slobin. 1982. Rules and schemes in the development and use of the English past. Language 58,2:265-289. Caudwell, C. 1973. Illusion and reality. London: Lawrence and Wishart. Cooke, D. 1959. The language of music. London: Oxford Universtiy Press. Danes, F. 1970. Zur linguistischen Analyse der Textstruktur. Folia Linguis tica 4:72-78. . 1970. FSP and the text organization. In F. Danes, ed., Papers on functional sentence perspective. Prague: Academia. Davies, E. 1979. The semantics of syntax. London: Croom Helm. Fawcett, R.P. 1980. Cognitive linguistics and social interaction. Heidel berg: Groos. . 1983. Language as a semiological system: a reinterpretation of Saussure. In J. Morreall, ed., The Ninth LACUS Forum 1982. Columbia, S.C.: Hornbeam Press. . 1987. The semantics of clause and verb for relational processes in English. In M.A.K. Hailiday and R.P. Fawcett, eds., New develop ments in systemic linguistics. London: Frances Pinter. Firth, J.R. 1950. Personality and language in society. In J.R. Firth, Papers in linguistics 1934-1951. London: Oxford University Press. Fonagy, I. and K. Magdics. 1972. Emotional patterns in intonation and music. In D. Bolinger, ed., Intonation. Harmondsworth: Penguin. Grimes, J. 1975. The thread of discourse. The Hague: Mouton. Hailiday, M.A.K. 1961. Categories of the theory of grammar. Word 17,3:241-291. . 1967-68. Notes on transitivity and theme in English. Journal of Linguistics 3,1:37-81 part 1; 3,2:199-244 part 2; 4,2:179-215 part 3.
440
ERICH STEINER
. 1970. A course in spoken English: intonation. London: Oxford uni versity Press. . 1985. An introduction to functional grammar. London: Edward Arnold. Halliday, M.A.K. and R. Hasan. 1976. Cohesion in English. London: Longman. . 1980. Text and context: aspects of language in a social semiotic perspective. Tokyo: Sophia University, Linguistic Institute for Interna tional Communication Sophia Linguistica. Halliday, M.A.K. and R.P. Fawcett. (Forthcoming), New developments in systemic linguistics. London: Frances Pinter. Ikegami, Y. 1981. 'Activity'-'Accomplishment'-'Achievement' — a lan guage that can't say T burned it but it didn't burn' and one that can. Trier: L.A.U.T. papers. Johnson-Laird, P.N. 1980. Mental models in cognitive science. Cognitive Science 4:71-115. Leont'ev. A.A. 1975. Psycholinguistische Einheiten und die Erzeugung sprachlicher Aussagen. Berlin: Akademie Verlag. . 1978. Activity, consciousness and personality. Englewood Cliffs, N.J.: Prentice Hall. Lowe, I. 1983. Two ways of looking at causes and reasons, papers given to the Hatfield (BAAL) Conference on Discourse. Luria, A.R. 1973. The working brain. Harmondsworth: Penguin. Lyons, J. 1977. Semantics, Vols. 1 and 2. Cambridge University Press. Malinowski, B. 1935. Coral gardens and their magic. New York: American Book Company. McLean, M. 1982. The 'rule of eight' and text/music relationships in tradi tional Maori Waiata. Anthropological Linguistics 24,3:280-300. Monaghan, J. 1979. The neo-Firthian tradition and its contribution to gen eral linguistics. Tübingen: Niemeyer. Nauck-Börner, . 1982. Struktur und Bedeutung in der Musik, der Beitrag Peter Faltins. Zeitschrift für Semiotik 4,1-2:139- 150. Palmer, F.R., ed. 1970. Prosodiac analysis. London: Oxford University Press. Parret, H. and J. Bouveresse, eds. 1981. Meaning and understanding. Ber lin: de Gruyter. Polanyi, L. 1982. Linguistics and social constraints on storytelling. Journal of Pragmatics 6,5-6:509-524.
