No title

JOURNAL OF

SEMANTICS Volume 3 no. 4, 1984

SWETS & ZEITLINGER BV LISSE - THE NETHERLANDS - 2000

JOURNAL OF SEMANTICS AN INTERNATIONAL JOURNAL FOR THE INTERDISCIPUNARY STUDY OF THE SEMANTICS OF NATURAL LANGUAGE MANAGING EDITOR:

Pieter A.M Seuren

EDITORIAL BOARD:

Peter Bosch Leo G M Noordman

REVIEW EDITOR:

Rob A. van der Sandt

ASSISTANT EDITORS:

G Hoppenbrouwers A. Weijters

CONSULTING EDITORS: J Alhvood (Univ Go'teborg), M Arbib (Univ Mass Amherst), Th T Ballmer (Ruhr Univ Bochum), R Bartsch (Amsterdam Univ ), J van Benthem (Groningen Univ), H H Clark (Stanford Univ ), G Fauconnier (Univ de Vincennes), P Gochet (Univ de Liege), F Heny J Hintikka (Florida State Univ ), St Isard (Sussex Univ ), Ph Johnson-Laird (Sussex Univ ), A Kasher (Tel Aviv Univ), E Keenan (UCLA), S Kuno (Harvard Univ ), W Levelt (Max Planck Inst Nijmegen), EDITORIAL ADDRESS:

J Lyons (Sussex Univ), W Marslen-Wilson (Cambridge Univ), J McCawley (Univ Chicago), B Richards (Edingburgh Univ), H Rieser (Univ Bielefeld), R Rommetveit (Oslo Univ), H Schnelle (Ruhr Univ Bochum), J Searle (Univ Cal Berkeley), R Stalnaker (Cornell Univ), A von Stechow (Univ Konstanz), G Sundholm (Nijmegen Univ), Ch Travis (Tilburg Univ), B van Fraassen (Princeton Univ), Z Vendler (UCSD), Y Wilks (Essex Univ ), D Wilson (UCL)

Journal of Semantics, Nijmegen Institute of Semantics, P O Box 1454, NL-6501 BL Nijmegen, Holland

Published by Foris Publications, P O Box 509, 3300 AM Dordrecht, The Netherlands Sole distributor for the USA. and Canada: Foris Publications USA, PO Box C-50, Cinnaminson NJ 08077 ©NIS Foundation ISSN 0167-5133 Printed in the Netherlands by ICG Printing

Volume 3, nummer 4

December 1984

JOURNAL OF SEMANTICS

CONTENTS

FRANS PLANK, Verbs and Objects in Semantic Agreement. Minor Differences between English and German that Might Suggest a Major One . . . . . 305 RICHARD RHODES, Lexical Taxonomies

.

361

Bookreviews AD FOOLEN Joachim Jacobs, Fokus und Skalen Zur Syntax und Semantik der Gradpartikeln im Deutschen

379

PETER BOSCH S.G.Pukmn,

392

Word Meaning and Belief.

Review Article CHARLES TRAVIS Nathan U. Salmon, Reference and Essense . . .

395

Publications Received. .

403

. . .

..

JOURNAL OF SEMANTICS CONTENTS VOLUME 3 (1984) Articles NICHOLAS ASHER, Meanings don't Grow on Trees

229

B. CORNULIER, Detachment ¥= "Meaning Detachment"

257

M J CRESWELL, Comments on Von Stechow

79

LARS HELLAN, Note on some Issues raised by Von Stechow

83

J HOEKSEMA, To be continued The Story of the Comparative

93

ROBERT R VAN OIRSOUW, Accessibility of Deletion in Dutch.

...

201

FRANS PLANK, Verbs and objects in Semantic Agreement Minor Differences between English and German that Might Suggest a Major One 305 RICHARD RHODES, Lexical Taxonomies

361

PIETER A M SEUREN, The Comparative revisited

109

LEON STASSEN, The Comparative Compared

143

ARNIM VON STECHOW, Comparing Semantics Theories of Comparison ARNIM VON STECHOW, My Reaction to Cresswell's Hellan's Hoeksema's and Seuren's Comments

1

183

ARNOLD M ZWICKY and JERROLD M SADOCK, A Reply to Martin on Ambiguity . 249 Review Articles PETER BOSCH Hans-Jurgen Worlds and Contexts

Eikmeyer and Hannes Rieser, Words,

GE CALIS John R Searle, Intentionality. ophy of Mind

261 An Essay in the Philos295

KIM STERELNY Andrew Woodfield, Thought and Object

277

CHARLES TRAVIS Nathan U. Salmon, Reference and Essense

395

Bookreviews PETER BOSCH S.G. Pulman, Word Meaning and Belief

392

AD FOOLEN- Joachim Jacobs, Fokus und Skalen: Zur Syntax und Semantik der Gradpartikeln im Deutschen

379

PUBLICATIONS RECEIVED

403

Journal of Semantics 3: 305-360

VERBS AND OBJECTS IN SEMANTIC AGREEMENT: MINOR DIFFERENCES BETWEEN ENGLISH AND GERMAN THAT MIGHT SUGGEST A MAJOR ONE FRANS PLANK "The soup sounds good." Maic Antony McGonigle ABSTRACT

1 ON THE ALLEGED UNSYSTEMATICITY OF SEMANTIC AGREEMENT

Argument expressions and predicate expressions both identify and classify aspects of the universe of discourse of the speech-event participants, including their own position within this universe. The number of these lexically and grammatically manifest classificatory aspects is obviously enormous, and may, in addition, differ from language to language or from culture to culture, although not a few of them tend to recur with more than

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Even though not boasting overt and systematically used noun classifiers of the variety known as classificatory verbs, languages may still have predicates (co-)signalling particular categories of nominal classification (outside syntactic agreement). Standard examples are English verbs such as to bark/neigh/gallop requiring subjects which refer to particular animals, or otherwise classify their subject referents a being in the relevant respects comparable to the animals in question. I hope to demonstrate in this paper that semantic agreement of this kind, which has often figured in theoretical discussions about the structure of the lexicon and the interface of semantics and syntax, is not as unsystematic as is commonly assumed. Although there may be considerable cross-linguistic variation, this variation at least appears to be quantitatively patterned insofar as some languages (such as German) have relatively more instances of semantic agreement between verbs and objects than others (such as English). I suggest further that the incidence of semantic verb-object agreement is not a minor, isolated, and entirely unpredictable difference between individual languages, but correlates with the typology of the grammatical core relations of subject and object, and in particular with the objectdifferentiation characteristics of a language verbs and objects seem to agree more commonly in languages which give morphosyntactic, and in fact lexical, recognition to at least two semantically relatively specific types of core objects (such as direct and indirect object).

306

(1)

a. The dog/fox/*lion/*eagle/animal/*sergeant barked b. The horse/*ostrich/animal/*sergeant galloped across the field and neighed

Violations of such agreement requirements typically do not result in ungrammaticality but in metaphor, in the starred examples of (1), the sound production or locomotion of the subject referents is compared to that of those animals which characteristically perform these activities. Of course there may be circumstances where it is difficult to draw a clear boundary line between literal and metaphorical uses of predicates, as metaphors are prone to fade, resulting, in the case at issue, in changes of the basic meaning of predicates, but even though occasionally blurred, and perhaps not in-

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

chance frequency, for reasons that presumably have to do with invariants of human perception and cognition. It seems significant that argument and predicate expressions often classify quite different aspects of the universe of discourse. To mention only a few of such apparent classificational preferences, arguments tend to classify what can be regarded as more or less individuated, whereas predicates tend to classify such 'things' as events, processes, activities, experiences, states, (clusters of) properties, and the like, which are hardly conceptualized as possessing individuality. Most attitudinal or evaluational aspects are preferably dealt with by mode categories of predicates, although a few of them also show an affinity to argument expressions (e.g. those concerning the relationship between the speaker and other individuals respect, politeness, endearment, etc.). Temporal deixis typically seems to belong to the domain of predicate classification, local deixis to that of argument classification.1 Many quantificational categories are associated primarily with arguments, but there are others, especially aspectual ones, which are typically expressed in the predicate. Such preferences notwithstanding, there certainly are classifications which are neutral with respect to the parts of predications they may be associated with; and there are, furthermore, categories which, although not necessarily conceptually neutral, may be encoded simultaneously by argument and predicate expressions, with the obvious consequence that such dual classifications have to agree with each other for the respective predication to be semantically coherent. Such classificational agreements are the subject matter of this paper, and therefore deserve some initial illustration. For example, certain English verbs classify activities (of sound production or of locomotion) as being characteristically performed by specific animals, and these verbs, thus, can only co-occur with arguments referring to animals of the appropriate kind or to appropriate classes including these specific animals.

307

variant for all members of a speech community, distinctions between literal and metaphorical meaning certainly cannot be argued away on the strength of some controversial instances. Needless to emphasize, zoological classifications are not the only ones featuring in such agreement requirements, nor are verbs the only kind of predicates potentially stipulating such requirements English, for example, has a verb of departure such as elope and an adjective of shape such as buxom which literally combine only with arguments referring to females rather than males. There also exist in English predicates which lexically classify the number of entities their arguments may refer to, or (and this is the interpretation of Sapir & Swadesh 1946) the number of performances of the action they denote*2 (2)

a. The dogs/*the dog/the crowd trooped away b. They massacred *Smith/the whole tribe Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

These verbs differ from such plural-only verbs as gather, assemble, surround, or separate (cf. Meyer 1909) insofar as their classificatory meaning component is semantically arbitrary rather than being a constitutive and integral part of their lexical meaning, it is possible to imagine verbs with roughly the same basic meaning as troop or massacre but lacking their quantificational aspect (and analogously for bark, gallop, neigh, elope, and buxom), whereas verbs like gather or assemble are such that this quantification cannot possibly be separated from their meaning - in fact not much meaning would be left behind if this component were removed. Of course it would again be unrealistic to expect the distinction between arbitrary and constitutive classificatory meaning components of predicates always to be clear-cut, as with the metaphor-literal distinction, some instances are bound to be doubtful, but these should not mislead us into abandoning the distinction as a matter of descriptive principle.3 Talking of distinctions, the agreement requirements in the cases that interest us here are due to the specialized meaning of predicates, hence differ from ordinary grammatical agreement between predicates and their arguments, where in principle any predicate combinable with an argument in a given construction is subject to agreement requirements. Again, this is not to say that the difference between classificational (semantic and grammatical (morphosyntactic) agreement is necessarily categorical, the categories utilized to establish syntagmatic coherence may be quite similar (cf. number or animacy),4 and it may not always be easy to decide whether an agreement regularity is still sporadic or already systematic. Returning to the arbitrary-constitutive distinction, it seems reasonable to expect, then, that the more arbitrary predicate-related classifications pertaining to properties of argument expressions are more likely candidates for cross-linguistic variation. To stay with our examples, one would

308

hardly be surprised to come across languages which have verbs of sound production, locomotion, departure, and killing and adjectives of shape without the zoological-species, sex, or number restrictions found with English lexical items such as bark, gallop, troop, elope, massacre, or buxom. In a recent fundamental work on semantic analysis, Viehweger et al. (1977:353), the likelihood of this kind of variation between languages is mentioned rather cursorily, on the basis of the following comparison between Russian and German (I have added the English equivalents), it is concluded that such 'sememic' agreements, as only observed in Russian, are language-particular idiosyncrasies. (3)

a. myt'.golovujico...,stirat':bel'e,brjuki....promyvat'.zoloto ... b. waschen.Kopf, Gesicht.... waschen.Wdsche,Hosen...,waschen: Gold... c. wash.head, face ..., wash.laundry, trousers ..., wash, gold... Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

It would indeed seem difficult to predict to what extent and in which lexical areas languages may vary in this respect, although it would also seem natural to expect, for example, that if certain animals play a considerable role in a culture, the language might have predicates for the exclusive purpose of denoting the culturally most salient activities of these animals. Thus, for example, a rich fund of specialized verbs of reproduction, with concomitant agreement requirements pertaining to the various species of animals (cf. Grimm 1853. 17ff.), is likelier to be found in rural livestock economies than in industrial societies of city dwellers. A formerly quite popular, and somewhat more general, approach to prediction at least ought to be mentioned here, according to which different predicates for different semantic classes of arguments are more likely to be found in languages of 'primitive' cultures, where people, struggling with the concrete and the particular, are allegedly unable to generalize. In this view, as espoused e.g. by Jespersen (1922: 429ff.,passim), a single, invariable predicate of washing, for instance, is characteristic of a superior, efficient language (like English), whereas three (cf. Russian) or even thirteen or fourteen verbs of washing (as, allegedly, in Cherokee) are symptomatic of inferior, more 'primitive', less efficient languages. Without denying the possibility of a cultural determination of grammatical and lexical patterns, it may be safely assumed that this general way of ethnopsychological reasoning is as obsolete as the belief that Cherokee really has thirteen or fourteen different verbs of washing chosen in agreement with the referent of the object (oneself, one's face, another's face, clothes, dishes, meat, etc.,cf. Hill 1952). Cross-linguistic unpredictability of such classificational agreements is, however, not the only problem facing linguistic theories that emphasize

309

(4)

a. ride (on a horse)/(auf einem Pferdj reiten b. *ride on a beam/auf einem Balken reiten - posture as in (a), but no movement c. ride on a bicycle/*auf einem Fahrrad reiten - posture, movement, control of conveyance as in (a), but less restricted choice of conveyance

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

the general rather than the particular in human language. Grammars of individual languages also often relegate such agreement to the domain of the accidental, genuine morphosyntactic agreement rules are supposed to take care of all co-signalling that is in any way regular in the syntagmatic combinations of a language, and if there are further combinatorial restrictions pertaining to predicate and argument expressions, they have to be registered individually as unpredictable lexical idiosyncrasies. This is not to say that syntagmatic lexical relations - variously known as "wesenhafte Bedeutungsbeziehungen" (Porzig 1934), "lexikalische Solidaritaten" (Coseriu 1967), "semanticeskoe soglasovanie" (Gak 1972, Leisi 1975 employs the corresponding German term), or selectional restrictions (transformational grammar) - differ from syntagmatic relations which are encoded by means of inflectional morphology in that they do not form patterns at all they, together with paradigmatic lexical relations, determine the structure of patterns known as lexical fields. But the possibility of lexical-field patterning still does not imply predictability. The field analyst may be able to discover some restrictedly regular lexical patterns once he knows the full set of lexical items of a language, but he should not be able to tell in advance which particular lexical items (denoting particular meanings) will bear which syntagmatic relations to which other items. Or is there any reason to believe that it should be predictable, on linguistic grounds, that in English a set of verbs of movement and sound production are (non-metaphorically) applicable only to particular animals (gallop, waddle..., neigh, bark, bleat...)? Predicate meanings may change in time with regard to classificational agreement, and given the post-hoc character of potential synchronic generalizations it is only natural not to expect such diachronic developments to be predictable (again on linguistic grounds). John Lyons (1977. 263 ff.) is one of the most recent advocates of this pessimistic view. Presupposing a very specific common original meaning of the English and German verbs ride and reiten, as defined by the syntagmatic lexical relation with horses (or animals very much like a horse),s he notes the different directions in which this meaning has been generalized in the two languages (cf. 4), and concludes that "there is no convincing evidence to support any kind of deterministic theory of semantic change" (1977 264).

310 d. ride in a carriage (in a bus, on a train)/*in einem Wagen (mit dem Bus, im Zug) reiten - only remaining condition (in English), being conveyed (by land) A diachronic theory could presumably be so deterministic as to preclude, for instance, developments whereby ride/reiten would contract syntagmatic relations, say, with temporal specifications (e.g. 'before noon') or with specifications of the colour of the conveyance - at least I think it could.6 But such negative statements on a common-sense basis probably are not very satisfactory if one is aspiring after a truly deterministic theory of lexical change.

2. COMPARING ENGLISH AND GERMAN Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Although it is often taken for granted that classificational agreements between predicate and argument expressions are essentially unpredictable as far as their cross-linguistic variability, the domains of their occurrence within individual languages, and their historical developments are concerned, it seems to me unduly pessimistic to accept unquestioningly the view that in this area the search for any generalizations is bound to be in vain. For example, if one were to investigate which arguments predicates are most likely to agree with in semantic classification, one might uncover a universally valid implicational generalization, agreement requirements are more likely to obtain between predicates and (direct, non-direct) objects and intransitive subjects than between predicates and transitive subjects, or, if the generalization is to be stated in semantic terms, in view of the possibility that genuinely grammatical relations such as subject and object cannot be defined in all languages (cf. § 3.2): the argument in an agentive role (at least with two-or-more-place predicates) is less likely to be in semantic agreement with the predicate than arguments in roles such as patient and instrument.7 Further lexico-grammatical regularities may then turn out to follow the same pattern, transitive subjects or agents, for instance, appear to be the least likely arguments, firstly, to form idiomatic expressions together with the predicate, secondly, to be incorporated into the predicate, not only in the classical noun-incorporating languages but also in compound types like bird-watching, noun-incorporation, birdchirping in other languages, thirdly, to be the point of orientation in derivational relationships between nominals and predicates (cf., with direct-object/patient orientation, washable, eatable, readable etc., with intransitive-subject/patient orientation, perishable). Rather than trying to sample and explain such regularities,8 I shall present in the following sections some data concerning only the relation of direct object which

311 bear on the issue of the alleged total unpredictability of semantic agreement between arguments in this relation and their predicates. When we compare English and German, two languages which are genetically, culturally, and areally closely related, we nevertheless observe considerable differences in semantic agreement between particular verbs, or verb groups, and their direct objects. After this survey, the question will be raised (in § 3) whether we really have to do with a random collection of idiosyncratic minor differences between the two languages, or whether these differences follow a pattern reflecting a characteristic major crosslinguistic difference. 2.1. With different nominals in the direct-object relation the English verb tell is translated differently in German. (5)

English in fact has more specialized verbs corresponding to erzdhlen, viz. relate or narrate, and there are also quite interesting constraints on what can be told and what can be said (cf. Taylor 1980, with further references). But the point here is that there is a set of object nouns, as suggested by the examples in (5), where tell may be used unrestrictedly, whereas even for this set there is no such general-purpose verb in German (cf. *femandem Lugen/eine Geschkhte/ ... sagen, *jemandem seinen Namen/seine Meinung/ ... erzdhlen). It may be difficult to characterize precisely the common semantic denominators of the object nouns in (5a) and in (5b), but I think it is sufficiently accurate to say that in (5a) the content of the verbal communication is an unembellished piece of information demanded by the addressee, whereas in (5b) what is communicated requires some creative effort on the part of the speaker.9 This formulation of the difference could seem to create problems for the semantic classification of arguments, since some of the object nominals of sagen/erzdhlen do not per se denote communicable contents or narrative genres (e.g. Uhrzeit, Abenteuer). However, these nouns can easily be interpreted as closely related to nominals which are prototypical objects of sagen ('information about the time') and erzdhlen ('the story of one's adventures'). Although sagen and erzdhlen denote different kinds of communicative activities, and their respective selection restrictions are not entirely in-

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a. to tell someone one's name/jemandem seinen Namen/die Wahrthe truth/the difference/the heit/den Unterschied/die Uhrzeit/ time/the answer/one's die Antwort/seine Meinung sagen opinion b. to tell someone lies/the jemandem Lugen/die Neuignews/a story/a tale/one's keiten/eine Geschichte/ein adventures Mdrchen/seine Abenteur erzdhlen

312 dependent of these differences in meaning, these verbs still have a great deal in common semantically, and on language-internal as well as comparative grounds it appears plausible to assume that what they have not in common are, essentially, the relatively arbitrary (vis-a-vis the common basic meanings of the verbs) restrictions on arguments occurring in the direct-object relation.

(6)

EFFECTED

AFFECTED

a

to dig the ground/potatoes den Boden umgraben/Kartoffeln ausgra ben to paint the wall die Wand streichen (an-/be-malen)

to dig a grave/a hole/a tunnel ein Grab/ein Loch/einen Tunnel gra ben b. to paint a picture/a landscape ein Bild/eine Landschaft malen c. to burn a hole/lime/bricks ein Loch/Kalk/Ziegel brennen

d. to conclude a treaty einen Vertrag (ab-jschliessen (but also: ein Geschaft abschliessen 'to secure a business") e. to force an entry/a confession Eintritt/ein Gestandnis erzwingen

to burn coal/the meat/one's mouth Kohle verbrennen ((mit) Kohle heizen)/ das Essen verbrennen (anbrennen)lsich den Mund verbrennen to conclude a lecture/a business eine Vorlesung beenden ((be-)schliessen)/ ein Geschdft beenden (also abschliessen)

to force someone (into doing something)/ to force war (upon someone) jemanden (zu etwas) zwingen/ (jemandem) Krieg aufzwingen

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

2.2. The difference between affected and effected objects, which may also play a marginal role in (5), is often said to be contingent on the governing predicates: some verbs govern affected objects (e.g. read a book), others effected objects (e.g. write a book). However, there are also verbs that can take affected or effected objects (e.g cook an apple vs. cook a meat), and the fact that misinterpretations do not usually occur in such cases demonstrates that the respective meanings (affectedness vs. effectedness) are not signalled by the verbs alone but by the verbs in conjunction with their object nominals. Thus, although it is not feasible to set up two semantic classes of inherently affected and effected nouns,10 there must be aspects of the inherent meaning of some nouns which are responsible for the different relational interpretations of these nouns under certain circumstances. Jespersen (1928 232-4) is one of those who at least mention the indeterminacy of some verbs with regard to the affectedness-effectedness distinction, and it is instructive to compare his (incomplete) list of pertinent English verbs with their translation equivalents in German.

313 f. to answer not a word kein Wort antworten

to answer a question eine Frage beantworten

Of course there are also verbs that are indeterminate in both languages (light a fire/a match - ein Feuer/ein Ziindholz anzunden, cook - kochen, etc ), M and closer scrutiny might also reveal instances where we get the reverse of what we have seen in (6). But is seems to me pretty safe to predict that if translation-equivalent English and German verbs differ with respect to the ability to take affected and effected objects, German is more likely to employ separate verbs, which may be morphologically related, with affected and with effected objects.

(7)

a. anziehen. Mantel, Handschuhe, Schuhe, Socken, Kleid, Hemd, Hose, Schurze, BH, Pullover, Anzug, Hosentrdger, Knieschu'tzer... (*Hut, *Maske, *Krawatte...) put on. coat, gloves, shoes, socks, dress, shirt, trousers, apron, bra, sweater, suit, braces, kneepads... (hat, mask, tie...)

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

2.3. Another example of one-to-many correspondences between English and German verbs, with more specific semantic agreement requirements in German, is provided by the verbs of dressing and also undressing, or, more generally, of putting on and taking off clothing and other articles one wears on one's body. In English one may resort to the stylistically marked verbs to don and to doff, and there are also available a number of specialpurpose verbs depicting the manner of putting on or taking off garments or garment-like kinds of body covering (e.g. to buckle on, slip on/into, get into, draw on/off, throw on/off, wrap oneself in/around oneself), some of which seem obligatory if the body covering is not culturally recognized as a standard garment (e.g. The Indian wrapped himself in/threw over his shouldersI *put on his blanket, blankets being not typically used as garments by Americans - except Indians, who therefore may disagree with the above grammaticality judgement). However, the cardinal and by far most common English predicates of this semantic domain have already been mentioned, viz. the generic phrasal verbs to put on and to take off. German parallels English insofar as it also has stylistically marked verbs (in particular an-/ab-legen) and numerous verbs focusing primarily on the manner of getting one's body (un)dressed or otherwise (uncovered. Nevertheless, although two further German verbs, viz. anziehen and ausziehen, may be regarded as the most generic expressions available to denote the activities of putting on and taking off articles of clothing, and in this sense as analogues of English put on and take off, these verbs cannot be used to translate each and any occurrence of put on and take off. Rather, one is obliged to choose among a number of more specialized verbs, depending in particular on what kind of article is put on or taken off.

314 b. aufsetzen: Hut, Krone, Penicke, Maske, Brille,11 Kapuze, Kopfhorer, Schleier, Stirnband, Horgerdt, Helm ... put on hat, crown, wig, mask, glasses, hood, headphones (earphone), veil, headband, hearing aid, helmet... c. anlegen: Robe, Ornat, Riistung (Panzer), Orden, Schmuck, Ohrringe... put on. robe(s), vestment, armour, medals, jewelry, ear-rings... d. umbinden. Krawatte, Kopftuch, Gurtel, Armbanduhr, Schal13... put on- tie, scarf, belt, wristwatch, scarf (muffler, comforter)... e. umlegen- Stola, Halsband.. put on: stole, necklace... f. anstecken: Ring, Brosche... put on: ring, brooch...

