<<   >>   TOC

10. Covert determiners


On the basis of evidence from languages with usually overt determiners, I claimed that NPs including proper names are always wrapped in a DP. The assumption of such a wrapper function does not only belong to the standard theoretical inventory of current syntactic theory, but also bolsters the argument made above that proper names and common nouns alike have to have the same semantic type in order to be selectable by another head. The claim is thus a two-fold one: all nominals, when they occur within a sentence (vocative use may be an exception), have the same semantic and syntactic type. Only on this basis can a truly general compositional grammar be formulated: just as rules of phonology apply to natural classes of sounds, rules of syntax/semantics apply to particular syntactic/semantic types. Whenever a subcategorization frame for a single lexical head has to be formulated using a disjunction, it is likely that a generalization was missed.

I am claiming here that the generalization missed so far is that proper names and common nouns form a natural class (together with other nominals) as far as syntax and semantics are concerned. I am not claiming that it is necessarily false to assume that proper names are identifiable as a separate class from common nouns. If someone is asked to list English proper names starting with the letter ‘s’, Sarah or Scott are likely to be included, but certainly not scissors or sandal. One may argue that proper names and common nouns are stored with separate access functions in the lexicon, but this is dangerously close to the original assumption of separate semantic categories. Or one may claim that the massively parallel processing power of the human brain allows to sort out randomly selected lexical entries according to their use in a fraction of a second. This is ultimately an empirical question to be answered by psycholinguistic experimentation. If there is a categorial difference between proper names and common nouns at all, it is, however, not relevant for the computational purposes of syntax/semantics.

   If it is accepted that nominals in sentences are always introduced by a determiner, one has to ask what the exact semantic function of a covert determiner may be. Contrary to common practice, I do not intend to invoke an empty category that nevertheless carries semantic content. Not only would such an assumption be difficult to falsify, but it would also suggest a ‘hard-coded’ functionality of the projection in question that is most likely inadequate cross-linguistically. A covert determiner in German is formal, polite standard in German, but informal, familiar style in Portuguese. It is thus hard to see what the lexicon entry of the given empty functional head should look like.

What I have in mind is much simpler: in languages that do emply overt determiners, an empty determiner can be treated as phonologically encoded, insofar as the lack of a lexical item is significant, if an overt item is normally to be expected. The covert determiner thus pragmatically gains meaning, because it is not overt.

Heim (1991) stipulates a similar negative semantics for the indefinite article, based on a conversational maxim like (32):

(32)       “Avoid the indefinite article if you can use the definite one.” (515)

When an indefinite is used, the listener will thus accommodate as follows: if the speaker wanted to refer to a discourse referent we already know, s/he would have used the definite article. Since s/he did not do that, s/he must be intending to introduce a new discourse referent.

Grice’s Maxim of Quality and Quantity allow to derive a conversational rule like (33):

(33)       “Avoid ambiguity where you can”.

From (33) follows (34) for naming strategies:

(34)       Use names such that unique identification is possible.

If a speaker now chooses to omit determiners completely, the listener can draw conclusions: if the speaker is trying to be as unambiguous as possible, but does not bother to even tell me whether the discourse referent is novel or familiar, s/he must assume that the intended discourse referent is straightforwardly identifiable. Thus, the DP must be a name (a pragmatic category) referring to a unique individual (or a unique set if the NP is marked for plural, as in Smiths). There are several scenarios for successful accommodation:

(a)                 The listener has one matching discourse referent available. If the sentence makes sense, s/he will assume that the speaker intended to refer to this discourse referent. If the sentence does not make sense, the listener may assume that the speaker intended to introduce a new discourse referent.

(b)                The listener has several matching discourse referents available. If all or at least several would provide a meaningful interpretation of the sentence, accommodation is interrupted and a repair question becomes necessary.

(c)                 The listener does not have matching discourse referents available. S/he will thus introduce a new discourse referent into the domain of discourse. If it seems that the speaker did not presuppose any knowledge about this discourse referent, simple introduction of a new referent already represents sufficient accomodation. If some knowledge seemed to be presupposed, a repair question is necessary.

        At first glance, the necessity of accommodation seems quite burdensome to the listener, so one may raise the question of why languages would withhold information in constantly reoccuring conversation types? I believe that the answer is very simple: the explicitness of language is determined in a constant struggle between lazy speakers who try to minimize their effort and lazy listeners who just hate to be accommodating. Conversations, for the most part, are about people and the set of possible discourse referents is usually fairly stable and limited in conversations between two given persons. Thus, speaker A may use a very common name like John in a conversation with speaker B and still have a good chance to be understood, since the number of Johns that could possibly come up in a conversation between A and B is most likely relatively small. Proper names function as descriptive shortcuts: they are predicates that are chosen because they yield exactly one discourse referent when computed in an ideal situation. Therefore, from an economical point of view, they provide the best of both worlds for speaker and listener. The speaker can rest her/his tongue, while the listener should be able to receive an unambiguous return value when the descriptive function supplied with the proper name is resolved. I have already discussed situations that deviate from the ideal state, and here I repeat that the possibility of failure to yield a unique discourse referent is evidence for the predicative nature of proper names. To most analysts, they do not feel like normal predicates, simply because the ideal situation for the use of a proper name is usually taken to exhibit the “true nature” of names. While abstraction may be a useful tool to approach and explore a particular topic, I doubt that a semantics that is based on idealizations can reach the optimal usefulness.     

<<   >>   TOC

© Philipp Strazny 1998