There will always be more than one way to skin a category.
4.1 On Categories
There is a sense in which all of linguistics is a categorisation problem of one kind or another; ‘[c]ategorization applies at every level of form and meaning’ (Bybee, Reference Bybee, Hoffmann and Trousdale2013: 50). Categories are, however, problematic, and linguistics is no different from other disciplines in being confronted with categorisation problems; ‘[w]e find entities that cannot be properly classified, either because they would simultaneously belong to two or more traditional categories or because they are too underspecified to put them in one of the traditional categories’ (van Lier and Rijkhoff, Reference van Lier, Rijkhoff, Rijkhoff and van Lier2013: 6). We have to categorise our objects of study, and yet the process of categorising is inherently problematic. However, categorisation should be an explicit part of what we do as linguists. In the previous chapter, we suggested that perhaps some of our grammatical categories need to tolerate a degree of underspecification and greater fluidity. According to Halliday and Matthiessen (Reference Halliday and Matthiessen1999: 193), ‘[t]he grammar imposes a categorization that is compromising, fluid, indeterminate and constantly in process of change, along with changes in the human condition and in the interaction of humans with their environment’. As humans, we do like categories, and they are essential in our everyday lives and our development both individually and socially. Categories involve boundaries, and therein lies the challenge. There are many areas of the grammar which have boundaries, but these boundaries are, to some extent, artificial even though there are normally good reasons for where and how a boundary is established. As Croft and Cruise (Reference Croft and Cruise2004: 89) explain, ‘a boundary is arguably the most basic of all the properties of a category. A category is like a container: one of its major functions is to divide the objects in the world into those things that are in it and those things that are not in it. This function cannot be fulfilled without a boundary’.
The title of this chapter suggests two categories: typical and atypical. While it is true that we want to make a distinction between the two, we are not proposing two closed, mutually exclusive sets. Much of the literature on reference has established our core understanding of acts of referring, and this body of work has made a significant contribution to how we view referring. However, in many cases, the evidence has been experimental, computational, philosophical, and/or based on small excerpts from discourse. We explained in Chapter 1 that our main interest in this volume was to present an integrated perspective on referring with particular attention given to what we have called atypical uses of reference. This chapter, then, sets out the distinction we would like to make on the division between typical and atypical uses. In our original plans, we anticipated having a chapter dedicated to each topic, but it became quickly obvious that it would be almost impossible to talk about one without the other. In a sense, we can think of typical and atypical as contrastive in that one gets its meaning from not being the other, but we would not like to conceptualise the two terms as antonyms. We have a good sense of the general use of ‘typical’, which means ‘having the qualities of a type or specimen; serving as a representative specimen of a class or kind’ (‘typical’, adj2). If someone drinks coffee regularly, we might say that coffee is this person’s typical drink. If one day they refuse a cup of coffee, we might notice that this is not very typical of that person. And we might even wonder if it means they are not feeling well. We have already suggested in various places so far, especially in Chapter 2, that some referring expressions are not typical, or that they are doing something different (e.g. my Canada). In this chapter, we will set the stage for both our set of chapters on typical reference (Chapters 5 and 6) and for the third section of this volume, which includes four chapters dedicated to more atypical uses of reference.
In the next section, we review what is meant in the literature by the concepts of typicality and atypicality by considering related terms such as prototypicality, markedness, core versus periphery, and norms versus exploitations. Having established how we will be using the terms in this work, Section 4.3 presents a brief overview of two cognitive theories of reference: the Accessibility Scale (Ariel, Reference Ariel1990) and the Givenness Hierarchy (Gundel et al., Reference Gundel, Hedberg and Zacharski1993). These models are acknowledged in the field of reference as setting the standards for typical uses of reference, and they have established the importance of the need for a ‘common ground’ between interlocutors. Finally, Section 4.4 takes a first step towards discussing atypical instances of reference. It critically examines what it means for a referring expression to be atypical, and what functions such atypical expressions might have in the discourse.