99

(8)

a. generally ausziehen (bur • Hosentrdger), in a few cases also the stylistically slightly marked ablegen (Mantel, 'BH, 'Hosentrdger, but *Schuhe, *Socken), seldom also abnehmen (Hosentrdger) b. generally abnehmen, occasionally also ablegen (Hut, Schleier) or absetzen (Helm) c. generally ablegen, with jewelry perhaps also abnehmen d. abnehmen or ablegen e. abnehmen or ablegen f. abnehmen

Disregarding (8) for the sake of simplicity, what are the criteria for the choice of the different German putting-on verbs listed in (7)? In (7d-f) we have quite transparent descriptions of the actual activities performed (7d: t o tie round', 7e: 'to lay round', 7f: "to pin on'), whereas in (7a-c) the choice of the predicate is determined by more general principles. Aufsetzen apparently may be used for anything worn on the head if it somehow rests on the head or on some of its prominent parts (e.g. nose, ears). Still partly reflecting the basic meaning of the simplex verb ziehen 'pull, draw', anziehen is appropriate if the extremities, preferably either both hands or both feet, are, in the process of dressing, put through or into the openings of the garment, and the garment or the appliance for keeping the garment in its position is thus pulled over the extremities.14 And anlegen, finally, seems to be the marked counterpart of anziehen, and particularly appropriate for festive and martial attire, although with less strict conditions concerning the manner in which articles are put on the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

With the German verbs corresponding to the uniformly used take off there may be additional complications, at least there is no neat one-to-one relationship to the six putting-on verbs in (7):

315

2.4. The well-known case of German stellen/setzen/legen corresponding to the single English verb to put (perhaps also place) is essentially similar insofar as it is not the inherent meaning of the object argument alone that determines the choice of the verb. With certain things, more than one of these German verbs may be used appropriately. (9)

a das Buch ins Regal stellen/legen 'to put the book on the shelves' b. einen Stein auf den anderen setzen/stellen/legen 'to put one stone on top of the other'

and the difference may then lie with the position of the object referent, with its orientation in space, with its relationship towards its environment, and perhaps also with the manner in which it is moved. However, there also are nominals which are inherently incompatible with one or the other verb. (10)

a. ein Tuch auf den Tisch legen/*stellen/*setzen 'to put a cloth on the table' b ein Glas Milch in den Ku'hlschrank stellen/*legen/*setzen 'to put a glass of milk in the fridge'

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

body. (Its stylistically marked English counterpart to don may be used much more liberally as far as kinds of garment or other body covering and manners of putting these on are concerned.) In some respects the relationship between these verbs and their objects in German, thus, seems somewhat different from the situation found in Japanese, where, according to McCawley (1971: 218), "the choice of verb is dictated not by the article of clothing but by the manner in which it is put on". ls If a shoe were to be put on the head, one would indeed use aufsetzen in German, but coats and robes are put on in more or less the same manner, and one still tends to use different verbs. And if articles which are not normally used as garments are put on instead of the appropriate garment, one may still not employ the verb normally used for the proper garment if the substitute garment lacks the criterial formal characteristics (e.g. if a beach towel is used in lieu of a bath-robe, or a blanket in lieu of a coat,anziehen would still be inappropriate). It is immaterial to our main point if we nevertheless grant that the verbs in (7), and also (8), have somewhat different inherent meanings and that the differences between them are not entirely a matter of restrictions on the choice of lexical items in the object relation. As a consequence of the greater differentiation of the inventory of verbs of getting dressed and undressed in German there certainly are stronger semantic associations between predicates and their direct-object arguments than in English.

316 nn

c. den Hut aufden Kopf setzen/*stellen/- 'legen 'to put the hat on the head'

2.5. A rather clear case of an exclusively object-related choice between arbitrarily differentiated verb alternants is the opposition between schiessen and erschiessen (cf. Leisi 1975 69) If the direct object refers to one of the animals or birds that are hunted for sport or food, schiessen is the verb to be used;17 if the victim is human or another animal (in particular an animal to which humans are in some way emotionally attached), erschiessen has to be used. The classification of an animal as ±game is, however, not absolutely invariant, which is reminiscent of semantically transparent systems of nominal classification or gender where particular nouns may be classified differently on different occasions - as, for instance, in Potawatomi, where nouns which are usually members of the inanimate class switch to the animate gender if their referents are somehow personified, as, e.g., if they are addressed by the speaker (cf Hockett 1966). Although partridges, for example, belong among the fair game and accordingly can co-occur with schiessen, one may keep a partridge as a pet, and if this pet is then killed by shooting, on purpose or accidentally, it would be perfectly appropriate to use erschiessen. Thus, it is really the predicate which signals how the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Whatever the correct conditions on the use of stellen/setzen/Iegen may be, we again observe a translation-equivalence between one English verb without semantic-agreement requirements concerning its objects and several German verbs all of which to some extent arbitrarily limit the class of object nominals they may occur with.16 But this is not yet the whole story. Firstly, English does have a differentiated set of verbs roughly corresponding to German stellen/setzen/legen, viz. to stand/set/lay. Notice however, that even if we take these verbs into account, we may still conclude that English has less semantic agreement between verb and object, but this time on account of having a richer rather than a poorer inventory of verbs. Relying on the multi-purpose verb put in English enables one to disregard the semantics of the object, whereas in German one has to choose between verbs none of which is absolutely neutral as to its objects. Secondly, there is, in fact, also a possibility of evading this decision in German, afforded by the multi-purpose verb tun 'to do', that may invariably replace stellen, setzen, legen in all relevant contexts, and hence seems a perfect analogue to put. Nevertheless, tun is more general than put by some degrees, because its meaning certainly transcends the semantic domain of moving something to a certain place and position. Within the confines of this domain, due to the availability of pur, English thus seems better equipped than German for talking about manipulating objects without paying much regard to their dimensional or other properties.

317 object referent is intended to be categorized. For the corresponding English verb to shoot such categories of semantic agreement are irrelevant. In some varieties of English, it is at best the related phrasal verb to shoot down which is comparable to erschiessen insofar as it is typically limited to human victims and certain animals close to humans (which must actually be killed rather than only be shot at), apart from being applied to any animate or inanimate objects in flight (and, non-literally, to entities such as arguments). Now, shooting is not the only method of killing where German encodes in the verb which semantic class the victims are supposed to belong to, while such differentiation is neglected in the corresponding English verbs. The following examples illustrate a fairly straightforward opposition between objects referring to people and to animals: (11)

There are of course many other ways of putting animate beings to death, and most of them appear to be designed especially for humans. There are, accordingly, further German and English verbs of killing which are used exclusively or at least preferably with human objects. We may conclude that the most significant difference between German and English in this semantic area is that English lacks verbs of killing exclusively used with animal objects. The only potential counterexample I was able to find is the phrase to put to sleep (German einschldfern), but even here some informants accept objects referring to people, although it is perhaps uncommon for people to be killed by being put to sleep. Destroy, unlike its German couterpart zerstoren, is used with animal objects (pets and others) but not with human objects (at least not when referring to individuals); on the other hand, it also combines felicitously with all kinds of inanimate objects, including (though perhaps non-literally) abstract ones - where destroy again parallels zerstoren (e.g. to destroy towns/hopes - Std'dte/ Hoffnungen zerstoren). Shoot down, as mentioned above, is likewise not limited to animal, or rather animate, objects. This general pattern is, incidentally, mirrored by intransitive predicates of dying. English has verbs that are neutral with regard to the humanness of their subject (die, perish)

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a. schbchten animals, abschbchten/niedermetzeln/massakrieren. people slaughter/butcher/pole-axe/massacre, animals or people b. abstechen. animals, erstechen (niederstechen). people stab (to death): preferably people (?) c. ersaufen preferably animals, ertrdnken preferably people drown animals or people d vertilgen certain animals (vermin) and plants (weed), ausrotten/ ausmerzen. more general exterminate: lower animals (esp. vermin) and people

318 and - often euphemistic or other indirect - predicates that require their subjects to denote humans (e.g. decease, pass (on/away), depart ((from) this life), go to one's account/Great Reward, peg out, kick the bucket). German, on the other hand, again provides verbs specifically for the death of animals, viz. verenden, verrecken (?) and eingehen (which may also be used of plants), in addition to neutral (sterben, but more common with people) and exclusively human verbs (yerscheiden, entschlafen, abkratzen).16

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

2.6. Having dealt with verbs of killing and dying, it is appropriate to consider also verbs relating to putting someone or something underground, as this is a quite usual method of disposing of dead bodies. Disregarding stylistically marked variants such as inter (perhaps inhume) and bestatten/ beisetzen, the verbs most commonly used for placing dead human bodies (or their remains) in the ground are bury in English and beerdigen and begraben in German. Unlike bury, beerdigen is restricted to (formerly) human objects, which must be actually dead (cf. *jemanden lebendig beerdigen/to bury someone alive). This agreement requirement of beerdigen apparently has to do with the ceremonial connotations of this verb: it seems that beerdigen cannot be used if the object is such that it would be inappropriate to perform the Burial Service over it. Begraben, on the other hand, is not so restricted, and could therefore seem to be exactly analogous to bury with regard to semantic object-agreement. Begraben, it is true, can be used with human objects who in fact need not be dead beforehand or even afterwards (cf. jemanden lebendig begraben), and it also occurs with objects which are, or were, not human (den Hund begraben/to bury the dog, cf. the set phrase Hier liegt der Hund begraben 'there's the rub', lit. here is the dog buried), animate (das Kriegsbeil begraben/to bury the hatchet, das Haus war unter dem Schnee begraben/ the house was buried under the snow), or even concrete (die Hoffnung begraben 'to abandon hope', einen Streit begraben 'to put an end to a dispute', etc.). Nevertheless, there is reason to believe that begraben still differs from bury in that it is most typically and productively employed only with human objects. With most animals (except perhaps pets) there is a distinct tendency to resort to other verbs such as eingraben, vergraben or verscharren devoid of any ceremonial connotations. (Verscharren indeed can also be used, depreciatingly, of dead humans jemanden wie einen Hund verscharren 'to bury someone like .a dog'.) The burieddog example is certainly idiomatic, and so are the most common uses of begraben with concrete inanimate objects at least begraben cannot be used to translate standard occurrences of bury as in to bury a treasure (einen Schatz *begraben/vergraben), and even sein Gesicht in den Hdnden begraben/to bury one's face in one's hands is now obsolete and sounds

319 much more natural with vergraben, the verb specialized for non-animate, or non-human, objects. (One could say die Leiche vergraben Ho bury the corpse', but only with the implication that the corpse is regarded as a mere thing rather than as a former human being.) As to begraben in construction with non-concrete objects such as 'hope' or 'dispute', these are clearly instances of non-literal uses of the verb, which are perhaps less characteristic of the English verb bury. Thus, we can conclude that German lacks a verb comparable to English bury with regard to the absence of agreement requirements, the German verbs most commonly corresponding to bury, viz. beerdigen, begraben vergraben, and also verscharren, all more or less severely restrict the set of (concrete) nouns that may occur in the direct-object relation.

2.8. The last few sections have demonstrated that animacy or perhaps individuality or personality, i.e. categories such as ±human, tanimal, ±thing, plays a significant role in semantic agreement between verbs and objects, and there are no doubt further verbs or verb groups that could be adduced to illustrate essentially the same point. Consider only the English verb to employ and its most common German translation-equivalents Assuming that employ basically means something like 'to make

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

2.7. The lexical field of verbs of teaching/training involves a number of semantic parameters in German as well as in English, see Schenkel (1976) for an attempt to differentiate the conditions of use of the pertinent German verbs ausbilden ('educate/train/drill/instruct', to give only some possible translation-equivalents), unterrichten ('instruct/teach/train'), erziehen ('educate/train/breed'), aufziehen ('bring up/rear'), anleiten ('instruct'), anlernen ('train/teach/break in'), drillen ('drill/train'), abrichten ('teach/train/break in'). All these German verbs are found to occur with direct objects denoting humans. Abrichten has rather negative connotations as a result of the educational process, which especially aims at skills intended to harm others, the trainees are supposed to act in absolute obedience to the orders of their superiors. Although abrichten is perhaps more commonly used in the case of certain animals, its object is not restricted to non-human referents (pace Leisi 1975 • 68). There exists, however, a verb exclusively used with objects referring to animals, viz. dressieren, and apart from this semantic-agreement requirement this verb, unlike abrichten, has no particular negative connotations or inherent meaning components radically different from educational verbs more commonly used with humans. In this semantic area English again seems to lack comparable agreement restrictions. To train (up), as dressieren is most commonly translated, may co-occur with objects denoting humans and animals, and even to break in does not seem to be restricted to horses.

320

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

use of, it is not surprising that it can be used with object nouns denoting persons (in which case 'making use of is tantamount to 'giving work to (usually for payment)'), animals, things (e.g. instruments), and even rather questionable 'things' such as time, methods and the like. If we disregard the more general multi-purpose verb einsetzen, closely corresponding to English use, there is no single verb in German that can be employed with the same range of object nouns as employ. The usual verb for employing persons is beschdftigen, for employing animals as well as things verwenden or benutzen, and with nouns with a questionable thingstatus verwenden, benutzen or nutzen would appear to be the most likely choices (e.g. to employ a certain method/eine bestimmte Methode benutzen or verwenden, How do you employ your spare timei/Wie nutzt du deine Freizeitl). If the animal-or-thing verbs are used with human objects, then it is only with an effect similar to the one we have observed earlier • Fur diese Art von Arbeit verwenden/benutzen wir Sklaven/Gastarbeiter for this kind of work we're employing slaves/foreign workers' suggests that the workers are classified as things rather than as autonomous persons. The situation is essentially the same with the semantically related verb to hire and the corresponding German verbs from the semantic domain of obtaining something or someone in return for fixed payment, normally for an agreed time and purpose. Even if we disregard a number of verbs from this domain that seem to be restricted to quite specific classes of employees (e.g. (an-Jheuern 'hire sailors', anwerben and the old-fashioned ausheben, which are particularly common in the military area (ausheben was in fact restricted to plural objects referring to soldiers) and thus roughly correspond to enlist, verpflichten or engagieren, which are preferably used in the case of non-permanent employment, especially perhaps in the show business and in professional sport), the more general German verbs corresponding to hire, viz. anstellen or einstellen and mieten, again turn out to show semantic-agreement requirements not found with hire anstellen and einstellen are used for hiring persons, and mieten (or perhaps anmieten) for hiring animals and things (such as buildings or parts of them, land, cars, horses and other property), thus roughly paralleling English to rent and subsuming the more specific verb chartern/to charter, which is limited to buses, aircraft, ships and other vehicles not operated and directed by the charterers themselves. The Worterbuch der deutschen Gegenwartssprache in fact still includes some uses of mieten with human objects (einen Knecht/Soldner mieten 'to hire a farm hand/mercenaries'), but rightly characterizes these as obsolete. What is again significant is the kind of human objects that were and perhaps still are marginally possible with mieten. humans who are not exactly paradigm instances of autonomous persons.

321 2.9. The English verb to avoid is most commonly translated as meiden or ausweichen, irrespective of whether someone or something is being avoided. However, there is a further common translation-equivalent of avoid, viz. vermeiden, and this verb cannot be used with human objects. As far as I am aware, there is no corresponding semantic - agreement requirement on the more common English verbs meaning roughly 'to keep away from'. 2.10. Under the rubric 'limitations to smallest classes of objects', Leisi (1975 70) mentions English to crack and German knacken, and claims that only the latter is restricted to objects referring to nuts. Although the conditions on the use of knacken are not, in fact, that strict, the English verb is no doubt employed more liberally (12)

However, I doubt whether this case is entirely on a par with those treated previously. There may be arbitrary object restrictions with knacken (cf. 12b), but crack and knacken would also seem to differ in their basic meanings, knacken basically refers to activities of breaking something open, whereas to crack is defined as 'to make a crack, i.e. a line of division where something is broken,but not into separate parts' (Advanced Learner's Dictionary).19 German has two stylistically differing nouns and participial adjectives to refer to this same state, viz. Sprung/Knacks and gesprungen/ angeknackst, but, as far as I know, no verb to denote the activity that produces it anknacksen sounds somewhat peculiar (except perhaps with limbs), sprengen or springen are impossible in this sense, and (zer-)brechen would imply that the cup, the window and the like are broken into separate parts. If to crack and knacken are, thus, dissimilar in basic meaning, this would weaken the argument that these verbs crucially differ only in that the German item has more specific semantic-agreement requirements. But what would still have to be accounted for are the co-occurrence restrictions of the German verb that are illustrated in (12b). 2.11 To stay in the same semantic field, English break and German brechen/zerbrechen are again clear cases where semantic-agreement requirements are more strict with the German verbs.20 Break and brechen can be used, non-literally, with more or less the same range of non-concrete

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a. to crack nuts/coconuts/a safe/ a code Ntisse/Kokosmisse/einen Safe/einen Kode knacken b. to crack the shell of an oyster/ a mussel/an egg/one's skull *die Schale einer Auster/*eine Muschel/*ein Ei/*sich den Schddel knacken c. to crack a cup/the window *eine Tasse/*das Fenster knacken

322

objects (e .g. strike, heart, will, law, word, record, silence, peace; a noticeable exception, however, is to break a journey - eine Reise *brechen/unterbrechen), but as far as literal uses with concrete objects are concerned, brechen and zerbrechen, unlike break, are restricted to things consisting of brittle and rather inflexible material: (13)

a. to break one's leg/neck/a branchfroma tree - sich das Bein/den Hals brechen/einen Zweig von einem Bawn brechen b. to break the teapot/the window-pane - die Teekanne/die Fensterscheibe zerbrechen c. to break the rope/a string - *das Seil/*eine Saite (zer-)brechen

2.12. Talking of picking flowers, it seems that there is an English verb that is exclusively used for flower-picking, viz. to cull, whereas German lacks a corresponding exclusively floral verb of picking: pflu'cken takes objects denoting flowers as well as fruits (cf. Leisi 1975 69f.). However, if we consider a larger set of verbs of gathering and harvesting, we probably end up with more semantic-agreement restrictions in German than in English. Unlike pick or pluck, pflu'cken for instance is restricted to flowers and fruits (e.g. apples, tomatoes, beans, strawberries, hops, chestnuts, also cotton and tea, but not plants with the fruit growing underground (such as potatoes) nor vegetables such as lettuce or cauliflower nor plants such as mushrooms, eine Henne rupfen/*pflucken 'to pluck a hen', *einen Faden vom Anzug pflu'cken Ho pick a thread from one's coat'). Sammeln

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Zerreissen 'to tear in two' is the verb to be used with things like ropes or strings. As to the difference between (13a) and (13b), the prefixed verb zerbrechen more strongly than the simple verb brechen suggests that the object is actually destroyed, broken to pieces (although, interestingly, brechen can occur intransitively with the same implication die Achse/ das Eis/die Teekanne bricht 'the axle/the ice/the teapot breaks' - • 'die Achse/das Eis/die Teekanne brechen 'to break the axle/the ice/the teapot'). There are a few set phrases with brechen where we would accordingly expect zerbrechen. den Stab liber jemand brechen 'to pronounce sentence of death on someone' (lit. break the stick over someone), eine Lanze fur jemand brechen 'to stand up for someone' (lit. break a lance for someone). Brechen in fact also occurs with objects referring to things which do not necessarily qualify as brittle or inflexible. Blumen brechen 'to break, i.e. pick, flowers', Brot brechen 'break bread', but these are also highly idiomatic expressions and certainly archaisms rather than standard, productive uses of brechen, and therefore do not invalidate the assumption that there are stricter verb-object agreement requirements with brechen/ zerbrechen than with break.

323

is much less general than gather, they share a large set of possible object nouns (e.g. gather firewood/Brennholz sammeln, gather followers/Anhdnger sammeln, etc.), but in the area of eatables, sammeln is used with little else but mushrooms and wild-growing berries, and with bees getting honey (die Bienen sammelten Honig), and it also cannot be used with flowers (unless you are a collector). For gathering hops (cf. *Hopfen sammeln) there is in addition to pflucken another verb, viz. zupfen, which does not seem to be used with other plants or fruits. The semantic domain of gathering and harvesting and also the entire area of botanical (folk-)taxonomy are too complicated to be dealt with in sufficient detail in this paper. It is obvious, at any rate, that it would be premature to conclude on the basis of the single verb cull that this is an area where semantic verb-object agreement is more characteristic of English than of German.

(14)

a. Elektrizitdt/Wdrme erzeugen - generate electricity/heat (also produce heat) b. Hass/Spannung erzeugen -generate hatred/suspense c. Kriminalitdt/Armut erzeugen - engender crime/produce poverty d. Milch/Eier/landwirtschaftliche Produkte erzeugen - to produce milk/eggs/agricultural products e. Zellstoff/Stahl... erzeugen - produce cellulose/steel...

(In German, hervorbringen or the most general produzieren would be alternative options in most of these cases.) Apparently there are kinds of products, especially non-agricultural and non-industrial ones, which are not particularly suitable objects of produce, although they are perfectly appropriate with, and in fact paradigm instances of the use of, erzeugen. Produce, on the other hand, also takes objects which are likewise not exactly industrial or agricultural products, such as success, sensations (in the sense of surprise and excitement), and films and plays, where erzeugen would be altogether inappropriate But this, it seems to me, is only an aspect of a more fundamental difference between English

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

2.13. Verbs of producing form another large semantic field which is not always neatly organized and the boundaries of which are occasionally difficult to determine. There are quite general and roughly corresponding verbs in English and German (e.g. produce/produzieren and herstellen, make/machen), and there are also corresponding verbs specialized for use with quite specific kinds of products (such as children: beget (or procreate)/zeugen).21 If we consider one of the more general German production verbs, erzeugen, and its English equivalents with various classes of object nouns, it looks like we have found here an area with stricter agreement requirements in English:

324

2.14. Schliessen (or zumachen) is usually translated as close, with some objects (cf. 15a), but not with others (cf. 15b), shut can be used as well. (15)

a. die Tur/das Fenster/die Augen/den Mund/die Schublade/das Buch/das Geschdft/das Theater schliessen (or zumachen) close/shut the door/the window/the eyes/the mouth/the drawer/ the book/the shop/the theatre b. den Stromkreisjdie Reihen/die Diskussion schliessen - close/ *shut the circuit/the ranks Ithe discussion

Now, are we entitled to conclude that object nouns such as those in (15b) are impossible with shut because of an object-related agreement requirement on this verb that does not apply to close nor to its German equivalent^)? I think not. Instead I would assume that shut on the one hand and close/schliessen on the other, although semantically closely related, do not share an important meaning aspect, and that it is this aspect which is responsible for the more or less limited choice of object nouns. Shut means roughly 'to move something in order to stop an opening', and its

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

and German in this semantic area with the exception of the non-native produzieren, there is no single German production verb that could be used with a range of objects as wide as that of English produce (notwithstanding the gap noted in (14a-c)), instead, there are a small number of high-frequency verbs, viz. herstellen, erzeugen (also hervorbringeri), and anfertigen (the prefixless variant fertigen is less usual), each of which is preferred or avoided with particular kinds of objects (although there are object nouns compatible with two or more of them). Herstellen, for example, is avoided with objects denoting goods or products not undergoing some manufacturing process (*Eier/*Milch/*Ol/*Weizen/*Strom herstellen 'produce eggs/milk/oil/wheat/electricity (generate)').22 Erzeugen is not used23 with objects denoting industrially manufactured or handmade goods such as hats, gloves, clothes, furniture, cars, books etc., which are the domain of herstellen and partly also (an-)fertigen. The objects of anfertigen are generally artifacts (e.g. clothes, pieces of furniture jewelry, portraits; cf. also eine Kopie anfertigen' 'make a copy', eine Liste anfertigen 'draw up a list'), with an emphasis perhaps on the individuality of the products. Clearly, a much more thorough comparative analysis of production verbs would be required to confirm the impression that with respect to this lexical field as a whole semantic verb-object agreement is again more characteristic of German than of English. But I think it has been shown that the object-related restriction on the use of produce illustrated in (14), which does not apply to erzeugen, is not necessarily representative of this entire field.

325

(16)

to close the road/the bridge/a river - *die Strasse/*die Brucke/*den Fluss schliessen

Sperren is the verb to be used instead with such objects.24 But what is the common semantic denominator of the nouns that cannot occur with schliessenl What these nouns have in common, as opposed to those which may be objects of schliessen, is that they refer to spaces which are normally passed through and/or within which someone or something moves in a

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

object identifies the opening or, more specifically, what is moved to stop it. If a box, an eye, a book or a shop is being shut, it is strictly speaking the lid of the box, the eyelids, the book covers or the shutters which are moved, but it seems that such nouns referring to wholes (box, eye, book, shop) simultaneously refer to the constituent parts of these wholes if these are moved from one position (open) to another (not open). Although close and schliessen, if used liteially, may be just as appropriate as shut under certain circumstances, their lexical meanings are more general, so that these verbs can also be used for activities where shut would not be appropriate. If close and schliessen can be taken as essentially referring to the activity of stopping (possibly by decree) something from being open, then it should be possible to use these verbs whenever the predicates 'open' and 'not open' are applicable to an entity, irrespective of whether this entity itself is actually being moved in order to stop the opening. 'Open/Not open' may be true of windows and doors as well as of the room or building they are openings of, of gates as well as the level crossing they may block, etc., and whenever the predicate 'not open' is true of windows, doors, gates, it is also true of the rooms, buildings, level crossings of which the windows, doors, and gates are constituent parts (unless of course there are further openings which are still open). Thus, if German lacks a verb corresponding to shut, this is presumably no lack in semantic-agreement requirements vis-a-vis English, but rather a lack of a verb to encode the meaning 'to move something in order to stop an opening' in addition to a verb with the more general meaning 'to stop an opening'. (In fact, it seems that zumachen is occasionally less appropriate if the object is not actually moved, unless the intended meaning is 'to close/shut down' die Schranke schliessen /zumachen 'to close/shut the gate' - den Bahnubergang schliessen/zumachen 'to close/*shut the level crossing', die Tu'r schliessen/zumachen 'to close/shut the door' -dasHaus schliessen/ zumachen 'to close/ *shut the house' (although zuschliessen/abschliessen would be preferred in this case if the door is actually locked).) There are, however, a number of object nouns which are compatible with close but not with schliessen (or zumachen), and this seems to point to a genuine semantic-agreement requirement with no parallel in English

326

2.15. We now turn to a case where English and German both seem to have equally well developed lexical fields, viz. to the notorious verbs of cooking.26 If we can assume that the number of cooking verbs in English and German is approximately the same (English in fact seems to have a few more), we are not likely to encounter a situation similar to those considered before, with a one-to-many correspondence between English and (selectionally more specific) German verbs. Nevertheless, English and German can again be shown to differ with respect to semantic agreement between verbs and direct objects, it is the internal structure of the lexical field of culinary verbs itself which manifests the difference. Presupposing that Lehrer's (1974 61 ff.) analysis is essentially correct, these are the parameters required to differentiate the main English verbs denoting ways of preparing food by heating

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

certain direction rather than to spaces (such as rooms, theatres, shops, stadiums, harbours, or envelopes, coats) which are entered, occupied, and left, without an emphasis on passing through them or on directed movement within them. Thus, if a noun refers to a space where 'not open' suggests that the passage through it or the directed movement within it is blocked, sperren is the appropriate verb of closing, whereas schliessen or zumachen do not admit objects of this kind. There are nouns which appear to be inherently indeterminate as to whether 'passage through/ directed movement within' or 'entrance into-stay-exit from' is their dominant meaning aspect, and which are therefore reasonably appropriate objects of both verbs. Tunnel 'tunnel' is a good example of this category (einen Tunnel schliessen /sperren) vehicles and people may pass through tunnels, which, on the other hand, closely resemble caves and similar habitats which are stayed in rather than merely passed through. And there are also nouns which are not really indeterminate or inherently vague (like Tunnel), but equally well admit of either classification,25 with the result that the use of schliessen {zumachen) or sperren correlates with a difference in meaning. Locks and borders are pertinent examples, a lock can be conceived of as a space through which boats pass (die Schleuse sperren, accordingly, implies that boats may not pass, i.e. not even enter, the lock), but also as a space where boats are staying after entering and before leaving (die Schleuse schliessen, accordingly, suggests that the gate is being closed after a boat has entered or left the lock), analogously, borders can be regarded as points of entrance into or exit from a country where one is going to stay or where one has stayed (die Grenzen eines Landes schliessen 'to close the borders of a country') as well as an area one has to pass through on one's way into or out of a country (die Grenze sperren).

