DISCUSSION* |
Meaning,
|
| Pavel Tichý's claim that linguistics turns
its back to the task of specifying and analyzing the concept of meaning, is true to a
large extent. However, there are trends in linguistics which attempt not only to handle
this concept, but to do so in close connection with Tichý's Transparent Intensional
Logic, to present a framework for describing how a given language maps his 'constructions'
by its sentences. In any case, it should be understood that this task is by far not
simple: the presystemic notion of meaning needs more than one explicatum; and the way from
sentences to constructions (and thus to propositions) consists in more than one step. Among
the linguists working in this direction, there is the research group of theoretical and
computational linguistics at Charles University, in which we have profited from regular
discussions with Pavel Materna and other Czech logicians, and especially from those with
Barbara Partee and her followers. I would like to summarize the views of this group, a
synthetic presentation of which can be found in Sgall, Hajičová and Panevová (1986);
specific contributions relevant for the present discussion were published by Vlk (1988),
Peregrin (1988; 1992), Petkevič (1987), Sgall and Hajičová (1992). 1. Linguists, and also logicians, often work with a notion of meaning as structured by an individual language, cf. such headings as |
Saussure's 'form of content' (versus 'substance of content'
in an ontological sense), Fillmore's 'internal semantics' (divided by a gap from 'external
semantics', or from the logical analysis of language), Chomsky's 'logical form' (against
'semantic interpre-tation'), Lewis type meaning (as compared with a formal output language
of the logical analysis). This dichotomy is based on the assumption that
the internal patterning of language comprises a disambiguated level, at which not only
ambiguity, but also synonymy and the idiosyncratic irregularities of the outer form of
natural language are absent. This level offers an appropriate starting point for semantic
interpretation. Let us note that it would be more precise to speak about a semantic-
pragmatic interpretation, since also such pragmatically based items are patterned by
language as tenses, modalities, indexicals such as here, yesterday, etc.
One of the assumptions usually connected with a procedure of interpretation is that its input lacks ambiguity, so that the procedure can be understood as specifying a function. Therefore, linguistic meaning in the just characterized sense may be viewed upon as an underlying level, or a set of representations of sentences, yielded by the linguistic analysis proper and used as an interface between linguistics on one side and the logical analysis of natural language (or another approach to cognition) on the other side. The sentence, with its representations on the different levels of the language system (from phonetics to the underlying structure), can be characterized as a unit of the system |
DISCUSSION/ PETR SGALL* MEANING, SENTENCE . . . |
82 |
| of language. Sentence has to be distinguished from an
utterance (token), i.e. from an occurrence of a sentence in a discourse. Let us illustrate
this difference: (1) My father has not read the book. In order to understand (1) qua sentence, i.e. to identify the meaning of this linguistic unit, it is necessary to get aware of such ingredients as a certain relationship between the speaker of (1) and a(nother) man, a certain kind of an object, a relation between the man and the object (an action), and the negation of this relation. 'Understanding' in this sense differs from 'understanding' an occurrence of this sentence when uttered in a discourse: the question "which book" is not relevant for understanding the sentence, although it is crucial for understanding the utterance. P. Tichý is right in stating that a class of underlying representations constitute another language (cf. D. Lewis on "markerese"), rather than the outcome of semantic interpretation. However, such an intermediate level can find its justification if it is shown to be useful as a stage in the overall process of interpretation. The above mentioned character of a disambiguated level (suitable as the input for the semantic interpretation) is convincing in this sense. The underlying level is an interface, allowing for an appropriate division of labour between linguistics itself (responsible for the formulation of the metalanguage of this level, specifying its units on the basis of operational criteria, and for the description of its |
relationship to the outer shape of sentences in all its
intricacies) and the interdisciplinary domains in which linguists cooperate with
logicians, psychologists, specialists in cognition science and others. It may be assumed
that the structure of the underlying level, and thus also the formal framework necessary
for its description, is common to most different natural languages; only the setting, the
repertoires of categories and their values (along with the basic order of sentence parts)
differ from one language to the other. Languages are much more similar to each other at
this level than in the outer shapes of their sentences. 2. If the underlying level as characterized in 1. above is understood as the first step in the series of explicata for 'meaning', then among the possible next steps there is that of specifying the reference of referring expressions present in the utterances. This step (the question "which book" in connection with (1) above) is necessary not only for the understanding of an utterance as a part of a discourse, but also for the specification of truth conditions. We do not understand this step as sufficient; certainly, it is also necessary, in a complete account of language comprehension, to handle several other, rather heterogeneous factors, among which there are the figurative (metaphorical and other) meanings and similar aspects studied by cognitive linguistics, stylistic and other connotations, as well as expectations given by cognitive frames typical of certain kinds of (chains of) actions, and so on. |
DISCUSSION/ PETR SGALL* MEANING, SENTENCE . . . |
83 |
| It should be noticed that truth conditions