4.2 What Is Typicality and Atypicality?
As will become clear throughout the following discussion, what seems to underpin all notions of typicality is the concept of frequency. Frequency, in turn, is intrinsically linked to usage. To better understand how typicality stems from frequency of usage, we will first briefly consider the role of usage. According to Bybee (Reference Bybee, Hoffmann and Trousdale2013: 50), ‘[i]t is repetition that leads to conventionalization of categories and associations, as well as to the automation of sequences’. The more a specific feature or pattern is repeated in use, the more conventional it becomes. Thus, by taking a usage-based approach to language, it is possible to make claims about ‘the extent to which linguistic patterns are common or rare’ (Biber, Reference Biber, Kemmer and Barlow2000: 287; emphasis added), that is, what is typical or atypical in language.
By looking at linguistic phenomena in use, we can identify association patterns, that is, ‘the systematic ways in which linguistic features are used in association with other linguistic and non-linguistic features’ (Biber, Reference Biber, Kemmer and Barlow2000: 289). With ‘linguistic features’, Biber is referring to both lexical and grammatical ones. In other words, an individual word can either frequently co-occur with other words (its collocations), or it can have specific structural preferences, such as ‘whether a particular adjective typically occurs with attributive or predicative functions, or whether a particular verb typically occurs with transitive or intransitive functions’ (ibid.: 290).
Furthermore, in addition to the variability of a linguistic feature in terms of its linguistic associations, research into its typicality also needs to consider its non-linguistic associations, that is, its distribution across registers, dialects, and time (Biber, Reference Biber, Kemmer and Barlow2000: 289). This flexibility of language is expressed as follows by Kemmer and Barlow (Reference Kemmer, Barlow, Kemmer and Barlow2000: ix), who discuss linguistic dynamicity and language users’ influence on each other’s linguistic systems:
[U]nits of language (from phonemes to constructions) are not fixed but dynamic, subject to creative extension and reshaping with use. […] Usage events are crucial to the ongoing structuring and operation of the linguistic system. Language productions are not only products of the speaker’s linguistic system, but they also provide input for other speakers’ systems (as well as, reflexively, for the speaker’s own), not just in initial acquisition but in language use throughout life. Thus, usage events play a double role in the system: they both result from, and also shape, the linguistic system itself in a kind of feedback loop.
Hence, everyday language input and usage shape the frequency patterns of individual language units and thus determine what language users consider a typical or an atypical use of language in a given context. Biber (Reference Biber, Kemmer and Barlow2000: 287–288) states that ‘a use perspective is required to investigate the stylistic preferences of individuals, the differing linguistic preferences of groups of speakers, and the ways in which “registers” (or “genres”) favour some words and structures over others’. It is thus evident that the notion of typicality is tightly interlinked with usage events and is dependent on the frequency of occurrence of individual linguistic units, which, in turn, is dependent on different registers of language.
In what follows, we first look at the concept of register as defined by Halliday in order to consider how its parameters interact with typicality. In the remaining sub-sections, we will discuss how (a)typicality relates to the concepts of prototypicality, markedness, and finally norms and exploitations. This section concludes with a summary of our position that both typicality and atypicality are best interpreted as usage-based and register-dependent properties of an instance of language use.
4.2.1 Register
Based on the discussion above, when we talk about typicality and atypicality, we are really talking about a kind of variation with a given context. Halliday (Reference Halliday and Fishman1968: 141) defines register variation as ‘variation according to use’, which he distinguishes from dialectal variation (variation according to user). Language in use is always in a particular context. Halliday refers to this as ‘context of situation’. Context is organised by the parameters of field, tenor, and mode of discourse, commonly referred to as the variables of register (see Lukin et al., Reference Lukin, Moore, Herke, Wegener and Wu2011). Field represents the social context, seen as ‘a field of significant social action’ (Halliday, Reference Halliday and Webster1977/2002: 55), in other words, what is going on. Tenor includes social relationships and roles, where text is viewed as ‘an intersubjective event’ (Halliday Reference Halliday and Webster1981/2002: 245). Finally, mode concerns the symbolic organisation of language, involving ‘the role assigned to the text, including both medium and rhetorical function’ (ibid.: 226). We will be discussing these parameters in more detail in Chapter 5 and Chapter 6 when we consider how they contribute to our understanding of referring in spontaneous and prepared discourse, respectively.
