Dissertation: Neale on the Referential/Attributive

In his book Descriptions, Stephen Neale argues that there is good reason to suspect that the pragmatic conception of the referential/attributive distinction is the right one, since positing a conversational implicature of a singular proposition while simultateously holding that the semantics of all definite descriptions are strictly Russellian allows us to make sense of a tension that we feel when evaluating mistakenly used descriptions. For example, imagine that Smith, who believes that Jones has killed Brown, utters (1) in a conversation about Brown's murder with a conversational partner who shares his beliefs.

(1) Brown's murderer is insane.

Now, if it turns out that Jones did not kill Brown, but he is in fact insane, we seem to end up with conflicting intuitions. We feel as though Smith has done something right - that his utterance is true in some sense, but we also feel as if he has done something wrong, and that his utterance is false in some sense. Neale claims that this tension is easily explained if the semantics of (1) are Russellian - that is, attributive - while what is pragmatically communicated by (1) is in fact the singular proposition containing Jones as a constituent. The tension occurs because both the semantics and the implicature are communicated, and since they have different truth values, we're not sure which one to pick.

Unfortunately, it's not clear to me that the referentialist, who holds that there is no implicature, but rather that the semantics of (1) are referential and express a singular proposition all on their own, cannot just as easily account for this tension. It's a matter of context. As the context changes - that is, as we describe the case - it becomes less and less clear what the proper referent of the description is. As we are constantly applying the principle of charity to speakers, it is not hard to see how we would want to favor a correct use of the description (one that attributively picks out the real murderer of Brown). But, as it turns out, the priciple of charity leads us in two different directions in such cases. For while the pricinple of charity pushes us to give the speaker the benefit of the doubt with regard to the correct use of the description in his utterance, it also tells us to interpret the speaker's utterances such that they come out true. To paraphrase Davidson, we must treat speakers as believers of the true, lovers of the good and admirers of the beautiful.

What relevance does this tug of war between two aspects of the priciple of charity have? Well, as we move through the context and more information is made salient, I would suggest that we get stuck at the level of semantic processing. That is, we are unsure which explicature is that which the speaker intended in making his utterance. The tension, then, is not a tension between two different propositions which have both been communicated, but is rather a difficulty in determining which proposition has been communicated in the first place. We know that the speaker has done something right and something wrong either way we go, and being bound as we are by the principle of charity, we don't know which proposition to accept as the complete and correct semantic evaluation of Jones' utterance.

Now, of course, it will be objected that the hearer in such a conversation is typically well aware of which proposition has been expressed by the speaker. This is certainly true, and evidence for this claim lies in the fact that if the hearer of (1) was aware that Jones did not kill Brown, his response would not be to say "That's false, he's not insane", it would rather be to correct the misuse of the description (i.e., by uttering something like, "Well, he is a loon, but actually he didn't kill Brown). But I am not holding that the hearer finds indeterminacy in the semantic processing of the utterance, but rather that we, the evaluators, cannot determinately process the utterance. The reason is that for the evaluator, information is being added to the context of evaluation in a way that it is not added to the conversational context in which the speaker and hearer are situated. The hearer and speaker enter the conversation with a set of background assumptions and the context slowly changes as conversation moves forward, with each conversational participant marking their mental conversational scorecard as things go forward. The conversational participants are situated within their subjective conversational context. The evaluator, however, is set outside this context, looking in as a third party observer, and slowly being fed information about the context as a whole. He is not adding this information to a conversational scorecard the way the conversational participants do. Rather, he adds each additional piece of information to his understanding of the conversational context in which the utterance occurred without viewing that context as having changed. As third party information is added to the evaluator's knowledge of the conversational context, he reassess his assessment of the semantics of any given utterance in a way that conversational participants cannot (since their view of the context is subjective).

A similar phenomenon appears to occur in the field of contextualist epistemology. The evaluator's context bleeds into the subject's context, causing the evaluator to improperly assess knowledge claims made in the target context. When evaluating what is going on in a particular conversational or epistemic context, we must be careful to remember that we evaluate from a context, one that can easily impact our evaluation.

So what is the upshot of all this? Well, if the tension that Neale describes can be explained as indeterminacy in semantic processing, rather than competing propositions both of which have been communicated, then this tension cannot serve as evidence for the view that there is a conversational implicature associated with referentially used definite descriptions.