THE INTERACTION OF LANGUAGE AND MUSIC
441
Pribram, K. 1971. Languages of the brain. Englewood Cliffs, N.J.: Prentice Hall. Quirk, R., S. Greenbaurn, G. Leech and J. Svartik. 1972. A grammar of contemporary English. London: Longman. Schank, R.C. and R.P. Abelson. 1977. Scripts, plans, goals and under standing: an inquiry into human knowledge structures. Hillsdale, N.J.: Lawrence Erlbaum. Serafine, L. 1983. Cognition in music. Cognition 14,2:119-184. Steiner, E. 1983. Die Entwicklung des Britischen Kontextualismus. Heidel berg: Groos. . 1985. Working with transitivity: system networks in semantic-gram matical description. In J.D. Benson and W.S. Greaves, eds., Systemic perspectives on discourse. Vol. 1: Selected theoretical papers from the 9th International Systemic Workshop. Norwood, N.J.: Ablex. . 1983. The text as reflection of and instrument in activity: a proposal for the analysis of spoken discourse in children's talk. Paper given to the Hatfield (BAAL) Conference on Discourse and published as L.A.U.T. paper 1984. . (Forthcoming). Review: preface to a neo- Firthian linguistics. Forum Linguisticum. Thogmartin, C. 1982. Age, individual differences in musical and verbal aptitude and pronunciation achievement by elementary school children learning a foreign language. International Review of Applied Linguistics 20,1: 66-72. Vygotsky, L.S. 1962. Thought and language. Cambridge, Mass.: MIT-Press (original 1934). . 1978. Mind and society. Cambridge, Mass.: Harvard Universtiy Press. Willis, P. and L. Jeffries. 1982 Participant (case) roles and lexical analysis. Nottingham Linguistic Circular 11,2: 1-19. Winograd, T. 1968. Linguistics and the computer analysis of tonal har mony. (Reprinted in M.A.K. Halliday and J.R. Martin, eds., Readings in systemic linguistics. London: Batsford, 1981. Wright, G.H. von. 1972. An essay in deontic logic and the general theory of action. Amsterdam: North Holland.
INDEX acting subject, 394 action, 35, 43, 47, 123-124, 134, 206, 222, 239, 247, 250-251, 253-255, 266, 270-275, 278, 281, 284-285, 295, 305, 307-308, 379, 395-404, 406-407, 409412, 416-423, 434, 438 evaluation of, 124, 395-396, 399-403, 407, 410, 417, 433 execution of, 395-396, 398-403, 407 motivation for, 395, 400-402, 405-406, 412 planning of, 395-396, 398, 400-402, 407 rank of, 395-404, 406-407, 409412, 416-423, 434, 438 activities, 138, 256, 394-398, 403, 405, 407, 422-423 activity, 17, 250, 253-254, 256, 287, 394398, 400, 402-403, 405, 407, 409, 411, 414, 416-417, 420, 422-423, 436 rank of, 395-398, 400, 402-403, 405, 407, 409, 411, 414, 416-417, 420, 422-423 actor, 35-36, 39, 42, 44-45, 251, 254, 259-260, 262-263, 265-266, 271, 274, 276, 278 actor focus, 251, 259, 274, 276, 278 adjective, 38, 65, 71, 78 predicative, 78 adverbial, 62, 66-68, 81, 84, 280, 415416 group, 62, 415-416 adverbs, 7-8, 14, 65, 415, 420-421 affective relevance, 160 agent, 34, 36, 271, 278, 305-306, 324, 367, 378, 394, 397, 399-407, 409-410, 412, 415-416, 419, 422, 433-434, 438 agent-attribuant, 405, 438
of action, 399-412, 422, 433-434 third party, 400, 405-407, 412, 419, 438 anaphoric, 9, 12, 21, 362, 409-410, 416 Arabic grammar, 28 aspect, 248-249, 250-251, 255-256, 279281, 307-308 system of, 250-251, 279-281, 307-308 assertiveness, positive, 22 association, 107-114 attitudinal meaning, 15-17 authority scale, 124-125 auxiliaries, 9, 66, 125 Biblical epic, 334, 355,382 Birmingham collection of English texts, 76 Bloomfieldian structuralism, 248 Boas, F., 38, 47 Borges, J.L., 99-100, 115 branching inquiry, 223-225, 228, 232233, 235 case grammarians, 53 case system, 190-197, 200-201, 208, 212, 239, 284 cataphoric, 12, 21, 410 causative voices, 248 Cervantes, M., 99-100, 115 Chao, Y., 35-36 character strings, 74 child language, 40 choice conditions, 224, 226, 228, 231, 234-235, 238 Chomskyian paradigm, 53 chooser and inquiry framework, 221223, 225, 234, 236-238, 240 chooser and inquiry semantics, 227, 236-237 chooser parcel, 235