Different constellations of field, tenor, and mode in different situations interact with the lexicogrammar through a relationship of construal in the sense that context of situation is realised in and construed by language. This relationship is mediated by the semantic layer, which sits between context and lexicogrammar in Halliday’s model. As Taverniers (Reference Taverniers2011: 1122) explains, it is the theoretical concept of construal that accounts for ‘the relationship between language and extralinguistic reality’. This relationship is not unidirectional. As Lukin et al. (Reference Lukin, Moore, Herke, Wegener and Wu2011: 207) explain, ‘[r]egister is perhaps best understood as a dialectic – between system and instance – since the two are never actually possible without each other’.
According to Bowcher (Reference Bowcher, Thompson, Bowcher, Fontaine and Schönthal2019: 146), ‘[l]anguage varies according to the occasions on which it is being used, and speakers of a language typically have no difficulty recognizing this fact and managing their “repertoires” of language use accordingly’. What we can glean from this is that speakers are aware, to some extent, what the expectations are of a given register and they normally are aware of any deviation from those norms. This view of register is not deterministic, nor is it static or stable; in other words, it is not the context that determines language use. O’Donnell (Reference O’Donnell2021) argues that ‘at each point of a text/interaction, we as interactants have the choice to affirm the contextual expectation, or to vary from it, either using novel means to achieve some situational goal, or by shifting to a distinct Context of Situation’.
Due to this dynamicity of language and due to register variation, what can be considered as typical or atypical differs from one linguistic situation to another. ‘[M]ost linguistic phenomena are not distributed in a simple binary opposition of “frequent” versus “rare”’ (Biber, Reference Biber2012: 11). Frequency and typicality of linguistic patterns should not be described across language use in general (Biber, Reference Biber2012). Some might even say that language is best viewed as a collection of registers, questioning whether a description of a language as a whole is even possible. Instead, register-specific accounts are seen as more informative, or at least, we could take the view that consideration of register parameters will lead to a richer description. For example, in differentiating between the patterns of use in speech and those in writing, which mostly involve differences in mode, we know that pronoun use increases in frequency considerably in spoken language. We will discuss pronoun use in spontaneous discourse in more detail in Chapter 5, and in Chapter 6 we consider referential choice in written genres.
4.2.2 Typicality and Prototypicality
As discussed above, typicality is usage-based and register-dependent. In addition, it is important to note that any notion of typicality in linguistic research would have to be based on empirical evidence, for example, from investigating a large corpus. Not only are speaker intuitions unreliable (Hanks, Reference Hanks2013), but ‘a corpus offers information that a native speaker cannot replicate: an indication of “central and typical” usage’ (Hunston, Reference Hunston2002: 42). We do want to be confident when we say something is a typical instance or typical use and, increasingly, big data is used to provide large-scale evidence. However, this is not to say that there is no place for individual perceptions of typicality. Intuitive ideas of what is typical in language are instead covered by the notion of ‘prototypicality’. The term originates from Rosch’s (Reference Rosch1975) Prototype Theory, which showed that people have an idea of what counts as a typical member of a category and which members are more peripheral. For example, in some places, such as North America, a robin might be considered a more typical member of the category of birdFootnote 1 than a penguin (also see Aitchison, Reference Aitchison1994: 55). Such prototypes are based on an individual’s judgement of what is typical and the idea that some instances are more typical than others.
For some in corpus linguistics (cf. Hunston, Reference Hunston2002: 43), the term ‘prototypical’ is used to mean what language users believe to be the most typical but which does not necessarily (although it might) align with corpus evidence (see also Hanks, Reference Hanks2013). Thus, while ‘typicality’ is closely linked with frequency and is based on corpus evidence, ‘prototypicality’ is more associated with a sense of ‘belief’ and of what ‘ought to be’, based on an individual’s experience of the world.
