The present paper has a double purpose. Firstly, it attempts to illustrate how ontology enters the realm of truth-conditional model theoretic semantics. To this effect I will use Nicholas Asher's theory of predication and shall discuss some linguistic examples that are relevant for assessing this approach to predication. Secondly, the paper is intended to take into account the theory of predication so defined within the Minimalism/Contextualism debate. I shall conclude by arguing that we can envisage in such a theory a profitable position which shares characteristics form both Minimalism and Contextualism, that may be called Enriched Literalism.
Minimalists like Borg (2004, 2012) and
Consider the following sentences: (1) a. John enjoyed the cigarette. b. The bottle froze.
c. John heard the piano.
Sentences (1a-c) have a literal meaning which is given by the truth conditions and
which distances itself from the intuitive content we usually would attach to them.
Thus, (1a) means that John enjoyed the cigarette, but what kind of eventuality links
John to the cigarette is not part of the literal meaning of the sentence. (1b) means that
the bottle froze, even if its intuitive truth conditions arguably regard the liquid inside
the bottle. And (1c) means that John heard the piano, and not that John heard the
sound emitted by the piano, like we would intuitively say. In other words, intuitive
truth conditions and satisfaction conditions are not something semantics ought to
account for
On the contrary, contextualists like Recanati maintain that “we should do our best
to account for the intuitive truth- and satisfaction-conditions of utterances, and to that
effect we may have to liberalize the notion of meaning/content to the point of blurring
the semantics/pragmatics distinctions”
I want to suggest a way for having both a theory of literal meaning, namely a semantics disentangled from pragmatics, and a theory which accounts for intuitive truth conditions without blurring the semantics/pragmatics distinction. The result would turn out to be an Enriched Literalism which shares with Borg, Cappelen and Lepore the need of distinguishing semantics and pragmatics, and with Recanati the wish of accounting for intuitive meanings. To this effect, I take on an extended version of the typed lambda calculus for modeling predication.
Let us consider a formal model of predication1 which incorporates new and complex
equipment to the standard typed lambda calculus
Firstly, it is worth saying from the outset that selectional restrictions of words in compositional processes are taken seriously in the theory, thus predication is modeled as type restricted β-reduction and, accordingly, lexical entries are modeled as typed lambda terms. The introduction of subtypes of E is justified by the following semantic anomalies: (2) a. # The number 2 weighs 1 kilogram.
b. # John is divisible by 5.
If we assume that ‘The number 2’ denotes a particular abstract object and that ‘John’
denotes a particular concrete person, then the predications in (2a-b) turn out to be
malformed. The reason “is that there is a conflict between the demands of the
predicate for a certain type of argument and the type of its actual argument”
(3) λP : E ⇒ T λx : P (W1kg(x) ∧ P (x)) It is said that lambda terms impose restrictions due to the fact that they express type presuppositions on their arguments. As it is clear from (3), such presuppositions are encoded in lambda terms via the standard colon notation. Numbers are of type A, and A P =⊥,2 but (3) requires an argument of type P. that is why (2a) results semantically malformed: an unresolvable type clash between the type presupposition of the verb and the argument’s type occurred during the β-reduction. Mutatis mutandis, the same considerations hold for (2b).3 2 This means that there are no objects of type A and P. In other words, such types are incompatible. To realize the importance of innovation (ii) let us consider now the following sentences: (4) a. My lunch weighs 1 kilogram.
b. My lunch lasts two hours.
Strangely, (4a-b) are both felicitous, even if the two lambda terms of the predicates
require arguments of different and incompatible types. ‘Weighs 1 kilograms’, as we
saw before, requires an argument of type P, whereas ‘lasts two hours’ requires an
argument of type EVENT (EVT); and P EVT =⊥. Accordingly, a complex type is
introduced: the dot type α • β. Such a type is an underspecified type containing two
incompatible types: “which is selected in a predication depends on the type
restrictions imposed by the predicate on its arguments”
b. John enjoyed the cigarette.
