Available online at www.sciencedirect.com
Journal of Pragmatics 42 (2010) 1786–1798 www.elsevier.com/locate/pragma
Indefiniteness marking and Akan bi Nana Aba Appiah Amfo * Department of Linguistics, University of Ghana, P.O. Box LG61, Legon, Accra, Ghana Received 23 May 2008; received in revised form 30 January 2009; accepted 3 February 2009
Abstract This paper examines the communicative role of the Akan (Kwa, Niger-Congo) existential quantifier bi both as a determiner and as a pronominal. In aiding the addressee to arrive at the intended interpretation of the utterance in which it occurs, determiner bı´ provides the procedural information that the addressee should construe bı´ as an existential quantifier that quantifies over tokens belonging to the type of entity expressed by the preceding nominal. In resolving the referent of pronominal bı`, the addressee construes bı` as an existential quantifier that quantifies over tokens belonging to the type of entity the antecedent expresses. When the definiteness marker no (the) precedes bı´ in a single noun phrase, the intended referent of that noun phrase is ‘a part, or member, of the whole’ or ‘one of this sort’. On the other hand, when bı´ follows no the noun phrase is to be understood as an echoic metarepresentation of an earlier mentioned noun phrase. Thus the syntactic configuration and the nature of the rightmost marker define the sequence as grammatically definite or not. # 2009 Elsevier B.V. All rights reserved. Keywords: Akan; Existential quantifier; (In)definite; Reference
1. Indefiniteness marking in Akan Akan (Niger-Congo, Kwa group) belongs to the group of languages in which definiteness versus indefiniteness is grammatically encoded.1 In Akan, a bare noun may be used to refer to something which is assumed to have the cognitive status of type identifiable on the Givenness Hierarchy (Gundel et al., 1993), as illustrated in (1), where the addressee is only expected to identify the type of object referred to by the nominal mpaboa (shoes).2
* Tel.: +233 26 4728324. E-mail addresses:
[email protected],
[email protected]. 1 Fretheim (in preparation) has questioned the appropriateness of the notions of definiteness versus indefiniteness as a universal semantic distinction and consequently also the validity of ‘indefinite’ and ‘definite’ used as attributes of nouns such as ‘entity’ and ‘reference’. I will not delve into the potential theoretical merits of possibly abandoning such a distinction cross-linguistically. However, I proceed on the basis that Akan belongs to the group of languages that encode a grammatical distinction between definite and indefinite as grammatical categories with precise semantic implications. 2 For convenience, I refer to the speaker as she and the addressee as he. The examples marked AK are from a text in the Akuapem dialect and those marked FT are from a Fante text (Anonymous, 1961; Coleman, 1995). The rest of the examples are in the Asante dialect. The abbreviations used in this paper are as follows: COMP = Complementizer; COMPL = Completive aspect; CM = Conditional marker; CONJ = Coordinative connective; CONS = Consecutive marker; DEF = Definite article; DEM = Demonstrative; FUT = Future tense; INT = Intensifier; MM = Modal marker; MP = Motional prefix; NEG = Negative morpheme; NP = Noun Phrase; PERF = Perfect aspect; PL = Plural; POSS = Possessive marker; PROG = Progressive aspect; REL = Relative clause marker; QUANT = Quantifier. 0378-2166/$ – see front matter # 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.pragma.2009.02.002
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
1787
However, linguistic signalling of indefiniteness is not restricted to the use of a bare noun; the Akan speaker may use a noun phrase containing the determiner bı´ to indicate indefinite reference. The difference between a bare noun and one modified by bı´ is that in the case of the latter, the speaker has a specific referent in mind, which is not already known or manifest to the addressee. It is not one which has to be uniquely identifiable, but N bı´ is a representation of a particular instantiation of the type of entity denoted by the head noun. Boadi (2005:127) states that the determiner bı´ gives the noun ‘‘a specificity-of-reference interpretation’’. In (2), the speaker has a specific pair of shoes in mind, and the addressee is expected to be able to form a representation of the shoes by the time the utterance is over.