4.2.3 Markedness
The terms ‘marked’ and ‘unmarked’ are often used to indicate that an instance or some feature stands out in some way. The term ‘marked’ is problematic for a number of reasons, the main one being that its use is not consistent, which has resulted in the term being highly polysemous. After examining twelve distinct senses of markedness across a variety of sub-fields of linguistics, Haspelmath (Reference Haspelmath2006) concluded that the term is misleading and unhelpful. He explains that ‘the “intuitive” shared sense of “marked/unmarked” is not distinguishable from the sense of everyday words like uncommon/common, abnormal/normal, unusual/usual, unexpected/expected’ (ibid.: 63).
Marked and unmarked are terms used by Halliday quite specifically to sub-categorise the theme of a clause. For Halliday (Reference Halliday1994: 33), an unmarked theme is one where ‘there is no prior context leading up to it and no positive reason for choosing anything else’. In other words, under normal circumstances, unmarked theme is the default pattern of the clause. A marked theme, according to Halliday and Matthiessen (Reference Halliday and Matthiessen2014: 97), adds ‘a feature of contrastiveness’. Marked in this sense means that the themes are ‘unusual enough to draw attention to themselves, and […] they only occur when contextual reasons overrule the unmarked choice’ (Thompson, Reference Thompson2004: 145). The distinction is illustrated in examples (78) and (79), taken from Halliday and Matthiessen (2014: 92). The assumption is that theme will be expressed, typically, by the subject of the clause, unless some other element occurs first (at least in declarative clauses).
in this job, Anne, we’re working with silver [marked theme]
the people that buy silver love it [unmarked theme]
This use of marked indicates a semantic distinction in the domain of a grammatical category. Although, by definition, marked themes are not necessarily ones that are rare, there is nevertheless an assumption that they are infrequent and/or that they stand out. Other uses of ‘marked’ in grammatical descriptions tend to carry a rarity or frequency distinction. In a sense, this is roughly what we are trying to capture with the terms typical and atypical.
4.2.4 Norms and Exploitations
The terms we have considered so far seem to be quite similar in terms of the roles played by frequency and, to some extent, expectancy. In Hanks’ (Reference Hanks2013) theory of norms and exploitations, he uses corpus pattern analysis to identify different patterns of use. A norm is defined as ‘a pattern of ordinary usage in everyday language with which a particular meaning or implicature is associated’ (Hanks, Reference Hanks2013: 92). Norms are not differentiated by frequency; ‘rare patterns are recognized as patterns, despite their rarity, because they have distinctive meanings’ (ibid.). In this sense, a norm has an established meaning. Exploitations are creative choices on the part of the speaker. Hanks (Reference Hanks2013: 212; emphasis in original) offers the following description of an exploitation:
An exploitation is a deliberate departure from an established pattern of normal word use, either in order to talk about new or unusual things or in order to say old things in a new, interesting way. Exploitations are part of the natural human habit of playing with language, but they can also serve a serious purpose, especially in cases where a more well-established way of talking about a particular event, situation, or entity is not available. […] exploitations are typically dynamic, creative, or graphic choices within the boundaries of possible language use.
Thus, exploitations are more deliberate and unusual uses of language, produced by individual speakers for creative purposes, including phenomena such as anomalous collocations, metaphors, metonymy, zeugmas, oxymorons, understatements, euphemisms, hyperboles, irony, sarcasm, and puns. We will return to the notion of exploitations in our discussion of atypical reference in subsequent chapters in order to illustrate that atypical use of reference is indeed often linked to a creative use of language.
As is the case with all categories, the boundary between norms and exploitations is not clear-cut. Instead, it is possible for exploitations to become norms, or an alternation of a norm (see Hanks, Reference Hanks2013), in their own right when used more consistently by a larger number of language users as this entrenches the meaning of the use. Exploitations are regularly the source of new meanings, especially by metaphoric extension. ‘Metaphorical variants are probably always marked close to the time of their coming into the language, but there is no reason why they should stay marked after that period’ (Steiner, Reference Steiner2004: 142). By this, Steiner means that they carry more meaning, in the sense that both speaker and addressee are likely to recognise that a new meaning is construed. Exploitations are atypical instances of language in the sense discussed above in that they are infrequent, but infrequency does not help us distinguish them from norms, which can also be infrequent. Instead, it is their creativity and the way they exploit meaning potential that defines them. Given our discussion of register as ‘variation according to use’ rather than ‘user’, we might expect that, due to the nature of exploitations, they are not influenced by register; however, it may be that certain tenor relationships (e.g. more formal relationships, such as interviewer/interviewee or doctor/patient) constrain creative and novel uses of language.