Typically, ‘enjoy’ collocates with nouns that denote events; accordingly, it requires
that its argument be of type EVT. Since ‘lunch’ is of type P • EVT, the predicate
‘enjoy’ combines with ‘lunch’ selecting the EVT aspect, and hence the predication in
(6a) succeeds: the type presupposition is satisfied. However, verbs like ‘enjoy’ “and
other verbs that take syntactically given arguments of type e but that presuppose
arguments of eventuality type sometimes force a predication over an eventuality that
is related in some way to the denotation of the syntactically given argument”
Why don’t we treat (6b) as an example of aspect selection? Because there are
compelling reasons for not considering ‘cigarette’ a dual aspect noun of type P • EVT.
Firstly, because it is unintuitive and misleading to consider cigarette as having a
second conceptualization in terms of events. Secondly, there are some linguistic
evidences
b. ?? The cigarette starts at 12:30.
If (7a) is felicitous on its own, (7b) needs an appropriate background of information in
order to express an eventuality reading of ‘cigarette’. Cigarettes are different in nature
from lunches. And the intuitions of native speakers seem to confirm that: if ‘cigarette’
“always has associated with it a possible event reading [i]t should then be possible to
access that appropriate event reading with other predicates that take events”
Dependent types come to the aid exactly at this point. When ‘enjoy’ takes an argument of type P a type conflict arises, and an adjustment in predication occurs. (5) c. The book is interesting.
d. The book weighs five pounds.
In (5c) ‘book’ expresses its informational aspect, whereas in (5d) its physical aspect.
There exist other ways in which a noun can be a dual aspect noun, like KIND • INDIVIDUAL
and KIND • MASS (see Asher (2011, Chap. 5)).
Therefore, verbs like ‘enjoy’ licenses a dependent type that is invoked when the direct
object is not of type EVT. Such a type ensures the shift in the predication from the
predication of a property to an object to the predication of a property to an eventuality
related to the object justifying the type presupposition. This is know in the literature
as coercion: “A function from one semantic value or one type to another that is
employed when some problem arises in the construction of meaning”
The possibility of such a recovery is determined by the existence of an ontological justification at the level of common sense. Basic and intuitive ontological considerations constitute the ground for aspect selection and coercion. Namely, cigarette are not events, but they typically involve the event of smoking them; and this is something which is deeply embedded in our common sense ontology we share as speakers of a language. Lunches are (also) events instead.6 However, one could say, the exact eventuality of (6b) might be sometimes underspecified. For instance, John might have enjoyed smoking, watching or smelling the cigarette. Nevertheless, the ontological information associated with the noun usually provide a specification. The fact that cigarettes get smoked – which is the main event involving them – is shared among speakers as an essential fact about cigarettes. If someone does not share such a knowledge, it would be hard to admit he really knows what a cigarette is. Consider the following sentences for other clues (from Asher (2011, p. 215)): (8) a. # Smith has begun the kitchen.
b. The janitor has begun (with) the kitchen. c. The cleaners have started the suits. d. The exterminator has begun (with) the bedroom.
e. The painters have finished the windows.
“Which eventualities end up being the internal arguments of the verbs begin, start,
and finish in [(8a-e)] is not just a function of the direct objects or the coercing
arguments themselves. The subject argument also plays an important role in
determining the eventuality internal arguments”
We have said that predications fail to result in a truth-evaluable sentence if a type
clash with no coercing recovery occurs during the β-reduction. Importantly, we cannot
say whether a given sentence is true or false if a type presupposition of one of its parts
is not satisfied or justified (like in the case of coercion). On the other hand, we have
also seen that sentences express truth conditions on the basis of a well-defined type
checking. So, a truth-conditional model theoretic theory needs types – or something
similar – to fix truth conditions and construct linguistic meaning if shifts in
predication and semantic anomalies are to be accounted for. Having said that, what
kind of truth conditions does such a theory express? Before answering this question
the notion of type has to be scrutinized a bit further. Once we enrich the type system
with subtypes of E ⇒ T (like the type of physical property P ⇒ T), we are forced to
abandon the classical set theoretic interpretation of types due to Montague’s influence
Since types are hyper-intensional entities, they have a proof theoretic semantics and not a model theoretic semantics; namely their content is expressed in terms of introduction rules and inference rules: the subtyping relationship (), for instance, which is hyper-intensionally understood, is explainable through an intuitionistic notion of deduction, ⊢∆ . And:
α β / α ⊢∆ β
7 “We can even distinguish between different conceptualizations of the same physical object or
of the same property, so in some respects the structure of types has the capacity to make finer
distinctions in meaning than intensional semantics can”
In this way our common sense ontology enters the realm of truth-conditional model theoretic semantics and allows us to calculate the shifts in meaning we have seen in dual aspect nouns and coercing predicates. (See Asher (2011, Chap. 4) for a technical demonstration).11 The shifts in meaning of the sort discussed above are difficulty treatable in a pure model theoretic semantics. (Among other things, a serious weakness is the impossibility of distinguishing logical from accidental polysemy.) By contrast, logical forms are in the present framework more informative, enriched and contain fine-grained shifts in meaning due to the introduction of our common sense ontology via type presuppositions.