Following Gundel et al.’s (1993) terminology, one could say that the cognitive status of the referent of a bare noun in Akan is at least type identifiable, whereas that of N bı´ is at least referential. For them, a form whose referent is judged by the speaker to have the cognitive status type identifiable for speaker and addressee invites the addressee to access a representation of the type of entity referred to. When the referent is assumed to have the cognitive status referential and this assumption is reflected in the speaker’s choice of referring expression, the speaker has a particular referent in mind, and the addressee is expected not only to access a type-representation. He must be able to mentally form a de re representation of the referent based on the descriptive content provided by a given utterance, rather than just a representation of a general type of entity conforming to the descriptive content of the noun phrase. Bi used as a determiner following a head noun (plus other optional modifiers such as adjectives) has a high tone, as in (3). When bi occurs with a low tone, it is an independent argument which does not have the support of a head noun.3 This is illustrated in (4).4
Irrespective of the tonal form that bi takes and the syntactic environment it occurs in, it fits the description of an existential quantifier, and it binds variables belonging to a conceptual domain restricted by a nominal which invariably occurs to its left, but as I am going to show, its lexical properties nevertheless make it rather different from a quantifier word like English some. The lexical meaning of bi will be argued to include a procedural component in addition to its logical function as existential quantifier, and this aspect of its lexical meaning is what deserves special attention, and in my opinion justifies writing a paper of the present sort. The lexical meaning includes information over and above the function of bi as existential quantifier. With regard to what I shall call the pronominal anaphor bı`, there must be an antecedent in the preceding discourse that directs the addressee to the denotatum that bı` represents, or bı` can be used deictically so that 3
In the Kwahu dialect of Akan, independent bı` when it occurs in a negative polarity sentence takes a high tone, otherwise it is low tone as in the other dialects. I owe this observation to Hudu-Lare Alhassan. 4 Depending on other contextual information, (4) could have other alternative translations such as (i) I would like some of those; (ii) I would like one similar to those; (iii) I would like some of it (where ‘it’ represents a non-count object and ‘some’ a proper subpart of it).
1788
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
identification of the denotatum depends on the presence of an extra-linguistic stimulus. However, the relation between bı` and its antecedent can be described as a weak one since the denotatum that bı` represents is not normally the same as the denotatum of its antecedent. Bı` causes the addressee to direct his attention to either a part of the referent of its antecedent or an entity of the kind signified by its antecedent (cf. section 4). In (4), the condition that must be met in order for bı` to be interpretable is that there exists some contextually established discourse entity, such that the speaker would either like an entity of the same type, or one of a (sufficiently) similar type, or a part of that entity. With regard to the high-toned determiner bı´, the properties defining the set are specified in the noun preceding bı´ as we see in (3), while in (4) the identity of the set, or category, must be inferred through an anaphorantecedent relation.5 The free choice universal quantifiers biara ‘any’, obiara ‘anybody/anyone’, ebiara ‘anyone (of something)’ and biribiara ‘anything’, all of which contain the morpheme bi, are used to quantify over an unrestricted domain constrained only by the conceptual semantics of the nominal in the scope of the free choice determiner biara and by contextual information when the quantifier is a morphologically complex pronoun like obiara, ebiara or biribiara. These are exemplified from (5) to (8).6
In this paper, I will be primarily concerned with the different processes involved in resolving the referents of N bı´ and pronominal bı`. I have already indicated the assumed encoded cognitive status of bi-nominals. Gundel et al. (2001:275) state that information about the cognitive statuses of nominal referents ‘‘serve as processing signals that assist the addressee in restricting possible referents’’. This suggests that information about the cognitive status of bi provides certain signals which aid the addressee in identifying the intended referent of N bı´ or pronominal bı`. This kind of information does not form part of the conceptual representations which feed the inferential process by providing logical constituents of the proposition expressed, rather they provide information about the kinds of inferences or ‘connections’ that the addressee needs to make in the interpretation process in order to arrive at the optimally relevant interpretation. This kind of information provided by certain linguistic expressions has been classified as procedural, as opposed to conceptual (cf. Blakemore, 1987, 2002; Wilson and Sperber, 1993). Procedural information is that kind of information encoded by a linguistic form which provides constraints as to how the inferential phase of an utterance interpretation process ought to proceed. In a process of nominal reference resolution, such information directs the search for the intended referent. The type of inference triggered by bi depends on the speaker’s assumption about 5 Also, bi is paradigmatically opposed to the definite marker no: bı´ versus no´ and bı` versus no`; high-toned bı´ and no´ are determiners and their lowtoned counterparts are independent phrases with the semantic function of an argument. 6 Note, however, that these forms also act as ordinary universal quantifiers. In other words, the distinction between a universal quantifier and a free choice (universal) quantifier is not lexicalized in Akan; the addressee relies on contextual information in determining which of the two uses is intended by a speaker on a particular occasion.