327

(17)

a. ± use of water or water-based liquid: + boil, simmer, stew, poach, braise, steam -fry, saute, deep-fry, broil, grill, charcoal, bake, roast b. ± use of oil (fat, grease). + braise, fry, saute, deep-fry - all others c. ± use of vapour rather than liquid + only steam d. amount of cooking liquid used: large deep-fry, small braise, saute e. cooking action: vigorous boil, gentle simmer, stew, poach f. kind of source of heat • radiant broil, grill, charcoal, roast conducted bake, roast g. cooking time long stew, short saute h. special utensil pot with lid braise; rack/sieve steam, frying pan fry, grill/griddle grill, oven bake i. special purpose: to soften stew, to preserve shape poach Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Unless one is a linguist or a gourmet, one may well be surprised that these are the parameters necessary and sufficient to distinguish the English cooking verbs one might naively have expected the kind of food being prepared to play a more prominent role in the choice of the appropriate verb. And there in fact are indications that the nature of the object referents also has to be taken into consideration in order to use some such verbs felicitously, to poach vegetables or to stew eggs sound odd although these collocations do not obviously contradict the meanings of poach and stew as specified in (17),27 considering that both rice and potatoes are compatible with fry, it seems a little surprising that only potatoes but not rice is compatible with saute, to toast is primarily applicable to bread, but also to wheatgerm, sprouts, and perhaps all other kinds of food which get brown and crisp when toasted (although they are also eatable untoasted), according to Lehrer (1974 34), the size and shape of the food are, among other factors, relevant for the choice of broil or roast; and the choice between bake and roast, if the cooking is done in an oven, is also to some extent contingent on the kind of food prepared (Lehrer 1974. 182 speculates that roast is favoured with juicy or moist foods, hence to roast meat/duck but to bake ham/fish, but it is still perfectly normal to roast chestnuts, coffee, potatoes etc.). All in all, however, these are still comparatively minor factors in the hierarchical structure of this lexical field, the higher-level differentiations are achieved by means of parameters which are of a more or less instrumental or environmental character. If some cooking verbs are associated with particular kinds of food, then this is only derivatively so, in cases where the criteria of how and where something is being cooked would already seem to exclude foods of a certain kind.

328

(18)

boil 1 gentle 2. slow, gentle

simmer

1 gentle

open fire/griddle spit frying pan much fat 5 oven

- (fritiereri) backen

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In her discussion of tentative cooking term universals, Lehrer (1974. 166) points out that languages may tend to use certain verbs for certain foods, although, interestingly, the shape of the food is unlikely to be employed as a criterion of classification And she also presents data from languages where the kind of food being cooked is relevant for the choice of a cooking verb on a much less peripheral level than in English. In Amharic, for example, the five cooking verbs, disregarding the general verb to cook', may only be used in connection with specific food categories, viz. with liquid food ('boil'), solid food ('boil'), bread ('bake'), meat (fry, roast'), and grain ('parch'). Lehrer (1974. 157f.) also briefly analyses the lexical field of cooking in German, and employs essentially the same environmental and instrumental criteria as in English. She mentions, however, that dunsten/schmoren/ddmpfen 'stew/steam/braise' (her translations) are only applicable to solid food (hence *Suppe diinsten/ schmoren/dampfen). Anyone who has ever had to rely on bilingual dictionaries in order to translate culinary verbs from German to English or in the reverse direction has presumably made two observations: firstly, there are hardly any one-to-one correspondences (apart from poachpochieren, saute-sautieren, if German dictionaries happen to include these verbs), and secondly, it often seems that in German one explicitly has to add instrumental or manner specifications, which in English appear to be part of the meaning of the verbs themselves. To present only some of the correspondences that I have encountered in some dictionaries

329

(19)

braten SOLID FOOD outranks -WATER, FAT, DIRECT HEAT e.g. -Eierj' Pfannkuchen/*Pfannkuchenteig braten - fry eggs/ pancakes/pancake batter (thus, braten seems to combine, though rather uneasily, with food that is transformed from a fluid to a solid state in the proces of frying) backen FOOD CONSISTING OF/COVERED BY DOUGH (but perhaps also chicken, fish, fruit) outranks -WATER, -OIL/-FAT, CONDUCTED HEAT e.g. *Wurste backen - bake sausages, einen Apfel backen - bake an apple sieden LIQUIDS outranks +WATER, -OIL/-FAT, -VAPOUR, VIGOROUS

e g *Kartoffeln sieden - boil potatoes rdsten.

SOLID FOOD outranks -WATER, ±FAT, FRYING PAN

e.g. *ein Ei rdsten - roast/fry an egg Thus, it may turn out that German has perhaps more in common with Amharic than with English concerning the relevance of semantic agreement between verbs of cooking and classes of foods cooked. 2.16 Consider, finally, another lexical field with an equally intricate structure, viz. verbs denoting certain kinds of activities of moving substances from one place, preferably from inside a container, onto the surface of another place. For syntactic and semantic reasons the field at issue is not easy to delimit appropriately, the list of verbs given by Leisi (1975 66f.) covers at best its core, sprinkle, pour - spritzen, giessen,

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Confronted with this fuzzy picture, one begins to wonder whether German really organizes this lexical field by means of semantic parameters similar to those found in English To be sure, in German just as in English (and apparently all other languages - see Lehrer 1974 164f.) the major distinction is between boiling and non-boiling (although schmoren probably shows up on both sides); but on the next lower levels contrast may be established by somewhat different criteria. And this is exactly what Leisi (1975 67, 83) suggests the condition of the food being cooked is more important for the internal structure of the lexical field in German than it is in English, where this parameter referring to object (or subject, if the verbs are used intransitively, i.e. with only one argument) nominal classes is dominated by instrumental, environmental, and manner contrasts. Even our simplified schema (18) points to a similar conclusion. It does not seem completely accidental that braten, for example, neutralizes a whole number of contrasts which crucially define part of the lexical field in English. And in the overall structure of the entire lexical field, the following feature weightings would seem appropriate for German, but not for English

330

(20)

a. +liquid object. giessen 'pour/shed/cast', trdufeln/tropfeln 'drop/drip/trickle', tropfen 'drop/drip', spnihen 'spray/sprinkle', spritzen 'spout/ splash/sqi'irt/spray/sprinkle', schwappen 'spill/slop', fluten 'flood' splash, slop, drip, flood, spurt, spout, decant, trickle (?) b. -liquid object: streuen 'strew/scatter/spread/dust', stduben 'dust', werfen 'cast/throw'29/fallen lassen 'drop'30 scatter (?), dust, strew c. ±liquid object: schiitten 'shed/cast/throw/poor', fullen 'fill', (aus-)spucken 'spurt/spit out' pour, fill, spray> sprinkle, spit (out), spill, shed, drop, spread, squirt, spatter, throw, shower, cast

An attempt to establish translation equivalences in a more precise manner would show that the English and German verbs which most closely corresponds to each other according to the above-mentioned criteria of quantity of substance and speed of movement, and perhaps further parameters differentiating manners of movement, very often differ with regard to the criterion of liquidness: the number of verbs in the neutral class (20c) is proportionately much higher in English than in German. Thus, although Leisi's choice of sprinkle and pour as his only examples may lead to a somewhat exaggerated view of the actual differences, his claim, if interpreted in relative rather than absolute terms, would seem basically correct. This lexical field is another instance where German exceeds English in semantic-agreement requirements on verbs and their objects.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

schiitten, streuen. If further verbs are included, the semantic features suggested by Leisi no longer suffice to distinguish the items in this field. My aim, however, is not to offer a more exhaustive analysis of the relevant semantic contrasts, nor to improve upon the characterization of the semantic features proposed by Leisi, vz. substance moved ±liquid, larger or smaller quantity (perhaps rather: surface covered ±completely), ±relatively quick movement. I only wish to support with a somewhat longer list of verbs Leisi's conjecture that, as in the case of cooking verbs, the condition of the substances referred to by the objects of these verbs is more important in German than in English. This is to say that in German liquid and non-liquid substances are more likely to require different verbs, whereas in English more verbs tend to be neutral with regard to this parameter of nominal classification. In the following lists, the possible translations provided for the German verbs are not exhaustive and, in particular, are not necessarily equivalent with respect to semantic agreement. Nonliteral meanings are to be excluded as far as this is feasible.28

331 3. ON THE SYSTEMATICITY OF SEMANTIC AGREEMENT

3.1. On the basis of this still rather limited set of data I should like to advance the hypothesis that German in general has more instances of semantic verb-object agreement than English. Note that this hypothesis is quantitative in nature, it does not predict that there will be no cases where English has verb-object agreement but German lacks it (carry esp. loads/wear esp. garments vs. tragen 'carry/wear' might be a pertinent example, breed esp. animals/cultivate esp. plants vs. zuchten 'breed/ cultivate/rear', and prune/lop trees and bushes vs. stutzen 'cut back trees/ bushes as well as hair and feathers' certainly are); the claim is only that on aggregate German numerically outranks English in instances of this kind of agreement. In principle it should be easy to find out whether this hypothesis is valid or not one would only have to survey the verbs, or the semantically coherent groups of verbs, which take objects in both languages and which are approximately translation-equivalent, and calculate and compare the proportions of verbs with agreement requirements. Of course there will be a number of practical difficulties; to mention only one, the notion of semantic agreement itself may have to be further clarified before we can distinguish more reliably between verbs differing in basic meaning and verbs with the same basic meaning but differing in agreement requirements (cf. crack-knacken discussed in § 2.10). However, there should be no insurmountable obstacles to a more extensive empirical analysis along the lines of our case studies in § 2, which would eventually provide a reasonably secure basis for the evaluation of our hypothesis. In carrying out this survey it would be an advantage if a qualitative dimension could be

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Let me summarize what has been shown in the preceding sections. In about fifteen instances of individual verbs or verb groups the semantic classification of direct-object arguments turned out to be relevant for the selection of particular verbs in German but not, or perhaps less so, in English. We can distinguish three manifestations of this difference in semantic-agreement requirements: firstly, two or more verbs in German may correspond to a single English verb lacking the object-oriented meaning component found with its German counterparts (cf. §§ 2.1, 2.2, 23, 2.5, 2.6, 2.7, 2.8, 2.9, 2.11, 2.14, perhaps 2.10, 2.16), secondly, English may have verbs with object-agreement requirements similar to those of their German counterparts, but may have additional, and perhaps more commonly used, verbs which neutralize this object-oriented meaning opposition (cf. §§ 2.4, 2.12, perhaps 23, 2.13), and thirdly, parameters referring to semantic classes of objects may have a more prominent status in the structure of lexical fields of predicates in German than in the corresponding fields in English (cf. §§ 2.15,2.16, perhaps 2.13).

332

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

added to our statistical prediction, not least because one would thereby avoid having to occupy oneself with large parts of the vocabulary of English and German which upon comparison may turn out to be irrelevant for this hypothesis. Are there, for instance, reasons to expect some semantic classes of two-or-more-place predicates to be more promising areas of comparison? And are some categories of argument classification more likely than others to be involved in the choice between particular predicates? My impression is that both questions can be answered in the affirmative. As to predicate classes, it is probably no coincidence that all cases analysed in § 2 involved only verbs of activity. If the generalization turns out to be valid that direct objects of verbs of activity are more likely to require semantically agreeing verbs than objects of verbs of experience or perception,31 the obvious question is. why should this be so? Given that agent-patient and experiencer/perceiver-stimulus role configurations are both construed syntactically as subject-(direct) object configurations, we would first of all have to observe that semantic agreement cannot be accounted for sufficiently in terms of grammatical relations as such, the relational-semantic content of the direct-object relation apparently has to be taken into consideration as well.32 I do not know whether it suffices to say that predicates are more likely to manifest semantic agreement with more patient-like objects than with more stimulus-like objects, or whether additional factors appropriately subsumed under the notion of semantic transitivity (e.g. aspectual or rather 'Aktionsart' differentiations of predicates such as ±perfective, tpunctual, individuation of arguments) or even finer distinctions of predicate classes (e.g. achievement, accomplishment, activity predicates) may also, and perhaps more crucially, influence the likelihood of the occurrence of agreement.33 But I think the examples considered tend to support the view that predicates where the object referents are prototypical patients, i.e. are under the influence/ control of their actively involved co-participant and are thoroughly affected/effected by what is happening to them, are the preferred domain of semantic agreement. Pointing out the relational-semantic conditions of verb-object agreement is not sufficient as an answer to the question raised above, what remains to be explained is why predicates should tend to agree with patient-objects (in highly transitive clauses) rather than with stimulus-objects. If pressed for an answer, I would speculate that what verbs of activity refer to, viz. particular kinds of activities, may vary considerably depending on what kind of entity is involved as a patient, whereas what verbs of experiencing and perceiving refer to as such would seem to be relatively constant and independent of who or what is the stimulus. For example, all acts which can be considered as instances of killing have no doubt something in common, but what a killer actually does may vary

333

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a lot if he is killing partridges and other fair game, or lice, or human enemies. On the other hand, sensory experiences like hearing, smelling, seeing etc. as such are exactly the same no matter who or what is heard, smelt, and seen (although they may of course differ in intensity and the like). This potentially greater variability of realizations corresponding to generic types of activity could, in my opinion, be responsible for semantic-agreement patterns evincing a higher degree of interdependence of the choices of verbs and nominals in the case of activity-patient relationships than in the case of experience/perception-stimulus relationships - provided such agreement requirements are characteristic of a language at all. As to relevant categories of argument classification, it is probably no coincidence that none of the cases analysed in § 2 have involved agreement with respect to categories such as the colour, smell, taste, or sound of the object referent I specifically mention these non-attested categories since a priori they would seem to be perfectly reasonable categories with which to classify a large number of argument expressions. And although they are perceptual categories, one could expect them to be relevant for semantic agreement regardless of whether the agreeing predicates denote activities or experiences and perceptions. If perceptual salience and classificatory usefulness alone are insufficient criteria for an appropriate delimitation of the range of potential semantic-agreement categories, the question is whether this range can be delimited at all in a principled manner. As an attempt to answer this question in the affirmative I suggest that only those kinds of categories which can be identified in the nounclassifier systems of recognized classifier languages may be relevant for the kind of predicate-object agreement that we are concerned with M As has recently been argued by Keith Allan (1977), noun classifiers in these languages fall into seven broad categories material, shape, consistency, size, location, arrangement, and quanta. Allan draws on John Locke's distinction between 'primary' and 'secondary qualities of bodies' in his attempt to explain why only this particular set of categories is employed for purposes of nominal classification, the characteristics referred to by classificational categories must be perceivable by more than one of the senses, and the senses that are especially relevant for the 'primary qualities' appear to be sight and touch. One could take issue with Allan concerning the perceptual basis of the category he labels 'material', which is supposed to include under the rubric 'inanimacy' such particular noun classes as 'tree/wooden object', 'body part', 'food', 'implement', 'boat/vehicle', and 'residual/general class'. However, rather than try to break down this obviously heterogeneous, to some extent functionally rather than perceptually based super-category, let me indicate very roughly how our cases of verbobject agreement fit in with this categorial system

334

§1 §2.1 §2.2 §23 § 2.4 § 2.5 §2.6 §2.7 §2.8

§2.13 §2.14 §2.15 §2.16 §3.1

Provided the categories available in principle for the purposes of verbobject agreement can thus be limited, although admittedly not very drastically, we could further ask whether the particular categories with respect to which certain predicates agree with their objects are arbitrarily chosen. Why, for example, are consistency and perhaps material rather than shape, size, or quanta the agreement categories for German verbs of cooking? Such preferences again do not seem to be fortuitous: arguments may be classified according to different categories, but predicates may agree with arguments only with respect to such categories as are particularly salient vis-a-vis the meaning of the predicate. Thus, food-stuff itself may very well be classified as to shape, size, quanta, or consistency, but, in view of our cooking and eating habits, the last mentioned category acquires particular salience in the context of cooking verbs. And from the considerations at the end of the preceding paragraph, concerning the variability of the realizations of particular generic types of activity depending upon the kind of patient involved, further criteria could be derived to limit the set of potential agreement categories If the categories available for semantic verb-object agreement are essentially similar in kind to those likely to be found in genuine noun-

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

§ 2.9 §2.10 §2.11 §2.12

massacre, quanta (number), myt'/stirat'/promyat'. material (body part ...) sagen/erzdhlen: material (artifacts?)3s affected/effected objects. material (artifacts) anziehen/aufsetzen .... material, shape, location (headwear, body wear) stellen/setzen/Iegen. size (?), shape (dimensionality) verbs of killing. material (individuality, animacy) begraben/beerdigen ...: material (individuality, animacy) dressieren .... material (individuality,animacy) beschdftigen/verwenden .... anstellen/mieten: material (individuality, animacy) meiden/ausweichen • material (individuality, animacy) knacken: shape, consistency brechen/zerreissen material, consistency verbs of gathering and harvesting, material, shape (?), location (?) verbs of producing* material (artifacts) schliessen/sperren: material, shape, arrangement (?) verbs of cooking consistency, material giessen/streuen .... consistency, size (?), quanta carry/wear, material, location; breed/cultivate, material (animacy); prune/lop. material (trees)

335

classifier systems, one must ask on what grounds recognized classifier systems are distinguished from the kind of agreement we have been considering here. The obvious criterion for classifier languages would seem to be that they possess a set of overt markers ofnoun classification, rather than simply a number of basic verbs with semantic co-occurrence constraints possibly involving the same classificational categories (cf. Allan 1977.289). Thus far we have been talking of semantic verb-object agreement as if its covert, purely lexical nature could be taken for granted. Although this is sometimes appropriate, notice that in several instances in German the choice was not between two (or more) completely different verbs, but between prefixed and prefix-less variants of the same lexical item: (21)

And even if there is no straightforward alternation between prefixed and basic verb, we often find prefixed verbs in cases of agreement: (22)

er-zdhlen (§2.1), be-enden (§2.2); an-ziehen, auf-setzen (§2.3), ab-stechen - er-stechen (§2 5), ver-scharren, be-/ver-/ein-graben, be-erdigen (§2.6), be-schaftigen, ver-wenden, an-/ein-stellen (§2.8),aus-weichen (§2.9);zer-reissen (§2.1 l),her-stellen (§2.13)

In fact one could easily multiply German examples where prefixed or basic verbs are chosen in accordance with the semantic class of the object, with animacy being a factor of particular importance (23)

storen 'disturb someone/something' - verstoren 'disconcert someone', drosseln/abdrosseln (inanimates) 'throttle (down), slow down (e.g. engines)' - erdrosseln (animates) 'throttle, choke, strangle', dnicken (animate and inanimate objects) 'press, squeeze' - bednicken (only animates) 'afflict, depress, oppress', merken 'notice something' - bemerken 'notice something or someone'

Thus, verb prefixes in German could be analysed as forming at least a residual system of overt noun classification. These verbs-cum-classifiers are reminiscent of, in particular, classificatory-verb constructions that are perhaps best known from the Athapascan languages, where a set of formatives in the verbal group varies with the class membership of, among

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

graben - um-/aus-graben, malen - an-/be-malen, brennen -an-/verbrennen, schliessen - ab-schliessen, zwingen - er-zwingen, antworten - be-antworten (§2.2), schiessen - er-schiessen, schlachten - ab-schlachten, (tilgen) - ver-tilgen (§2.5), nutzen - be-nutzen (§2.8), meiden - ver-meiden (§2.9); brechen -zer-brechen (§2.11), zeugen - er-zeugen, fertigen - an-fertigen (§2.13)

336

3.2. Having put forward the hypothesis that semantic agreement between verb and direct object in general is less characteristic of English than of German, and having briefly considered which predicate classes and which categories of argument classification are most likely to be involved in semantic agreement, we are still left with the question of whether or not the differences between English and German are fortuitous, in the sense that no universal or typological theory of grammar and lexicon should be able to predict which languages are likely to manifest more such agreement than others. In the case that our quantitative hypothesis should prove correct, we could still do no more than conclude that a whole lot of minor lexical (plus word-formational) differences between English

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

others, direct-object arguments (cf. again Allan 1977, with further references). Disregarding genuinely morphosyntactic gender agreement, comparable systems of noun classification by means of formatives placed in the verbal group have also been reported from Arawakan languages such as Palikur, Terena and Waura (Derbyshire 1982) and from Iroquoian languages such as Mohawk or Onondaga (Bonvillain 1974, Woodbury 1975), where the classifying formatives are analysable (synchronically in Iroquoian and at least diachronically in Arawakan) as incorporated nouns specifying the general class intended to subsume the referent of the autonomous noun-phrase argument (e.g. 'I vehicle-bought a bike'). Also comparable to German verb prefixes are the preverbs of the South Caucasian language Georgian, which, although usually encoding perfectivity, direction, orientation etc., sometimes function as number classifiers (singularity vs. plurality) of 'goal' or 'patient' arguments (cf. Schmidt 1957). One need not go so far as to claim that German really has a sufficiently productive verb-prefixal system of overt noun classification to find it highly significant that this particular type of verb-related formatives, which to a considerable extent participate in the encoding of aspectual or 'Aktionsart' categories (cf., for example, schlafen 'sleep' - einschlafen (ingressive) 'fall asleep', bliihen *be blooming' - auf-/er-bhihen (ingressive) 'blossom' - verbliihen (terminative) 'wither', schneiden 'cut' - zerschneiden (completive) 'cut up') and of variations of object selection and transitivity (cf. examples such as horen with accusative 'hear' -gehoren 'belong to''/gehorchen 'obey' with dative, dienen dat. 'serve' - bedienen ace. 'serve', folgen dat. 'follow, obey' - verfolgen ace. 'pursue'Ibefolgen ace. 'obey', wohnen in live in, reside in/at' - bewohnen ace. 'inhabit'), should also be able to help encode contrasts pertaining to the agreement between verbs and semantic classes of objects.36 It will become clearer in the following section that these multiple functions of verb prefixes are not entirely coincidental.