444
INDEX
circumstances, 249, 307, 415-416, 420 clause, 5-7, 43-45, 208, 223-224, 232, 327-329, 339, 393, 405 clause-complex, 59 declarative, 17-19, 23, 25, 41, 68, 119120, 173, 193, 232, 306, 309-310, 327-328 defining, 6-8 non-defining, 6, 8 rank-shifted, 61-62, 71, 127 cluster, 56, 60-61, 70-71, 107, 116, 195 genitive, 60-61, 70 co-text, 126, 159-160, 167, 336, 379 cognitive content, 107-108, 110 cohesion, 73, 311-312, 334, 339, 352, 363-364, 384, 405, 407, 409-410, 416420 cohesive link, 13 collocation, 54, 69, 73-74, 82, 85, 116, 338-339, 342-343, 363 command, 119, 134, 137, 142, 144, 150, 223-228, 239, 266, 306, 309-310 non-command, 223, 225-227 communicating community context, 302-303 communication linguistics, 302 competence, 58, 121, 382 complement, 58, 62, 208, 249, 260, 277, 328 componential semantics, 234 conceptual schemata, 101 conduit metaphor, 29-30, 33, 50 congruent, 8, 13, 18, 23, 25, 119-120, 125, 245, 247, 286, 309, 419 context, 47, 112, 119-121, 125-126, 133, 138, 157-166, 172-173, 176-178, 221, 227-228, 284, 293, 295-296, 302-304, 315, 331-333, 335-336, 340, 345, 350351, 353, 379-381, 418-423 contextual theory of meaning, 113 co-operative principle, 123, 162, 182 core meaning, 31, 428 core vocabulary, 73, 288 covert categories, 244, 251 cryptotypes, 38, 244, 246, 248-249, 251-
252, 274, 335 cultural classification network, 204 culture, 37-38, 40, 45, 47, 204-205, 245247, 261-262, 282-288, 290-291, 293297, 315-316 declarative, 17-19, 23, 25, 41, 68, 119120, 173, 193, 232, 306, 309-310, 327328 deep grammar, 249 degree of delicacy, 3-5, 28, 236, 314, 318 deletive, 169-170, 172, 175-176, 179,181 deliberative oratory, 342, 351-353, 356358, 364, 379-380, 384 delicacy scale, 2 deontic logic, 155 non-deontic, 163-164, 166, 181 dependency, 108, 189, 194, 207, 235, 240 determiners, 53, 293 deviance, 83 dialectical relationship between code and behaviour, 302 dialogism, 332-333, 338, 344-346, 351354, 359, 363-364, 376, 378-379, 381384 dialogic quality, 344-345, 363 dialogic speech, 344 dialogue, 10, 43, 46, 110, 228, 291, 312, 336, 338-342, 344-345, 352, 359, 364, 384 diatypic language variety, 302 diminuitive address forms, 245 directives, 119-150, 284, 308 discourse, 1, 3-15, 17-19, 30-31, 38-39, 100, 155-158, 161-170, 175-181, 303, 307-312, 351-354 fields of, 308, 317, 357, 363, 365, 367, 369, 371, 373-374, 380 meaning, 155, 161-166, 168-169, 172173, 175-181, 183 modes of, 311-312, 317 negation, 156, 161-162, 175 participants, 5, 15, 17, 159-160, 290, 307-308, 312
INDEX spoken, 1,9,38-39, 137,311-312 tenor of, 294-295, 309-310, 314-315, 338, 351-352, 354-356, 358-359, 363, 365, 367, 369, 371, 373-374, 377, 380 distributional criteria, 192 elements apex, 64-65, 416 clause, 61-63, 66-68, 306, 405, 414417 completive, 65, 67, 307 deictic, 60-61, 421 finisher, 64-65 negative, 7-8, 16, 66 structural, 56, 63, 422 syntactic, 7, 55-56, 58-59, 66-67, 100, 308-309 discontinuity among, 7 ellipsis, 58, 67, 175, 232, 259-262, 264, 284, 309-312, 327 topic ellipsis, 259, 262 enclitic, 265-266, 270, 272 enclitic particles, 272 respect enclitics, 265, 270 endophora, 259 entry conditions, 128, 196, 210 conjunctive, 198 disjunctive, 197 environment, 16, 36, 41, 43, 63, 80, 85, 109, 227-228, 234, 237, 239, 289, 302 epistemic meaning, 131 equipollent systems, 226, 239 ergative system, 222 exophora, 259-260 experiential role, 249-250 explanatory power, 190, 211 explicitness, 189, 191, 211, 214, 216 exponence, 28, 54,56, 63-64 expository prose, 105 extra-textual, 110 face, 243, 252, 262, 264, 267-270, 284286,290-291,293-297 face conspiracy, 262, 264, 267-270, 296 grammatical metaphors related to, 284