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
1789
the cognitive status that the referent has for the addressee at the time of utterance. Information relating to the cognitive status of referents constrains the addressee’s search for the referent, as the search space will not include stored information on referents whose cognitive status is lower than what the referring term demands. This is procedural information, as it triggers certain types of inference that the pragmatics system is to perform in order to arrive at an optimally relevant interpretation. Here I follow Scott (2008), a paper which argues that both pronouns and determiners encode procedural information that places constraints on the truth-conditional content of an utterance, as they ‘‘narrow down the set of potential referents until the most accessible potential referent remaining is (if all goes well) the intended referent.’’ (Scott, 2008:11). I am going to show that when N bı´ is used, the reference is to an instantiation of a type of object. The intention is for the addressee to focus on the type properties described by the head noun in question, or by the head noun plus possible modifiers, including restrictive relative clauses. In the case of bı`, its discourse antecedent must be highly accessible, because its referent must be represented in the interactants’ short-term memory, the general condition for appropriate use of pronominal forms. A given token of bı` either refers to a part of the entity denoted by the antecedent or to a similar entity, one which belongs to a set made up of members of the same type of entity. Second, I analyze the collocation of bi with the definiteness markers yi/no and dem/saa and address the issue of how the order in which these determiners occur within the noun phrase reflects a syntactic dominance relation which in turn determines the intended interpretation. The rest of this paper is organized as follows. The following section provides a brief summary of the Givenness Hierarchy. Section 3 focuses on the determiner bı´, while section 4 is concerned with the pronominal bı`. In section 5, I address certain syntactic constraints which occur when bi is in subject position and is thus preceded by a nominal prefix. The issue of the co-occurrence of bi with other markers of definiteness is addressed in section 6. Section 7 is the conclusion. 2. Givenness Hierarchy In their influential paper, Gundel et al. (1993) proposed a set of purportedly universal implicationally related statuses of discourse referents. These statuses, known as cognitive statuses, are considered necessary for explaining the relation between referring forms and conditions for their appropriate use across languages. They are correlated with different types of referring expressions. Below are the six cognitive statuses aligned with some typical English forms; the most restrictive status (in focus) is on the extreme left and the least restrictive status (type identifiable) on the extreme right. Each cognitive status is a necessary and a sufficient condition for the appropriate use of the form(s) aligned with it, in the same column in (i); any cognitive status lower than (i.e. to the right of) a given cognitive status CSi is a necessary but not a sufficient status for use of a referring expression aligned with CSi, while any cognitive status higher than CSi is a sufficient but not a necessary condition for use of a referring expression aligned with CSi. (i) The Givenness Hierarchy (for English)
Due to the implicational nature of the Givenness Hierarchy, each status by definition entails all lower statuses; a referent that is activated is thus automatically familiar, uniquely identifiable, referential and type identifiable. When a form associated with the cognitive status ‘type identifiable’ is used, all that is expected of the addressee is for him to access a representation of the type of object referred to by the noun phrase. This cognitive status is necessary for the use of all nominal expressions and it is sufficient for the use of the indefinite article in English. The use of the nominal phrase a schoolbag in (9) is appropriate so long as the addressee is assumed to be able to identify what type of object the noun phrase denotes.
1790
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
Use of an expression associated with the cognitive status ‘referential’ signals to the addressee that the speaker intends to refer to a particular object or objects. The addressee is expected to go beyond accessing an appropriate type-representation and retrieve an existing representation of the intended referent, or construct a new representation by the time the utterance has been processed. According to Gundel et al, this cognitive status is necessary for the appropriate use of all definite expressions and it is sufficient for the use of indefinite this in colloquial English. The noun phrase this boy in (10) is appropriate since the addressee is expected to construct a representation of the particular boy referred to by the speaker. The addressee may have no prior representation to access, and (10) could still be felicitous.
The addressee is expected to be able to identify the referent of a nominal associated with the cognitive status ‘uniquely identifiable’ on the basis of the meaning of the nominal alone. Unique identifiability is a necessary condition for the use of all definite expressions and it is sufficient for the appropriate use of the definite article in English. In (11), previous knowledge about the pipe in question is not a prerequisite in order for the addressee to form a unique representation of the reference of that nominal.
The addressee is expected by the speaker to already have a representation in his memory of a nominal associated with the cognitive status ‘familiar’. It is in his short-term memory if the object has recently been mentioned or perceived, otherwise it is in his long-term memory. This cognitive status is necessary for the use of all personal pronouns and demonstratives and it is sufficient for the appropriate use of the demonstrative determiner that. The noun phrase that man in (12) is felicitous only if the addressee has a representation of the man being referred to in his short- or long-term memory.
An ‘activated’ referent is represented in current short-term memory. It may arise from the immediate linguistic or extralinguistic context. The cognitive status activated is a necessary condition on the referent for the appropriate use of all pronominal forms. It is a sufficient condition on the referent for the use of the demonstrative pronouns this and that, all stressed pronouns, and the demonstrative determiner this. The use of this baby in (13) is felicitous only if the baby (or its photograph) is present in the extralinguistic context, or has been very recently and quite prominently mentioned (linguistic context).
An ‘in focus’ referent is at the current centre of attention. The status ‘in focus’ is necessary and sufficient for the use of unstressed pronouns. The use of she in (14) is appropriate since Jemima is the current centre of attention by the time the first utterance has been processed by the addressee.