cannot be attributed directly to a sentence (as a unit of the system of language), but
rather an utterance (token), if we do not have in mind just mathematical assertions. The
truth conditions of (1) differ according to who utters the sentence (referent of the
indexical part of the meaning of my), which book he has in mind, and what is the
time point of the utterance (relevant for the value of the tense). The notion of reference should not be understood so that we would be committed to the naive ontology, assuming that "the objects" of the world are immediately accessible to human cognition. It is sufficient to assume that natural language and its use are based on a kind of naive ontology being accepted in the basic layers of language use (everyday conversation, narration, etc.). In philosophical texts it is well possible to discuss the patterning of our image of the outer world to be conditioned by language. Reference assignment does not belong to the system of language; this step takes us from language itself to the domain of language use. In identifying the referents of expressions, the participants of a discourse take recourse to their non-linguistic mental capacities. A preliminary formulation of a finite mechanism allowing the hearer to determine reference is based on the degrees of salience of the elements of the stock of knowledge (information) shared by the speaker and (according to her/his assumptions) by the hearer. This psychological mechanism includes the hypothesis that the most activated item at a certain time point of the discourse |
prototypically is that which has been mentioned in the
immediately preceding utterance as its focus proper. If the underlying structure
enriched by reference assignment is called the sense of the sentence, this notion can be
compared to Frege's 'sense' in that it constitutes a more subtle patterning of the content
of sentences than the patterning determined by pro-positions. If a proposition is defined
as a function from possible worlds into truth values, then our notion of sense comes close
to Tichý's constructions, which also constitute such a more subtle patterning. 3. The third step of our analysis then may concern propositions, truth conditions, the layer of intension. The sense of a sentence (more precisely, a given sense of an underlying structure of a sentence) determines a proposition. Even if another approach to meaning is chosen than that of truth- conditional semantics, a counterpart of the concept of truth conditions is needed, since otherwise it would not be easy to distinguish a semantically valid difference between two words or morphemes from other (syntactic, stylistic) differences. We can state that a difference between two word forms (or their collocations, word order positions, etc.) A and B is semantically relevant if there is a (non-quotational) context in which A and B are not freely substitutable 'salva veritate', i.e. if there is a pair of sentences S and S', |
DISCUSSION/ PETR SGALL* MEANING, SENTENCE . . . |
84 |
| where S differs from S' only in that S contains A in the
position where S' contains B and the truth conditions of S and S' are not identical (which
can be checked operationally); if no pair of S and S' with these properties is found, then
A and B may be assumed to be synonymous. 4. The next step can then be seen in the transition from intension to extension, from truth conditions to truth values as specified for a given possible world (or situation, and so on). The sentences of natural language, such as (1), are true in some situations, false in other ones, and a third possibility is that of a presupposition failure, when a possible world is assigned no truth value. As E. Hajičová has pointed out, in some cases a definite noun group triggers a presupposition if it belongs to the topic of the sentence, whereas a noun group belonging to the focus triggers no such presupposition: (2) (a) The king of France is bald. (b) Yesterday Prague was visited by the king of France. If the sentences are pronounced with normal intonation, then (2)(a) cannot be appropriately used about such possible worlds as the actual one (according to what we assume to know of it), with the obvious reference assignment, since neither the sentence itself, nor its negative counterpart are true. On the other hand, (2)(b) can be used as a false statement, since a |
prototypical token of its negative counterpart holds true.
Instead of a pre-supposition, another specific kind of entail-ment is present here, called
'allegation' by Hajičová (see Sgall et al., 1986, Chapter 3 and the writings quoted
there). It seems that this characteristics is appropriate for definite noun groups
rendering individual offices in Tichý's sense, along with nominalized groups such as in
the following example: (3) (a) Our victory was caused by Harry. (b) Harry caused our victory. Here the negative counterpart of (3)(b) is not connected with any entailment concerning the victory, whereas the positive sentence entails that the victory took place. Thus the series of explicata for the notion of meaning can be understood as completed: from the disambiguated underlying level we came by means of reference assignment to sense, which determines the intension, possibly relativized to a class of possible worlds to yield the extension of the sentence. The corresponding steps for predicates and terms can then be reconstructed on this basis in a way similar to Carnap's elaboration of the intension, using the mentioned criterion of ex-changeability for synonymy and other known criteria for checking the boundary line bet-ween the ambiguity of linguistic expressions and the vagueness of literal meaning. 5. A few remarks may be added, the first of which concerns the relationships between syntax, semantics, and pragmatics. The semiotic trichotomy has its significance as an analytic abstraction, but it would be |
DISCUSSION/ PETR SGALL* MEANING, SENTENCE . . . |
85 |
| detrimental to assume that natural language is patterned
along its lines. We have already seen that both purely semantic and pragmatic aspects are
present at all the stages of the analysis of meaning. There are pragmatic moments
patterned by the system of language, as well as purely semantic items not patterned by it.