445 family, 243, 252-253, 257, 259, 262-264, 267-268, 284-288, 290-293, 295-297 family conspiracy, 259, 263-264, 268, 284, 286-288, 296 grammatical metaphors related to, 284 fashions of speaking, 247, 252, 293, 335 fatalism, 278, 281 fate, 243, 252, 270, 274, 281, 284-287, 290-293, 295-297, 350, 365, 374 fate conspiracy, 270, 274, 281, 296 grammatical metaphors related to, 284 Fawcett, R.P., 55-57, 59-63, 66, 68, 186, 188, 197, 204, 207, 210, 213, 216, 222, 231,234,236,238,394,437 features, 1-2, 23-24, 29-31, 34, 43, 4546, 53, 55, 120, 126-129, 131, 136-137, 161, 167, 181, 186, 190-191, 193-198, 200-210, 213-218, 222-224, 226, 229, 233-235, 238, 305, 309-311, 419-420, 427 felicity conditions, 234 feminists, 245, 290 Firth, J.R., 37, 47, 113-114, 187, 248, 394 Firthian, 248, 294 foregrounding, 333, 337-338, 343 form, 2, 9, 28, 31, 43, 75, 77, 81-82, 85, 104-105, 120-121, 133, 135, 137, 139, 144, 169-173, 175, 185-187, 201, 213215, 253-257, 259, 263, 276, 280, 293, 309, 436 formal network, 199, 205 Foucault, M. 103, 335 function, 1, 4-9, 11-15, 17, 20, 23, 30, 35, 37, 43-44, 105, 109, 114, 119-120, 135, 223-225, 266, 293, 296, 301, 304307, 309-310, 312, 314-315, 339, 404405, 436 gender system, 194-195, 201-202, 212 generic situation, 301, 303-305, 307-308, 310, 314-316 genre, 48, 332-333, 335-336, 338-339, 341-343, 351-356, 358-359, 364-365,
446
INDEX
380-381, 383-384 geographical dialect, 302-303 geographical provenance, 302 Givón, T., 156, 161, 175 gnostology, 303-304 goal, 36, 41, 48, 249-251, 254, 260, 262263, 265-266, 269, 271, 274-275, 278, 398-403, 407, 412 gradience phenomenon, 59 grammar, 2, 16-17, 28, 30, 33-37, 39-43, 45-46, 53-56, 63-64, 69-70, 73, 79, 185-189, 191-193, 200-202, 209-215, 221, 227-228, 232-233, 236-240, 244, 249, 254, 275, 296-297, 338-339, 428 grammarians, 2, 28, 36, 53, 64, 66 Greek, 28 grammatical categories, 27-49, 54-55, 62, 119, 244247, 251-252, 294, 296 conspiracies, 243, 245-246, 252-253, 264, 282-286, 288, 293, 296-297 units, 4-6, 8, 62-63, 66, 70 grammatically complex, 65, 68 graphology, 304 Grice, H., 123, 156, 158, 162, 181 group, 3, 5-6, 56, 61-62, 64-66, 71, 158, 176-179, 181, 192, 203-204, 208-210, 215, 224, 257, 307, 309, 326, 329, 358, 415-416, 432-433 discontinuous, 65-68 group-complexes, 59 Halliday, M.A.K., 1-3, 9,13, 16, 19, 24, 38, 54, 62, 69, 119-120, 127, 137, 155, 181, 186-187, 190, 200, 216, 222-223, 228-229, 236-237, 247, 249, 261, 274, 276-278, 289, 294-295, 304-305, 308, 311-312, 315, 332, 334-337, 339, 341, 343, 352, 354-355, 364, 379, 393-394, 410, 422, 425, 433 Hasan, R., 1, 9, 73, 261, 311, 333-335, 337, 343, 354-355, 364, 393, 410 head, 3, 55, 61-63, 71, 85, 192, 203-204, 209, 307 heteroglossia, 332-333, 335, 337, 341, 345, 351, 353-354, 359, 378, 383
