A problem with stipulation
3. An alternative model of indexical interpretation
4.1 Kaplan*s objection
In response to Kaplan Nunberg comments:
Kaplan discusses the case o f someone pointing to a flower and saying "He has been following me around all day," and observes in passing that a "background story can be provided that will make pointing at a flower a contextually appropriate, though deviant, way o f pointing at a man; for example, if we are talking about a great hybridizer." (p 30).
He complains that Kaplan has loaded the dice by his choice of what is, prima facie, a bizarre example. However, Nunberg counters this with the observation
that although it may seem odd to point to a flower while uttering 'he', 'there is no bizarreness in pointing at a painting' and uttering (8) (reproduced again below for convenience) with the intention o f referring to the artist.
(8) Now he knew how to paint goats.
Nunberg further - and correctly - observes that if we wish to exclude Kaplan's flower example from the analysis o f indexicals on the grounds that it is deviant, then we must also exclude a great many other, apparently acceptable, analogous examples, such as those already discussed above. One further example, much discussed in the literature, is that o f the enormous footprint in the sand at which the speaker points while uttering (25).
(25) He must be a giant.
(This example is taken from Schiffer, 1981, p 49.^"^ )
33
See Green (1989), p 24.
It does, indeed, seem to be the case that pointing to a footprint to refer to the owner of the foot or to a picture to refer to the artist - or possibly even to the subject - of the painting is far from deviant, but this does not, o f itself, show that these are examples of deferred reference. Once again, some further investigation of how such reference actually works seems called for.
With respect to (8), according to the account developed above, we get the following schema:
(i) First, the deictic component must be resolved in order to determine the index. In the case of nonparticipant pronouns such as Tie' the deictic component may be identified with the accompanying demonstration. Thus, in this example, the deictic component is the gestural ostension and the index determined by this gesture - the demonstratum - is the picture.
(ii) The classificatory component must next be resolved. In this case it comprises the semantic features ANIMATE, MALE, SINGULAR. It is associated with the interpretation or referent, and does not relate to the index. This is why it makes sense to point to an inanimate, gender-neutral, object such as a painting while uttering Tie'.
(iii) Finally, as 'he' is a nonparticipant term not subject to the constraints of NRC, we must resolve GRC’ to determine the interpretation . GRC’ stipulates that the index o f any utterance of 'he' must stand in some contextually salient relation to a singular, male, entity, and, that that entity - if it exists and is determinable - will be the referent o f the utterance o f 'he'. Alternatively, if no such entity exists or can be determined, then the relevant description that would be satisfied by such an entity, if it exists, will be the interpretation of that utterance.
This is still broadly in accordance with Nunberg’s own suggestion, for with respect to the operation of deferred reference in the interpretation of nonparticipant terms he writes:
... used indexically [nonparticipant terms] can contribute any individual that corresponds to their indices in some salient way. (p 25).
The notion o f salience remains regrettably vague, and once again I fall back on the assumption that salience, or possibly relevance, in a context may be defined in terms o f pragmatic, or even cognitive, principles. However, as has already been noted, a discussion of what those principles might be, or of mental models of the utterance situation, is well beyond the scope o f this investigation.
Now Kaplan's complaint is that in obvious cases o f deferred reference, such as the flower example, the semantic features o f the classificatory component do not describe the demonstratum. He sees such a mismatch o f features as deviant. However, as we have seen, indexical utterances in which the semantic features of the classificatory component do not describe the entity ostended are extremely common and are readily accommodated within Nunberg's model o f deferred reference. Indeed, they may even constitute its motivating force. It may, therefore, be concluded that Kaplan's objection poses no very serious threat.
4.1.1 Cross-sortal mismatches
Interestingly, Nunberg sees in these mismatches o f features evidence o f a distinction between participant and nonpartcipant terms which he believes further supports his postulation of NRC:
Inasmuch as the indices o f nonparticipant terms need not instantiate their interpretations, the two objects [i.e., index and referent] will often differ with respect to the properties associated with the classificatory component o f the expressions,... (p 25).
and :
The absence of [an] explicit relational component in nonparticipant terms makes it possible to exploit ... these extravagantly cross-sortal referents and indices,... (p 27).
To put it another way, Nunberg appears to assume that it is the presence of an 'explicit relational component' (NRC) that prevents such mismatches in the interpretation o f participant terms. However, it seems to me that there is another more obvious and simpler explanation o f this apparently invariant concordance.
A nonparticipant term such as 'he' has no linguistically encoded deictic component whatever. The sortal features of the index o f an utterance of 'he' are, therefore, totally unspecified. Even in the case of the nonparticipant terms 'this' and 'that' where proximality and distality are encoded in the deictic component, these properties are not sortal features. However, all indexical terms - both participant and nonparticipant - have an encoded classificatory component which does specify sortal features, and these always relate to the referent or interpretation, whether or not they truly describe the index. Therefore, in the case o f nonparticipant terms where there is no linguistic constraint concerning the sortal features of the index, no semantic or logical contradiction or linguistic infelicity is entailed if the features which that index turns out to have are non- cotenable with the linguistically determined features of the interpretation.
Conversely, the deictic component o f a participant term, by definition, encodes sufficient information for successful determination of that index to occur. It is this that obviates the need for an accompanying demonstration. Moreover, it still remains the case that the classificatory component stipulates the sortal features that define, or help to determine^ the referent. It would, therefore - at the very least - be linguistically infelicitous for such a term to stipulate through the agency of encoded features that the index, say, shall be feminine (female) while the referent must be masculine (male). How, for example - without apparent
contradiction - might the features MALE and FEMALE or ANIMATE and INANIMATE be conjoined in the same term?
I suggest that this need for semantic coherence might prevent cross-sortal relations from holding between index and referent - if indeed such relations really are prevented - in the case o f participant terms. There is no need to posit a specific relational component to explain this postulated effect.
However, it is not clear to me that such mismatches do not occur in the case o f participant terms also. Consider example (18) again.
(18) In a couple o f years we'll probably all be women.
The index in this example is singular and male. The interpretation is plural and female. Yet 'we' is a participant term.
In fact 'we' presents further problems that might seem to contradict my claim that a term cannot felicitously encode mutually exclusive features. The index o f exclusive 'we' is always the speaker, and hence - presumably - singular, whereas the classificatory component encodes the feature PLURAL. Furthermore, 'we' is a participant term and the information that identifies the singular index is also encoded. Does not this indication of both singularity and plurality within a single term amount to semantic infelicity?
I don't think so, because it seems to me that singularity is not encoded in the deictic component of 'we'. That is to say, although it very frequently turns out that the index of'w e' is just the speaker, and hence singular, it is not semantically required that this should be so. In cases o f joint authorship and choral recitations or singing, the index of 'we' turns out to be plural. Therefore, I suggest that the deictic component of 'we' is silent as to number, and that the singularity or plurality of the index is pragmatically, not semantically, determined. If this is correct, no logical contradiction or semantic infelicity will arise in connection
with those cases where the index o f 'we’ is singular while the interpretation expresses plurality. Nunberg is, therefore, mistaken to claim that such mismatches only occur in the case o f nonparticipant terms, and the erroneously postulated asymmetry cannot be used as an argument in favour of NRC.