4.2.5 Typicality and Atypicality as Usage-Based and Register-Dependent
To bring this discussion of (a)typicality to some sort of conclusion and to try to consolidate the various perspectives we have touched upon, let us return to Bybee’s (Reference Bybee, Hoffmann and Trousdale2013) point about repetition and its role in conventionalisation. It is like a well-trodden path in the forest; it is just easier than making a new trail. The more we are used to a way of speaking, the easier it is for us and for our collocutors. Automation, as Bybee (Reference Bybee, Hoffmann and Trousdale2013: 50) points out, ‘allows the units of language to be combined in conventional ways that lead to fluency in both production and perception’. However, the point we discussed above by O’Donnell (Reference O’Donnell2021) is important; as speakers, we can opt out of conventionalised ways of speaking at any moment. And, as Hanks (Reference Hanks2013) claims, if we want to express something new, a new meaning, and if our conventions are not fit for the purposes at hand, we can be creative and exploit existing conventions to do something different. How, then, can we settle on a view of typicality? While we acknowledge that any boundaries we try to place around a given term will likely end up overflowing, we propose that we view typicality as conventionalised language use which is reasonably expected in a given context of situation. The corollary, then, is that atypicality is infrequent, unconventionalised language use for a given context of situation. As Hanks’ (Reference Hanks2013) theory of norms and exploitations states, norms may be infrequent; in other words, frequency is not a property of a norm. A low-frequency instance of a language pattern may be a norm (typical) or an exploitation (atypical). As a final point, and we would not like to put too much emphasis on the terminology here, we simply do not know if Hanks’ (Reference Hanks2013) model is suited to reference, given that it is designed for lexical analysis. For this reason, we prefer to use ‘typical’ and ‘atypical’ as a means of identifying the types of reference that we are primarily interested in discussing in this volume. Therefore, there are a variety of factors which contribute to determining typicality. It may well be that the use of these terms is only of temporary value, for the specific purposes of this volume. There may not be any functional value to them in the broader aim of developing our understanding of reference.
The two remaining sections of this chapter will focus, respectively, on what we understand by typical reference and atypical reference. This discussion will be necessarily brief since these concepts and the evidence for them will be considered in detail in Chapters 5 and 6, respectively.
4.3 Typical Reference
In Chapter 1, we emphasised the importance of viewing reference from multiple perspectives, including discourse, functional, and cognitive approaches, and informing our understanding with empirical research, for example, from psycholinguistics. In this section, we will consider what we learn about typical reference from these areas. Traditionally, reference has been viewed as an anaphoric pairing of a referring expression with a lexical antecedent, that is, a previous lexical instance referring to the same referent (see Chapters 1 and 2). This traditional view of referring has been found to be unsatisfactory since a referent seldom remains static throughout discourse. Instead, a discourse-functional perspective has the advantage of being able to take into account the ‘evolving’ nature of a referent throughout the discourse (Cornish, Reference Cornish2010: 226) in ways in which the antecedent models cannot. As an illustration of what is meant by ‘evolving reference’, we can track the mental representation we have of the referent (potato) of the expressions highlighted in bold in the recipe given in example (80), from Jones (Reference Jones2014).
i) Peel 1kg of Maris Piper potatoes and cut each into 4 even-sized pieces if they are medium size, 2–3 if Ø smaller (5cm pieces). ii) Drop the potatoes into a large pan of water. iii) Add salt, then wait for the water to boil. iv) Simmer the potatoes uncovered, reasonably vigorously, for 2 mins. v) Meanwhile, put your choice of fat into the hot roasting tin and heat it in the oven for a few mins, so it’s really hot. vi) Drain the potatoes in a colander. vii) Carefully put the potatoes into the hot fat and roast them in the oven until they are golden and crisp.