As
Apart from these clues, there are interesting linguistic tests supporting the idea that certain information belong to the language itself. Coercing inferences are tied to particular predicates and they are not always available: (9) a. The painter enjoyed his brush.
b. ?? The brush has just started.
The eventuality reading of ‘brush’ is available in (9a), and not in (9b). This is a good reason for believing that the information is tied to the verb ‘enjoy’: it a matter of conventional, linguistic meaning. If in (9b) the word ‘brush’ has not an eventuality reading is because the linguistic environment, which is driven by the type presuppositions, does not allow for it. The same holds for the sentences we started with: (1) a. John enjoyed the cigarette.
b. The bottle froze.
c. John heard the piano.
In (1b) the predicate ‘froze’ coerces the meaning of the sentence into the liquid inside the bottle froze. That is because ‘freeze’ requires an object of type liquid, and since bottles are ontologically associated to their liquid contents, we can derive an enriched interpretation of (1b) which belongs to the language system itself – if we agree on the fact that language relies on ontological features of reality speakers tend to presuppose in conversation. Also in (1c) the literal meaning turns out to be enriched. Given the ontological nature of the event of hearing and the object piano, we derive the intuitive 12 Until recently, there has been little interest in such a view. The research has tended to focus on the pragmatic explanation. meaning according to which John heard the sound emitted by the piano. Only sounds can be literally heard, and since ‘piano’ is something essentially emits sounds, the selectional restrictions of the language operate to yield the intuitive meaning autonomously. Other readings are derivative and belong to side of speaker’s meaning.
Let me now go back to the above question: what kind of truth conditions does such a semantic theory express? Arguably, they are richer and more fine-grained. They are intuitive, like Recanati justly requires them to do, in the sense that they match our intuitions. In such an approach, however, logical polysemy and coercion is formally treated and subtle meaning shifts are accounted for in a crisp and profitable way. Semantics turns out to be beefed up. The common sense ontology behind the language increases the explanatory capacity of model theoretic semantics.
In this fourth section I turn to consider a metatheoretical question: where does the perspective here presented stand with respect to the Minimalism/Contestualism debate?
Minimalism and Contextualism are two big opposing research programs with
several different approaches inside them. However, for clarity’s sake, it is possible to
single out the pivotal points of disagreement. The dispute between Minimalism and
Contextualism centers on the scope of a theory of linguistic meaning and on the
notion of truth conditions
Minimalists maintain that every sentence says something which is truth-conditional (after linguistically triggered saturations), whereas contextualists maintain that no sentence says anything which is truth-conditional, only the speaker does. Accordingly, two different notions of truth conditions have to be disentangled. Minimalism accounts for liberal truth conditions, which are separated from speakers’ intuitions; whereas Contextualism accounts for intuitive truth conditions, which, on the contrary, express speakers’ intuitions.
Accordingly, besides the two notions of truth conditions, two different notions of what is said emerge from such a picture. A contextualist, typically, means by what is said something like what has intuitive truth-conditional content (WISint). And such a truth-conditional content is deemed to be affected by free pragmatic processes which are not linguistically triggered. On the other hand, a minimalist refers to another notion of what is said, namely what is literally said without intuitive contents (WISmin with liberal truth conditions).
For Borg and Cappelen and Lepore there is a sharp distinction between semantics and pragmatics, and a semantic theory is deemed to account for liberal truth conditions: “the constraint we want are pretty minimal” (Borg 2005, p. 3). Semantics is disentangled from both metaphysical constraints and communicative skills. On the contrary, for contextualists like Recanati and Jaszczolt our theory has do to account for intuitive truth-conditions of utterances. To that effect, they add, we may have to abandon the notion of literal meaning and weaken the semantics/pragmatics distinction.