337

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

and German are mysteriously following the same pattern, should we not succeed in finding additional properties of the grammar or lexicon of English and German which can be shown to correlate with an aversion or a propensity to semantic agreement in a cross -linguistically significant manner. Without such correlates, incidentally, it would seem rather difficult for language learners to figure out the agreement requirements of predicates: in principle, a learner may always assume that any verb-object collocations that (s)he has not happened to come across so far in the speech of his/her models are nevertheless potentially well-formed, as long as (s)he lacks explicit evidence to the contrary. In other words, (s)he is hardly likely to hypothesize relatively specific constraints on the use of verb-object combinations unless (s)he already expects that there must be such constraints in the first place, having been alerted to them by individual features or the overall structure of the language to be acquired. Apart from offering many acute observations, Leisi (1975: 77-9) in fact also comments on semantic agreement as a characteristic feature of particular languages, and on the possible forces behind this overall preference for linking verbs and objects through common classification. Leisi calls those verbs 'expressive' which may only be used in combination with subjects or objects of particular semantic classes, and those verbs 'rational' which are independent of any arbitrary semantic conditions referring to subject or object. Verbs may be expressive to varying degrees, depending on the specificity of their agreement conditions. In accordance with the predominance of expressive or rational verbs, whole languages may be characterized as expressive or rational, and Basic English and French are adduced as paradigm cases of rational languages, whereas Modern English and German are claimed to be more on the expressive side. That in Leisi's opinion the preference for rationality or expressivity is not an entirely arbitrary choice becomes obvious if one takes his view of the properties, of the advantages and disadvantages, of rational and expressive verbs seriously. Leisi holds that the separation of argument and predicate expressions, of argument-specific and predicate-specific classificational categories, is in part an artifact of linguistic representations; what is perceptually given are events or states as a whole. But since this linguistic distinction of autonomous argument and predicate expressions is to some extent arbitrary to begin with, different linguistic representations may also differ in the ways in which, consequently, they carry out this separation. Expressive verbs, then, are the result of an inconsistent, non-rational (vis-a-vis the logic of linguistic representation) association of conditions on the use of arguments with predicates. But there are some factors which, according to Leisi, favour this inconsistent attitude: expressive verbs, emphasizing the unity of predicate and argument, provide a more adequate, sensuous, poetic expression of the basic perceptual unity,37 and

338

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

also allow for the creation of 'indirect' metaphors (e.g. The bike capsized). Rational verbs have their advantages too: they are more logical concerning the functional separation of predicate and argument, and they are useful from the viewpoint of economy since one can make do with fewer verbs if these can be employed without regard for the semantics of their arguments. Given these properties, the choice of-rational or expressive verbs may also involve quite subjective stylistic considerations, and thus may be an area of interindividual variation. If one language in its entirety, then, is more rational than another, this must be due to the fact that its speakers, for whatever reason, value logic, rationality, economy higher than expressivity, sensuousness, poeticality. I doubt that Leisi's account, here briefly summarized, can be taken seriously as an explanatory theory. What at first sight may look like a somewhat impressionistic characterization of different 'cognitive styles' (Hymes 1961) motivating a well-defined difference in linguistic structure, quickly turns out to hinge entirely on the labelling of the two kinds of verbs. The whole psychological motivation of the predominance of one or the other of these verb types is convincing only to the extent that the labels 'rational' and 'expressive' are appropriate for the two verb types. Leisi in fact adds two further labels for the expressive type, viz. 'primitive' and 'archaic'. Quoting once more the thirteen or so different verbs for washing in one primitive language, he concludes that in our own languages there are agreement requirements in the case of expressive verbs which are equally primitive. The manifestation of cognitive styles in language is certainly an important question, and it may indeed be appropriate to treat the linguistic phenomenon at issue, semantic agreement between predicates and arguments, under this heading. Leisi's psychological notions of rationality and expressivity per se, however, would not seem to be able to contribute a lot to an explanatory account of why the incidence of verbobject agreement differs in two languages such as English and German. Thus, rather than speculate about the possibility that the English (and not only the speakers of Basic English) might be more rational than the Germans, I suggest we ought to look at relations between predicates and arguments from a wider perspective in order to explore whether the frequency of verb-object agreement is an independent variable or rather a concomitant feature of other cross-linguistic differences in this area. I contend that verb-object agreement does not in fact belong to the realm of minor, fortuitous, and cross-linguistically unpredictable differences between individual languages, but correlates with the typological parameters of subjectivity and, in particular, of object-differentiation. Since these parameters, although occasionally employed in typological research past and present, are not necessarily self-explanatory, I shall briefly outline how languages may differ in the manifestation of the

339

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

grammatical core relations of subject, direct, and indirect object.38 Typologically, the essential point is that the notions of grammatical subject and direct and indirect object cannot be supposed to be equally relevant for all human languages: some languages manifest such genuinely grammatical categories, some more prominently, some less so; others do not manifest them at all. An important determinant of this kind of typological variation seems to be the lexicon, viz. particular aspects of the meaning of predicate expressions. An argument of a predicate may have various statuses, which can be distinguished as relational-semantic and pragmatic. The semantic relations of arguments can be characterized on various levels of abstractness there are relatively abstract role types such as agent, patient, experiencer, stimulus, instrument, less abstract, i.e. more intimately tied up with individual predicates, are role specifications such as killer - victim (associated with a predicate meaning 'to kill'), lover - beloved ('to love') and the like Interrelated but not necessarily coincident with such statuses are such (again more abstract) relational-semantic notions as 'more/less/ least influential participant', 'primarily responsible participant', 'participant initiating an event', 'participant in control of what is happening', 'more/less thoroughly affected (or effected) participant'. Whereas the agent - patient and killer - victim role-type configurations may not vary with a given predicate, the last-mentioned kinds of relations are not uniquely determined by predicates but are to some extent variable, e g the participants playing the roles of agent/killer and of patient/victim with a predicate denoting a relationship of killing may alternatively be regarded as responsible for what happened (compare A lorry killed five pedestrians/Five pedestrians were killed by a lorry and Five pedestrians got killed by a lorry) The pragmatic statuses may be distinguished as indexical and informational To mention only the most straightforward aspect of indexical-pragmatic structure arguments may not only refer to entities external to the speech-event (third 'persons'), but the speechevent participants themselves, viz speaker and addressee(s), may simultaneously be involved as participants in the event, process, state etc. denoted by the predicate The informational-pragmatic organization of utterances includes statuses such as old/new information (i.e. already/ newly activated referent), definiteness (uniquely identifiable referent), focus of attention (topic)/comment, and probably focus of contrast and frame of predication (background/foreground information). These pragmatic distinctions all constitute inherently asymmetric oppositions, arguments with the statuses of speech-event indexicals (speaker/addressee reference), already activated referents, definiteness, focus of attention, and probably frame and focus of contrast, regardless of their semantic relations, are in some sense primary constituents of discourse, arguments

340

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

not having one of these statuses are in this respect secondary constituents. Since the various pragmatic statuses need not necessarily coincide, different arguments of a predicate may be primaries in different respects; nevertheless, largely on account of the obviously egocentric, or at least speaker/addressee-centric, bias of human discourse (speaker and addressees are prototypically already activated, definite, the focus of attention), there is a tendency for single arguments to assume the status of pragmatic primaries in general. Now, grammatical rules and regularities may be stated with reference to all these argument statuses including all possible combinations of semantic and pragmatic statuses, and the (empirical) question then is under which circumstances an additional structural level, that of genuinely grammatical relations such as subject and direct and indirect objects, has to be assumed. Of course there is no logical or empirical necessity to recognize such additional relational concepts, but it seems that for certain languages at least some grammatical rules and regularities can be formulated more perspicuously in terms of grammatical relations. These, even if they can be shown to be empirically necessary, have to be defined in terms of the patterning of relational-semantic and pragmatic statuses rather than be considered basic and undefined categories only coincidentally related to semantic roles and pragmatic primaries. There is no need to recognize subjects if the choices of semantic-role configurations and the assignment of pragmatic primehood statuses according to the requirements of discourse are in principle independent of one another, i.e. if the choice of a role configuration associated with a particular predicate has no implications, other than perhaps statistical ones (experiencers and probably agents are frequently speech-event indexicals and the focus of attention etc.), for the distribution of pragmatic statuses among the referents in these roles. A primary grammatical relation, or subject, has to be recognized only if pragmatic primehood is integrated with relational-semantic structures in a particular way. there must be preferences for the distribution of pragmatic primehood statuses determined by individual predicates. That is, given a particular predicate and the corresponding semantic-role configuration, there must be one argument which is designated as the preferred candidate for the assignment of pragmatic primehood irrespective of discourse considerations, so that the relations holding between argument(s) and predicate are no longer purely semantic but an amalgam of relational-semantic content and pragmatic-primehood privileges. An argument with these lexically predetermined constant pragmatic privileges need not be chosen as actual primary on each occurrence of the respective predicate in discourse. Any assignment of actual primehood to an argument in a semantic role without these lexical privileges is, however, marked vis-a-vis the choice of pragmatic primary in accordance with the lexical preferences stipulated by a

341

(24)

a. He opened the door with a key b. A key opened the door (*by him) c. The door opened (*by him)

(25)

a. He sold the book b. The book sold well

(26)

a. He hung pictures on the walls b. Pictures were hanging on the walls (*by him) c. The walls were hanging with pictures (*by him)

(27)

a. In 1979 we witnessed twenty big firms go bankrupt b. 1979 witnessed twenty big firms go bankrupt (*by/*to us)

There are, or at least were, a number of predicates in English which do not qualify as subjective, all of their arguments being equally eligible for pragmatic primehood and the alternative choices of primary, therefore, not being conditional upon semantic or other constraints (cf. e.g. This dress becomes her/She'd better become her dress if she .... This theory faces some problems/Some problems face this theory, A minister edified

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

basic predicate. Marked constructions such as passives and antipassives are, accordingly, not found with predicates where all arguments (and even terms not holding a predicate-determined argument relation) are equally eligible for pragmatic primehood statuses. Thus, predicates can be called subjective if their semantic argument roles are not equivalent as to the availability for pragmatic-primehood statuses; and (syntactic) subjects are those arguments which are chosen as actual pragmatic primaries with such predicates,39 basic or unmarked (lexical) subjects being those arguments with lexical primehood privileges. Languages can also be characterized in toto as more or less subjective depending on the proportion of subjective predicates and perhaps also on the number of grammatical rules and regularities referring to subjects rather than to relational-semantic or pragmatic statuses as such. For the present purpose we need not go into the question of the generalizations underlying the selection of basic subjects in different types of languages (such as the ergative and accusative types). What ought to be mentioned, however, is the possibility of marked subject choice without also employing a marked, non-basic (passive, antipassive) form of the predicate, which seems to be available to different degrees in different subjective languages. There are a great number of predicates in Modern English which nicely illustrate this possibility of second-option subjects.

342

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

them/'A minister ... by whom they can edifie (1657), He dwarfs great PompeylBy him great Pompey dwarfs (1833)). However, predicates such as open, sell, hang, witness would still seem to differ from such nonsubjective verbs, insofar as they designate one argument role as lexically preferred pragmatic primary (cf. the unmarked constructions 24a-27a): the alternative choices of primary in conjunction with an unmarked, basic verb-form (cf. the b- and c-constructions in 24-27) are all subject to more severe semantic and/or configurational constraints than the lexically preferred distribution of primehood statuses In some of these cases, one argument role (the second-option subject) can only be chosen as primary if another argument role (the basic subject) has no overt representation and is at best semantically implied by the predicate (cf. 24b, 26b, 27b), and in some, if not in all, cases the alternative constructions of a basic predicate also differ semantically in that the second-option subject role, unlike the same role in an unmarked basic-subject construction, must be filled by an argument with particular semantic-role properties such as those mentioned above, viz. responsibility or capability (cf. 25b and perhaps 24b,c) and thorough or total involvement (cf. 26c, 27b). Although it seems justifiable, in view of such considerations of semantic and configurational markedness, to regard predicates admitting of second-option subjectivization as still subjective, it ought to be noted that the relational-semantic content of the grammatical relation of subject decreases in specificity to the extent that arguments in various rolerelationships (rather than, say, only agents) may assume the subject relation with basic predicates. Turning to objects, we must again ask under which circumstances we are entitled to recognize direct and indirect objects (and only these will concern us here) as distinct grammatical relations rather than as merely semantically and perhaps pragmatically differentiated argument roles. Objects may be distinguished according to a variety of parameters all of which, independently or in combination, may be reflected in the patterning of grammatical rules and regularities. To mention only some of these potential distinctions, there are separate abstract role types such as patientobjects and recipient- or beneficiary- or also experiencer-objects, provided such distinctions actually correspond to different kinds of participation in situations rather than being dependent on the syntactic rendering of situations (as examples like She dealt him (recipient?) a blow/She hit him (patient?) could seem to suggest). There are distinctions of objects according to semantic properties of the respective referents such as animacy (e.g. person-objects vs. thing-objects), or according to informationalpragmatic statuses such as definiteness, or according to some syntactic feature of the respective arguments (such as nominal vs. pronominal character). Objects may also be distinguished according to semantic factors

343

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

such as degree of involvement (cf. / drank the wine vs. / drank of the wine), or - and this is a more comprehensive parameter - according to the degree of semantic transitivity of the clause in which they occur (cf. especially Hopper & Thompson 1980). Clausal transitivity in this sense is a complex property referring to factors such as kind of predicate (action predicates being more transitive than, say, experience or state predicates), number of participant roles, wilfully responsible participation of an agent, thorough affectedness or effectedness of its opposite number, individuated referents, aspect and 'Aktionsart' (perfective aspect and punctual verbs being more transitive than imperfective aspect and non-punctual verbs), affirmation of propositional content, and mode (realis being more transitive than irrealis). There is a further quite general basis for differentiating types of objects both paradigmatically (i.e. as objects of different single-object clauses) and syntagmatically (i.e. in clauses with two or more objects), and, although not unrelated to some of the other factors just mentioned, this general differentiation of objects according to the degree of opposedness of the arguments of a predicate is in my opinion crucial for an eventual distinction of direct and indirect objects. The arguments of many twoplace verbs of activity, for example, almost by necessity refer to participants that are diametrically opposed to one another with regard to the relationship denoted by the predicate, one referent is most actively involved, the other least actively, the latter is most thoroughly affected/ effected by what is happening to him, and is thus seen as being completely under the control and influence of the former. Typical instances of polar opposedness are activities such as killing an enemy, destroying a building, building a house, writing a letter, throwing a stone, chasing rabbits, or eating haggis. The successful performance of such activities as digesting or chewing haggis, on the other hand, may already involve the food in a slightly less uninfluential, less 'passive' or inert capacity, at least in comparison with simply eating food. In general, in activities with two participants which are less than diametrically opposed, there again is a most active participant, but its opposite number is more appropriately characterized as less active vis-a-vis the least active participant of the polar opposites, as less completely under the influence and control of the agent. Answering, obeying, following (as opposed to pursuing or persecuting), helping, thanking, meeting (with), avoiding or giving way to someone are typical examples of activities where one would not normally think of the participants as polarly opposed to one another. In the case of syntagmatic differentiation, the relative differences in meaning are the same, viz. 'less active, less completely under the control/influence' vs. 'least active, completely under the control/influence', only the polar and non-polar opposites occur in one and the same participant configuration.

344

(28)

polar, jemanden rufen 'to call/summon someone (ace.)' - non-polar: jemandem rufen 'to call/shout to someone (dat.)'

As far as German is concerned, this is clearly the minority pattern: most German verbs, unlike rufen, lexically stipulate that their objects can only be of one particular type, and in this sense they can be said to govern polar-opposite and nonpolar-opposite objects, which are encoded with the accusative and dative respectively (just as in the case of non-governed objects of the appropriate semantic type). There frequently are pairs of verbs with a common basic meaning but differing with respect to the lexical determination of their objects as polar or non-polar opposites. The

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

For example, with activities such as sending someone a letter, telling someone a story, stealing someone a horse, giving someone an apple etc., the speaker has to make a choice as to which of the two object-referents present is to be rendered as the polar opposite of the most active participant. Although in most situations it could seem that this will not be the person involved as addressee, recipient, beneficiary, or victim, this is by no means a foregone conclusion: in principle, either choice appears to be possible in such cases, given the appropriate circumstances (e.g. if an addressee is literally flooded with letters, he rather than the letters could appropriately be represented as the polar opposite of the sender(s), as most completely under the influence of his/their activity). Now, presupposing that objects can be differentiated in this manner in a particular language, what are the additional conditions under which polar-opposite and nonpolar- opposite objects may acquire the grammatical status of direct and indirect objects? Although we have so far avoided the notion of government, this is not to say that predicates can be entirely disregarded as determinants of the (semantic/pragmatic) status of objects; the suggestion merely is that predicates alone do not, or do not necessarily, determine this status. I have mentioned examples of predicates implying argument relationships which are almost by necessity and unalterably either of the polar or of the nonpolar opposedness type; but these object statuses in such cases are not due to arbitrary lexical properties of the predicates: the relational meanings encoded by such predicates simply are compatible only with polarly or with non-polarly opposed arguments. However, predicates can also encode relational meanings which do not uniquely require the argument relationships compatible with them to be of the polar or of the nonpolar opposedness type. The lexical meaning of the German verb rufen, for example, is such that its two arguments may either be polar or nonpolar opposites, and it would therefore be inappropriate to assume that this verb inherently governs a particular type of object:

345 members of such verb pairs may be formally unrelated (cf. 29a), but more often they turn out to be morphologically transparent variants involving verb prefixes, which were seen to be employed also for purposes of verbobject agreement and of aspectual or 'Aktionsart' differentiation (cf. 29b,c) (29)

The English translations are intended to bring out the semantic differences between the members of such pairs; but of course in order fully to justify the contention that the essential distinctive feature is that accusative objects are invariably more polar opposites than dative objects, a much closer analysis of these verbs would be required, focusing attention in particular on their typical contexts (including the kinds of nominals they typically occur with, etc.). But this illustration must suffice for the present purpose, and we can now attempt to characterize the grammatical relations of direct and indirect object: given that a language differentiates objects as polar and non-polar opposites, an argument is in the grammatical relation of direct object if it is predetermined by the predicate to be a polar opposite, and in that of indirect object if the predicate requires it to assume the status of a non-polar opposite. It is also impossible here to discuss in detail the generalizations potentially underlying the selection of direct and indirect objects. Although these selections always depend on the meaning of individual predicates (e.g with a predicate meaning roughly 'use-for-killing' the instrument is a much better candidate for the status of polar opposite than with a predicate like English to kill), there are no doubt certain factors, such as 'referent undergoing a change of state or location' or 'total involvement', which in general increase the likelihood of an argument being chosen as polar opposite/direct object. For instance, the relationship between a person

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a. jemanden/etwas unterstiitzen 'to support someone/something (ace)' - jemandem helfen 'to help/give help to s.o. (dat)',/emanden/etwas meiden 'to avoid s.o./s.th. (ace.)' - jemandem/ etwas ausweichen 'to give way to s.o./s.th., parry s.th. (dat.)' b. jemanden/etwas bedienen 'to serve/wait on/attend on s.o. (ace.), operate/manipulate/handle s.th. (ace.)' - jemandem dienen 'to serve/be a servant to/perform duties for/be of service to s.o. (dat.)', jemanden/etwas verfolgen 'to pursue/persecute/prosecute/trace/trail s.o./s.th. (ace.)' - jemandem folgen 'to follow/ succeed/obey s.o. (dat.)' c. jemandem etwas liefern 'to deliver s.th. (ace.) to s.o. (dat.)' jemanden mit etwas beliefem 'to supply s.o. (ace.) with s.th.', jemandem etwas rauben 'to rob s.th. (ace.) from s.o. (dat.)' jemanden um etwas berauben 'to rob s.o. (ace.) of s.th.'

346

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

and the place where he happens to live is not normally regarded as one of polar opposedness; but if the relationship of 'living-in' holds between a set of persons and the domicile they fully occupy, it is much likelier to count as one of polar opposites. Given the contextual variability of such factors increasing the degree of opposedness of an argument configuration, it seems reasonable to assume, just as in the case of predicate-determined subject selection, that with particular predicates the statuses of polar and non-polar opposite may not necessarily be assigned to individual arguments once and for all, but may be assignable to different arguments at different times - to the arguments, that is, which under the given circumstances are the best candidates for the statuses of polar and non-polar opposite (e.g. which denote the referents undergoing a change of state/location or being totally involved). This variability may be reminiscent of the situation where we have a purely semantic differentiation of types of objects independent in principle of the predicate, but it is in fact not like this situation as long as we can draw a distinction between unmarked (basic, lexical) and marked object selection. If a language is grammatically objectdifferentiating, its predicates cannot be neutral with regard to the opposedness value of their arguments. They would not be neutral, however, if they designated particular arguments merely as lexically preferred rather than as absolutely obligatory polar or non-polar opposites, allowing for the possibility that other arguments not so preferred may assume these statuses in their stead in a marked construction, provided they have the appropriate semantic properties (e.g. change of state/location, total involvement). Since the choices of direct and indirect object would then still be governed by predicates, on a preferential rather than categorical basis, we could expect the markedness of a construction, i.e. the choice of an argument as direct/indirect object which is not the lexically preferred candidate for the respective relations, to be registered by the predicate. If we recall the German examples presented above, where arguments are alternatively construed as direct or indirect objects, we notice that in many cases (29b, c) there are not entirely different verbs corresponding to the choices of direct/ indirect object but rather basic and morphologically marked, i.e. prefixed, verbs. After this very summary account of two fundamental parameters of relational typology, viz. subjectivity and (direct/indirect) object-differentiation, the next question to ask is whether English and German indeed differ with respect to these typological criteria. As to subjectivity, English and German both have grammatical subjects, but English much more liberally admits what we have called above second-option subjects. Where English subjective predicates can be freely construed with second-option subjects without any formal verb-marking (cf. 24-27), the corresponding German verbs often appear in a marked (pseudo-reflexive) form (compare

347

(30)

a. They sent letters to the President b They sent the President letters

Different people have held quite different views about the relational status of the prepositional phrase in (30a) (indirect object or non-direct, oblique object), the immediately postverbal noun-phrase in (30b) (indirect object or direct object or none of these), and the sentence-final bare noun-phrase in (30b) (direct object or non-direct, oblique object or none of these)41 which suggests that there must be considerable terminological or conceptual confusion, or else it would be difficult to understand this profusion of seemingly contradictory analyses of what looks like a relatively straightforward set of data Although I sympathize with those who deny that there are any rules in the grammar of English which have to refer to a non-subject argument in constructions like (30) in terms of indirectobjecthood, this issue is too complex to be settled in this paper. But what ought to be pointed out here is that if verbs like send should eventually turn out to govern an indirect object, they would still differ from the translation-equivalent German verbs, many of which have a three-way contrast in the construction of the argument which is potentially analysable as an indirect object, rather than only a two-way contrast as in English.42 (31)

a. Sie schickten Briefe an den Prasidenten (=30a) b Sie schickten dem Prasidenten Briefe (=30b, with dative object) c Sie beschickten den Prasidenten mit Briefen "They Z>e-sent the President (ace ) with letters'

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

with 25b. Das Buch verkaufte sigh gut lit. 'the book sold itself well'), or an entirely different verb or a modal construction has to be chosen (compare Five cars park in this garage/This garage parks five cars with Fiinf Wagen parken in dieser Garage/Diese Garage fasst (*parkt) fiinf Wagen 'this garage holds five cars' or In dieser Garage kdnnen fiinf Wagen parken 'in this garage may park five cars'), or the second-option subjectivization is at best marginally acceptable (compare with 27b: '1979 sah zwanzig grosse Firmen pleite gehen). Thus, compared with German, the grammatical subject relation in English is semantically not very specific but instead approaches what V Mathesius has called 'subjects with a purely thematic function'.40 As to object-differentiation, since German examples were used to illustrate this typological parameter, what remains to be determined is whether English predicates can likewise be claimed to govern direct and/ or indirect objects That this is no easy task as far as verbs like give, sell, send, etc. are concerned, ought to be obvious considering the perennial controversies about the 'correct' relational analysis of the two kinds of constructions exemplified in (30):

348

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In (31b)-(31c) we have clearly a contrast between a basic verb governing a nonpolar-opposite indirect object (in addition to a polar-opposite direct object) and a marked verb with the same argument now in polar opposition (thoroughly affected) to the subject. The two-way contrast found in English, on the other hand, does not seem to involve such variation of an object as to its degree of opposedness,43 but rather seems to be entirely a matter of variable informational-pragmatic status. But setting aside the question of syntagmatic object-differentiation in English, it is much less controversial to conclude that English lacks paradigmatic object-differentiation, or at least has considerably fewer two-place verbs potentially analysable as governing an indirect object than German. One might wish to claim that verbs such as belong (to), listen (to), object (to), reply (to), agree (to/ with) indeed govern an object which is non-polarly opposed to the subject. Even here it could be argued, however, that these objects after all do not behave any differently than bona fide direct objects, on account of the 'preposition' marking these objects being (re-)analysed as proper constituent part of the verbs. In other cases objects do seem to be encoded differently depending on their status as polar or non-polar opposites: He swam across the Channel - He swam the Channel, He fled from the city - He fled the city, etc.; but note that it is not the verbs per se which govern one (polar) or the other (non-polar) status of the objects. It is true, there are some verbs which are only compatible with one or the other object-status (cf. The Eskimos live in the Arctic/The Eskimos inhabit the Arctic), but on the whole this situation is again more likely to be found in German (cf. Er schwamm liber den Kanal/*Er schwamm den Kanal/Er iiberschwamm den Kami 'he swam (across) the Channel', Er floh aus der Stadt/Er entfloh der Stadt, but also Er floh die Stadt 'he fled (from) the city'). In general there is thus almost no lexical basis for claiming that English has grammaticalized (i.e. lexicalized) a paradigmatic object-differentiation as we find it in German. Rather we may recognize at best one uniform grammatical core-object relation in two-argument configurations, and on account of the lack of a systemic opposition with an indirect object, this general object relation is semantically much less specific than the directobject relation in German: an argument need not be in a relationship of polar opposedness in order to be eligible for objecthood. And this, I think, has implications even for the primary grammatical relation of subject. Recall that we observed above that the availability of second-option subjects with many predicates is much more restricted in German than in English. We can now see that this seeming difference in the grammar of subjects is presumably related to differences in the semantic constraints on the selection of direct objects. Here are a few examples of 'secondary' subjectivization in English without a German equivalent.44

349 (32)

a. The car burst a tyre - *Der Wagen (zer-jplatzte einen Reifen (instead- Dem Wagen (dat.)platzte ein Reifen (nom.)) b. The roof was leaking water - *Das Dach tropfte Wasser (instead: Vom Dach tropfte Wasser 'from the roof was leaking water') c. This caravan sleeps five persons - *Dieser Wohnwagen schldft fiinf Leute (instead: In diesem Wohnwagen konnen fiinf Leute schlafen 'in this caravan may sleep five persons' or Dieser Wohnwagen fasst fiinf Leute 'this caravan holds five persons') d. The latest edition of the bible has added a chapter - *Dieftingste Ausgabe der Bibel hat ein Kapitel hinzugefugt (instead: Der jtingsten Ausgabe wurde ein Kapitel hinzugefugt 'to the latest edition was added a chapter') e. Keegan's second goal ended the match - *Keegans zweites Tor beendete das Spiel*5 (instead- Mit Keeganszweitem Tor endete das Spiel 'with Keegan's second goal the match ended') Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

It seems that the absence of such second-option subject constructions in German is not entirely due to lexical constraints on subject selection per se, but is due to lexical constraints on direct-object selection: the predicates concerned (or, where appropriate, their non-basic variants, such as zerplatzen (32a), vertropfen (32b) beenden (32e)) require polar opposites as direct objects, and the direct objects which would correspond to those found in the English second-option subject constructions, where directobjecthood is not restricted to polar opposites, simply do not qualify for this status for semantic reasons. In conclusion, I think it is safe to assume that English and German differ significantly with respect to object-differentiation unlike German, English almost certainly lacks (at least paradigmatically differentiated) indirect objects, and correspondingly has a (direct) object relation lacking in semantic specificity. This result, then, clearly suggests that the existence, or at least the frequency, of semantic agreement between verbs and (direct) objects46 is not the only difference between English and German as far as the relations between predicates and their arguments are concerned: we find more agreement in the language, German, where the grammatical differentiation of direct and indirect objects is certainly better developed and where the grammatical relation of subject is also semantically more specific. Too little is known really about the phenomena at issue, in particular about semantic agreement, in a sufficiently wide range of languages to be able to conclude with some degree of certainty that this co-occurrence of verb-object agreement and object-differentiation (and perhaps also semantic specificity of subject) ought to be elevated to the status of a universally valid typological correlation. Pending further empirical examination of this typological hypothesis, it nevertheless seems quite reasonable