A reader of this recipe would very likely not hold the same mental representation for potato in vii) as the initial mention of potato, the original unpeeled potatoes in i). Throughout the recipe, potato has evolved as a referent. According to Jones (Reference Jones2014: 37), this example ‘clearly illustrates the need to take into account the discourse dimension of a text as the antecedent’s referent does not remain static’. The intended referent has developed from a non-edible vegetable to a delicious culinary delight. We saw in Chapters 1 and 2 that accounts from philosophy and from textuality (i.e. textual co-reference) are limited when it comes to a chain of referring expressions like those in (80), and yet anyone used to reading recipes, and especially anyone who is used to roasting potatoes, will find these references entirely typical.
From psycholinguistic research on reference, we have discovered how interlocutors in any communicative event collaborate in order to refer successfully (see Clark and Wilkes-Gibbs, Reference Clark and Wilkes-Gibbs1986; Schober and Clark, Reference Schober and Clark1989; Brown, Reference Brown1995; Brennan and Clark, Reference Brennan and Clark1996; Bezuidenhout, Reference Bezuidenhout, Gundel and Abbott2019). Specifically, Schober and Clark’s (Reference Schober and Clark1989) collaborative view of referring suggests that participants in a conversation actively work together to ensure that understanding takes place and do not proceed with the conversation until they are satisfied that they have understood each other. We will take a close look at examples where this happens in Chapter 7. Interrupting your speech partner when you have doubts about identifying the referent is also typical. In fact, speakers often catch themselves before the addressee even has a chance, and they rephrase the expression to make it more specific if they think the addressee might not be able to identify their intended referent.
Finally, cognitive research on referring considers the interlocutors’ mental representations of a particular referent and whether they are matching at any given point in the discourse. Chafe (Reference Chafe and Li1976: 27) discusses the various statuses that nouns and referents may have in terms of ‘the speaker’s assessment of how the addressee is able to process what he is saying against the background of a particular context’. He suggests that not only are we able to store a great deal of knowledge, but we also have ‘temporary states’ with relation to that knowledge (ibid.). Therefore, speakers must adapt what they are saying to fit in with what they assume the addressee is thinking at that moment, and only then will the message be assimilated (unless they assume incorrectly of course, which might lead to a misunderstanding, requiring more collaborative work to reach a common ground). Two theories that deserve special attention in this field are Ariel’s Accessibility Theory (Reference Ariel1990) and Gundel et al.’s Givenness Hierarchy (Reference Gundel, Hedberg and Zacharski1993) as they will help us understand the form (i.e. the referring expression) that we conventionally expect most typically in order to successfully pick out an intended referent. These two theories are discussed in more detail in the following two sub-sections.
4.3.1 The Accessibility Scale
For Ariel (Reference Ariel1990, Reference Ariel1994), the status of referring items is a matter of their degree of accessibility. In fact, at any given point in the discourse, the speaker or writer must assess how accessible the intended referent is to their addressee and then choose an appropriate referring expression that corresponds to the addressee’s mental representation of the discourse under construction. Indeed, given the dynamic nature of discourse referents, as we discussed above, the speaker must carry out a new assessment of the referent’s accessibility each time the same referent is referred to. Thus, the assumed accessibility of a referent will affect the type of referring expression a speaker chooses to use, and in turn, a chosen referring expression indicates how accessible the speaker deems the referent to be.
Ariel (Reference Ariel1990) proposes an Accessibility Marking Scale, which we illustrate here in Figure 4.1, where a referring expression of type (a), that is, a full name + modifier, indicates a low degree of accessibility, whereas pronouns and ellipses of type (k) to (o) suggest high accessibility.
Thus, when choosing a referring expression, its degree of accessibility as given on Ariel’s scale should match the accessibility of the referent it is used to refer to. Ariel (Reference Ariel1990: 32) argues that the degree of accessibility is dependent on three overlapping criteria: ‘informativity’ (the amount of lexical information), ‘rigidity’ (the ability to select a unique referent, based on the form), and ‘attenuation’ (phonological size). The more informative, rigid, and unattenuated a referring expression, the less accessibility it encodes. Examples (81) and (82), taken from Jones (Reference Jones2014: 162), illustrate this.