Hjemslev, L., 114 homonymous, 58 Hopi, 36, 48, 216, 246-247, 294 ideational, 31, 34, 44, 47, 155, 160-175, 177-181, 294, 305, 307-308, 315, 324, 352, 357, 369, 371, 394, 400, 402, 431 ideology, 335 idiolect, 59, 302 idiomatic expressions, 68 illocutionary act, 120, 124-125 imperative, 119-120, 134-135, 137, 142, 144, 146-147, 149, 222-223, 225-229, 232-233, 239, 260, 263, 266, 308-310, 327-328, 400 implicature, 121, 158-159, 182 implicitness, 259, 261-262, 264, 284-285 indicative, 43, 160-161, 173, 176, 222223, 225-227, 229, 232-233, 306, 308310, 327, 402 individual dialect, 302-303 individual provenance, 302 information system, 306, 323 information unit, 5-9, 12-14, 24-25 information-type structure, 5 innuendo by denial, 156 inquiry presentation conditions, 230-231 instantial situation, 303-304, 312, 314 intensive pronouns, 207 interactant role, 202-203, 206 interaction, 17, 23, 39, 46, 73, 123-125, 128, 136, 189, 221, 232, 254-255, 314-315, 336, 339, 345, 381-382, 393, 431 interactive function of tones, 17-23 interrogative, 17-18, 119-120, 133, 135, 144, 146-147, 193, 232-233, 266, 306, 309-310 interpersonal meaning, 305 interpersonal rhetoric, 123, 358 intertextuality, 332, 341, 353, 358 intervals, 138, 427, 430-431, 434-435 intonation, 1-25, 55, 137, 150, 195, 206207, 309, 400, 425-426, 433-434 Jespersen, ., 34-35 Lamb, S., 202, 215-216, 232, 303
INDEX langue, 103 Leech, G., 17, 122-124, 127, 129, 134136, 211,215 lemma, 74-76, 91, 174 lexical cohesion, 73, 352, 363-364, 384, 405, 407, 409-410, 416-420 items, 12, 14, 22, 39, 53-70, 106-108, 116, 245, 253, 277, 283, 288-289, 339, 358-359, 363, 365, 405, 409410, 416, 432 macrostructure, 74, 109-110 relations, 54, 311 structure, 74, 115, 245 words, 8-14 lexicalization, 68 lexicogrammatical organization, 312 lexicography, 74, 86-87 lexico-syntactic form, 121 lexis, 53-71, 73-96, 186-187, 244-246, 289-290, 296, 363, 366-372, 374, 377 sexism and, 289 linguistic code, 304 communication, 304 description, 2, 74, 82-83, 404 relativity, 243, 245 local-circumstance, 421-422 locational, 324, 400, 411-412, 414, 419, 422, 438 marked, 9-15, 23-25, 37, 44, 121, 156, 160-161, 165-167, 177-178, 181, 248, 252, 275, 310-311, 326, 328-329 maxims, conversational, 121-122, 124, 156-158, 160, 181-182, 341 meaning potential, 46-47, 187, 303, 307, 315, 352, 429 melody, 427, 426-427, 430-431, 433-434 mental process, 249-252, 274-278, 284, 411,419 message, 13-14, 30, 46-47, 104-105, 216, 302, 305-306, 309, 311-312, 318, 345, 383, 394, 419, 435, 438 metafunctions, 46-47, 238, 240, 294, 305, 315, 318, 324-329
447 metalanguage, 29, 31-34, 36, 45, 47-49, 239 metalinguistic, 32-33 metonymy, 30, 47, 49, 332 Milton, John, 331-387 mimetic genre, 383 modal modalized directives, 119-150 predicate, 128 probability, 169-170, 175, 179, 183 verbs, 16, 68, 120, 125, 127, 131-137, 140-150, 155, 163-183, 229, 308310, 327 modality, 36, 120, 125-150, 155, 163183, 223, 229, 232, 272, 281, 285, 309, 369,371,373-375,402,426 mode, 294-295, 311, 314-315, 338, 351, 354-356, 358-359, 363-365, 367, 369, 371, 373, 375, 377, 380, 384 modification, 126-127, 133, 149, 307 modifier, 55, 85, 88, 90, 422 modulation, 128,155,181, 250, 371, 374 Mohan, ., 121, 134, 144, 146, 149 mood, 43, 45, 119-120, 125-126, 131, 150, 193, 195, 223-227, 232-233, 239240, 261, 264, 266, 270, 272, 308-309, 323, 326-327, 339, 352, 367, 369, 371, 373-374, 426 morpheme, 29, 54, 56-61, 63-65, 122 class labels, 64 complexes, 59 co-ordinated, 58 free, 58-59, 61 morphology, 31, 57, 187, 244-246, 250, 264 derivational, 57 morpho-syntax, 304, 307, 324, 326-328 music, 48, 294, 297, 393-394, 423-437 narrative, 22, 44, 310, 332, 334, 351, 355-356, 359-360, 362, 364, 367, 369, 376, 378-379, 381-382 networks, 36, 46, 55-56, 63, 70, 107-116, 119, 126, 128, 130-131, 175, 179, 185191, 193, 195-218, 222, 227-229, 231233, 235, 238, 240, 305, 314-315, 323,