Gundel et al. suggest that these cognitive statuses are universal; however, not all languages code all six statuses and the natural language forms aligned with the various statuses may vary from language to language. In (ii), we find the Givenness Hierarchy alignments for Akan adapted from Fretheim and Amfo (2008:362).7
7
The Givenness Hierarchy alignments for Akan given in fig. 2 have been modified to include the pronominal bı` and the null pronoun which was omitted in the original proposal in Fretheim and Amfo (2008).
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
1791
(ii) The Givenness Hierarchy (for Akan)
Even though bı` and no` are both indexicals which require the presence of an anchor or discourse antecedent, they occur in quite widely different slots (referential and in focus respectively) because unlike no` the referent of bı` in many cases is not coreferential with the antecedent. The antecedent only provides an indication of the kind of entity bı` refers to (see section 4). 3. Determiner bı´ When bı´ is used as part of a noun phrase, the speaker expects the addressee to be able to associate the type of object described with a specific token. The use of bı´ is quite common in introductory sentences where the speaker introduces a referent that will be mentioned recurrently in the ensuing discourse. Example (15) is the first sentence of a story.
By modifying da ‘day’, bea ‘woman’ and ba so denfo ‘stubbornn child’, each with bı´, the speaker introduces new referents into the discourse, and she indicates that she intends to talk about a particular time, woman and child, which the addressee is able to form a representation of, by the end of the utterance. Bı´-modification of the various nouns in (15) encourages the addressee to form a representation of what is typically the relation between a mother and a stubborn child even before the story unfolds, as he focuses on the type-features associated with these nouns. Bı´modified noun phrases are very often followed by relative clauses which provide more specific information about the intended referent, as in (16). This linguistic expansion further constrains the addressee’s search for the singular referent of the direct object argument in (16), which is described by means of the linguistic material on either side of bı´, the adjective and head noun to the left and the relative clause to the right of the indefinite determiner.
As indicated earlier, the referent of a bı´-modified noun phrase has the cognitive status referential. Akan bı´ shares with English certain and Turkish accusative noun phrases (see Enc¸, 1991) the semantic feature of specificity. Specificity, according to Enc¸ (1991:21), ‘‘involves linking objects to the domain of discourse in some manner or other’’. Akan bı´, like English certain, is a relational specific; in identifying the exact referent of N bı´ the addressee is expected to consider the type features associated with the preceding noun.8 The referents of a bı´-modified noun phrase are novel entities as far as the current discourse situation is concerned; they are distinct from any previously established discourse referents. Their specificity lies in the fact that the speaker has a particular referent in mind and expects the addressee to be able to form a representation of this referent provided there are enough clues in the utterance itself and in an accessible context. In sum, the referent of a bı´-modified noun phrase is not one which the addressee already has a representation of, but he is able to form a representation of a token-level referent based on the type-features provided by the nominal. 8
Another way of ‘‘linking objects to the domain of discourse’’ suggested by Enc¸ (1991:21) is through a subset relation. This is the kind of linkage that is often observed when the pronominal bı` is used, (see section 4).
1792
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
The procedural information provided by bı´ can be stated as follows:
Given that the cognitive status of the referent N bı´ is no more than referential, the addressee is expected to identify ‘one of a kind’, that is, a token typified by the head noun and its potential modifiers, rather than a uniquely identifiable token. Even though the Givenness Hierarchy does not place an upper bound on the cognitive status of the referent, the fact that the cognitive status will be no more than referential follows from the interaction of ‘‘the Givenness Hierarchy (and the form-status correlations in the language) with general pragmatic principles’’ (Gundel et al., 2001:276), and the particular assumption that the speaker is providing as much information as is required. 4. Independent bı` Bı` occurs independently of a head noun. Its distribution is restricted to the object position in the clause; it never occurs as a subject of a given clause.9
