Asserting Grammatically Subordinated Information

In uttering the sentence

(8)	An innocent man was hanged today.

it is quite likely that the speaker means to convey primarily not the fact of the hanging, which is probably mutually known, but rather the innocence of the victim. That is, the new information is not, as in the canonical case, the predication associated with the main verb, but a predication associated with a grammatically subordinated element, a prenominal adjective. That is the primary assertional claim of the sentence.

The logical form of the sentence, without coercion, will contain the predications

$hang'(e_{1},x,m) \& innocent'(e_{2},m)$

That is, e₁ is a hanging of the man m by someone x, and e₂is m's innocence. The entire sentence would normally be described by the Syn predication

Syn(``An innocent man was hanged today.'',e₁,v,-,-,-,-)

where the sentence is taken to be a description of the hanging event e₁.

In Section 2, it was stated that in the IA framework, the job of an agent is to interpret the environment by proving abductively, or explaining, the observables in the environment, thereby establishing that the agent is in a coherent situation. When the observable is an utterance by a speaker i to a hearer u of a string of words w, the most plausible explanation is that w is a grammatical, interpretable sentence (or an otherwise coherent text) describing an eventuality that the speaker wants the hearer to believe or to adopt some other cognitive stance toward. For the purposes of this chapter we will take the goal to be the hearer's belief. Thus, the linkage between syntax and compositional semantics, represented with Synpredications, and pragmatics, involving the predicates goal and believe, is effected by axioms of the following flavor:

Syn(w,e,v $,-,-,-,-) \& goal(i,e_{0}) \& believe'(e_{0},u,e) \, \supset \,utter(i,u,w)$

That is, if w is a grammatical, interpretable sentence describing the eventuality e and a speaker i has the goal e₀ that a hearer u believe e to obtain, then (defeasibly) i will utter to u the string of words w. This axiom is used to explain the occurrence of the utterance.

In the case of sentence (8) the pragmatic part of the interpretation would seem to be

$goal(i,e_{0}) \& believe'(e_{0},u,e_{1})$

involving a belief in the hanging event, whereas what is wanted is

$goal(i,e_{0}) \& believe'(e_{0},u,e_{2})$

involving a belief in the innocence.

This again can be seen as an instance of metonymy where the explicit content of the sentence is used as the coercion relation. The desired top-level Syn predication is

Syn(``An innocent man was hanged today.'',e₂,v,-,-,-,-)

indicating that the innocence is what the sentence asserts. The metonymy axiom (1) decomposes this into

Syn(``An innocent man was hanged today.'',e₁,v,-,-,-,-)
$\& rel(e_{2},e_{1})$

The first conjunct is proved as it is normally, yielding the parse tree and the logical form of the sentence. The transitivity of reldecomposes the second conjunct into

$rel(e_{2},m) \& rel(m,e_{1})$

The first conjunct is established using innocent'(e₂,m), and the second conjunct is established using hang'(e₁,x,m) and the symmetricity of rel.

In a sense, we have coerced the sentence ``An innocent man was hanged today'' into the sentence ``The man who was hanged today was innocent.'''

Similarly, in

(9)	I have a sore throat.

it is not the possession of a throat that is being asserted, but the soreness of the throat the hearer already knows the speaker has. This can be viewed as an instance of metonymy as well. The explicit assertion of the sentence, the possession, is coerced into the soreness of what is possessed. The possession is related to the throat and the throat is related to the soreness, both by properties that are explicit in the logical form of the sentence and are thus emminently accessible.

This example requires that both possession and soreness be possible coercion relations, that is, instances of axiom schema (4):

$have'(e_{1},z,x) \, \supset \,rel(e_{1},x)$
$sore'(e_{2},x) \, \supset \,rel(e_{2},x)$

These coercion relations compose through the transitivity and symmetricity of rel.

Figure 10 shows an abbreviated version of the proof graph for the interpretation of example (9). The having e₁ is taken as the eventuality conveyed by the verb phrase, but that is coerced into the soreness e₂, using as a coercion relation a composite of the having and the soreness.

The sentence ``I have a sore throat'' is coerced into the sentence ``My throat is sore.''

These two examples used Metonymy Axiom (1). The eventuality is coerced, rather than one of its arguments.

  

Figure 10: Interpretation of ``I have a sore throat''.

I have not said what constraint forces this coercion, but it could be the constraint that what is said should be informative, an instance of the more general principle that one does not usually have the goal to achieve a state that already holds.

A similar story can be told about examples in which high stress changes what is asserted or alters our interpretation of predicate-argument structure. For example, in

John introduced Bill to MARY.

the assertional claim of the sentence is ``It was Mary that John introduced Bill to.'' In

John didn't introduce Bill to MARY.
John only introduced Bill to SUE.

the high stress forces a coercion of the arguments of not and onlyfrom the e such that introduce'(e,j,b,m) to the e₀ such that Mary'(e₀,m). It was not Mary that John introduced Bill to. It was only Sue that John introduced Bill to.

High stress indicates new information, and the new information in a sentence is generally what the speaker wants the hearer to believe. The coercion is one way to bring the intonation and the rest of the interpretation into correspondence with one another. Similarly, in example (8) the word ``innocent'' is likely to be given high stress.

In Hobbs (1995) it is shown how a similar move is a key part of an account of how the correct interpretation of monotone-decreasing quantifiers can be extracted from a flat logical form. Essentially, the sentence ``Few men work'' is reinterpreted as ``The men who work are few.''

<1212>>