448
INDEX
415-416, 421, 429-430, 434-435 neutralization, 202, 210, 214, 228, 231, 240 Nigel grammar, 221, 227 nominal group, 192, 203-204, 208-210, 215, 224, 257, 307, 358, 416 non-topic complements, 249, 260 norms, 79, 83, 246-247, 282, 317, 333, 336-337, 342, 344-346, 351-353, 356, 365, 378-379, 381-384 establishment of, 79, 82, 317 explicit reference to, 79 tentative, 83 noun, 6, 28-31, 36, 43, 58, 60-61, 63, 65, 68, 79, 85, 88-90, 111, 204, 247, 258 number system, 28, 31-32, 43, 63, 195197, 200-204, 207, 212 object, 34, 36, 43, 80-82, 84, 258-259, 263, 276 direct object, 36, 81 language as, 30, 34 obligation, 129-130, 133, 136, 326 orthographic word, 60-61 Palmer, F.R., 127, 129, 134, 181, 426 paradigmatic, 185, 201, 207, 215, 228, 334, 363 Paradise Lost, 331-387 parallel categories, 302 parole, 104 participant roles, 305, 308, 401, 405, 414-415 particles adverbial, 67-68 prepositional, 67 enclitic, 272 passive participle, 248 passive voice, 33-34, 36, 67, 71, 84, 248, 261, 275, 278-279, 281 passivization, 67 past tense, 134, 305 past time, 129 Penman project, 36 person system, 192-206, 211 personal tenor, 315 phase, 318
phatic communion, 105, 160, 181 phenomenological reality, 102 phenotypes, 251-252 phonaethesia, 283 phonemes, 426, 428, 431, 435 phonetic substance, 1-2, 16 phonology, 2, 5, 27, 33, 187, 283, 304, 306, 329, 425 phrase, 5-8, 80-82, 85-87, 178, 257, 358, 415, 427, 431-435 pitch, 1, 3-4, 8, 16, 24, 104, 329, 427, 433 movement, 4, 8, 427 variation, 1-2 planning, 221, 293, 395-396, 398, 400402, 407 plurifunctionality, 435 polarity, 132, 137, 139-140, 161-163, 232-233, 240, 327, 345, 352, 358-359, 366, 368, 370, 372, 374, 377-378, 434 polysemous, 37, 58, 135, 377-378 possibility, 129, 155-156, 158, 163, 166, 168, 176, 180, 272, 274, 281, 285, 325 power, 123-124, 190, 211, 295, 297 pragmatics, 15, 17, 121, 159, 181-182, 393 Prague school linguistic and literary theory, 338 praxis, 316 predicate, 34-35, 127-130 prefix, 58, 250-251, 253-256, 272, 274276 prepositional group, 64, 415-416 present, 42, 45-46, 229-233, 248-249, 280-281, 308, 311 in present, 249, 311 simple, 248-249, 281 presupposition link, 9, 11-13 primary tense system, 229, 233 privative systems, 226, 239 probability, 31, 40, 138, 155, 163-181, 183, 325 probability modals, 169-170, 175, 179, 183 process, 41, 229-230, 238, 249-252, 254-
INDEX 255, 258-259, 272, 274-278, 284-285, 305, 307-308, 339, 400, 404-405, 407, 411-412, 414, 416, 419, 426, 433, 438 pronouns, 9, 185-215, 248, 257, 261, 263-265, 284-285, 289, 358, 409-410 personal, 185-215 reflexive, 207 respect, 264-266, 289-290 propositional content, 5, 135, 160-162, 171, 176 prosodic features, 1-2, 24 psychological reality, 114, 189, 211, 382 qualifier, 55, 61 question contour, 18 questioning function, 20 rank, 54, 62-63, 237, 262, 268, 309, 314, 396, 400-401 rank scale, 54, 62, 237 rankshifted predication, 127 reactance, 251, 274-276 realization, 43, 46, 189-190, 195-196, 201, 203, 207-216, 228, 239-240, 305306, 314-315, 318, 325, 328, 395, 400402, 407, 412, 414-415, 420-421, 427, 430, 432, 435 realization rule, 189, 195, 209-210, 215 referent situation, 208, 230 referent thing, 208 register, 37, 43, 46-48, 59, 294-296, 301, 303, 307, 309-310, 314-318, 335, 345, 351-352, 354-355, 357 description, 303, 306, 314, 316-318 dialect and, 304, 317 realization, 314-315, 318 registerial provenance, 317 theory, 315-317 relation attributive, 308, 323, 357 interstratal, 56, 214, 226, 239-240 relational predication, 308 relational verbs, 258-259, 264, 285 relationship interaction relationship, 304-305 medium relationship, 304-305, 310312, 314