Example (17) is a complex clause with an embedded complement clause, and an adjunct clause introduced by the serial verb ma (to give).10 The complement clause has as its object edziban ‘food’, and the object of the adjunct clause is bı`. Since bı` is a pronominal anaphor, the addressee identifies the referent of bı` by beginning the search for an antecedent in the immediately preceding discourse. The search is governed by the principle of relevance (Sperber and Wilson, 1995), and aided by reliance on contextual information. In this case, the antecedent of bı` is edziban ‘food’, the grammatical object of the complement clause. However, the use of bı` suggests that the speaker only intends to refer to a part of what the identified antecedent refers to, and not to all the food. In other words, this use of bı` results in a partitive interpretation. The speaker refers to some amount of an already linguistically or contextually available entity, an entity which is mutually in focus by the point in (17) where bı` appears. Edziban (food) is the linguistic antecedent of bı`, but the intended referent of bı` is not identical with the referent of the antecedent. In Enc¸’s (1991) terminology, the relation between bı` and its antecedent can be described as a ‘weak’ and ‘loose’ one.11 This use of bı` thus contrasts with the way that the universal quantifier nyinaa ‘all’ is used, as well as with a null pronoun12, which would have otherwise represented the inanimate antecedent edziban ‘food’. 9 What accounts for the syntactic positional restriction of bı` is not clear to me. What is clear, though, is that bı` and its definite counterpart no` form a paradigm with ebi and . Whereas bı` and no` are restricted to the object position, ebi and occur in the subject position (see section 5). 10 The verb ma means ‘to give’. It, however, has other functions in the language: Lord et al. (2002) explains the relation between the ‘give’ meaning of ma and its causative marker function; Osam (1998, 2008) suggests that ma sometimes functions as a complementizer. 11 The kind of interpretation that arises with this particular use of bı` can be likened to a subgroup of specific noun phrases in Turkish as reported by Enc¸ (1991). In Turkish, when an indefinite noun phrase is marked with the accusative case, it may result in a partitive interpretation. For example, it is only (a) below which can be interpreted as ‘I knew two of the girls’, where ‘girls’ is an entity which has already been introduced into the discourse. a) Iki kiztan yordum Two girl-ACC I-knew ‘I knew two girls’ b) Iki kiz tan yordum Two girl I-knew ‘I knew two girls’ (Enc¸, 1991: 6) 12 In Akan, reference to an inanimate entity syntactically represented as a grammatical object is done by means of a null pronoun, unless the object argument is followed by an adverbial. Even though such an entity is unexpressed, it is understood from the context. (See Stewart, 1963 and Saah, 1992, for more on this subject.) Null objects are not uncommon in Kwa languages. Ga, another Kwa language also spoken in Ghana, exhibits the same phenomenon. Larson (2002) reports that null objects exist in Baule, a Kwa language spoken in Coˆte d’Ivoire.
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
1793
In certain cases, the interpretation which is made available by the principle of relevance is ‘one of this kind’. This is so when the antecedent of bı` refers to a kind of entity that cannot be divided into smaller parts, like reference to a mass, or when it is physically impossible to extract a sub-part of the entity referred to by the antecedent. In (18), B’s intention is not to have a half pair of the shoes A had just shown to her; the only relevant pragmatic interpretation is that what she wants is another pair of shoes which look like what A has, the same model.
A’s intention is to use the phrase mpaboa bı´ ‘some shoes’ to refer to a type of shoes, exemplified by those shoes she is displaying. She intends to highlight a higher-order entity, a type instantiated by a particular token that meets the condition specified in the following relative clause. The reference is to shoes of that make - to such shoes - rather than to the particular left and right shoes displayed. If the speaker intends to refer to the exact pair of shoes displayed, she would have to use the proximal demonstrative determiner yi, as in (19), which is the appropriate determiner for an activated (token) entity. Whereas English may use the same linguistic stimulus, in the form of a demonstrative determiner and a noun to refer to a specific activated token as well as a kind, Akan systematically distinguishes between reference to a type (a general concept), as in (18), and reference to a spatio-temporally bounded entity (a singular concept), as in (19).
When the use of bı` is meant to result in a ‘one of this kind’ interpretation, its function can be described as that of an identity-of sense anaphor, and it is reminiscent of the Norwegian identity-of-sense anaphor det as described by Borthen (2004), and of English that, as in Oh yes, we have that (e.g. central heating), which is grammatically definite even though its referent is an instantiated member of a designated type. Akan bı`, like Norwegian det, needs an antecedent located in the immediate linguistic or extra-linguistic environment to resolve its referent. However, since the referent of bı` is not exactly the same as the reference of its antecedent, the anaphor-antecedent relation can still be described as a weak one. Unlike what Borthen reports for Norwegian det, Akan bı` can result in a partitive interpretation as we saw in the case of (17) above.13 A third use of bı` is exemplified in (20) and (21).14 Here the subject of the bı`-clause is (or is presented as being) part of an event, which has been mentioned earlier in the discourse, or can be inferred from a preceding utterance.
13 Borthen (2004:11) states that the type anaphor det signals ‘‘a focus on the type of thing designated’’ and that ‘‘such type-emphasizing phrases can never be referential or partitive. . .’’. For example, the direct object det of Jeg kjøpte det (I bought it/that) invariably denotes a property (a type or general concept) rather than a sub-part of the nominal referent of the anaphor det. 14 Even though this use of bı` has its peculiarities, possibly due to its collocation with (n)so (also/too), it can be linked to the other two uses in that it may refer to a kind of an already activated entity/situation, or it may result in a partitive interpretation.