350

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

to suspect that this correlation will eventually turn out to be typologically significant and that the differences found between English and German, therefore, are more than language-particular accidents. The correlation suggested here should indeed not be too surprising, in view of our emphasis on the lexical determinants of relational typology. I have argued that it is a matter of the lexical entries of predicates whether arguments can be said to bear the grammatical relations of subject and of direct and indirect object, subjectivity and object-differentiation thus have their roots in the meaning of predicates. The grammatical relation of direct object, where it can be defined with particular predicates, has a specific semantic content ('polar opposedness') by virtue of being in contrast with the likewise semantically specific indirect-object relation ('nonpolar opposedness'), and of course with further non-direct or oblique relations. The important point is that in direct/indirect-object differentiating languages there are thus strong semantic constraints on the choice of direct objects, only arguments in relationships of polar opposedness qualify as candidates for direct- objecthood. On the other hand, if the semantic differentiation of degrees of opposedness is not systematically relevant in a language, or if it is not lexically governed by individual predicates, arguments can assume the core object relation (which may still be labelled 'direct' for the purpose of contrasting it with non-direct, oblique, less verb-dependent objects) even if they are not in a relationship of polar opposedness. In the absence of a systemic contrast between direct- and indirect-objecthood the semantic content of this kind of 'direct' object relation is correspondingly less specific. But why should predicates taking 'direct' objects of this latter kind be rather unlikely to agree semantically with them, and why should predicates governing direct and indirect objects tend to agree with these governed objects, at least with the direct ones (under the conditions outlined in §3 1)? The answer could be that governing polar-opposite (direct-object) and nonpolar-opposite (indirect-object) arguments and requiring (direct-)object arguments to be members of particular semantic classes ultimately are not entirely different properties of predicates. If predicates actually govern direct and indirect objects, they indeed require the respective arguments to be of a particular kind, arguments, and the relationships among them, must meet certain semantic conditions in order to qualify as polar or non-polar opposites. This classificational aspect of object-government is particularly obvious with a factor that has often been regarded as essential for the differentiation of direct and indirect objects, viz. animacy. Although it would be wrong to simply define direct and indirect objects as thingand person-objects respectively, animacy plays a considerable role in

351

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

determining the degree of opposedness of an argument configuration: animate beings, and especially persons, are the prototypical instances of non-polar opposites, whereas inanimates (things) are more easily conceived of as being under the influence and control of an agent and as least actively involved in activities, and thus are prototypical polar opposites. In general, with predicates whose lexical meaning is specific enough for them to be able to govern semantically specific kinds of objects, the predicate-object syntagm is thus semantically more cohesive than in the case of predicates whose valencies may be filled irrespective of finer semantic properties of arguments and argument relationships. And even if nominal classes are involved which are not, or not directly, exploited for the purpose of object-differentiation, the classificational agreement of predicates and object arguments would seem to be another manifestation of this same semantic coherence of the parts of verb-object syntagms, where the construction of cohesive wholes is oriented towards, and lexically controlled by, verbs. It may be appropriate, therefore, to characterize languages as verb-centred if relational clause structures are fundamentally determined by verbs, with verbal constituents in fact incorporating large parts of the relational frame of clauses including at least partial categorizations of the referents in relation. Object-differentiation and classificational agreement thus attest to a high degree of verb-centredness. Interestingly, overt systems of nominal classification associated with the verb group have been claimed to be particularly common in languages of the 'active' (rather than accusative or ergative) type, whose hallmark is the differential coding of intransitive arguments depending on their active/dynamic or inactive/static involvement, as e g in 'He/Him fell down', meaning 'he threw himself down' or 'he fell down inadvertently' (cf. Klimov 1977). In §3.1 Athapascan, Arawakan, Iroquoian and Georgian were mentioned as languages or language families exhibiting such classification or functionally equivalent noun incorporation; and precisely these languages, families, or the stocks or phyla containing them (Na-Dene and Equatorial in the cases of Athapascan and Arawakan respectively) are among the favourite, if not always uncontroversial, examples of the active type If the active type were defined exhaustively in terms of the intransitive alternation just illustrated, it would deserve but little interest in the development of holistic typologies. There are indications, however, that this intransitive alternation is merely a diagnostic, pointing to a more fundamental typological determinant in fact all relationships between predicates and arguments that are morphosyntactically recognized in active-type languages appear to be semantically very specific, even more so than in subjective and object-differentiating languages, with semantically relatively opaque grammatical relations such as subject and core object,

352

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a la English, playing no significant role. To the extent that this characterization proves correct, the pervasively semantic nature of relational clause structures would, thus, be a trait active-type languages share with a language such as German,47 which otherwise prefers accusative patterns although it also shows traces of the active-style intransitive alternation (Ich laufe/ protestiere/hungere etc - Mich friert/hungert etc. 'I (nom.) run/disagree/ don't eat - I (ace.) am cold/hungry'). Certainly it is not implausible to expect that semantic specificity of relations might ultimately turn out to be more crucial than grammatical object-differentiation (and subjectivity) per se as a common trait of all languages that may be described as verbcentred' the more relational and referential meaning components the verbal group incorporates, the narrower is the range of its applicability, hence the basis for generalizations of semantically unspecific relations from semantically diverse relations contracted by a verb in its various occurrences. Active-type languages are not noted for an abundance of case marking, if not entirely innocent of nominal cases, they typically appear to get along with no more than two cases, active and inactive (cf. again Klimov 1977). The present notion of verb-centredness may in fact provide a functional rationale for the scarcity of relational coding directly associated with the expressions holding the relations to be encoded, viz. the arguments of a predicate. If the functional goal is to avoid relational ambiguities, and if predicates incorporate at least partial categorizations of the referents of their arguments, as they typically do in verb-centred languages, further encoding of the relations of arguments on the arguments themselves is actually superfluous from a functional point of view. For example, encountering a predicate meaning 'to kill a human victim', rather than simply and more generally 'to kill', in construction with two arguments one of which refers to a person and the other to a wild animal, one can unambiguously compute the meaning of the whole clause ('the wild animal killed the person' rather than 'the person killed the wild animal') without the assistance of further indications of the grammatical relations or semantic roles of either argument. On the other hand, the case of German demonstrates that some degree of verb-centredness (manifested in predicate-governed object-differentiation, semantically relatively specific grammatical relations, and an inclination to verb-incorporated noun classification) is not absolutely incompatible with relational coding on arguments themselves either. Rather than going on to speculate that this German state of affairs, with semantically relatively specific relational coding distributed among verbs and their arguments, might reflect an uncertainty of typological allegiance, deviating from the pure (or ideal) type characterized by semantic and morphosyntactic agreement marking gravitating towards predicates, I prefer to wind up with a largely un-

353

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

annotated list of further differences between German and English. All of them seem to me to deserve to be taken into account as potential correlates of the high or low incidence of semantic verb-object agreement, or of the more fundamental typological parameter implying the agreement differences. First there is the 'indirect' passive {He was sent many letters), which construction is rather untypical for languages with semantically specific object relations such as German (*Er wurde viele Briefe geschrieberi). In these latter languages object distinctions tend not to be neutralized in the passive, by means of employing differential case marking {Ihm wurden viele Briefe geschrieben 'him were written many letters'), differential verb marking in direct and indirect passives {Er bekam viele Briefe geschrieben 'he got written many letters'), or resumptive pronouns (approximately as in 'He had many letters sent to him').48 Second, Raising-to-Subject/Object is more common if the relations lower-clause arguments are to be raised to are semantically not very specific (e.g. Bloggs is likely to come/*Bloggs ist wahrscheinlich zu kommen, I expect him to come/*Ich erwarte ihn (zu kommen). Instead of argument raising one tends to find lowering of attitudinal, epistemic etc higher-clause predicates to the rank of nonsubordinating adverbs, particles or parentheticals. Third, there appear to be stronger constraints on the movement of core arguments out of and/or into finite clauses if the language has semantically specific verb-governed relations (cf. The hat which I believe that he is always wearing is red/*Der Hut, den ich glaube, dass er stets trdgt, ist rot), the closest analogue one tends to get to such argument shifts are constructions with the argument concerned in a peripheral relation in the superordinate clause and with a resumptive pronoun in the subordinate clause (Der Hut, von dem ich glaube, dass er ihn stets trdgt, ist rot 'the hat of which I believe that he is always wearing it is red'). Fourth, pronominal objects appear to delete more easily, under pragmatic or syntactic control, if the language has a semantically unspecific core object relation (cf. He knows that the earth is flat but she doesn 't know (it)/Er weiss, dass die Erde flach ist, aber sie weiss *(es) nicht, I bet he's forgotten/Ich wette, er hat "(es/darauf) vergessen) Fifth, if two-place predicates are used with only one argument, i.e. intransitively, languages with semantically specific grammatical relations tend to require pro-forms, which often resemble reflexives (cf. The door opened/Die Tu'r offnete *(sichj, He and she met/Er und sie trafen *(sich), He shaved/Er rasterte *(sich), He behaved/Er benahm *(sich)*9) Sixth, instrumental objects tend to be less common in languages with semantically specific object relations - but this may simply be one manifestation of the polar-opposedness constraint on subject/direct-object configurations (cf to wag one's tail vs. mit dem Schwanz wedeln 'wag with the tail', She played the grand piano/*Sie spielte (den) Fltigel

354

Fachgruppe Sprachwissenschaft Universitdt Konstanz Postfach 5560 D-7 750 Konstanz I West Germany

NOTES 1 A bolder generalization would be to associate spatial categories with arguments, and temporal categories with predicates in general Probably this is what Leisi's (1975: 58) distinction between static and dynamic properties ultimately amounts to 2 Compare Sapir & Swadesh's interpretation with Lyons' (1968: 281) assertion that number is necessarily a category of the noun 3 Cf eg Friedrich (1970) for some discussion of inherent vs arbitrary classification, or Viehweger et al (1977: 353) on 'semic' vs 'sememic* compatibility, or also Coseriu's (1967) distinction of "Affinitat" and "Selektion" (arbitrary) vs "Implikation" (inherent) I agree with the critics of McCawley's (1971) view that there are no lexical items with identical meanings but differing only in selection restrictions, although I recognize that it is often difficult to motivate one's decision one way or the other. See also Lehrer (1974: 180ff ) on these issues 4 Although the distinction may to some extent be a matter of the integration of such categories into a system of obligatorily signalling syntagmatic relatedness, the way particular categories are utilized would still seem to depend at least partly on their conceptual structure (cf Plank 1981: ch 2)

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(...auf dem Flu'gel 'on the grand piano'); This was the first time she played on a piano/a piano/piano vs. Sie spielte zum erstenmal auf einem Klavierj*ein Klavier/Klavier). And, for the time being finally, if the core grammatical relations of a language are semantically specific, their encoding seems likelier to be synthetic, and prototypically perhaps polysynthetic, rather than analytic Thus, the semantically rather unspecific subject and object relations in English are encoded without much morphological assistance, whereas German employs both nominal (case) as well as verbal (prefix) morphology to take care of its subjects and objects. As most of these differences involve rules or regularities governed by individual predicates or classes of predicates, it is not so surprising that they should correlate with subjectivity and object-differentiation, or with the semantic specificity of grammatical relations these latter properties themselves are contingent on predicate conceptualizations. However plausible they are in principle, all of these correlations, and especially those linking traditional morphological typology to more fundamental syntactic and lexical typologies (as suggested in our last correlation), still need to be empirically examined in a much wider range of languages, before we can rule out the possibility that we are faced after all with an arbitrary set of minor and accidental differences between German and English, rather than with a system "ou tout se tient".50

355

nn

(9b): • -Em Stein sitzt aufdem anderen) 17 Or of course the slightly more technical erlegen, which is also restricted to animals but does not necessarily imply shooting (at least for many native speakers) 18 In English, incidentally, die and perish are used with plants as well, although there also are specifically botanical verbs German absterben also has interesting agreement requirements: it seems to be used with plants, limbs, and perhaps certain lower animals 19 Nevertheless, there are similar extended meanings in German and English pertaining to the mental sphere: crack-brained, crackers - beknackt. Knacks

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

5 Notice, incidentally, that this is not the original meaning of this verb in Germanic; it acquired this specific meaning after it had previously denoted a more general kind of movement 6 But I wonder whether one should also be prepared to exclude in principle that ride/reiten could ever be restricted, for instance, to subjects denoting horse-women, or to subjects denoting a whole group of riders 7 This universally ergative pattern has been recognized eg. by Gak (1972) and Moravcsik (1978, 1984) Although dealing specifically with noun incorporation, Mardirussian (1975: 387) also mentions the possibility that such patterns have a semantic rather than genuinely syntactic rationale 8 Cf Plank (1980b), in particular on ergative patterning in word formation Concerning explanations, Leisi (1975: 65) speculates that transitive subjects (agents), referring almost exclusively to humans and perhaps animals, are less likely than are objects to vary a lot in material, shape, and weight, and that objects (patients) therefore play a more crucial role in semantic verb-agreement Cf also §3 1 9 Cf Ludwig (1979), who, drawing on notions suggested by Labov and Waletzky, argues that the constituent parts 'complication' and 'resolution' are necessary for a narrative to count as an Erzahlung; if this is correct, my 'creative effort1 should be interpreted accordingly Some of Taylor's (1980) remarks on telling vs saying are also relevant here 10 Although a noun class 'artifacts' would properly delimit the set of nouns likely to be used as effected objects Human, or animate, nouns are likely to occur as effected objects, except with subjects referring to some superhuman creator and the appropriate verbs of (pro)creation (on problems attending these see Plank 1982) 11 Cf Paul (1959: 239ff ) for an almost exhaustive list of German verbs whose objects may be affected or effected 12 Interestingly, my German and English informants could not make up their mind about which verb would be most appropriate with monocle IMonokel to put on/aufsetzen was usually rejected, which may point to a different categorial status of glasses and monocles One informant tells me that the verb to use with monocle is insert 13 If a Schal is not exactly tied around the neck, it is rather difficult to find an appropirate verb; most informants did not accept einen Schal anziehen Pace Stern (1968: 380), anziehen is not used with ties and belts, unless they are pre-tied or prefastened so that one can actually slip on these appliances 14 Accordingly one ought to be able to use anziehen with Muff 'muff; but this is another one of the few garments (or is it not considered a garment?) where German seems to lack appropriate verbs of putting on and taking off 15. More detailed accounts of Japanese dressing verbs may be found in McCawley (1978), Backhouse (1981), and Kameyama (1983) 16 Cf also the corresponding intransitive verbs stehen/sitzen/liegen, of which at least sitzen would seem even more specific than its causative counterpart (cf with

356

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

20 Cf also Leisi (1975: 64), who ignores, however, the difference between brechen and zerbrechen 21 Kinder erzeugen used to be possible, but is now definitely obsolete 22 There are indeed further uses ofherstellen where produce would be inappropriate, cf. eine Verbundung/das Gleichgewicht herstellen 'to establish a connection/the equilibrium' 23. At least not in Standard German In Austrian German, erzeugen seems to be used more liberally 24. Sperren again seems to be used more liberally in Austrian German; den Laden sperren 'to close the shop (temporarily)' is the Austrian equivalent of Standard German den Laden schliessen 25 This is of course reminiscent of dual classifications such as those considered in earlier sections (cf partridge game/pet - schiessen/erschiessen) 26 On these cf Leisi (1975: 65,83), Lehrer (1974), Newman (1975), to cite but a few pertinent references 27 If the meaning of poach, however, is as given in the Advanced Learner's Dictionary, viz 'cook by cracking the shell and dropping the contents into boiling water', it is not surprising that this verb is not applicable to vegetables 28 This means that spurt flames/Flammen spucken, for example, is to be excluded as non-literal Some of these decisions are surely debatable, but I do not think this affects the point being made here 29 I thus do not agree with Leisi (1975: lOOf), who holds that werfen, unlike throw, is necessarily "akt-bedingt"; for throwing dice on the table, using a dicebox, werfen seems to me perfectly appropriate Notice, incidentally, that in spite of the complex nominal Wasserwerfer 'water-cannon' (lit. 'water-thrower'), *Wasser werfen is definitely impossible 30 Eine Trine auf den Brief fallen lassen 'to drop a tear on the letter' looks like a counterexample since tears are obviously +liquid That a tear is an individuated unit of liquid cannot be the explanation because einen Tropfen Wasser fallen lassen 'to drop a drop of water .' is still odd 31 There were also no verbs of possession among those with object-agreement, which may be an accidental gap in our data In general, verbs of possession may shade off into the domain of activity as well as into the domain of experience and existence, and since most languages have more than one way of expressing possessorpossession relationships, it seems reasonable to expect that different kinds of possessions may require different kinds of expression (e g different possessive verbs) 32 Cf the introduction to §2 concerning other patterns (idiom formation, derivation, noun incorporation) which may also have to be stated in semantic terms 33 On semantic transitivity see especially Hopper & Thompson (1980), and Plank (1980a: §2 2) for problems this conception encounters with grammatical objectdifferentiation 34 Probably semantic agreement in other types of constructions is subject to similar restrictions 35 In fact noun classifiers are on record which distinguish specific literary forms or forms of oral or written presentation (cf Adams & Conklin 1973) in a manner that is reminiscent of the sagen/erzahlen distinction 36 An affinity between aspect (or 'Aktionsart') and object-agreement-like encoding is by no means a German idiosyncrasy. Since the Athapascan languages have been mentioned as an example of a particular kind of noun-classifier languages, it is worth mentioning that classificatory verb stems in Navajo have been assumed to be actually bi-morphemic, consisting of an aspectual and a classificatory element (cf

357

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Landar 1965) And the Georgian verb prefixes also attest to the naturalness of the association of aspectual and classificational functions 37. Recall what was said above about the relative semantic coherence of verbs of activity and their patient-objects 38 My views on subjects and objects are presented in more detail, including comparisons to other views, in Plank (1979, 1980a, 1982, 1983, 1985) Fuller treatment of the attendant typological correlations has to be deferred to a forthcoming monograph 39 Being a subject can then still be a matter of degree, depending on the number of pragmatic-primary statuses assigned to an argument 40. Cf. Mathesius (1929, 1975) Nevertheless, subject in English is still not a purely thematic notion according to our conception of subjectivity; if it were, the notion of a grammatical subject would actually be superfluous, and all pertinent regularities would have to be stated with reference to the informational-pragmatic status of theme (old information, focus of attention) 41 To mention only some recent studies reaching quite different conclusions: Perlmutter & Postal (1977), J. Anderson (1977, 1978, 1984), DeArmond (1978), Ziv&Sheintuch(1979) 42. Cf also the examples in (29c); the third, prepositional, alternative in these cases is: etwas an jemanden liefern, etwas von jemandem rauben 43 It does, however, in the case of constructional alternatives such as load hay on the wagon/load the wagon with hay 44 For detailed empirical analysis of these and other pertinent cases see Rohdenburg (1974) 45 This example is not strictly ungrammatical, but its only interpretation is that Keegan's second goal was somehow responsible for the end of the match, rather than merely coinciding with it - and this is an instance of polar opposedness! 46 We have not considered the possibility of semantic agreement with indirect objects Since only particular classes of nominals are suitable for this relation anyway (essentially humans), I doubt that there will ever be much agreement variability with predicates governing indirect objects 47 Aronson (1977) tries to associate Modern English, as opposed to Old English, with languages of the active type, on account of the common lack of a classification of verbs as transitive or intransitive With regard to many other typologically significant features, including classificational agreement between predicates and arguments, this characterization of Modern English seems to me rather unfortunate But then Aronson modifies the traditional concept of the active type (due especially to Klimov 1977) considerably, so that his alignment of English with this type may not really mean much 48 Cf Givon (1979: ch 4) for a similar interpretation of such passive variants 49 It is perhaps somewhat inappropriate to include absolute reflexives among twoplace predicates, but the point is that translation-equivalent verbs of this kind tend to differ, just like the other verbs mentioned, as to whether they require an overt 'reflexive' marker Note also that the correlation in the case of object pro-forms may primarily be with subjectivity rather than with object-differentiation as such. 50 This paper, based on Chapter 3 of my M Litt thesis (Plank 1980a), dates from August 1980, and underwent some minor cosmetic surgery in September 1984 Over the last five years, the manuscript has been used for numerous oral presentations and has been circulated among various colleagues, by myself and others Some of the reactions to it were sympathetic and/or helpful (those of Johanna Nichols, Edith Moravcsik and the editor and referee of this Journal, for example), others were neither

358 REFERENCES

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Adams, KL & N F Conklin, 1973: Toward a theory of natural classification Papers from the 9th Regional Meeting, Chicago Linguistic Society. CLS, Chicago. Pp 1-10. Allan, K., 1977: Classifiers. Language 53: 285-311. Anderson, J M , 1977: On Case Grammar Prolegomena to a Theory of Grammatical Relations Croom Helm, London Anderson, J M, 1978: On the derivative status of grammatical relations. In: W Abraham (ed ), Valence, Semantic Case, and Grammatical Relations Benjamins, Amsterdam Pp 661-694 Anderson, J.M , 1984: Objecthood In: Plank (ed ): 29-54. Aronson, H 1, 1977: English as an active language Lingua 41: 201-216. Backhouse, A, 1981: Japanese verbs of dress Journal of Linguistics 17: 17-29 Bonvillain, N , 1974: Noun incorporation in Mohawk In: M K Foster (ed ), Papers in Linguistics from the 1972 Conference on Iroquoian Research National Museum of Man, National Museums of Canada, Mercury Series, Ethnology Division. Paper No 10, Ottawa. Pp 18-26 Coseriu, E., 1967: Lexikalische Solidaritaten. .Poerica 1: 293-303. DeArmond, R C, 1978: On the indirect object in English In: J E Hoard & C Sloat (eds.), Proceedings of the 8th Annual Meeting of the Western Conference on Linguistics Linguistic Research Inc , Carbondale, 111 Pp 7-13 Derbyshire, D C, 1982: Arawakan (Brazil) morphosyntax Work Papers of the Summer Institute of Linguistics, University of North-Dakota Session, 26: 1-81. Friedrich, P., 1970: Shape in grammar Language 46: 379-407 Gak, V.G , 1972: K probleme semanticeskoj sintagmatiki In: SK Saumjan (ed ), Problemy struktumoj lingvistiki 1971 Nauka, Moscow Pp 367-395 Givon, T , 1979: On Understanding Grammar Academic Press, New York Grimm, J., 1853: Geschichte derdeutschen Sprache. Hirzel, Leipzig (2nd ed ). Hill, A A , 1952: A note on primitive languages UAL 18: 172-177 Hockett, C.F , 1966: What Algonquian is really like UAL 32: 59-73 Hopper P J & S A Thompson, 1980: Transitivity in grammar and discourse Language 56: 251-299 Hymes, D H., 1961: On typology of cognitive styles in language (with examples from Chinookan) Anthropological Linguistics 3.1: 22-54 Jespersen, O., 1922: Language, its Nature, Development and Origin Allen & Unwin, London Kameyama, M., 1983: Acquiring clothing verbs in Japanese Papers and Reports on Child Language Development (Stanford) 22 Klimov, G A , 1977: Tipologija jazykov aktivnogo stroja Nauka, Moscow. Landar, H , 1965: Class co-occurrence in Navajo gender UAL 31: 326-331 Lehrer, A, 1974: Semantic Fields and Lexical Structure North-Holland, Amsterdam Leisi, E , 1975: Der Wortinhalt Seine Struktur im Deutschen undEnglischen Quelle & Meyer, Heidelberg (5 th ed) Ludwig, O , 1979: Berichten und Erzahlen - Variationen eines Musters Mimeo Lyons, J , 1968: Introduction to Theoretical Linguistics Cambridge University Press, Cambridge Lyons, J , 1977: Semantics Cambridge University Press, Cambridge McCawley, J D , 1971: Where do noun-phrases come from? In: DD Steinberg & L A Jakobovits (eds ), Semantics An Interdisciplinary Reader in Philosophy, Linguistics, and Psychology Cambridge University Press, Cambridge Pp 217-231