449 relationship of instruction, 310 rheme, 28, 30, 35,311,413-420 rhetoric, 123, 316, 358-359, 366, 368, 370, 372, 374, 379 rhetorical eulogy, 100 rhythm, 2, 5, 427, 426-427, 431, 435 root modals, 120, 127 Saussure, F., 101-104, 114-115 science text, 107, 110, 114, 116 segmental features, 2, 23 selection expression, 123, 189, 195-196, 202, 208-210, 216, 430 semantic feature, 2, 120, 126-127, 161, 167, 181, 201-210, 213, 216, 222-224, 234, 427 force, 126-127, 132-133, 137, 139-140, 142, 144 force markers, 126 networks, 46, 55-56, 126, 128, 187, 193, 207-208, 210, 213-214, 429 predicates, 127-128 semantics, 15, 17, 30, 33, 37, 41, 43, 4647, 55-56, 104-106, 110-111, 119-150, 159, 167, 169, 171, 174-175, 180-181, 185, 187, 192-193, 202-216, 221-240, 248, 332, 334-337, 339, 342-343, 345, 351-354, 359, 362-365, 377-381, 393, 407, 430 generative, 190-193, 211-213, 215-216 procedural, 235 sociological, 205, 236-237 semantico-collocational constraints, 58 semiotics, 47, 332, 335-336, 341, 359, 364, 377, 379-380, 382-383, 393-437 of deceit, 336, 359, 377, 379-383 semiotic system, 34, 37, 42, 47, 239, 393-437 semogenesis, 39, 354, 366, 368, 370, 372, 374 semology, 37, 187, 304, 325 semological choices, 305, 308-309 focus, 306, 311, 328 rank of proposition, 314
450
INDEX
roles, 305, 308-309, 323 speech function, 306, 314, 323 sense and syntax, 87-89 sentence, 5, 17, 28, 30, 34-35, 38, 47, 53-54, 62, 121-122, 249, 309, 311-312 sexism, 245, 289 signified, 101 signifier, 101 Sinclair, J.McH., 69, 104-106, 114, 215 singulary branching, 62-63 situation, 9-10, 12, 14, 19-21, 25, 31, 39, 42, 47, 208-210, 221, 230, 238-239, 262, 295, 301, 303-305, 307-308, 310312, 314-316, 329, 394-396, 398, 400, 403-404, 409-410, 412, 414, 416, 420421 social communication, 301-318, 335 context, 119-121, 125-126, 133, 138, 157-166, 227-228, 284, 293, 295296, 302-304, 315, 331-333, 335336,345,350-351,353 neutral, 125, 133, 138, 161 dialect, 302-303 distance, 123-125, 134, 150, 206, 284285, 294, 309, 314, 325 horizontal, 124, 150 scale of, 123, 135,206 group, 157-158, 243, 252, 290-291, 293, 309, 431-433 provenance, 302 socialism, 303 socio-cognitive, 301, 315 socio-linguistic variation, 68 sociological semantics, 205, 236-237 speech act, 17, 120-121, 125, 132-134, 136-137, 139, 141-142, 150 indirect speech acts, 121 statement contour, 18 stratal polysemy, 30 stratificational theory, 232 structural realization, 305-306 structure, 3, 5-8, 18, 35, 41, 45, 53, 5557, 61-66, 70-71, 74, 79, 82, 100-101, 103-104, 109, 115-116, 192-193, 244-
246, 249, 253, 256-259, 261-262, 266268, 271, 275, 279-280, 284, 296-297, 305-306, 334, 339, 343-344, 352-353, 358, 363-365, 378, 380, 382-383, 395396, 400, 403, 405, 407, 409, 412, 414416, 420-423, 430, 435 stylistics, 54, 332 subject, 34-36, 39-45, 192, 208-211, 261, 263-264, 277-278, 298, 306, 309-310, 328, 394 theme, 33-36, 39, 41-42, 44-45, 208210, 328, 405, 407, 409-410, 414, 416 syntax, 17, 53-71, 82, 87-89, 104, 137, 149, 155-183, 186-187, 210, 214, 309, 371, 393 syntactic meaning, 15, 67-68 structuring of discourse, 5 system networks see "networks" systems aspect, 250-251, 279-281, 307-308 case, 190-197, 200-201, 208, 212, 239, 284 equipollent, 226, 239 ergative, 222 gender, 194-195, 201-202, 212 information, 306, 323 number, 28, 31-32, 43, 63, 195-197, 200-204, 207, 212 person, 192-206, 211 primary tense, 229, 233 privative, 226, 239 qualification, 307 semantic, 30, 37, 43, 46-47, 55-56, 180-181, 192-193, 202-207, 222, 226-227, 234-240, 334-337, 393 semiotic, 34, 37, 42, 47, 239, 393-437 syntax, 55-56, 309, 393 tense, 229, 233, 279-280 tact maxim, 124 tag, 16, 41, 120, 126-127, 132-133, 135137, 139-140, 142, 144, 149, 327 Tagalog, 243, 246, 249-298 tempo, 2, 6, 427, 426-427, 431, 435