There are two roads to eternity, a straight and narrow, and a broad and crooked.
After the news broke that comedian and actor Robin Williams had been found dead at his California home, tributes began to pour in from friends and colleagues.Footnote 2
The sentence in (81) is an example of a high degree of accessibility, where the gaps created by the unexpressed categorisation (e.g. one or road) in a straight and narrow [one/road] and a broad and crooked [one/road] are lexically uninformative, phonologically unattenuated, and lacking in rigidity (i.e. the use of ellipsis is flexible and can be applied to a limitless number of referents). In example (82), on the other hand, the noun phrase comedian and actor Robin Williams is an instance of a full name with pre-modification. It is lexically informative, highly rigid in that it can only apply to one individual, and therefore indicates that the speaker or writer does not consider the referent to be currently accessible to the addressee; instead, the expression is used to introduce the referent into the discourse.
Interestingly, Ariel (Reference Ariel1990) does not include indefinite noun phrases in the scale, even though they are probably the most typical means to introduce a new entity into the discourse. While the Accessibility Marker Scale may be more ‘reference’ oriented, indefinite expressions should still have a place on the scale as they can be used to refer to a specific entity, newly introduced into the discourse. Consider, for example, the utterance in example (83), which was said in a scenario where a mother was looking for her son at a supermarket.
Here, the mother clearly has a specific referent in mind, even though she does not assume that the addressee knows the identity of her son (although she certainly hopes they can at least identify him). There is, then, a case to be made for indefinite noun phrases to have a place on the Accessibility Scale, although it should be noted that Ariel’s model only accounts for the accessibility status of antecedents (Reboul, Reference Reboul, Connolly, Vismans, Butler and Gatward1997). Other similar scales do indeed include them (see Givón, Reference Givón and Givón1983; Gundel et al., Reference Gundel, Hedberg and Zacharski1993), and we will turn to one such scale in the next sub-section.
To sum up, speakers (including writers), in their choice of referring expression, are influenced by how accessible they deem the intended referent to be for their addressee (Vogels et al., Reference Vogels, Krahmer, Maes, Abbott and Gundel2019). Different types of referring expressions mark for a higher or lower degree of accessibility. Based on this scale, which type of referring expression is typical in a given context is dependent on the level of the intended referent’s accessibility.
4.3.2 The Givenness Hierarchy
Similar to the Accessibility Scale, the Givenness Hierarchy (Gundel et al., Reference Gundel, Hedberg and Zacharski1993, Reference Gundel, Hedberg, Zacharski, Gundel and Abbott2019) evaluates which type of referring expression is deemed appropriate based on the intended referent’s status in the interlocutors’ shared common ground. As opposed to Ariel’s mutually exclusive levels of accessibility, however, this hierarchy comprises six cognitive statuses that are implicational, which means that each status entails all the lower statuses (Gundel et al., Reference Gundel, Hedberg and Zacharski1993: 275). The six statuses of the Givenness Hierarchy are displayed in Figure 4.2 alongside the hypothesised English forms they are assumed to encode (see also our discussion of Givón’s (Reference Givón1993a) continuum of referential intent in Chapter 2).
Figure 4.2 The Givenness Hierarchy
As is explained by Gundel et al. (Reference Gundel, Hedberg, Zacharski, Gundel and Abbott2019: 69), by using a particular form, a speaker/writer indicates that the intended referent matches the respective cognitive status of that form, and that the hearer/reader is able to correctly identify the referent. Moreover, because all lower cognitive statuses in the hierarchy are entailed in the higher ones, a speaker/writer also implies that all the lower cognitive statuses are met. For example, a referent that is in focus is necessarily also activated, familiar, uniquely identifiable, referential, and type identifiable.
In the following, each of the six cognitive statuses will be considered in turn. The lowest cognitive status of a referent is type identifiable, which means that the addressee is able to understand what type of entity is being referred to (cf. type specification as discussed in Chapter 3). For instance, in example (84) the addressee is expected to know the denotation of dog and use that knowledge to categorise the instance of that class. This type of expression was discussed in Chapter 3 in terms of individuative indefinite reference by Radden and Dirven (Reference Radden and Dirven2007).