359

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

McCawley, J D , 1978: Notes on Japanese clothing verbs In: J Hinds & I Howard (eds ), Problems in Japanese Syntax and Semantics Kaitakusha, Tokyo Pp 6878 Mardirussian, G, 1975: Noun-incorporation in universal grammar Papers from the 11th Regional Meeting, Chicago Linguistic Society. CLS, Chicago. Pp. 383-389. Mathesius, V , 1929: Zur Satzperspektive im modernen Englisch Archiv fur das Studium der neueren Sprachen und Literaturen 155: 202-210 Mathesius, V , 1975: A Functional Analysis of Present-Day English on a General Linguistic Basis (Edited by J. Vachek ) Mouton The Hague Meyer, R M , 1909: Verba pluralia tantum Indogermanische Forschungen 24: 279288 Moravcsik, E A , 1978: On the distribution of ergative patterns Lingua AS: 233-279 Moravcsik, E A., 1984: The place of direct objects among the noun phrase consituents of Hungarian. In: Plank (ed.): 55-85. Newman, A , 1975: A semantic analysis of English and Hebrew cooking terms Lingua 37: 53-79 Paul, H, 1959: Deutsche Grammatik Band3 Niemeyer, Halle (5th ed ). Perimutter, D M. & P.M. Postal, 1977: Toward a universal characterization of passivization Proceedings of the 3rd Annual Meeting of the Berkeley Linguistics Society BLS, Berkeley. Pp. 394-417 Plank, F , 1979: Ergativity, syntactic typology and universal grammar Some past and present viewpoints In: F Plank (ed ), Ergativity Towards a Theory of Grammatical Relations Academic Press, London Pp 3-36 Plank, F , 1980a: About Subjects and Objects M Litt thesis, Edinburgh University. Plank, F , 1980b: Die ergativische Orientierung der Derivationsbeziehungen zwischen Adjektiv und Verb im Deutschen und ihre typologische Relevanz Paper read at the 8 Jahrestagung Osterreichischer Linguisten, Salzburg. Plank, F , 1981: Morphologische (Ir-)Regularitdten Aspekte der Wortstrukturtheorie Narr, Tubingen Plank, F , 1982: Coming into being among the Anglo-Saxons Folia Linguistica 16: 73-118 Plank, F , 1983: Transparent versus functional encoding of grammatical relations: a parameter for syntactic change and typology Linguistische Berichte 86: 1-13 Plank, F (ed ), 1984: Objects Towards a Theory of Grammatical Relations Academic Press, London Plank, F , 1985: Die Ordnung der Personen Folia Linguistica 19: 111-176. Porzig. W, 1934: Wesenhafte Bedeutungsbeziehungen Beitrage zur Geschichte der deutschen Sprache und Literatur 58: 70-97 Rohdenburg, G , 1974: Sekundare Subjektivierungen im Englischen und Deutschen Vergleichende Untersuchungen zur Verb- und Adjektivsyntax Cornelsen, Bielefeld Sapir, E & M Swadesh, 1946: American Indian grammatical categories Word 2: 103-112 Schenkel, W , 1976: Zur Bedeutungsstruktur deutscher Verben und ihrer Kombinierbarkeit mit Substantiven Enzyklopadie, Leipzig Schmidt, K -H, 1957: Eine sudkaukasische Aktionsart? Mu'nchener Studien zur Sprachwissenschaft 10: 9-24. Stern, G, 1968: Meaning and Change of Meaning With Special Reference to the English Language Indiana University Press, Bloomington (3rd e d ) Taylor, M , 1980: Saying Studia Anglica Posnaniensia 12: 27-46 Viehweger, D et al, 1977: Probleme der semantischen Analyse Akademie-Verlag, Berlin

360 Woodbury, H , 1975: Onondaga noun-incorporation: some notes on the interdependence of syntax and semantics IJAL 4 1 : 10-20. Ziv, Y & G. Sheintuch, 1979: Indirect objects - reconsidered. Papers from the 15th Regional Meeting, Chicago Linguistic Society CIS, Chicago Pp. 390-403

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Journal of Semantics 3: 361-377

LEXICAL TAXONOMIES RICHARD RHODES

ABSTRACT

If it were the case that only the terms of biology participate in conceptual and lexical taxonomies, then facts about them would merely be an interesting quirk of semantics like cyclically organized lexical fields (Fillmore 1978). However, I will present evidence here that both conceptual taxonomies and the lexical taxonomies that label them are common throughout the lexicon In fact I will show that lexical taxonomies exist for verbs as well as for nominals. In this paper, however, I will go beyond presenting evidence that lexical taxonomies have an important place in the lexicon, and argue that the semanticist's traditional means of accounting for taxonomic relations, namely semantic features (Katz and Fodor, 1963), do not account for the observable properties of lexical taxonomies. Instead I will propose that the organizing principle of lexical taxonomies is the lexical relation type of. The structure of this paper will be first to discuss the properties of lexical taxonomies known from ethnobiology, then to show that there exist semantic fields with taxonomic structure besides those in the realm

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In the course of my work on a dictionary of Qjibwa, an Amerindian language of the Great Lakes region, I have come across a number of interesting lexical facts which constitute an important part of what speakers know about the reference of words but which are not generally included in semantic analysis as it is currently practiced.1 In particular, while working on plant names, it came to my attention that there exist lexical taxonomies which have an existence based on, but separate from, the conceptual taxonomies they label (and which, in turn, are separate entities from the Linnean taxonomies which the Western mind tends to believe are the 'real' truth about biology). Even a cursory examination of the literature on ethnobiology will quickly reveal the language and culture specific nature of these lexical taxonomies and the conceptual taxonomies which underlie them. Berlin, Breedlove, and Raven (1973) (henceforth BBR) outline the basic principles governing the structure of such taxonomies which they refer to as folk taxonomies.2

362

(1)

a. Taxonomies are at most 5 levels deep.4 b. Some taxa may be unlabelled. These taxa are called covert categories. c. The inclusive (the most inclusive taxon) is frequently covert.5 d. In general the morphological complexity of a taxon label is related to the position of the taxon in the taxonomy - the more complex the label the less inclusive the taxon it labels. e. It is frequently the case that some generic taxa do not belong to any kind, and f. Generic taxa have a cognitively privileged status.

Property (If) is often taken as the single most important fact about ethnobiological taxonomies. It is to the generic taxa that the philosophical notion, natural kind, corresponds

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

of biology and exhibiting the same sorts of properties as the biological ones; and finally to argue that the lexical relation type of accounts for the properties of lexical taxonomies better than features do. First let us define our basic terminology. The most basic unit of a taxonomy is a taxon. Each taxon represents a grouping of objects which are treated as being of the same conceptual type. Although Hunn (1983) has pointed out some important problems with BBR's pioneering work, something very close to their general approach will suffice here as a first approximation. Therefore we will take taxonomies to be organized such that a grouping represented by a higher level taxon constitutes the (set theoretical) union of the groupings represented by a subset of the lower level taxa. BBR worked in terms of a fixed level system. One of their results was to discover that there are at most five levels in any folk taxonomy so that five terms are needed to label the levels of inclusiveness of taxa. BBR used the terms, unique beginner (the most inclusive), life form, generic, specific, and varietal (the least inclusive). Because some of these terms are so heavily biological, we will use the term inclusive instead of unique beginner and kind instead oflife form.3 BBR present evidence that taxonomies have a number of characteristic properties which hold over a wide range of data. The primary relationship between conceptual taxonomies and lexical taxonomies is that the lexical taxonomies consist of taxon labels. In fact, some of the characteristic properties of taxonomies involve the nature of the relations between the conceptual taxa and the lexical items that label them. Now let us review what is known about lexical taxonomies from BBR's work. The basic properties of non-technical biological taxonomies are listed in (1).

363

(2)

a. Taxa, particularly at less inclusive levels, frequently participate in competing conceptual taxonomies. b. There is often significant personal variation in the structure of conceptual taxonomies and in taxon labels, particularly at less inclusive levels c. Taxon labels, under certain conditions, can move one level up (generalize) to label otherwise unlabelled taxa?

All of these properties are ultimately based on the notion that taxa are of two types, core and peripheral (Hunn 1983). The recognition that the apparent inconsistencies of non-technical taxonomies have regular sorts of properties is important, but, for the most part, beyond the scope of this paper.7 The fact that the properties listed in (1) are generalizations over a large body of data makes it very hard to present compelling arguments in support of them at all, let alone within the limits of a paper of this size. Therefore, for the most part, we will restrict ourselves to the presentation of typical examples which illustrate the properties in question in nonbiological semantic fields Property (lc) is the best place to start, since the examples supporting (la) depend on (lc) in a way that may not be immediately obvious, and since (lb) is logically subsumed in (lc). Property (lc) says that many taxonomies lack a label for the inclusive Let me list a few such taxonomies First consider the class of objects which we use to keep time. An exerpt of this taxonomy is shown in (3).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Two comments about technical sophistication are in order. First, where there is extensive technical expertise, there are frequently technical terms which can be used to label otherwise unlabelled taxa, but the knowledge of such terms depends on the degree of special knowledge on the part of the speaker. Second, there also exist technical taxonomies which speakers with expertise in a particular field will have competing with (and for many purposes preferred to) the non-technical taxonomy. It is important to recognize this difference between the non-technical taxonomies that are of interest here and technical taxonomies that are not. Technical taxonomies do not necessarily share the properties outlined in (1), e.g. their depth is not limited, they do not have unlabelled taxa, nor do they have priveleged levels Such technical taxonomies reflect more accurately the nature of the world and less accurately the nature of language and the mind. As mentioned above, more recent work (e g. Hunn 1983) has shown that non-technical taxonomies have somewhat different properties than those which BBR thought, and, in the last analysis, these will be of importance in our inquiry. They are outlined in (2).

364 (3)

clocks alarm wall clocks clocks

watches grandfather clocks

wrist watches

pocket watches

stop watches

Or consider the class of objects one provides where one moves into unfurnished living quarters. The kinds include furniture and appliances as shown in Fig. 2. At first one might be tempted to call such things furnishings, but I find it hard to call furnishings things like stoves and refrigerators on the one hand, and TVs and stereos on the other. Or consider the class of places where plants are cultivated. The kinds are farms, gardens, nurseries, and orchards, as seen in (4). Again there is no superordinate term. (4) farms pig dairy farms farms

nurseries

orchards

gardens

apple cherry vegetable orchards orchards gardens

flower gardens

Finally consider the case of those things which you find in people's wardrobes. An excerpt is given as Fig. 3. The kinds include clothes, underwear, and the unlabelled kind consisting of the class of things one wears to bed. While clothing might be considered a good candidate for a superordinate term, I again find it hard to call underwear clothing and downright strange to call the things one wears to bed clothing. The other possible collective terms wardrobe.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The major kinds of this inclusive are clocks and watches, but there is no superordinate term. Even words that at first blush look like they should be such a term all fail, for example, timepiece. I can hardly call any digital time-keeping device a timepiece 8 Next consider the class of objects used in the preparation and consumption of food in the typical household 9 The kinds of this inclusive include dishes, silverware, and the interesting term pots and pans. A portion of this taxonomy is shown in Fig. 1. Again no superordinate term exists.

platters

Fig. 2. Household items

kitchen chairs

specific

salad bowls

plates

saucers N

cups

sofas/ couchs

'

/I

love hide-a-beds [ seats

tables

furniture

dishes

folding chairs

serving bowls

chairs

easy rocking chairs chairs

soup bowls

bowls*

generic

kind

inclusive

Fig. 1. Food utensils

specific

genenc

kind

inclusive

£—A

glasses dessert forks/ salad forks

appliances'"

\

lamps

serving forks

forks

T

silverware

refrigerators stoves

/

knives

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

[

clocks

spoons

366 Levels

underwear shirts

— •

=

.

.^__— -. pants

blouses A — shorts —-T— . knickers ~-~— \ short bermuda (regular) shorts shorts shorts

A

-^^^^S

-

—7—

1

clothing ^

*

M'

dresses

•

^

-

.

—

==::

.

~

•

skirts

?sleepwear 2 *. sweaters vests 3

/v

(regular) pants ^-—-__ ^

^

/ chinos good pants

•

cords

^4

S

Fig 3

(5)

a. Taxon labels are cited in the plural if they are count so that they parallel labels which are collectives b. Square brackets indicate covert categories, i.e. unlabelled taxa c. A triangle indicates there are taxa omitted from the example for the sake of space10 d. An asterisk after a taxon label indicates that the taxa subordinate to it do not exhaust the members of its taxon.11 e. A form in square brackets with a preceding T indicates that the taxon is unlabelled except for a well-known technical term f. A slash is used to separate taxon labels which are synonymous from the point of view of the taxonomy.

Now we are ready to reexamine some example portions of the taxonomies of artifacts just presented to see that their height is five levels or fewer, parallel to BBR's observation for non-technical taxonomies in nature As the first example consider Fig 1 again which displays part of the taxonomy of kitchen items mentioned above. Fig. 1 shows only four levels, from the covert inclusive down to specifics The same thing can be seen in Fig 2, the excerpt from the taxonomy of household items.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

attire, and apparel are related to lower level taxa in a rather complicated way, but one which is based on part-whole relations rather than type of relations. As you can see from these examples, covert inclusives are reasonably common among taxonomies of artifacts as well as among taxonomies of objects in nature At this point let me digress briefly to indicate the notational conventions I have been using in presenting my taxonomies. These are outlined in (5).

367

snap

burst

shatter

smash2 crumble crack

chip splinter rip

(Note: there is one unclassified term split) Fig 4a Intransitive verbs of the impairment of physical integrity

shear

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Fig. 3, part of the taxonomy of items of apparel, can also be seen to have only five levels. It also shows a typical problem with varietals One might argue that brands and brand types, like Levis or designer jeans, and styles like bell bottoms, are subtaxa of varietals (here jeans). I doubt this conclusion for two reasons. If these are actually part of the taxonomy, such distinctions verge on the technical, but more importantly, they constitute cross classification whose hall mark is simple (i.e. compositional) modification (often with head deletion). Thus Levis correspond to plaids, designer jeans to white jeans, and bell bottoms to khakis. These examples are typical. All the taxonomies we have looked at are, like these, mostly four levels deep. Those points that have a fifth level normally have the exact same sorts of complications - questions of technicality and/or cross classification. Property (Id) regarding the morphological complexity of taxon labels can also be seen in the figures Specific labels are generally more complex than generic or kind labels, although kind labels are occasionally more complex than one might expect This probably relates to (If) regarding the privileged status of generics. Finally these figures also show generics which belong to no kind as indicated in (le). In Fig. 1 glasses and cups are generics, as are lamps and clocks in Fig 2, but none of these belongs to any kind. Similarly Fig. 3 shows shoes, socks, hats, coats belonging to no kind. In addition to the nominal taxonomies given in Figures 1-3, there are also taxonomies of events, processes, states and the like which are normally expressed in verbs n For example the English verbs expressing impairment of physical integrity are given in Figs 4a and 4b.

smashn

chip rack

^--'^tlatten / crush \ mash dent squash crumple

Fig. 4b. Transitive verbs of the unpaument of physical integrity

(Note: there is one unclassified term split.)

snap burst

mangle

np

\ slash shred

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

/ slit \ sever slice dice

^

shea dissect

oo

ON

369 In Figs. 4a and 4b we see the same sorts of things evidenced as in the previous taxonomies. The inclusive is covert There are less than five levels, and there are generics that do not belong to any kind. In fact, the main difference is that these taxonomies are much smaller. Nor is the taxonomy of verbs of breaking the only verbal taxonomy. Consider the verbs of rotary motion shown in Figs. 5a and 5b.

turn*

rock swirl

spin' whirl

wheel

swivel

pivot

Fig 5a Intransitive verbs of rotary motion

rock spin*

twirl

swirl

whirl Fig 5b Transitive verbs of rotary motion

Again the same properties obtain, covert inclusive, other covert taxa, and a five level limit. The biggest issue with verbal taxonomies appears to be the fact that they have relatively fewer taxa which makes it harder to see if a particular field is taxonomically organized or not The pervasiveness of verbal taxonomies can perhaps best be seen in Algonquian languages. In these languages the vast majority of verbs contain a verbal morpheme of very general reference which classifies the type of process, event, or state expressed by the verb. The kinds of meanings expressed in these morphemes are commensurate with English verbs like cause, fall, push/pull, move, use, grasp, and the like. These morphemes, known to Algonquianists as finals, belong to several taxonomies two of which are given in Fig. 6a and 6b

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

gyrate

370

-bi-* "push/pull'

-apt'tie'

-boo'move back and forth'

-wi-aha'transport' 'follow'

'drop/ place'

-aas'move in the wind'

-ishkoo'(indirect) cause'

-isah'(direct) cause'

Fig 6a Ojibwa (transitive) concrete finals of motion

(6)

coat

raincoat trench topcoat suit sport tuxedo smoking spring winter coat coat coat jacket coat/jacket jacket b

clothes coat suit coat

sport(s) coat

tuxedo

smoking jacket

wind breaker

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

These taxonomies represent a refinement of Rhodes (1980), where the semantics of the individual terms and their relationships within the semantic field are explained in detail. Of course we have only briefly mentioned and mostly glossed over the complexities in studying taxonomies that arise from the fact that nontechnical taxonomies are organized not only hierarchically but also into a core and a periphery In ethnobiology this shows up as residual categories, tied up with a complex of related notions called usefulness. The major premise of usefulness is that if a plant or animal/insect/bird is useless, then it is relegated to a residual category which is therefore not directly morphological (in the biological sense). The English term weed is such a residual category In dealing with taxonomies of artifacts the question of usefulness does not arise, Useless artifacts (in this special sense) are not made. There is, however, a corollary, given as (3a) Taxa of artifacts are often classified differently for different purposes. For example the taxon labelled knives as a type of tool is rather different than the taxon knives as a type of silverware. A similar thing is shown in (6) for coats as a type of garments versus coats as a type of clothes. On the other hand, for verbs there do seem to be residual categories, (cf Figs. 4a and 4b).

-am-* 'use the mouth -m'grasp

-ldah'use a stick'

- ^ -lzh'use a blade'

-agah'use a axe'

1 ^

-ah-* 'use an instrument'

-1Z-*

-aas- -aakiz'use the 'use fire' sun'

'use heat'

-oom'carry on the back'

Fig. 6b. Ojibwa (transitive) concrete finals of action

(Note: the triangle represents three ungrouped generics: -aganaam-, 'pound on';-/to-, 'cut'; and -ma-, 'pursue')

-ik-un-anem'wear' 'use 'act on s.t. speech' in the mouth'

-ishk-* 'use the body

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

-daabii'drag'

-nigaa'carry on the shoulders'

u>

372

Penultimately, let us turn to a consideration of what theoretical mechanism will account for the properties we have observed for taxonomies. As a matter of normal practice, semantic features are used to distinguish the content of forms. Such features would seem at first to provide a mechanism for building taxonomies. However, a closer inspection reveals a number of significant problems First, features are classificatory, thus the normal case is that a feature will be 'used up' - it will classify all items to which it is potentially relevant, and hence by implication non-hierarchical. But taxonomies are by nature hierarchical, and moreover, they are more restrictive than simple classification. There is no expectation that a descriptor which distinguishes the taxa of certain branches will have any relevance to taxa on other branches. That this is the case for taxonomies is illustrated in (7) where some of the descriptors for verbs of rotary motion are outlined. a. [

fixed axis

roll

turn spin whirl swivel etc. b turn

fast

u

m

swivel pivot

spin whirl wheel gyrate very fast

c. spin u

m

wheel gyrate

whirl

This means that the property of classification is not among the properties of the descriptors needed to distinguish taxa. therefore classificatory features are inappropriate to characterize taxonomies.13 A similar argument is given in Aspects of the Theory of Syntax (Chomsky 1965: 79ff) where it is argued that the opposite is true - the notion feature as it is understood from phonology should be applied in syntax.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(7)

373

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

It would be possible to have hierarchical semantic features which would support the structure of the taxonomy as well as defining the content of forms, but then there would be a need to have both simple classificatory semantic features for places where classificatory properties are needed (Fillmore 1978), and hierarchical semantic features for taxonomies. Furthermore, the two kinds of taxon label shifts, up and down taxonomies, also provide grounds for arguing against using hierarchical features to do the work of supporting the hierarchical structure of the taxonomy. Taxon labels may move down the taxonomy (specialize) when there is a subordinate covert taxon. Thus the type of undershirt which has straps rather than sleeves and is technically known as a vest has no non-technical name, but is conventionally referred to as a (regular) undershirt. If semantic features were used to support the hierarchical structure, this would require a special optional status for the relevant hierarchical feature in such cases. Similarly, the label of the least marked immediate subordinate of a covert taxon may be used to label that taxon under certain circumstances (generalize). For example, silverware and glasses are not dishes, but in expressions like wash the dishes it is the superordinate taxon which includes silverware and glasses as well as dishes that is referred to. Again, if the structure of the taxonomy were supported by hierarchical features, there would need to be a special optional status for the relevant hierarchical feature just in case it belonged to a taxon whose coordinate taxa all had longer feature lists or a higher markedness assigned by whatever the relevant measure is The second and probably stickiest problem with hierarchical features has to do with the relatively common level skipping noted in (le). Hierarchical features provide no motivated way to do this. Third, hierarchical features do not predict restrictions on depth of classification. Fourth, there is a problem with features as they are currently used relating to the fact that they do not have enough values of the same logical type to account for the variety of contrasts seen in taxonomies. There need to be at least the values + and - and the values m and u. In addition, I would argue that there is also a value am (anti-marked) needed, i e. marked as not having a descriptor The examples in (7) show some of these values Some of these problems can be solved by marking the notion type of a lexical relation and not using descriptors to establish the hierarchical relations. This will provide principled means to account for most of the properties of taxonomies discussed above with the exception of restricting depth of taxonomies to five levels. It is possible that this restriction has to do with the kinds of cognitive limitations that are usually (but I think wrongly) referred to as "performance" limitations and are therefore not to be solved by formal means

374

Before I conclude, some brief mention of pro-forms should also be made. At first blush one might think that these forms are hierarchical and therefore participate in taxonomies However, more careful consideration reveals that pro-forms are, in fact classificatory rather than hierarchical This includes such forms as stuff, junk, and probably goodies, and the like Thus man is not a type of he/him nor are books or papers a type of stuff In conclusion I would like to raise a problem of lexical analysis which I think a more developed understanding of taxonomic relations in the lexicon might solve. Among several classes of verb-object collocation there is one in which the verb may take any of a class of. loosely speaking, semantically related objects, as in the examples in (8). (8)

law rule regulation ordinance etc

b. practice

law medicine dentistry etc.

c. incur

wrath injury a penalty etc.

d. invoke

god(s) a theorem a rule etc.

I suspect these classes of being taxonomically organized in at least some cases, but I currently lack sufficient corroborating evidence, and strong enough explanations for the inconsistencies Intuitively, laws, regulations, ordinances, and commandments are all types of rules. But no two verbs collocate with them exactly the same way as can be seen in (9)

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a enforce a

375 (9) keep break obey violate enforce

law

rule

regulation

ordinance

7 OK OK ? OK

?? OK OK ?* OK

?* OK OK ?? OK

* ? OK OK OK

The problems are even tougher for some of the other examples in (8), like the class of things one can incur, or one can invoke. Nonetheless my feeling is still that taxonomic relations will play a crucial role in resolving the collocational mysteries of these expressions University of Michigan and Eastern Ojibwa Dictionary Project Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

NOTES 1 Earlier versions of this paper were read by Alexis Manaster-Ramer, John Lawler, and Fied Lupke I would also like to thank D A Cruse for his observations on the issues in this paper The paper has benefited greatly from their comments. The preparation of this paper was supported in part by the National Endowment for the Humanities grant RT20086-81-2179 2 While BBR are not explicit in separating the linguistic entities - the lexical taxonomies - from the cultural entities - the conceptual taxonomies, their argumentation makes it clear that these two things must be kept separate 3 Varietals are rare enough that the term is not a problem. 4 There is some discussion of the possibility of a sixth level in rare cases, although it is possible that such a level does not constitute a true level which raises the depth to six, or whether such cases simply have a sixth type of level which constitutes a conflation of properties of other levels with the upper limit on the depth of taxonomies remaining at five Technical taxonomies can, of course, be deeper 5 Even in those cases where the inclusive of a field is labelled, it is very often a borrowed term This probably reflects a history of the dissemination of technical knowledge as noted below. 6 Because of the hierarchical nature of taxonomies, it is always possible in principle for more inclusive taxon labels to be used to refer to items belonging to less inclusive taxa Berlin (1966) argues that the generic level is cognitively the most salient and therefore the most common, but as Hunn (1983) argues, things are not so simple I would argue that in such cases, not only are there Gricean principles at work governing the choice of taxonomic level from which a term is drawn, but there are language and culture specific conventions against which such principles operate, and these take more into account than just the taxonomic level at which a term occurs 7 Hunn uses a notion activity signature to characterize his (core) taxa The type of taxonomies we are discussing here can be used to test such notions, because they are native to us So, for example, the taxon linen (including towels and sheets) can be identified by an activity signature of where they are stored, but only marginally by one of how they are washed, because they are often washed with clothes

376

BIBLIOGRAPHY Berlin, Brent, Dennis E. Breedlove, and Peter H Raven, 1973: General Principles of Classification and Nomenclature in Folk Biology American Anthropologist IS: 214-242. Chomsky, Noam, 1965: Aspects of the Theory of Syntax MIT-Press, Cambridge. Fillmore, Charles, 1978: On the organization of semantic information in the lexicon. Chicago Linguistic Society Parasession on the Lexicon Pp 148-173. Hunn, Eugene S , 1983: The Utilitarian Factor in Folk Biological Classification American Anthropologist 84: 830-847

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

8. My use of an expression like time-keeping device might lead some to think that such phrases might constitute inclusive labels This is not the case In order to count as a taxon label a term must be lexicalized This does not, of course, mean that taxa can't be labelled with phrases (in some cultures this is common), it just means that the phrase must be conventionalized as the taxon label The same argument can be made for the expressions which title Figures 1, 2, and 3 These expressions are not lexicalized and therefore not candidates for taxon labels 9 This class of objects does not, as my description might suggest, include those devices which do the work of cooking - stoves, ovens, microwaves, grills, hot plates, etc. It is unclear to me whether blenders, juicers, food processors, and the like are in the class, but electric can openers (being can openers), and warming trays definitely are in it 10 Taxonomies which are large enough to show the properties we are interested in are all too large to be shown in full here Some contain hundreds of taxa. Most contain around 200 11 Contrary to BBR's notion of a taxonomy, there is a contrast between covert subordinate taxa and elsewhere case items belonging only to the superordinate taxon. This fact also follows from the recognition of the core/periphery distinction among taxa. A covert subordinate taxon is often ad hocly labelled in English as a regular X where X is the superordinate taxon label. It seems likely that the elsewhere case items are themselves taxa If this is the case then it represents a point in the taxonomy where an intermediate taxon is missing, which would suggest that property (le) may be simply a common special case of a more general principle. 12 It is harder to think about verbs being in hierarchical relationships to one another While asking a question like: 'Is X a type of Y?' works for nouns, verbs often sound quite strained in corresponding contexts. Only in the best cases does asking: 'Is Xing a type of Ying?' not sound strange So far I know of no other tests While it might at first seem possible to question entailments, particularly of states or resultant states, this only guarantees a hyponymic relationship, not a taxonomic relationship (D A Cruse, p.c.) So the relationship between a pair of verbs like shatter and break has to be determined by observing that shattering is a type of breaking 13. Of course it remains an empirical issue whether the descriptors which constitute the semantic description should also bear the burden of the hierarchical structure. My present opinion is that these are separate matters and that this accounts for the difference between taxonomy and hyponymy Hyponymous terms differ from their superordinates only in having more descriptors, while taxonomically subordinate terms both have more descriptors than their superordinate terms and bear the semantic relation type of to their superordinates.