1794
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
The excerpt from which (20) is taken is a discussion between a husband and his wife. It is about whether their children should attend a community school which they had been involved in building. Other children in the community had already signed up to be among the group of pupils in the school, and in (20), the husband insists that their children should also attend this school. They are to be ‘a part’ of the school that their parents helped to build. Example (21) should be interpreted within a context where the interactants are enumerating the people billed for a particular trip, and Kwesi will be ‘a part’ of that trip. Notice that the use of bı` as exemplified in (20) and (21) is often (though not always) reinforced by the additive focus marker nso.15 In other instances, as in (22), the subject referent, who is also the speaker, does not intend to be a part of the specific event mentioned, but she intends to engage in an activity of the kind mentioned in the preceding utterance. The speaker of (22) intends to do some things typically associated with people on vacation, and one common denominator is obviously that they do not show up at work. The addressee is expected to be able to form a representation of the kind of vacation desired by the speaker, given the type-features associated with a successful vacation that the antecedent akwama ‘vacation’ of the anaphor bı` is likely to trigger.
In (23), what Kofi has bought is something of the same type that Kwesi has bought–a house.
As illustrated in examples (22) and (23), the communicative function of bı` goes beyond that of an existential quantifier. This is reflected in the inappropriateness of translating (22) and (23) as ‘I will also go some’ and ‘Kofi has also bought some’. Bı` first directs the addressee to an antecedent which provides the basis for deciding how the pronoun is meant to direct the hearer to the intended propositional content of the utterance. We have seen that bı` does not refer to the specific entity represented by the antecedent; it may refer to what one might consider as a part (of the whole) of the referent of the identified antecedent, ‘one of a specific kind’, or ‘one of a similar kind’ of entity. Omitting bı` from the examples provided in this section will result in the creation of a strong anaphor-antecedent relation, with the null pronoun representing a uniquely identifiable referent. I illustrate with (18), repeated below as (24), but this time with two alternative responses from B.
The response in B1, as has already been mentioned, suggests that the speaker wants a pair of shoes which has features similar to the one mentioned by A in (24), and not those particular shoes displayed by A. On the other hand, the response in B2 suggests that what B wants is precisely the pair of shoes that A talked about. 15
A detailed discussion of the marker nso (also/too) is found in Amfo (2007, to appear).
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
1795
The procedural meaning of bı` which is used by the addressee in the reference resolution process can be formulated as follows:
The actual referent of bı` could be a part of the entity denoted by the antecedent. For example, bı` in (17) refers to a part of the amount of food that Araba helps cook. The antecedent of bı` is familiar to the addressee. However, the actual referent of bı` is not already familiar to her, since bı` strictly refers to a portion (not the whole) of the food described earlier. 5. Subject position bi I mentioned earlier on that independent bı` can only occur in object position. Bi can occur in subject position only if it is preceded by the prefixes e- (inanimate) and o- (animate), resulting in the forms e´bı´ and o´bı´ with a high tone on both syllables. Even though it is apparent that both forms e´bı´ and o´bı´ originate from bi, with a prefix marker of inanimacy and animacy, respectively, they behave quite differently. Affixing the animate prefix o- results in the animate existential quantifier o´bı´ ‘someone’.16 This form can be used in both subject and object positions and like English someone it can be used at the beginning of a discourse. It is not an anaphor that must be saturated through association with an antecedent. On a mother’s return from work, her child who has been at home for some time can utter (25).
Assuming I am in need of some research assistants, I can utter (26) to one of my colleagues who I believe is in a position to help me find someone.
´ bı´ does not even necessarily represent a specific individual (with the cognitive status referential of the Givenness O Hierarchy). On the other hand, inanimate e´bı´ cannot be used felicitously without the presence of some form of antecedent in the preceding discourse, so unlike o´bı´, it cannot be used at the beginning of a discourse. In (27), the antecedent of the anaphoric e´bı´ in B’s utterance is paanoo ‘bread’, which was mentioned earlier by A.
16 What can be appropriately described as the inanimate equivalent of obi is biribi (something). The morphological components of this word are not clear, even though it is reasonable to assume that either the first or last syllable (or both) is the quantifier bi, which we have been concerned with here. Like obi, there are no syntactic occurrence restrictions on biribi.
1796
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
E´bı´ and bı` are in complementary distribution. While e´bı´ is ungrammatical in object position, bı` is ungrammatical in subject position, as shown in (28) and (29).
Bı` in contradistinction to e´bı´ and o´bı´, is underspecified in relation to animacy; it can be used in reference to animate and inanimate entities. Imagine that in a conversation with one of my colleagues, she tells me that there are some research assistants available, and I happen to need some. I could utter the sentence in (28a), to indicate that I need some (or one) of the people she has just talked about. 6. Bı´ and markers of definiteness We have established in the previous sections that the cognitive status of referents of both determiner bı´ and independent bı` is referential; that is, they refer to a non-familiar, non-uniquely identifiable instantiation of an entity denoted by the head noun in the case of the determiner, and to some subpart or token instantiation of the type of object denoted by an antecedent when bı` is an independent argument. Now I am going to consider the fact that bı´ may co-occur with the definiteness markers no (the), yi (this) and the prenominal demonstrative saa/dem in a single noun phrase, as illustrated in (30) and (31).