We have here a methodological alternative: if a theory of meaning accounts for liberal truth conditions (WISmin), then it allows for a theory-internal distinction between semantic, literal meaning and pragmatic, speaker’s meaning. On the contrary, if a theory of meaning accounts for intuitive truth conditions (WIS int) such a distinction is doomed to fail, since every intuitive, truth-conditional content is affected by pragmatic processes which are not linguistically triggered. I argue that this is not the case and that the alternatives presented by Recanati are not so as binding as they seem. Indeed, I maintain that the distinction between semantics and pragmatics is still tenable in a theory of meaning that addresses intuitive truth conditions, provided some constraints given by the present proposal.
The introduction of the ontology-based approach inside the model theoretic conception of meaning allows us to construct enriched logical forms which include intuitive contents derived by the conceptual structure embedded in language. Words which express type presuppositions sometimes force interpretations and language itself provides to give enriched meanings. Therefore, this does not prevent us from considering such a theory overtly a theory of literal, encoded, linguistic meaning of sentences. As far as this fact is concerned, I agree with minimalists, since a theory of linguistic meaning separated from a theory of speaker’s meaning is achievable in the way proposed above in section 2. At the same time a theory of linguistic meaning, expressing enriched meanings, de facto expresses intuitive truth conditions, as spelled out by WISint: coercionsm and aspect selections can be considered to be part of semantics.13 13 Such a theory works insofar as words are linked to conceptual material speakers share as essential ontological features of reality. Consider the following sentences: (10) a. John is ready.
b. The shooter is ready. As we have seen, different ways of conceptualizing the same object are deeply interwoven with semantics and predicational processes. What derives from this perspective is thus a wide idea of semantics: a semantics that spells out intuitive, rich contents and encodes conceptual patterns. However, we may keep a useful distinction between semantics and pragmatics. It looks as though we have found an Enriched Literalism between Minimalism and Contextualism.
The considerations in the this paper have led me to some conclusions. As we saw discussing an extended version of the typed lambda calculus for predication, the mechanism of compositionality has to be enriched and transformed in a sophisticated device capable of spelling out subtle shifts in meaning formally. I contend that the semantic approach to aspect selection and coercion fares better then the pragmatic one. Cheafly, I clam that the pragmatic approach fails to give the right importance to crucial intuitions of common speakers. The sole relation of words within sentences could be the sole cause of an intuitive truth-conditional shift in meaning produced in an automatic and systematic fashion. We do not need to trot out complex pragmatical explanations here.
Crucially, this implies the reformulation of the classical principle of
compositionality. A widespread belief within the contemporary philosophy of
language is that the linguistic meaning of a complex expression is function of the
meanings of its parts and depends on the way they are structured. Therefore, the
meanings of the parts, whatever they are, do not change depending on the elements
they combine with
Coercions and aspect selection are examples of lateral influences that seem to compel us to give away the classical principle of compositionality. The account of subtypes, dot types and dependent types in structuring the meaning of our words permits to organize a formal treatment of subtle semantic contents that are strongly tied to sentences within a sophisticated principle of compositionality.
If in (10a) words do not suffice to determine intuitive truth conditions – we do not know what John is ready for – in (10b) things change. Due to the ontological nature of shooters, who essentially shoot, language forces the interpretation over such an event, since ‘is ready’ requires an object which is ready for doing something. Of the name ‘John’ we do not share enough information for determining what John might be ready for. On this view, semantics comes maximalized. To sum up. Words exhibit pre-conditions which produce presuppositions. Type presuppositions are ontological features words deploy in order to collocate in wellformed predications. The meaning which emerges from a predication is affected through restrictions and forced modifications by the shared cognitive material which reflects our common sense ontology. The main consequences of this are the supplementation of the classical Minimalism/Contextualism debate with Metaphysical Semanticism and the enrichment of the classical principle of compositionality with lateral influences of type presuppositions. A new scenario with new equipment seems possible. It takes the best from both Minimalism and Contextualims, and represents a valid alternative for the study of meaning in composition.