377 Lakoff, George, 1982: Categories and Cognitive Models Berkeley Cognitive Science Report No 2 University of California, Berkeley Rhodes, Richard, 1980: On the semantics of the instrumental finals of Ojibwa. Papers of the 11th Algonquian Conference Pp. 183-197 Carleton University, Ottawa Rhodes, Richard, 1981: The semantics of the Ojibwa verbs of breaking. Papers of the 12th Algonquian Conference Pp. 47-56 Carleton University, Ottawa Rhodes, Richard, in press: Some comments on Ojibwa ethnobotany Papers of the 14th Algonquian Conference Carleton University, Ottawa.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Journal of Semantics I: 379-391

BOOK REVIEWS

Joachim Jacobs, Fokus und Skalen Zur Syntax und Semantik der Gradpartikeln im Deutschen (•- Iinguistische Arbeiten, 138). Niemeyer, Tubingen, 1983 Pp ix + 298. Ad Foolen

(1)

dass X Luise X der Polizei X ein Bild von Peter X zeigte that Luise the police a picture of Peter showed

Fokus und Skalen (henceforth FuS) is Jacobs' second book. Although FuS has been conceived as a self-contained book - and is very well readable as such - it may be helpful to the reader to read his dissertation Syntax und Semantik der Negation (Jacobs 1982) first, since many of the principles developed by the author in his 1982 study on negation are now applied to FA's I already mentioned Altmann's book from 1976 on FA's, and could further mention Altmann (1978), in which special attention is paid to the gerade (particularly) -group within the FA's. The work of Altmann on FA's is now considered as a classic in the German FA research, and it is very common for more recent work to take that of Altmann as a point of departure Altmann's work can be characterized as strongly data-oriented the distributional properties of FA's are investigated, and a first semantic and pragmatic description is given Jacobs takes on the one hand the many observations of Altmann as a point of departure, and on the other the formal framework that he developed in his own 1982 work on negation The result is a formal grammar (with a syntax and semantics) for FA-sentences Whereas Altmann does not choose one particular formal model to account for his observations, Jacobs clearly opts for the framework of Montague-grammar Readers who are not familiar with this formalism might have a handicap here, but not an insuperable one the author has

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

As the subtitle indicates, the book under review is on Focusing Adverbs (henceforth FA) in German. Jacobs refers to this category with the German term "Gradpartikeln", which he has taken from one of his teachers at the University of Miinchen, Hans Altmann. Altmann introduced the term in German linguistics in his dissertation, published in 1976, that had the FA's in German as its subject Auch (also), nur (only) and sogar (even) can be considered as typical representatives of this category provided they appear in certain syntactic positions, like those marked with X in (1) (with subordinate word order)

380

(2)

Das Buch ist interessant, nur miisste es kiirzer The book is interesting, only should it shorter (nur functions here as a kind of conjunction)

(3)

Werdas auch getan hat, erwird dafiir biissen Who that also done has, he will for that pay 'Whoever has done that will pay for it" (auch is a modal particle here)

sein be

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

either done his best to avoid unnecessary formalisation, or has relegated the formalisation to the Appendix (= chapter 6). I would like to mention another difference between Altmann and Jacobs, both authors mention in the subtitles of their books the syntactic and semantic components, but the pragmatic component is only mentioned by the first. It is not that Jacobs denies the existence of pragmatics, but he prefers to describe as many meaning nuances as possible within formalized semantics. A certain scepticism against pragmatics is discernible in various places between or even on the lines, as is shown from formulations like the following "Bleibt immer noch der Griff in die pragmatische Wundertute. Wenn man sich nicht scheut, sich dabei die Finger schmutzig zu machen (diese Wundertute fungiert ja bekanntlich auch als Abfalleimer)..." ("There always remains the possibility to dip into the pragmatic grab-bag Provided you are not afraid to dirty your fingers (for, as is well-known, this grab-bag functions also as waste-basket)" (p. 167) Up till now I have only mentioned publications of Jacobs and his teacher Altmann as relevant in the context of FuS From this one should not conclude, however, that the book is restricted to a 'local' discussion Although Jacobs has not taken every possible reference in the FA-literature into consideration, his list of references does contain the main titles. As to German, and German-English contrast, several articles by Konig are mentioned, among others Konig (1981), and for English especially Karttunen & Peters (1979) is taken into consideration. Jacobs also uses insights that have been developed in the recent literature on polarity-items, particularly in Ladusaw (1979) and Linebarger (1980). Although he does not opt for the framework of generative grammar, he makes use of it wherever it seems to offer insightful analyses. So far on the context of research in which FuS is embedded I will now turn to the structure and content of the book itself FuS opens with three short introductory chapters (0, 1 and 2, together 21 pages) In ch. 0 the author states that he will concentrate on auch, nur and sogar as far as they are used as focusing adverbs, disregarding other syntactic functions as for example in (2) and (3)

381

doctor will be considered as a constituent in the syntactic scope of every, but the rest of the sentence will be considered as lying outside that scope b. The semantic scope of an expression X is reflected in a logical representation of the sentence Almost every logical analysis of (4) will start with For every y, if y is a doctor . , and then an analysis of the rest of the sentence. That is why everything that comes in (4) after every is in its semantic scope In general, the semantic scope can be inferred from the left-to-right word order disregarding certain constructions, it can be said that everything that is to the right of X lies within the semantic scope of X This statement is meant for German, and might not be equally useful for English or other languages c The focus of an FA is that part of the sentence with which it is strongly associated That part must also contain the main accent of the sentence, compare (5) (5)

Auch jeder Arzt besitzt ein Auto Also every doctor possesses a car

In (5) the focus of auch can be Arzt but also jeder Arzt The context often helps to identify the focus, by mentioning 'alternative values' for the focus (5a)

Nicht nur jeder Lehrer, auch jeder Arzt besitzt ein Auto Not only every teacher also every doctor possesses a car

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In ch 1 it is shown that FA-sentences can vary considerably with respect to the position of the FA, the position of the main and secondary accents, and the contribution of the FA to the meaning of the sentence It is this strong variability that constitutes the descriptive and theoretical challenge of FA's Ch 2 is very useful as a preliminary chapter, because it tries to clarify such difficult notions as 'syntactic scope', 'semantic scope' and 'focus'. Just like quantifiers, FA's have 'scope' or a 'domain', and according to Jacobs it is necessary to give an explanation of this notion on at least three levels of analysis a. The syntactic scope of an expression X is fixed by the tree-diagram for the constituent structure of the sentence. All constituents that are dominated by the first branching node above X are within the syntactic scope of X. So, for example, in almost every analysis of (4) (4) Every doctor has a car

382

In (5a) Arzt is the focus of auch, and teacher is a contextual alternative for the focus. (5b)

Nicht nur Lehrer, auchjeder Arzt besitzt ein Auto Not only teachers, also every doctor possesses a car

Here, jeder Arzt is focus of auch, and Lehrer is the contextual alternative. The focus is indicated by coindexation with the FA and lower placed semi-brackets, as in (6) (6)

Luise onlyi showed the police a photo of

Li Peter , J

(7)

Peter bewiindert Luise nicht. sondern er beneidet sie Peter admires Luise not, but he envies her

The complex negative expressions nicht einmal (not even) and nicht nur (not only) can also be grouped with the FA's in their syntactic and semantic behaviour. Ch. 6, 36 pages, is meant as an appendix. It gives a formalized Montague-grammar for the fragment of German that is analysed in this book (with syntactic rules, lexicon, translation rules etc ), and the complete 'calculation' of example sentence (8) (8)

Auch einige Botaniker beneiden nur Luise Also some botanists envy only Luise

Ch. 6 ends with some problematic examples that the grammar presented here cannot easily handle. Added to the book there are three pages of errors, only a part of all the typing errors in the book. I will now discuss somewhat more extensively the central chapters, which present the original proposals of the book

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Chapters 3 and 4, about 100 pages each, are the central ones in the book They discuss the syntax and the semantics of the FA-sentences, respectively I will return to these chapters below Ch 5, 15 pages, on FA's and negation, can be regarded as an extra. Jacobs discusses the question of the extent to which FA's and expressions of negation show similarities in their behaviour, thereby referring back to his research on negation from 1982 It appears that 'contrastive-mc/jf' (not) behaves in most respects as FA's. A contrastive nicht can be recognized by the possibility to continue the sentence with a sondernconjunction, as in (7)

383

In ch. 3, Jacobs presents a syntax that can be qualified as relatively rich and also as relatively independent of the semantic component, at least in comparison with many other descriptive proposals in a Montague-framework To give an impression of this richness, I will mention some of the properties of the syntax presented" a. Different types of rules for derivation, which account directly for the different distributional possibilities for an FA in relation to its focus, as are visible in (9)-(l 1)Sogar Johann kommt Even John comes

(10)

Johann sogar kommt John even comes

(11)

Johann kommt sogar John comes even

b. Categories of the A/B-type, so that operator - operand relations between constituents are directly visible in the tree-diagram. c Insights from valency-grammar are built in by branching the VP per argument A sentence as (13) is thus assigned a structure as in (14) (13)

(dass) Luise ihrem Arzt ein Auto vermachte (that) Luise her doctor a car left

(14)

whereby V0 = V-zero = S, T = Term = NP, and V0, VI, V2 and V3 refer to verbal groups with respectively 0, 1, 2 and 3 open places for arguments I have elaborated a bit on this aspect, because I will need it in the discussion of the relation between syntactic scope and focus of FA's. d. Subcategorisations, which account for restrictions on the syntactic distribution of certain members of a category, for example the restriction

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(9)

384

for FA's as a subcategory within adverbs to stand on their own in the first position of a sentence: (12)

*Sogar kommt Johann Even comes John

From the aspects mentioned I will discuss one a little bit further, namely the relations between syntactic scope and focus of the FA At this point the analysis of Jacobs deviates considerably from more traditional ones like that in Altmann (1976) The syntactic rules are formulated in such a way that the FA is adjoined with its syntactic scope, and at the same time the focus of the FA is marked (by semi-brackets) In many cases the focus is smaller than the syntactic scope, and this is partly caused by the fact that in Jacobs' analysis the syntactic scope is often wider than in traditional analysis, especially when the FA focuses on one of the V-arguments. Let's have a look at sentence (15) (15)

(dass) Luise nuri Ljihrem Arzti J ein Auto vermachte (that) Luise only

her

doctor a car

left

In the FA-literature, Altmann (1976) or Konig (1981) for example, the FA and the focus-NP are usually analysed as co-constituents in such cases In Jacobs' notation, this would look like the analysis in (16)

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

e So-called meta-rules, which provide for the possibility that expressions from a certain lexical category can be used in other functions also. FA's are lexically categorized as VO/VO, i.e. as sentence adverbs, but by the meta-rules their use in ad-verbial, ad-adnominal, and ad-predicative functions is made possible (the ad-article functions seems to be accessible for FA's only in a restricted way). f. Syntactic and semantic filters on the output There is, for example, a syntactic filter that restricts the possible foci for an FA in certain syntactic positions, and a semantic filter that accounts for certain cooccurrence-restrictions. For example, it is not allowed that sogar lies in the semantic scope of sogar, nur, auch, gern, or niemand. g. The syntax contains at least one (semantically neutral) transformation, namely Verb-second. Sentences are generated in the subordinate clause word order and in case of main clauses transformed by Verb-second.

385 (16)

Luise

nur

ihrem Arzt

ein Auto

vermachte

Jacobs, however, proposes a different analysis, in which the FA is attached to a separate V-node, so that the whole part of the sentence that is to the right of the FA comes in the syntactic scope of it, as in (17) (17)

(18)

Nur Luise vermachte ihrem Arzt ein Auto Only Luise left her doctor a car

and also when a non-subject-argument and the FA are topicalized, as in (19) (19)

Nur ihrem Arzt vermachte Luise ein Auto Only her doctor left Luise a car

Jacobs presents a number of distributional data in support of his analysis, like the following a. Nur ihrem Arzt and other such supposed NP's cannot occur as complement of a preposition

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

At first sight, this analysis might look a little bit strange. It has the unpleasant consequence that contrary to the traditional view, it allows two constituents before the verb in main clauses. This situation can occur when an FA has the subject as its focus, as in (18)

386 (20)

*von nur ihrem Arzt from only her doctor

This observation is interpreted by Jacobs as an indication that nur ihrem Arzt might not be an NP-constituent at all. b. The focus can be placed sentence-initially without its FA(21)

L Ihrem Arzt J hat Luise nurj ein Auto vermacht Her

doctor had Luise only a car

left

If we consider (21) to be the result of a transformation from a structure like (16), then we should accept a violation of the generative A-over-Aprinciple. c How would, in a traditional view, the analysis of a sentence like (22) look like, where the FA has multiple focus (Luise vermachte nicht ihrem Astrologen ihr Haus, (Luise left not her astrologer her house), sie vermachte nurj L j ihrem Arzt j J Lj ihr Auto j j she left

only

her

doctor

her car

Besides the validity of these kinds of supporting arguments, it can be said that the proposed syntactic analysis of FA-sentences provides for an elegant link between syntactic and semantic scopes It is also nice to see that in this analysis FA's show greater similarities with other subclasses of adverbs than the recent literature tends to suggest. Future research will have to show whether Jacobs' analysis will retain its credibility if more extended fragments of German and other languages are taken into consideration. In ch. 4 the semantics of FA's is presented. The first section is devoted to the lexical meaning of the different FA's, and the second section to the meaning composition of sentences that contain one or more FA's. To start with this second aspect: it turns out that in general the principles that were developed in Jacobs (1982) for the meaning composition of negative sentences can be applied to FA-sentences as well. As a first step in the execution of the meaning composition, the semantic scope of the different scope-bearing expressions must be established. Scope-bearing expressions are, besides FA's, other adverbs, quantifiers, etc. At least for German, it seems possible to derive the scope relations from the syntactic component (left-to-right word order and constituent structure) and the phonological component (information about sentence

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(22)

387

accent). In this respect FA's do not behave differently from other expressions with scope. A special property of FA's, however, is their ability to subdivide the semantic scope into a focus part and a background part. In fact, the focus part is already marked by the syntax, but what part of the sentence constitutes the background can only be established after determining the semantic scope. The background is not automatically the whole sentence minus the focus An example can illustrate this (cf. p. 202)(23)

Peter bezweifelt nurj, dass Peter doubts bestehen wird pass will

(24)

only whether Gerda the

Peter bezweifelt, dass

oral

exam

Gerda nur j die L^mundliche^J Priifung

whether Gerda only the

oral

exam

In these two sentences, the focus of nur is the same, but not the background in (23) the whole sentence is in the semantic scope of nur, so that the background pertains to the question of what Peter doubts, whereas in (24) the semantic scope is restricted to the subordinate clause, so that the background pertains to the question of which exams Gerda will pass. I will not go into the formal treatment that Jacobs gives for FA-sentences, nor will I discuss his excursions on polarity items and other related subjects On the whole, this part of the semantics seems to me rather uncontroversial This cannot be said of section 4.1, in which a formal theory is developed for the explication of the lexical meaning of the different FA's. In this theory, a central place is assigned to the notion of 'scale', because Jacobs contends that FA-sentences must always be interpreted in terms of scales. This view is controversial, in so far as it is usual in the literature to divide the FA's in one group that indeed must be interpreted in terms of scales, and another group for which that doesn't hold. Konig (1981), for example, includes as members of the first group sogar (even), erst (temporal only) and schon (already), among others, and as members of the second group auch (also), nur (only) and eben (just). For an appropriate discussion of this part of the book, it is necessary to summarize here the scale theory of Jacobs Leaving aside the formalisation, we can say that a scale is a sequence of cardinal numbers, on which both the focus of the sentence and the contextual relevant alternatives get a value All alternatives, including the focus, are allowed to receive the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Peter doubts bestehen wird pass will

Gerda die L mvindliche J Priifung

388

(25)

Auch der Klassenprimus Also the best pupil of the class

bestand diese Priifung nicht passed this exam not

A 'non-real' ordering is most plausible for cases like (26) (26)

Auch Hans kommt Also Hans comes

In such a case, all alternatives have the same position on the scale of the people who come. Up to this point the analysis of Jacobs does not deviate substantially from the traditional analyses for sogar and auch. The main problem concerns the analysis of the restrictive FA nur (only), not only in the book of Jacobs, but in the FA-literature in general (cf., among others, Van der Auwera 1984, and Foolen 1983). The difficulties that nur poses are - the question of whether a semantic unification is possible, i e. can we restrict ourselves in the lexicon to specifying one meaning for nur or not; - secondly, the question of whether the evaluative component is specified for nur (as it is for sogar) or not (as it is for auch)

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

same number. Two of the numbers on the scale are marked as upper and lower 'boundary value'. These numbers and the numbers above and below them have the status of relatively high and low values respectively. The dimension of the scale is constituted by the background part of the semantic scope. Now, the explication of the meaning of FA's consists in stating conditions for the focus and the alternatives. If an FA-sentence obeys these conditions, it can be considered as a well-used, adequate, sentence These conditions pertain to - the question of whether the background proposition should or should not hold for the focus and/or the alternatives (this condition constitutes the so-called 'quantifying' component of the meaning), - secondly, the place of the focus on the scale (the so-called 'evaluative' component of the meaning) Take, for example, an FA-sentence with sogar (even). If the sentence is well-used, then, first, the background proposition should hold for the focus and for at least one alternative, and secondly, the focus should not have a number that lies under the upper boundary value. In the semantic description of auch (also) the evaluative component is not specified, i.e. FA-sentences with auch are compatible with different orderings of focus and alternatives on the scale. A 'real' ordering might be assumed in the interpretation of (25)

389 A positive answer to the first question has, of course, its attractions for FA-authors, and has accordingly been defended several times, in several versions, whereby a certain tour de force has not always been avoided. Jacobs also proposes a unified description, but he admits that there are some uses of nur as FA for which he cannot provide a formally controlled account within his theory. Compare for these problematic cases section 4.2.5, in which examples are discussed as (27) (27)

An diesem Abend wurden uns nur Spitzenweine kredenzt At this evening were us only top-wines offered

The positive connotation of (27) seems not to be very compatible with the evaluative component that is proposed for nur (see below). As to the second question, there is a lot of confusion, not in the last place because of the co-existence of data such as (28) and (29). Nur Hanskommt Only Hans comes

(28) suggests that the evaluative component does not play a role {Hans and the excluded alternatives of persons who don't come do not seem to stand in a real rank order relation on a scale), whereas in (29) (29)

Peter besitzt nur zwei Hemden Peter has only two shirts

the number of two is evaluated as relatively low in relation to other contextually relevant values. These and other observations have resulted in a lack of agreement in the literature. Altmann (1976) opted for a polysemyanalysis of nur, Konig (1981) for an analysis in which the evaluative component is not specified, whereas a third group of authors, to which Jacobs belongs, do specify an evaluative component as part of the meaning of nur In the unified descriptions, i.e. the non-polysemy analyses, the following strategy can often be observed an underlying unified meaning is assumed, from which various different meaning nuances are derivated with the help of pragmatic, Gricean, principles. Such a model of description is followed for example in the monograph of Lerner & Zimmermann (1981), that is completely devoted to nur. Jacobs turns himself rather sharply against such 'pragmatic repression processes', and tries instead to derive the different observed meaning nuances within the formally controlled semantic component. For example, the suppression of the evaluative component in one of the interpretations

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(28)

390

It might have become clear that my overall opinion of the book under review is positive. When I was reading FuS, I encountered more or less at the same time Taglicht (1984). The latter book treats almost the same subject, but for English. Comparing the two works, one can say that Taglicht is much more data-oriented (extensive use is made of the Survey of English Usage) and full of interesting observations and analytical distinctions. From a theoretical point of view, however, Jacobs is clearly superior. Everyone who is interested in the syntactic and semantic description of focusing adverbs, or, more in general, of 'function words', can without doubt profit by reading this book Nijmegen University Department of Linguistics Erasmusplein 1 6525 GG Nijmegen The Netherlands

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

of (28) is, informally stated, accounted for in the following way. if we assume for that interpretation of (28) a scale on which focus and alternatives all take the same position, then all alternatives are excluded by nur, because, by definition, nur excludes all alternatives that are equally or higher ordered in relation to the focus. If such scaling plays a role in the interpretation, then the evaluative component is cancelled, because it is fulfilled in a trivial way, i.e. because it has no informative value anymore. This cancelling runs as follows in the meaning explication of nur the evaluative component says that the focus must have a relatively low value. But if focus and alternatives take the same value on the scale, as is assumed in the envisaged interpretation of (28), the upper and lower boundary values also take that same value. In that case, the evaluative component is fulfilled automatically. It might have become clear, that the semantic flexibility of a word like nur is made possible in this theory by allowing contextual variation in the type of scale that plays a role in the interpretation process of FA-sentences. Especially the allowance of a scale type in which focus and alternatives take the same value makes it possible to describe all FA's as scalar. One could ask oneself however, whether it is not the formal trick that provides here for a unified treatment of the FA's, more than a new substantive analysis of the data. Jacobs would gain more credit on this point, if he could show that his approach can handle elegantly semantic variation of lexemes in other parts of the language also. However, I expect that the discussion of the nur-problem, and of the whole FA-problem, will profit considerably from the thorough formalisation that Jacobs has given. In this respect the theory of Jacobs can be seen as the most advanced one at this moment.

391 REFERENCES

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Altmann, H , 1976: Die Gradpartikeln im Deutschen Untersuchungen zu ihrer Syntax, Semantik und Pragmatik Niemeyer, Tubingen Altmann, H., 1978: Gradpartikelprobleme Gunter Nan, Tubingen. Eikmeyer, H-J & H Rieser (eds), 1981: Words, Worlds, and Contexts New Approaches in Word Semantics De Gruyter, Berlin Foolen, A , 1983: Zur Semantik und Pragmatik der restriktiven Gradpartikeln: only, nur und maar/alleen In: Weydt (ed ), pp 188-199 Jacobs, J , 1982: Syntax und Semantik der Negation im Deutschen Fink, Miinchen Karttunen, L & S Peters, 1979: Conventional implicatures In: Oh & Dinneen (eds ), pp 1-56 Konig, E , 1981: The meaning of scalar particles in German In: Eikmeyer & Rieser (eds), pp 107-132 Ladusaw, W , 1980: Polarity Sensitivity as Inherent Scope Relations IULC, Bloomington Lerner, J-Y & Th Zimmermann, 1981: Mehrdimensionale Semantik Die Prdsupposition und die Kontextabhdngigkeit von nur SFB Arbeitspapier No 50, Konstanz Linebarger, M , 1980: The Grammar of Negative Polarity Ph D. thesis, MIT Oh, C & Dinneen, D (eds), 1979: Presupposition (= Syntax and Semantics, Vol 11) Academic Press, New York - San Francisco - London. Taglicht, J , 1984: Message and Emphasis On Focus and Scope in English Longman, London Van der Auwera, J , 1984: Maar en alleen als graadpartikels In: Van der Auwera& Vandeweghe (eds.), pp 103-118. Van der Auwera, J & W Vandeweghe (eds), 1984: Studies over Nederlandse Partikels APIL No 35, Antwerp Weydt, H (ed ), 1983: Partikeln und Interaktion Niemeyer, Tubingen

Journal of Semantics 3: 392-394

BOOK REVIEWS

S.G. Pulman, Word Meaning and Belief. Croom Helm, London, 1983. Pp. 179. Peter Bosch

" the relation between lexical items X and Y is analytic if and only if: (i) X entails Y; (ii) It is not possible to consistently imagine a situation in which X does not entail Y; (iii) Y provides a necessary condition for being X; (iv) native speakers can and do use Y as a (partial) criterion for being X"

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The area of lexical semantics is one of the under-developed regions of the world of theoretical linguistics. An important reason for this can probably be found in the philosophical doctrine that the truth of a statement may be due either to the facts or to the meanings of the words (and constructions) that occur in the statement, and that meanings and facts must be kept apart. Quine argued more than thirty years ago that this distinction may not only be difficult to draw in practice, but that it does not make sense even in theory. - Semanticists in both philosophy and theoretical linguistics have largely ignored Quine's argument and have instead carried on attempting the construction of theories of meaning. The results have not been encouraging, certainly not when we take empirical success into account. Pulman takes the interesting position of both wanting to take Quine's arguments serious and wanting to develop a theory of meaning. The book sets out with an account of Quine's arguments against the analytic-synthetic distinction and his theory of indeterminacy of translation. This is a clear and fundamentally sympathetic account. But Pulman misconstrues Quine, not immediately noticable in the beginning, as putting forward a fundamentally empirical point, that because there is in fact no true synonymy, there is no way of construing word meanings. But Quine goes much further, his argument is that the notions of the theory of meaning, including, synonymy, analyticity, and necessity, do not make empirical sense. - Unsurprisingly then, Pulman can come up with a "solution" to Quine's problems with the notion of word meaning. Although Pulman concedes that there may be no true synonymy relations, he insists that there are truely analytic conditional relations, and proposes the following staggering definition:

393

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Usually, "P entails Q" is taken to mean that the inference from P to Q is logically valid, in other words, analytic. Similarly, it would seem that condition (iii) cannot be made sense of either without relying on the notion of analyticity in one form or the other, and the definition ends up in circularity also on this point As for clause (ii), I am uncertain that there is any way of checking upon the consistency of imagination. The only clause that could offer any help at all, to me seems clause (iv), provided the redundant and uncontrollable "can" is thrown out and there is a procedure for checking if speakers use particular conditions as criteria (which could conceivably be developed). But since clauses (i) to (iv) are offered, or so it seems, in conjunction, the hopeful fourth clause does not save the definition. Whether or not there is true synonymy, and hence word meaning, is of course a purely empirical question if we assume a definition like the above. For once we have got the notion of analyticity we can define synonymy as the analyticity of the biconditional. - Still, there are sufficiently serious problems with the above definition and the central notions of the theory of meaning make as little sense as before. But Pulman may have something else in mind, although he does not make it explicit. The fact that he never worries about entailment or necessity suggests that he may be thinking of operational definitions of clauses (i) to (iv) in terms of native speakers' judgements. "P entails Q" then simply means, native speakers assent to 'T entails Q", and analogously for clauses (ii) to (iv). Of course, such tests have nothing to do with entailment, necessity, or analyticity in their ordinary philosophical sense. There is no reason to think that native speakers should be competent to judge any of these relations or to have an implicit or 'tacit' knowledge of them. No form of language use requires any knowledge of entailment, analyticity or necessity. - None the less: if a significant proportion of native speakers assent to something like "for someone to be called 'a bachelor' it is a necessary condition that he be unmarried" (or analogously for the other conditions), then these are interesting data in themselves and one could hardly argue that such data are irrelevant to lexical semantics. These data are data, however, that pertain to native speakers' reflection upon their usage and we must not be surprised if we find our subjects falsifying their own pronouncements in their own linguistic practice. In order to accommodate both actual use and reflection one must be aware that there are also other parameters involved in actual use, and not only those that are investigated via reflection judgements. Such other parameters are concerned, for instance, with situational similarities and contrasts in particular contexts of language use. Even if I am convinced that pencils are for writing with, I will still refer to a blown up giant pencilshaped balloon above the entrance of a stationary shop as "a pencil". And

394

Nijmegen University Department of Philosophy PO Box 9108 6500 HK Nijmegen The Netherlands

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a wind-up toy mouse I will still call "a mouse", although I will occasionally refuse to call it "an animal". It just depends on what I want to say on what aspects of the thing I am focussing and what other things I want to contrast it with. Pulman, unfortunately, does not consider this discrepancy between reflection based data and data of actual usage. In Chapter Four he introduces, in a very clear exposition, the line of research on natural classification systems and prototypes, which was started by Berlin and Rosch. Prototypes, in this sense, are founded on out of context judgements on the use of words, and as such they are one of the important ingredients of a theory of lexical semantics. In Chapter Five, Pulman extends prototype theory to verb-prototypes and reports experimental results of his own supporting this extension. All this constitutes an interesting and important contribution to lexical semantics. The last Chapter is largely an exposition of ideas of Kripke, Putnam, and Schwartz on rigid designation and reference to kinds. Much of the argument, as in the authors discussed, relies on questions like whether or not we could discover that cats are not animals, or whether they are necessarily animals Not having made intelligible the relevant metaphysical notions of necessity and aprioricity, all this discussion remains, as far as I can see, quite irrelevant to lexical semantics. Pulman hopes, via this discussion, to isolate fundamental semantic categories (like the category of natural kinds). This path might indeed seem promising, provided the relevant distinctions are drawn empirically, i.e. in terms of linguistic behaviour (including native speakers' reflection). If Pulman had here applied methods as in his argument for verb-prototypes, we could now discuss the no doubt interesting results. On the whole, I like the book very much, despite what I mentioned above in terms of shortcomings or disagreements. Pulman makes a start that is needed in lexical semantics: trying to make sense of the developments of speculative philosophical semantics for a linguistic theory of word semantics and considering also what there is in terms of empirical approaches in other disciplines. Of course there is much more, also in psychology and artificial intelligence, but then this should not remain the only book of its kind. A little more interdisciplinary work of the sort of Pulman's would considerably widen the perspective and improve the results of the linguistic work in this field.

