Referring expressions and accessibility

Prince (1981) and Ariel (1988, 1990) propose that different types of referring expressions are used by producers according to their assumptions about how accessible the referent is in the mind of the target. Ariel (1988) proposes an Accessibility Marking Scale, along which referring expressions are ranked according to degrees of accessibility. Thus, at one end producers use zeros (gaps), pronouns and demonstrative pronouns for highly accessible referents, while at the other end they use long definite expressions and full names to refer to less accessible referents. For example, in saying have you got it with you? the producer assumes that the referent of it is highly accessible to the target, one does not have to spell it out. In contrast, in saying have you got your classic car with you?, the producer assumes that the referent is less accessible to the target and so the fuller your classic car is required. The accessibility of referents is said to be determined by four factors: the saliency of the referent, the recency of a previous mention, the number of other referents competing for the role of antecedent, and the degree of cohesion between the units that the expression and its antecedent occur in.

Gundel et al.’s (1993) Givenness Hierarchy ranks referring expressions along a scale according to the assumed cognitive status of the associated referents for the target. Cognitive status refers to the assumed attention state of the referent (e.g. something in focus right now, something in memory but not currently active, something completely unknown). However, unlike the Accessibility Marking Scale, the thrust of the Givenness Hierarchy is that particular referring expressions provide information to the target addressee about where in memory (or not in memory at all) the representation of the referent is expected to be found. For example, using the pronoun it (indeed, any unstressed third person pronoun) to refer to a classic car signals to the target that they do not need to search widely for the referent classic car, but that it is in focus, as in currently activated in short-term memory and the centre of attention (e.g. it is the immediate topic of conversation). In contrast, a classic car signals merely that the target knows the category “classic car” described by the expression – they can identify the type. Somewhere in between would be, for example, the expression that classic car, which signals that the target is familiar with the referent: they not only know the category but are able to identify the specific intended referent because they have a representation of it in memory.

Each type of referring expression is conventionally associated with a particular cognitive status; any use of that expression signals a particular cognitive status which the target can then take account of in their search for an interpretation. The cognitive statuses discussed in the work of Gundel and colleagues are as follows:

Figure 2.1 The cognitive statuses of referring expressions (solid black arrows can be taken to mean “and thus also”, i.e. entailing)

A key point to note about these cognitive statuses is that they are not mutually exclusive. The statuses are related such that a particular cognitive status entails lower statuses. For example, an item that is in focus is necessarily activated, familiar, uniquely identifiable, and so on. The importance of this scale, Gundel and colleagues argue, is that it is a basis for implicational meanings, via Grice’s (1975) Maxim of Quantity (to be discussed later), whereby using a form with a weaker status implicates that a stronger status does not obtain. Thus, saying a classic car implicates that the speaker believes that the addressee cannot uniquely identify the referent; whereas saying that classic car implicates, for instance, that the speaker believes that the referent is not in focus (i.e. activated in short-term memory and the centre of current attention).

Let us briefly illustrate some of these issues with the following joke:

At the heart of the joke are the differing understandings arising from the demonstrative this in the second utterance of the receptionist. Most likely for the receptionist (and indeed the reader) the referent of this has activated status, as the baby was mentioned in the immediately previous sentence, though it was not the topic (hence it was not in focus). In the receptionist’s second utterance, her first baby is the topic, and thus it is brought into focus. According to the Givenness Hierarchy, we might have expected a husband to reply “no, it’s not her first baby” or “yes, it’s her first baby”. However, the husband constructs a radically different understanding, namely, that this refers to the identity behind his voice. To sustain this, we must assume that he thinks that the receptionist thinks that this refers to the voice on the phone and that the voice is that of her first baby (i.e. in some bizarre world, the fi rst baby is telephoning to ask for an ambulance!). In this scenario, this merely has the status of being referential, that is, referring to a particular voice, but not one that is even uniquely identifiable. This, of course, is completely improbable, because their respective identities have in fact been established (the social deixis of the wife implies that he is the husband): they at least have the cognitive status of activated.

The Givenness Hierarchy usefully throws light on how the choice of referring expression works. It is not, however, a complete account of referring expressions. It is based on the notion of givenness, understood as assumptions about accessibility for the target (we elaborate more on givenness). However, we saw pronouns used in contexts where the referents’ accessibility was not the only issue: it was not simply a matter of where to access something in one’s mind. Epstein (1999), whose work is focused on the definite article the, was very aware of this, and hence defined accessibility simply as being “available for interpretation” (1999: 67). He concludes from his analyses of real data that “the article seems to be an instruction which prompts the addressee into making the appropriate conceptual constructions for interpreting the nominal in discourse” (ibid.). In other words, such prompts go beyond helping identify where in the mind to access a referent. A prompt can introduce new referents, and do so in such a way that facilitates interpretation of the discourse by, for example, creating psychological prominence, expressing a viewpoint or triggering frames.