3. Definites and indefinites

When is a or the appropriate? One can see the contrast in (14).

The example in (14) illustrates the role played by uniqueness, i.e., the existence of one and only one entity meeting the descriptive content of the noun phrase. Definite NPs are used appropriately when they refer to a referent who is unique in the context. Indefinite NPs refer to a referent that is non-unique.

But what do the determiners a/the mean? In effect, they signal the speaker's assessment of the uniqueness of the referent. In (15) below, the speaker explicitly asserts that they met someone who teaches on 2B; the use of an signals additional meaning -- that there is more than one instructor for 2B. This extra meaning is not part of the asserted content of what the speaker said (for that, they'd need to actively assert "There is more than one instructor for 2B"). Similarly for (16), the speaker explicitly asserts that they met someone who teaches on 2B, but they also convey the additional meaning that there is one and only instructor for 2B (a fact that is at odds with the real world).

The fact that a sentence with a definite determiner carries extra meaning is an example of the way that sentence meaning goes beyond what is explicitly said. The sentences in (15-16) assert content about the speaker meeting someone but they presuppose the existence and uniqueness status of that person.

How do we know that the use of a referring expression yields presupposed meaning rather than asserted meaning? Well, if that meaning happens to be false, it's hard to deny, and this is one of the hallmarks of backgrounded pragmatic meaning. If a speaker says they met the instructor for 2B and another speaker expresses a denial, what are they specifically denying?

A speaker's denial denies the asserted meaning (that a meeting happened), not the presupposed meaning that the instructor is unique. This behavior should sound familiar from the discussion on the first page of this reading. The case there involved someone hearing the sentence "Stefano went to Paris" and responding "No, that's not true", in which case their denial seemed only to be a denial of the explicit literal meaning (that Stefano went to Paris) rather than a denial that there exists a person named Stefano or a place called Paris. In other words, the use (or misuse) of referring expressions is a pragmatic phenomenon in that it goes beyond what a speaker simply asserts.

For definites and indefinites, the property of uniqueness explains many of the usage patterns. It's a property that depends on one type of context -- the real world (is a referent the only one in the relevant context?). But there are cases like (18) where the referent of a definite NP need not actually be unique.

If (18) is felicitous (which most native speakers tend to agree that it is), maybe it's enough if there is a single most salient entity that satisfies the description. The dog who is referred to as "The dog" at the start of the sentence may be the salient one in the context, and that alone can be enough to license the use of the definite. This notion is captured in the distinction between familiarity/novelty. Separate from uniqueness, the familiarity constraint on definites ("the dog", "it") posits that such forms are permitted when the existence of the referred-to entity has been established in the particular discourse. The novelty constraint on indefinites ("a dog", "some dogs") posits that these forms presuppose that their referent is being introduced into the discourse for the first time. In that sense, which form a speaker uses depends on what they want to convey to the listener about the status of the entity (is it familiar or is it novel). Similar to uniqueness, these constraints depend on context, but in this case it is the preceding discourse context.

Familiarity definitely helps explain speakers' choices, but there are some contexts in which a speaker can use a definite NP even if the referent is unfamiliar. Consider the cases in (19).

Another oft-cited exception is the variation between British and American English since speakers vary in whether they produce "to be in hospital" versus "to be in the hospital". If these patterns are really about uniqueness/familiarity constraints, then dialect variation shouldn't be possible (since uniqueness and familiarity are not language-specific concepts).

All of these exceptions suggest that we need to think more about a usage-based account and to ask which contexts (with which kinds of speakers and which types of addressees) license which forms. Our course project will be about mapping usage across contexts in the MapTask corpus, which has the advantage of containing naturally occurring (unscripted) dialogue that was produced in a constrained environment which systematically varied certain properties of the context.

To go on to section 4 on the "MapTask corpus", click here.