INDEX temporal dialect, 302-303 temporal provenance, 302 tenor, 294-295, 309-310, 314-315, 338, 342-343, 351-352, 354-356, 358-359, 363, 365, 367, 369, 371, 373-374, 377, 380, 384 tense, 42, 45-46, 134, 155, 223, 228-233, 239, 247-249, 279-281, 305, 307-308, 311,418,421 primary tense, 223, 228-233, 239 tense system, 229, 233, 279-280 text, 1, 4, 30-31, 36, 40, 42-43, 46-47, 73, 79-82, 99-116, 157, 160, 167, 175, 187, 190-191, 221, 227, 234, 236-239, 259, 262, 279, 293, 303, 308-309, 312, 314-315, 331-346, 351-354, 362-365, 376-378, 380-384, 393-422, 431-432, 436 generation, 36, 221, 227, 235-236, 238 texture, 311, 393 theme, 33-36, 39, 41-42, 44-45, 186-187, 195, 208-210, 253, 267, 270, 279, 284, 287, 290, 296, 306, 311, 323, 328, 355, 363, 400, 402, 405, 407, 409-410, 414, 416 thematic progression, 407-410, 412, 414, 417, 420 theory, 107, 113, 121, 185, 188, 191-192, 194, 197, 201, 208, 215, 232, 248, 301, 315-317, 336, 338, 364-365, 376, 393396, 422-423, 428 of activity, 394-396, 422 token, 28-29, 32-33, 202, 204, 207, 334, 342, 351-355, 359-362, 377, 379-380 classification network, 204 tonal harmony, 393-394, 423, 426, 428430, 435 tonality, 3,5,8-9, 428 tone, 1-5, 8, 12, 14-25, 130, 311, 329, 427, 426, 430-435 compound, 3-4, 14-15, 22-23, 25 falling, 3, 16, 18-19, 22, 25, 427, 433434 groups, 1, 3, 5, 24-25, 432 rising, 3, 18-20, 427, 433-434
451 tonic, 3-5, 8-15, 21-22, 24-25, 305-306, 329, 427, 426-427, 430-433 non-tonic, 11, 25, 427 post-tonic, 4, 11 pre-tonic, 4, 24, 427, 432 syllable, 3-4, 305-306, 329, 427, 431, 433 word, 1, 4-5, 8-14, 21, 24-25, 427 tonicity, 3-5, 8-9, 11-13, 15, 18, 23-25 topic, 13, 35-36, 157-158, 163, 176, 181, 249-250, 259, 261-265, 271, 279, 357, 363-365, 368, 380 comment, 13, 35, 157 total accountability at all ranks, 62 'traditional feature' system network, 193 transduction, 345-346, 351, 353, 378, 381,383 transformationalists, 53, 59 transformational generative grammar, 216 transitivity, 38, 41, 48, 80, 84, 186-187, 222, 249, 295, 307, 323-324, 339, 357, 359, 363, 365, 367, 369, 371, 373-374, 377, 379-380, 419-420 translatability, 317 translation, 48, 99, 254, 317 usage, 56, 77, 81-82, 84-87, 166-167, 316-317 valeur, 103-104 value, 28-29, 32-33, 103, 105, 115, 127, 142, 176, 178, 180-181, 185, 189, 224, 233, 290, 325, 342, 351-352, 359-362, 376-377, 379-380 verb, 16, 28, 30, 33-34, 36, 38, 43, 48, 65-68, 79-80, 83-86, 88-90, 105, 120, 125-127, 129, 131-137, 140-150, 155, 163-183, 192, 216, 229, 246, 248-250, 258-259, 262, 264, 270, 276, 280, 285, 308-310, 327, 415, 419, 422, 438 extension of, 66-68 intransitive, 66, 79-81, 86, 248 object focus, 258 particle, 66, 80 perception, 79, 250-252
452
INDEX
reciprocal, 254 verbal group, 5-6, 56, 66, 307, 309, 326 vocabulary, 33, 73, 102, 104, 107, 109110,288 vocative form, 206 voice, 84, 278-279, 281, 284-285, 295, 331-332, 336, 343, 349, 351, 356-357, 363, 366, 370, 374, 376, 378-382, 438 volition, 36, 129-130, 273 Whorf, ., 29, 37-39, 47-48, 243-249,
251-252, 281-283, 286, 293-294, 296, 335 Wilkins, J., 101 words, 1, 4-5, 8-14, 21, 24-25, 27, 30-31, 33-34, 39, 42-44, 46, 53-55, 57-58, 6061, 65-66, 68, 74-76, 81, 104, 106-111, 283-284, 287-290, 339, 362-363, 427 internally inflected, 66 non-lexical, 9, 11-12, 14, 21