How is the referent of a noun phrase which contains bı´ as well as one of the definite markers resolved? The clue lies in a reasonable syntactic analysis of such complex noun phrases. In (30), for example, nkor fo (people) and no (the) form a constituent at a lower level, and the larger phrase nkor fo no bı´ (some of the people) is a constituent associated with a higher node whose constituents are nkor fo no and bı´, as indicated by the bracketing in (30).17 There is a mother node which is grammatically indefinite and a daughter node which is grammatically definite. No and yi belong to the daughter constituents in (30) and (31), respectively, which are syntactically embedded under the larger indefinite phrases, bounded on the right by bı´. In all likelihood, the referents of nkor fo (people) in (30) are already known to the addressee, the additional information provided by bı´ is that the whole noun phrase nkor fo no bı´ is in reference to some (and not all) of the members of that group. Even though the addressee is already familiar with this whole group, 17 The constituent structures assumed for (30) and (31) are based on native speakers’ intuitions and the fact that they permit compositional derivation of the intended interpretations. The constituent breakdown of the daughter node in (31) is more complex, and may be represented linearly as follows: [Dem N N Det]. It contains a nominal head, plus another nominal modifier, a pre-head demonstrative, which cannot occur without the help a post-head determiner. In Akan, the usual position for determiners including articles, demonstratives and quantifiers is post-nominally. It is only the demonstrative dem and its dialectal variant saa which occur pre-nominally, but they co-occur mandatorily with a supporting post-nominal determiner. As a result, the post-nominal determiner cannot occur outside of an NP which contains only the pre-nominal determiner dem/saa and head noun. This obviously poses a challenge for a strictly binary analysis.
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
1797
he is not familiar with the dissenting sub-group, but he is able to form a representation of them and construe them as the subject argument of the predication. In (31), the interpretation derived from the combination of the definite markers dem and yi and a following bı´ is ‘one of this sort’. The referent of the subject noun phrase is an entity of the kind described by the descriptive adjective and the head noun. When the definiteness marker precedes bı´ so that the larger complex noun phrase syntactically dominates a lower definite constituent, the intended referent of that noun phrase is ‘a part of the whole’ or ‘one of this sort’. This syntactic configuration allows the communicator to introduce a topic in the subject position, the topic in (31) being a type of occupational activity whose identity is specified in the post-copula predicate. ‘Modern jobs’ is the general concept which the speaker intends the addressee to concentrate on. However, the discourse-new exemplification of such jobs is found in the nominal predicate phrase after the copula: ndowanyen ‘bee rearing’. In (30), the grammatically definite daughter constituent refers to a set of tokens, an identifiable group of persons who have been introduced earlier on in the discourse. In such complex constructions, the definiteness marker (yi/no´) which form part of the grammatically definite daughter constituent underdetermines the answer to the question whether that constituent denotes a type, as in (31), or refers to a token (or a set of tokens), as in (30). When the definiteness marker follows bı´, the noun phrase is presented as an echoic metarepresentation (cf. Noh, 2000) of an earlier use of a given phrase N bı´. In (32), the indefinite phrase maame bi (a certain woman) belongs to a daughter node which is embedded under a mother node marked at the right edge with the definiteness marker no. The form maame bı´ (a certain woman) should be an activated discourse entity to warrant the use of the string maame bı´ no (that certain woman). In other words, there should have been an earlier reference to maame bi (a certain woman) and for some reason this needs to be repeated in the present discourse as an echo of the earlier occurrence. The lower phrase maame bi is embedded under a higher one whose function is simply to signal that the embedded element is a metarepresentation – an entity which has been mentioned earlier on and is only being repeated.
7. Conclusion The present paper has been concerned with the communicative role of Akan bi both as a determiner and a pronominal. Bi provides information about the cognitive status of the relevant nominal referring expression and thus aids in identifying the intended referent The information provided in the form of the cognitive status of the referent of a bi-NP is procedural rather than conceptual; this is because such information does not contribute directly to the conceptual representations which serve as inputs into the inferential process of utterance interpretation. Rather, the semantic information encoded by bı` or bı´ provides crucial information which leads to the resolution of the intended referents of a bı´-modified noun phrase and pronominal bı`. For bı`, the establishment of a referent begins in the immediately preceding discourse, where an antecedent is located. The addressee then considers a token which introduces a type of entity specified in the antecedent, to which the token belongs; this token could be just a portion of, or a specific representative of, the type of entity the antecedent represents, or it could belong to a type which is similar to the one represented by the antecedent in that it shares certain properties with it. Akan grammar thus linguistically encodes a distinction between reference to a general type and reference to a token or an instantiation of the general type. The use of bi serves as a crucial indicator that the addressee needs to construct a representation of a specific token, while concentrating on the properties that characterise this token as a representative of a specific type of entity. Acknowledgements I am grateful to Kaja Borthen, Thorstein Fretheim and an anonymous reviewer for very helpful comments. References Amfo, Nana Aba Appiah, 2007. Explaining connections in Akan discourse: the role of discourse markers. Languages in Contrast 7 (2), 119–136.