Journal of Semantics 3: 395-402

BOOK REVIEWS

Nathan U. Salmon, Reference and Essence. Basil Blackwell, Oxford, 1982 Pp. xvi + 293. Charles Travis

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Salmon's book divides into two parts. The first is concerned to expound and defend something which he calls the 'direct theory if reference'. The second has two goals first, to argue that no non-trivial essentialist theses follow from the direct theory alone, or together with non-controversial empirical discoveries; and second to take certain philosophers (notably Hilary Putnam and Keith Donnellan) to task for thinking otherwise. These are quite separate projects, as I shall presently argue. In what follows, I will save most of my attention for the second half of the book, though this should by no means suggest that the first half is any less useful or valuable. Indeed, by giving a clear exposition of a theory that has been in the air for some time now, Salmon renders it eminently ripe for attacking. But I must defer doing so for another occasion. The first half of the book again divides in two. The first half concerns singular terms, its second half general terms, or anyway some of them. The theory of direct reference is first explained, then, with regard to singular terms. It is important to keep in mind, however, that in both cases, the direct theory by itself is a purely negative thesis, the denial of an 'indirect', or perhaps better, 'mediate' theory of reference. On the mediate theory, referring terms - e.g., names such as 'Aristotle' or 'Shakespeare' may be understood independently of knowing to whom or to what they refer. Such a proper understanding fixes a condition for being the referent of the reference in question On Salmon's version, the mediate^heory further asserts that something is the referent of that reference if it (uniquely) satisfies the condition thus fixed. Such needn't be a feature of all recognizably mediate theories.1 But it is an important part of what the direct theory means to deny, on Salmon's account. This general account of a mediate theory is generally made more specific by Salmon by equating such theories with what might be called description(al)ism. This is the thesis that associated with a name such as 'Shakespeare' is a set of descriptions such as 'author of Hamlet and MacBeth ' such that someone is the referent of the name if those descriptions (uniquely) fit. For the most part, Salmon takes the association in question to be synonymy. Here, there is an equivocation in the book which comes to take on an absolutely crucial role. Salmon's official view is that what refers is a name on a use, or a speaking. So, what a set of descriptions would have to be associated with is a name, e g , 'Shakespeare', as spoken

396

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

on an occasion. And perhaps different sets of descriptions might be associated with different speakings of the name. Clearly, there is not much future for the view that a set of descriptions is synonymous with a name on a speaking of it. For the most part, though, Salmon speaks as if what refers is the name 'Shakespeare' as such, or perhaps in English. It is through this equivocation that the idea of synonymy creeps in - an idea that casts mediate theories in a much less favorable light. Against mediate theories of singular terms, Salmon advances three arguments. These he calls the modal argument, the epistemological argument and the semantical argument. Here I will mention a few features of each. A full evaluation must be deferred for another occasion. The idea of the modal argument, then, is that if D is one of the descriptions associated with the name on a given mediate theory, then it is supposed to be a consequence of that theory that necessarily N is D. For example, if one of the descriptions associated with 'Shakespeare' is 'the author of Hamlet', then necessarily Shakespeare is the author of Hamlet. If that is a consequence of the theory, then, since it clearly isn't so, the theory is wrong. The modal argument is the most clearly mistaken of the three listed. I will briefly indicate why. First, note that it does put the above equivocation between words in a language (or names as such) and words on a speaking to work Strictly speaking, what Salmon is entitled to conclude from a mediate theory (given his official views on where references are made) is this, if 'the author of Hamlet' is one of the descriptions associated with a given speaking of 'Shakespeare' - e.g., in some utterance of 'Shakespeare was fat.' - then necessarily no one is the referent of that reference unless he is the author of Hamlet. The conclusion Salmon wants (because it is the only obviously absurd one in the region) is that Shakespeare could not but have written Hamlet.2 He could not, for example, merely have become the village cobbler (if Shakespeare existed). But this is a vastly different conclusion, and there is no reason to take it to be entailed by the first. For example, if Shakespeare had restricted himself to cobbling, and if, under those circumstances, the above mentioned reference could have been made, perhaps it just would not have referred to Shakespeare Mediate theories essentially make two claims: first, that associated with each reference via a name is some set of descriptions, and second, that these descriptions are, in fact, a path to the referent. As we are construing the second claim so far, this means that for it to be the case that someone is, in fact, the referent of a given reference, it must be the case that he satisfies the associated descriptions. There is simply nothing in this about how the referent would be to be identified under counterfactual circumstances. That is a problem, about identity conditions under counter-

397 factual circumstances' appealing to the mediate theory (suppose), we locate the actual referent. We then face the question what makes someone in some counterfactual story him? As far as the mediate theory is concerned, you can tell any story about this you like.3 So, having located the referent of the above reference via 'Shakespeare' (thus, the author of Hamlet), we can then go on to ask whether he could have confined himself to cobbling. And that question we may answer however such questions are to be answered, whether we hold a mediate theory of reference or not.4 Let us turn to the epistemological argument. I let Salmon set it out. Note that he makes controversial assumptions about which descriptions would be associated with the name 'Shakespeare' on a mediate theory:

(Bracketed numbers are my addition.) The argument consists in denying 2 and 3, and pointing out that, since these are (allegedly) consequences of mediate theories, such theories are false. But are 2 and 3 consequences of mediate theories? I think that it is not easy to settle that issue. Here I merely note two points. First, Salmon helps himself to some certainly questionable classifications of knowledge as to source - e.g., a priori' versus 'empirical'. Second, again, he puts the equivocation over words to work. To see what the problem ought to be, we might consider the following. Suppose that you and I, in reading the Blackwell's catalogue, learn that someone named Saul Kripke has written a book called Naming and Necessity. Suppose I now say to you, 'Kripke must be either a Millian or an existentialist.' Conceivably, one of the conditions a mediate theory would associate with that speaking of 'Kripke' is 'the author of Naming and Necessity.' If the theory is right, then, then from a proper understanding of the utterance (and perhaps from that alone) one could glean the information that Kripke is to be understood to be the author of Naming and Necessity. We do understand that in understanding what I said. Now, is our knowledge a priori or

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

"Consider again the two sentences, Shakespeare, if he exists, wrote Hamlet, MacBeth, and Romeo and Juliet and If anyone is an English playwright who is sole author of Hamlet, MacBeth and Romeo and Juliet, then he is Shakespeare (1) As we know the orthodox theory alleges that these sentences are analytic in the traditional sense (2) it would follow that both of these sentences should convey information that is knowable a priori, i e , solely by reflection on the concepts involved and without recourse to sensory experience (3) it should be impossible to conceive that Shakespeare existed though he did not write any of these works, or that some one Englishman other than Shakespeare was responsible for each of these works " (p 27)

398

"As in the case of singular terms, the central and primary thesis of a strong version of the direct reference theory of general terms can be simply stated thus: Certain single-word common nouns which are alleged by the orthodox Fregean theory to be descriptional are in fact non-description^ " (pp 43-44)

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

empirical? Derived from conceptual reflections, or from the BlackwelTs catalogue? What kinds of choices are these? More generally, in understanding a (speaking of) a name, and thereby, on the mediate theory, associating some set of descriptions with it, why should we not rely on our shared background knowledge, derived, perhaps, from all sorts of sources, to see which descriptions these ought to be? Finally, the semantical argument. This, again, relies on the premise that it is an analytic truth on a mediate theory, e.g., that Shakespeare wrote Hamlet. Again, that is putting the equivocation to work. Without equivocating, though, we can put the argument as follows. Consider a speaking of a name - 'Shakespeare', say - and let D be a description associated with the name on that speaking. Then on the mediate theory it is necessarily so that nothing could be the referent of that reference unless it satisfied D. Further, if D is the only (or total) description associated with the name on that speaking, then nothing could satisfy D (uniquely) and fail to be the referent. Against this, Salmon claims that, no matter what D may be, something may fail to satisfy D and be the referent, or satisfy D and not be the referent (even if D is the total associated description). About this, two comments. First, if Salmon's claim is correct, then the mediate theory as considered so far must at least be modified. What we might want, for example, is a new view of the uses to which associated descriptions are to be put. In any case, they cannot quite play the role assigned them thus far both being associated with a reference tout court and serving as necessary and sufficient conditions in the way described above. Which way to jump here is a long story which I propose to tell elsewhere. But second, whether Salmon is right may depend a great deal on which descriptions get associated with a name on a given mediate theory. Here it becomes crucial that Salmon generally just assumes that the right thing to associate with a speaking of 'Shakespeare', for example, will be something like 'Author of Hamlet and MacBeth' - something a mediate theorist may well regard as not a plausible candidate at all. Salmon recognizes that when you think you have discovered a phenomenon of singular terms, you ought to check to see to what extent the phenomenon is due to special features of singular terms, and to what extent it might also arise, e.g., for general terms as well. Thus, Salmon devotes some considerable attention to the question whether terms such as 'tiger' or 'water' can be defined in terms of descriptions. His conclusion is that some, at least, cannot. In his words,

399

" it is conceivable that water isn't H , 0 It is conceivable, but it isn't possible! Conceivability is no proof of possibility a statement can be (metaphysically) necessary and (conceptually) contingent.", or more picturesquely, "nothing counts as a possible world in which water isn't H, O" (both quotes from Putnam's "The Meaning of 'Meaning' ", quoted by Salmon on pp 98-9.)

He then claims that such conclusions depend on some hidden essentialist premise, exemplified by "Given any possible chemical structure F and any possible substance Z, if it is merely possible that some sample of substance Z have the chemical structure F, then it is necessary that every sample of Z have the chemical structure F " (Salmon's formulation, p 184 )

Given the wholly negative character of the (official) theory of direct

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Salmon takes this to be a consequence of the applicability of the above three arguments to (some) general terms My only question here is why Salmon restricts the thesis to some general terms. It seems to me that, in the case of general terms, at least, roughly two and a half millenia of trying ought to have convinced us of the general thesis that lexical simples are not semantically decomposable - i.e., not logically equivalent to any complex analysis. So, true enough, 'water' is not definable in terms of descriptions, but neither is 'chair'. If we recall that, for present purposes, the theory of direct reference is a purely negative (i.e., anti-descriptionalist) thesis, then the above negative fact seems enough to account for its correctness not only in the case of 'water', but in the cases of 'bachelor', 'chair' and general terms in general. Considering the connections between general and singular terms, this observation might lead one to raise the following question: Is there anything more to direct reference in the case of names than the fact that they, being lexically simple, do not semantically decompose? If not, then to what extent is there any room for special 'theories of reference' - particularly, 'theories of reference to individuals'? The second half of Salmon's book is devoted to demonstrating two theses. The first is that the theory of direct reference, together with noncontroversial and non-essentialist premises, such as discoveries in chemistry or biology, does not entail 'non-trivial essentialist' conclusions. (As we shall see, the idea that discoveries in chemistry and biology would be noncontroversial and non-essentialist already runs against the grain of what Putnam is trying to do.) The second thesis is that certain philosophers, notably Putnam and Donnellan, have mistakenly thought otherwise. For example, Salmon cites the following passages from Putnam:

400

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

reference, Salmon's first thesis strikes me as fairly trivial (bracketing problems about what is and isn't an essentialist additional premise). The theory has it simply that referents are not fixed by sets of associated descriptions. So much leaves it entirely open how they are fixed. As far as the official theory is concerned, there might be an oracle living atop Mt. St. Helens, such that, e.g., every time we want to know of someone (under actual or counterfactual circumstances) whether he is the referent of some speaking of 'Shakespeare', we consult the oracle (whose word is to be taken as binding). Whether essentialist conclusions follow from that depends on the behavior of the oracle. What Salmon has in mind, I think, is that essentialist theses do not follow from the purely negative point plus a causal theory of reference. Given the vagueness of the idea of a causal theory of reference (an intentional vagueness, at least in case of Kripke), I would not wish to dispute this either. In finding Putnam guilty as he does, however, Salmon has evidenced misunderstanding of what Putnam is up to. Or so it seems to me - perhaps Salmon has Putnam right, and I have him wrong. We can still ask. To show the misunderstanding as I see it, I now want to point to a contrast between reference via singular terms and that via general terms. Beginning with singular terms, let us consider problems of identifying a referent. In general, these can be factored into two parts. First, beginning with a reference via a name, N, we can try to identify something, J, which is the referent of N. Second, for any arbitrarily selected candidate, K, we can try to determine whether K is the referent. Suppose the first problem solved, with J identified as the referent. Then, for arbitrary K, K will be the referent just in case K is (identical with) J. For practical purposes, there is rarely a problem about whether this is so or not. Hence, it is easy to overlook this possibility of factoring. Suppose, for example, that a causal theory of reference is true. Suppose somewhat fancifully that following a causal chain back from a speaking of 'Shakespeare' leads us to a certain baby being christened Shakespeare. Now we want to know whether that reference referred to the author of Romeo and Juliet. The answer is yes, if that author is that baby. Normally, there is no problem about how to determine this. If, however, Shakespeare were to have gone to bed one night, and in the morning there awoke someone looking and sounding exactly like Richard Nixon, there might be a problem about whether that person is the referent. It would then be a propos to ask what kind of philosophical story there is to tell about how problems of this sort are to be solved. The theory of direct reference is first and foremost a solution to problems of the first sort above. Or rather, given its wholly negative character, it is not a solution, but a thesis about solutions. But suppose we have a solution to that problem compatible with the direct theory -

401

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

e.g., through adopting a causal theory. That still leaves various possibilities open for solutions to the second problem. For example (to consider one extreme), one might hold that conceptual analysis of 'Shakespeare' tells us that the referent had better be a man (such is not yet descriptionalism), and that conceptual analysis of man tells us that someone will be the same man as the baby above if he shares with it properties pj ,--.,p n . Or, we might hold that if, as a matter of fact, the causal chain from 'Shakespeare' leads us back to a man, then conceptual analysis of man tells us that someone is that referent if he shares with it properties p^ pn. In the case of general terms, the same kind of factoring of problems can be done verbally, at least. We can ask, for example, how we identify something as the referent of 'gold' - how we determine which substance gold is, if indeed, it is a substance. We can then go on to ask how, for a given sample of stuff, we determine whether it is what 'gold' refers to whether it is a sample of the right substance, if gold is a substance. Here, however, our answer to one question had better not be independent of our answer to the other. Suppose, for example, that we adopt the direct theory vis-a-vis the first problem. Then we had better not make a solution to the second problem a matter of conceptual analysis. Suppose we maintain, for example, that conceptual analysis tells us that something is a sample of what 'gold' refers to if it is identical in color, density and malleability with what 'gold' refers to. The problem is that if, e.g., it is up to causal chains to lead us to what 'gold' refers to in the first place, and if these are not constrained by a descriptional account (e.g., 'whatever has the color, density and malleability we usually associate with 'gold' ') then there is always the possibility that what 'gold' turns out to refer to does not have any particular color or density or malleability - such just aren't characteristics of the referent. So there is the possibility of our solution to the first problem and that to the second being inconsistent. If one is a direct theorist with respect to general terms, then, how should the second problem be solved? As I understand it, Putnam's 1975 answer to this question (which he has since modified somewhat) was: let science give us the answer. So, e.g., it is up to chemistry and physics to tell us what a sample has to be like to be a sample of what 'gold' refers to. It is important to realize that if this is the solution to the problem posed, then, when science has spoken on the subject, unless there are grounds for science correcting itself, it has given us the last word on what makes a sample a sample of gold. It may do this through scientific discoveries 'of the normal sort' (whatever that is) But the consequence is that ordinary discoveries in physics, chemistry, biology or whatever may have modal, hence essentialist import. That thesis on how to read the content of (some) scientific statements is, I take it, part of Putnam's (1975) theory of the reference of general terms.

402

NOTES 1 Cf my "How To Be A Referent" H Pairet, M SbisaandJ Verschueren (eds ) In Possibilities and Limitations of Pragmatics John Benjamins, Amsterdam, 1981 Pp 713-747 2 There is also, perhaps: Hamlet could not but have been written by Shakespeare This, too, fails to follow from mediate theories, though even it is less obviously absurd 3 The point would become sharper if we examined the actual motives for holding a mediate theory The most important concern the exigencies of disambiguation But there is no space to go into that here 4 In the worst possible case, suppose that every reference ever made by anyone in saying 'Shakespeare' up to now has 'the author of Hamlet' as an associated description Why should we, now, not make a reference via 'Shakespeare', with the associated description, 'the man all those people were talking about', e g , in asking, 'Could Shakespeare have refrained from writing HamletV Such a reference might not pick out anyone, but then again, it might

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Suppose, for example, that someone tells a counterfactual story about a substance (which he may call 'gold') with an atomic number other than 79. (Assuming that that is what science says to be the essential or identifying property of gold.) It may be that, in the story, science counterfactually tells us that that substance is gold. But is this a story about gold? On Putnam's theory, that is something for, actual science to tell us. And, on the above assumption, actual science, if it does settle that question, tells us that the answer is no. If that is correct, then there are no counterfactual stories about gold in which it has an atomic number other than 79. So having atomic number 79 is an essential property of gold. And such is a physico-chemical discovery. Of course, such does not follow from the bare direct theory. But it was never part of Putnam's project that it should. The above criticisms notwithstanding, I think Salmon has written a good, and a very useful book. It provides a very clear, concise and up to date account of one strain - perhaps the dominant one - in current discussions of reference. In philosophy, setting out a theory clearly is more often than not the first step towards refuting it. Whether such should prove true in this case or not, Salmon has provided a useful service.

PUBLICATIONS RECEIVED

Foley, W.A. and R.D. van Valin Jr., Functional Syntax and Universal Grammar (Cambridge Studies in Linguistics 38) Cambridge UP, Cambridge, 1984. Pp. xii+416 Gazdar, G., E. Klein, and G. Pullum, Order, Concord, and Constituency. (Linguistic Models 4). Foris, Dordrecht, 1984. Pp. 219, DFL 40.00 (paper) Grazer Linguistische Studien, 21 (1984) Hintikka J and J Kulas, The Game of Language. Studies in Game-Theoretical Semantics and its Applications (Synthese Language Library 22) Reidel, Dordrecht, 1983 Pp. xii + 324, DFL 115.00 (cloth). Jones, A J I., Communication and Meaning An Essay in Applied Modal Logic (Synthese Library 168). Reidel, Dordrecht, 1983. Pp. xii + 160, DFL 78 00 (cloth). Joseph, B.D , The Synchrony and Diachrony of the Balkan Infinitive. (Cambridge Studies in Linguistics Supplementary Volume) Cambridge UP, Cambridge, 1983 Pp. xiii +314

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Bauerle, R., Ch. Schwarze, and A. von Stechow (eds.), Meaning, Use, and Interpretation of Language. (Foundations of Communication. Library Edition), De Gruyter, Berlin, 1983 Pp ix + 490, $ 29.00 (cloth). Bennis, H., and W.U.S. van Lessen Kloeke (eds.), Linguistics in the Netherlands. (Publications in Language Sciences 12). Foris, Dordrecht, 1983. Pp. 199, DFL 25.00 (paper). Bosch, Peter, Agreement and Anaphora. Academic Press, New York, 1983. Pp xiii + 260, $35.00 (cloth). Carlson, L. Dialogue Games. (Synthese Language Library 17). Reidel, Dordrecht, 1983. Pp. xxiii + 310, DFL 115.00 (Cloth). Communication and Cognition, 16 4 (1983), 17.1,2/3 (1984). Cook, E. and D.B. Gerdts (eds.). Syntax and Semantics 16. The Syntax of Native American Languages. Academic Press, New York, 1984. Pp. xiii + 324, $ 59.00 (cloth). Dik, S C (ed.), Advances in Functional Grammar. (Publications in Language Sciences 11). Foris, Dordrecht, 1984. Pp. x + 415, DFL 40.00 (paper)

404

£15.00 (cloth) Texas Linguistic Forum 22 (1983), 24 (1984). Van den Broecke, M., V. van Heuven, and W. Zonneveld (eds.), Studies for Antonie Cohen. Sound Structures (Publications in Language Sciences 13). Foris, Dordrecht, 1983. Pp. xxvi + 318, DFL 40.00 (paper). Werth, P., Focus, Coherence, and Emphasis. Croom Helm, London. Pp viii+93, £16.95 (cloth). Zaefferer, D. Frageausdruecke und Fragen im Deutschen Zu ihrer Syntax, Semantik, und Pragmatik (Studien zur Theoretische Linguistik 2) Wilhelm Fink Verlag, Muenchen 1984 Pp. 208, DM 68.00 (paper). Zuber, R. Non-declarative Sentences. (Pragmatics and Beyond IV 2) John Benjamins, Amsterdam, 1984. Pp. 123, DFL 45.00 (paper).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Landman, F. and F. Veltman (eds.), Varieties of Formal Semantics. Proceedings of the Fourth Amsterdam Colloquium, September 1982. (GRASS 3). Foris, Dordrecht, 1984. DFL 85.00 (paper) Language and Communication 3 2(1983). Language Research 19 2 (1983), 20 1 (1984). Marcus, S., Contextual Ambiguities in Natural and Artificial Languages 1/2. (Studies in Language). Communication and Cognition, Ghent 1981/83. Pp. 123 + Pp. 130, BF 350 + 350 (paper). Martinich, A.P., Communication and Reference (Foundations of Communication Library Edition). De Gruyter, Berlin, 1984. Pp. xiii + 205, DM 84.00 (cloth). Moortgat, M., H. van der Hulst,and T Hoekstra (eds.), The scope of Lexical Rules. (Linguistic Models 1). Foris, Dordrecht, 1981. Pp. ix + 295, DFL40.00 (paper) Nef, F (ed.), L'Analyse Logique des Langues Naturelles (1968 -1978) Centre National de la Recherche Scientifique, Paris, 1984. Pp. 239, FF 95 (paper). The Prague Bulletin of Mathematical Linguistics 40 (1983), 41,42 (1984). Reports on Mathematical Logic 16 (1983). Saile, G., Sprache und Handking. (Schriften zur Linguistik 10). Vieweg, Braunschweig, 1984. Pp. 285, DM 48.00 (paper). Sanford, A. and S.C. Garrod, Understanding Written Language. Explorations in Comprehension Beyond the Sentence. Wiley, Chichester, 1981