Interpretation

Semantic tests are one source of behavioral data for which a theory can be made to account. More narrowly then, it should predict certain native-speaker responses for given tests. Beyond that, it is assumed that responses reflect the relevant portions of speakers’ codes and that the theory will also account for everyday language behavior.

It is important to stress the premise that a hearer will try to interpret or at least to account for a speaker’s utterances. He will often accept and make sense even out of sentences that initially appear odd or contradictory when presented to him on semantic tests or in actual discourse, or he may reject them and then proceed to interpret them anyway. On actual tests, informants’ responses have shown greater agreement on the interpretations, matchings, and rankings of sentences than on judgments in terms of grosser discriminations of unacceptability, meaninglessness, anomaly, etc. Such discriminations often depend on context and what criteria are used to make them may be inaccessible and possibly irrelevant. Judgments in terms of such discriminations are heuristically useful in analysis, but the central part of a semantic description is constituted by the definitions of lexemes and other syntactic elements. From these, the meanings of compound expressions of unlimited complexity can be correctly inferred. Whether particular combinations, however, are ‘impermissible’ or ‘anomalous’ is left to individual speakers - including logicians - to decide. It is not for the description to make such judgments but at most to account for how they may be made.

It appears, then, that more reliable indices of semantic content are provided by tests which require informants to perform tasks derived from the more usual, every-day activity of trying to understand sentences, such as giving interpretations, ranking for interpretability, matching for similarity in meaning, but a theory which uses responses to such tests with discrimination must also account for them by replicating the hearer’s manipulation of the code in the process of interpretation. Besides the fact that we make sense of utterances in spite of constantly occurring lapses, the code can be manipulated creatively to produce sentences that are more than every-day prose. Such an advertising line as ‘ If you were a banker, would you lend you money?’ (N.Y. Times 1967) successfully communicates what ‘. . .would you lend yourself money?’ does not. Weinreich (1966) constructed a semantic theory which explicitly includes deviant utterances. Replicas of code manipulation in (re)interpretation of sentences were given in 8.1 and 8.4.1-3 above. Two more examples follow.

Further above, He has the book that is his illustrated the independence of the accidental and inherent relations. However, He has the book that he has presents a different case. Let us say that sentence structure, as in He has the book, conveys the meaning ‘this is new information’. On the other hand, the construction definite article plus relative clause of the book that he has seems to identify the referent as the one about which the previous new information ‘he has a book’ has already been given. So it appears ‘A has B’ is and is not new information. In the but-test, irony was one alternative of reinterpretation given to resolve a contradiction; so here also: ‘ one of these contradictory statements is a lie, namely that this is new information, in order to let you know that I do not want to give you new information in the matter’. The first interpretation of two such occurrences of A has B as found here would be that they are tokens of the same type and not of two different, homonymous types. Thus another reinterpretation might be that the second occurrence is a token of the inherent A has B.

Consider this causative-have construction with embedded passive sentence: John had the walls whitewashed by the painter, i.e. A has B V_tr + en by C. For embedded passives, the causative-have construction includes the restriction A ≠ C. To read the semantic import of such a restriction, the theory will include the following rules:

(24) (i) Let: (a) a Phrase be a member of the class of structural symbols, such as S or NP [here A, B, C represent NPs]

(b) a Type be a member of the class of formal symbol types, such as any particular morpheme symbol or string thereof

(c) a Token be a member of a class of symbol tokens which class is symbolized by a Type.

Then: for any particular utterance, the referent of Phrase_i = the referent of the Token whose Type is labeled Phrase_i in the structural description of the utterance.

(ii) A rule ‘Phrase_i = Phrase_j’ should be read as ‘referent of Phrase_i = referent of Phrase_j’.

(iii) Referent_i of Phrase_i — referent_j of Phrase_i.

Rule (24 iii) states that each Token has only one referent (and is not an allusion to the singular-plural distinction). Thus the restriction A ≠ C should be read as ‘referent of A ≠ referent of C’, i.e. ‘John is not the painter’. Now assume that someone utters a sentence to us which appears to have the structure A has B V_tr + en by C in its first portion: The painter had the walls whitewashed by himself and not by his assistant. This utterance appears odd, and we might interpret it in at least two possible ways.

On the one hand we might take it as a lapse and understand that the speaker meant ‘The painter himself, and not his assistant, whitewashed the walls’. The account of this interpretation would run as follows: (a) The occurrence of the reflexive seems to indicate that the noun phrases the painter and himself are derived from underlying constructions in which one equals the other, i.e. A = C, giving them the same referent, (b) The utterance seems to represent the causative-have construction with the restriction A ≠ C , indicating, on the contrary, that the two noun phrases do not have the same referent, (c) We resolve the contradiction by ignoring the causative-have construction, thus eliminating the rule A ≠ C.

On the other hand, we might take the speaker more seriously and keep the causative have, crediting him with meaning something thereby. We could then suspend rule (24 iii) and assign two referents each to the two noun phrases: ‘A1’, ‘A2’ and ‘ C1’, ‘ C2’. We equate ‘A1 = C1’, satisfying the implication of the reflexive, A = C, and apply the causative-have restriction as 'A₂ ≠ C₂'. This leaves the painter as one individual but two personae, and, using the occurrence of assistant as a cue, we might credit the speaker with having meant to convey some such message as ‘The painter in his capacity as boss had the walls whitewashed by himself in his capacity as worker and not by his assistant’.