2. Reference

To see the role of context in reference, we'll start by stepping through a set of examples. In each case, the acceptability of the underlined referring expression depends on a particular aspect of the context of its use.

In (5), the kind of context that is at play is the context of the sentence itself. The appropriateness of using herself/her/she depends on constraints that are dictated by the grammar (and on Hannah's preference in pronouns). Chomsky's rules for Binding Theory dictate that a pronoun in that syntactic position that is c-commanded by its antecedent must appear in the reflexive form (herself). The use of the bare accusative form her or the nominative form she is governed by syntactic rules, which are independent of other context-driven effects. The use of herself reflects context, but not in the pragmatic sense.

Here in (6), what matters is who is speaking. The referent of pronouns like I and you are of course highly context-dependent.

In (7), the use of it is formally ambiguous since there is more than one singular non-human referent in the discourse context (cup, table). In this case, the recovery of the speaker's intended referent depends on real-world knowledge, i.e., the knowledge that cups typically go in sinks (and fit in sinks), whereas tables do not. The use of general knowledge and the ability to reason about plausibility and relevance is part of pragmatics.

The example in (8) is a reminder that utterances can be grammatically well-formed but semantically meaningless without context. If I started a lecture or a web reading with the opening line She did it again! listeners would understandably be confused. But if two good friends meet up and one starts the conversation this way, it is entirely possible that the referents of She and it would be recoverable. The study of this reliance on shared common ground is part of pragmatics.

In example (9), the three underlined expressions are intended to refer to the same referent. The felicity of the changing forms reflects linguistic constraints, but unlike in (5), these constraints are not syntactic. Rather, the felicity of the indefinite "a burnt forest", the definite "the burnt forest", and the pronoun "it" depends on the status of the referents in the discourse model. The discourse model represents, among other things, what has been said and what referents have been introduced. The discourse model is part of the shared common ground between the two interlocutors, and Speaker B's use of the expression "the burnt forest" depends on there being a salient referent that is unique and/or familiar to the other speaker.

One could say that reference is the most fundamental aspect of pragmatics -- which is the kind of thing someone says if it's their own research topic... Regardless, it is the case that the narrowest interpretation of pragmatics is the study of indexicals. In order to assign meaning to a sentence, expressions like the following indexicals must be resolved:

• Indexical pronouns: I, you
• Adverbs: here, yesterday
• Tenses: will, had been
• Demonstratives/pronouns: this/that/him/her

Without context, a sentence containing an indexical cannot be understood as true or false. With context, a listener can establish the meaning.

A note on terminology -- what does "it" mean? When we talk about the meaning of a referring expression, we really end up talking about usage: When do speakers produce it and how do listeners interpret it? You can see the flexibility of "it" in (10) where it (it!) refers to an entity (10a-c), an entire situation (10d), or nothing at all (10e). Notice how you use real-world knowledge to resolve reference in each case.

In (10a-d), "it" is used to refer to something either concrete like an entity or something abstract like a situation. The last case in (10e) is called the non-referential use of "it" (as in "it's raining").

Even with an understanding of possible usage, we need to be clear on how we talk about reference. The convention in pragmatics is to use the verb refer to talk about the way a referring expression picks out a referent in the real world. We say:

Two NPs are said to co-refer if they pick out the same referent. For example, the NP "a dog" and the NP "the dog" refer to the same referent in (11) so they co-refer.

Note that, according to these conventions, it doesn't work to say that the NP "The dog" refers to the NP "a dog" because an NP refers to an entity (or any referent, like an event or abstract idea), not the linguistic expression describing that entity.

A question one might be asking at this point is 'Why care about reference?'. Just because your instructor Hannah seems to have deep affection for this topic, why should you as a linguistics student be interested? One reason is typology (i.e., research questions about what commonalities there are across languages and what systematic differences there are). Understanding differences across languages requires understanding not just what's possible in a given language (its grammar) but also what preferences speakers and listeners have (the distribution of usage). In the domain of reference resolution, even a casual survey of languages offers reminders about the different kinds of systems in play -- e.g., Italian and English differ in that Italian permits both null and overt pronouns. Given these differences, another research question that arises is whether second language learners draw parallels where possible between their first and second languages and what do those parallels look like (does an Italian speaker map the usage of their overt pronoun to the English overt pronoun or do they link their most reduced form, the null pronoun, to the English reduced form, the overt pronoun)?

In a more applied domain, there's a research question for computational linguists of how to get computers to produce and interpret natural speech. For example, in a Google news summary or an automatic translation, what generalisations does the computational system need to learn in order to produce (12a) and not (12b-d)?

Studying reference is about studying what forms of reference are appropriate and will allow communication to succeed. At stake is a fundamental issue in communication, namely ambiguity (or ambiguity avoidance). For writers (of textbooks, course essays, NHS forms, etc.), one of the main goals is to be clear and informative. Understanding how reference works and what expectations readers have can guide authors decisions about how to communicate most efficiently and effectively.

In a context where Hannah is speaking, she typically refers to herself as "I", but we also need to think about how context guides the use of other referring expressions.

For the LEL2B course project this year, we're going to be analysing the referential choices speakers make in the MapTask corpus. This is a well-studied corpus that was created in Edinburgh in the early 90s. It has been widely used to explore a range of linguistic questions but there remain many open research questions about the use of referential forms, which this corpus can help answer. The corpus consists of dialogues between pairs of speakers who engage in a cooperative task involving navigating a map. The tricky part is that the maps are not identical and so part of the dialogue is the negotiation of what is present on the map and then where the path should be drawn. Here's an example:

Observations about speakers' referential forms (as in (13)) have led to generalizations and hypotheses about key relationships between form and function.
 Experimental pragmatics aims to test those claims empirically in the lab or in large corpora. Our LEL2B project will be an opportunity to learn a bit about such theories and to see how empirical data can be used as a testing ground to see how these theories hold up.

To go on to section 3 on "Definites and indefinites", click here.