1798
N.A.A. Amfo / Journal of Pragmatics 42 (2010) 1786–1798
Amfo, Nana Aba Appiah, to appear. Lexical signaling of information structure in Akan. Linguistics. Blakemore, Diane, 1987. Semantic Constraints on Relevance. Basil Blackwell, Oxford. Blakemore, Diane, 2002. Relevance and Linguistic meaning: The Semantics and Pragmatics of Discourse Markers. Cambridge University Press, Cambridge. Boadi, Lawrence A., 2005. Three Major Syntactic Structures in Akan: Interrogatives, Complementation and Relativisation. Black Mask Limited, Accra. Borthen, Kaja 2004. The Norwegian type-anaphor det. In: Branco, A., McEnery, T., Mitkov, R. (Eds.), Proceedings of the 5th Discourse Anaphora and Anaphor Resolution Colloquium (DAARC 2004). Edic¸o˜es Colibri, Lisboa, pp. 9–15. Enc¸ Mu¨rvet, 1991. The semantics of specificity. Linguistic Inquiry 22, 1–25. Fretheim, Thorstein, in preparation. Reference and Linguistic Underdeterminacy. Monograph, Norwegian University of Science and Technology. Fretheim, Thorstein, Amfo, Nana Aba Appiah, 2008. Reference, determiners and descriptive content. In: Klinge, A., Høeg-Mu¨ller, H. (Eds.), Essays on Nominal Determination. John Benjamins, Amsterdam, pp. 339–366. Gundel, Jeanette, Nancy, Hedberg, Ron, Zacharski, 1993. Cognitive status and the form of referring expressions in discourse. Language 69, 274–307. Gundel, Jeanette, Nancy, Hedberg, Ron, Zacharski, 2001. Definite descriptions and cognitive status in English: why accommodation is unnecessary. English Language and Lingusitics 5 (2), 273–295. Larson, Martha, 2002. The semantics of object drop in Baule. In Nissim, M. (Ed.), Proceedings of the Seventh ESSLLI Student Session, pp. 1–10.
. Lord, Carol, Foong Ha Yap, Shoichi, Iwasaki, 2002. Grammaticalization of give: African and Asian perspectives. In: Wischer, I. (Ed.), New Reflections on Grammaticalization. John Benjamins, Amsterdam, pp. 218–235. Noh, Eun-Ju, 2000. Metarepresentation: A Relevance-Theory Approach. John Benjamins, Amsterdam. Osam, E. Kweku, 1998. Complementation in Akan. Journal of African Languages and Linguistics 19, 21–43. Osam, E. Kweku, 2008. A note on the cross-dialectal diffusion of grammaticalization: the case of ma in Akan. Acta Linguistica Hafniensa 40, 111– 119. Saah, Kofi K. 1992. Null object constructions in Akan. In Collins, C. and V. Manfredi (Eds.), Proceedings of the Kwa Comparative Syntax Workshop. MIT Working Papers in Linguistics, vol. 17, pp. 219–244. Scott, Kate, 2008. Reference, procedures and implicitly communicated meaning. UCL Working Papers in Linguistics vol. 20, pp. 1–17. Sperber, Dan, Wilson, Deirdre, 1995. Relevance: Communication and Cognition, 2nd edition with a postface. Blackwell Publishing, Oxford. Stewart, John M., 1963. Some restrictions on objects in Twi. Journal of African Languages 2, 145–149. Wilson, Deirdre, Sperber, Dan, 1993. Linguistic form and relevance. Lingua 90 (1/2), 1–25.
Akan Texts Anonymous, 1961. Ananse Akuamoa. Bureau of Ghana Languages, Accra. Coleman, S.K., 1995. Woana nye Araba nyinsenii K,. Sam-Woode Limited, Accra. Nana Aba Appiah Amfo received her PhD from the Norwegian University of Science and Technology, and is a Senior Lecturer at the Department of Linguistics, University of Ghana. Her research focuses mainly on Akan (a Kwa language), and it includes the semantics and pragmatics of discourse markers, reference resolution, information structure, and gramaticalization. She have recently published in journals such as Languages in Contrast (2007), Studies in African Linguistics (2007) and Lingua (2007). Together with Thorstein Fretheim, she contributed chapters to the books Languages and Cultures in Contrast and Comparison and Essays on Nominal Determination: From Morphology to Discourse Management, both published by John Benjamins (2008).