The intrinsic interconnections of pragmatics with semantics on a single level can be
illustrated by such lexical units as yesterday, with which the pragmatic
(indexical) and the purely semantic (=day) parts are integrated in a single
lexical meaning. As P. Tichý states, also the relationship between semantics and syntax is more complex than is usually realized. We may add that, since every representation of a sentence has its syntax, we have to work with syntax at every stage of semantic-pragmatic interpretation. Syntax is present not only as the class of relations between linguistic expressions, but also in every system used to render the output of the interpretation. Having this in mind, we should ask how to describe the differences between the syntax of a formal system and that of natural language, which has no explicit means for variables, for operator scopes, no fixed arity of predicates. Thus, if Tichý points out that e.g. a prepositional group has no meaning of itself, this can only stress that in a representation rendering the syntactic patterning of a sentence, the counterpart of the prepositional group has to occupy a certain syntactic position (as a |
certain kind of obligatory or optional complementation of the
verb or of another word, with a classification of the complementations and with their
transduction to arguments or adjuncts in the output of the interpretation); this also
concerns word order positions in a language displaying a certain degree of
configurationality. In the recent versions of several linguistic
theories, including Chomsky's 'minimalist program' and also our approach (Sgall, 1992),
the assumption emerges that it is not necessary to differentiate between two levels of
sentence structure (surface and underlying syntax); this would mean that what was stated
above on the underlying structure as the first element of the series of explicata for
'meaning' would rather concern syntactic structure as such. We would then have to work
with disambiguated representations of sentences as with the only representations of their
structure. This may simplify the whole procedure leading from the outer form of
expressions to truth conditions, although the necessity of elaborating an interface level
appropriate for the collaboration of linguistics and logic would not diminish. A closing remark may be devoted to the fact that a suitable formal framework for the semantic-pragmatic interpretation may include partial functions. We have mentioned that a proposition as a function from possible worlds does not always assign every world a truth value. With presupposition failure a subclass of possible worlds is assigned no truth value, the given sense of a sentence cannot be appropriately used in the given time point of the discourse. With such a contradiction as in (4), we even get no truth value for any |
DISCUSSION/ PETR SGALL* MEANING, SENTENCE . . . |
86 |
| possible world (with that sense of the sentence in which the
words round and quadrangle are understood in their literal meanings): (4) This round quadrangle belongs to Jim. However, also the assignment of a proposition to a sense of an
utterance should be understood as a partial function only: (5) This utterance is not true. As P. Tichý has shown, such a paradoxical assertion as that of the Liar's sentence in the form of (5) cannot be viewed upon just as corresponding to a proposition that assigns no possible world any truth value. Thus, an
utterance of (5) in its self-referring sense corresponds to no proposition. |
REFERENCES Peregrin J. (1988): Intersubstitutivity scales and intension. In: Proceedings of the
'87 Debrecen Symposium, ed. by I. Rusza and A. Szabolczi. Budapest, 185-197. Peregrin J. (1992): Sprache und ihre Formalisierung. Deutsche Zeitschrift für Philosophie 40:237-244. Petkevič V. (1987): A new dependency based specification of underlying representations of sentences. Theoretical Linguistics 14: 143-172. Sgall P. (1992): Underlying structure of sentences and its relations to semantics. In: Wiener Slawistischer Almanach, Sonderband 33, Wien, 273-282. Sgall P. and E. Hajičová (1992): Linguistic meaning and semantic interpretation. In: Current advances in semantic theory, ed. by M. Stamenov. Amsterdam. Sgall P., Hajičová E. and J. Panevová (1986): The meaning of the sentence in its semantic and pragmatic aspects, ed. by J. Mey. Amsterdam - Prague. Tichý P. (1993): Scandal of Linguistics. In: From the Logical Point of View, 3/92 Vlk T. (1988): Towards a transduction of underlying structures into intensional logic.
Prague Bulletin of Mathematical Linguistics 50: 35-70. Petr Sgall Charles University |