I couldn’t sleep last night. A dog (next door) kept me awake.
For the second cognitive status, referential, the addressee is not only expected to know the type dog, but also to ‘either retrieve an existing representation of the speaker’s intended referent or construct a new representation by the time the sentence has been processed’ (Gundel et al., Reference Gundel, Hedberg, Zacharski, Gundel and Abbott2019: 69). The use of this dog in example (85) illustrates how the speaker signals to the addressee that they are talking about a particular dog, but that the addressee is not expected to be able to identify the referent as distinct from other members of the class.
I couldn’t sleep last night. This dog (next door) kept me awake.
The referent, however, is only signalled as having the status, of uniquely identifiable when the definite article is used, as shown in (86), although the referent is not necessarily familiar to the addressee. They can make sense of the referring expression and identify it as one particular dog as long as enough descriptive material is provided to single out the referent, that is, next door in this case (see Gundel et al., Reference Gundel, Hedberg, Zacharski, Gundel and Abbott2019: 70).
I couldn’t sleep last night. The dog next door kept me awake.
In instances where the status is familiar, as in example (87), the addressee is assumed to be able to not just recognise the referent as uniquely identifiable, but also use knowledge the speaker believes they have previously gained in relation to this particular referent (e.g. perhaps as part of their long-term memory, or if the referent has recently been mentioned). The demonstrative pronoun that indicates that the speaker assumes that the addressee already knows of their neighbour’s dog. In this model, the status of familiar is a necessary condition for all personal pronouns and definite demonstratives (Gundel et al., Reference Gundel, Hedberg, Zacharski, Gundel and Abbott2019: 71).
I couldn’t sleep last night. That dog (next door) kept me awake.
Building on this shared common ground between the interlocutors even further, the next stage in the hierarchy is ‘activated’. As we read example (88), we might be left wondering what that refers to. We feel like some information is missing. However, if we imagine that the speaker and addressee are together at the speaker’s home and they can both hear the sound of a dog barking, then it becomes much easier to understand the connection between the two sentences. Here the speaker is referring to something from the context of situation (recall our discussion of register above), that is, exophoric reference, through the sound of dog barking. The activated status is assumed to always apply to the speech participants themselves, and it is also required for all pronominal forms (Gundel et al., Reference Gundel, Hedberg, Zacharski, Gundel and Abbott2019: 71).
I couldn’t sleep last night. That kept me awake.
Finally, the highest cognitive status in the Givenness Hierarchy is when the referent is in focus. In other words, the referent is not just activated ‘in short-term memory, but is also at the current center of attention’ (Gundel et al., Reference Gundel, Hedberg, Zacharski, Gundel and Abbott2019: 72), that is, it is the current focal topic in a conversation.
The Givenness Hierarchy identifies the speaker’s assessment of the addressee’s level of activation for the referent in question and associates the various levels with certain lexical forms. Such a correspondence does assume a one-to-one relation, which is an appealing idea, but in practice we often find that linguistic forms are recycled for more than one function. Nevertheless, these form predictions do provide us with a general description of typical reference, depending on the activation level. In the next section, we will turn our attention to what it means for a reference to be considered atypical, although based on what we have said here, we can expect that this will involve a mismatch of one kind or another.
4.4 Towards an Account of Atypicality in Reference
This section examines uses of reference that deviate from their expected forms as predicted by the Accessibility Scale or the Givenness Hierarchy. Both of these theories predict the form of a referring expression that is needed in order to make an intended referent identifiable for the addressee. We first discuss reasons why identifiability might not be the only function of a referring expression, and that when such other functions are at play, the form of the expression may deviate from what is typically expected in the context. Following this, we examine what it means for a referring expression to be atypical, drawing on the discussion of typicality and atypicality in earlier sections. We also consider various examples of referring expressions that deviate from the norm in order to fulfil further pragmatic functions.
4.4.1 The (In)Stability of Identifiability
As we have seen in previous sections, one of the functions of referring expressions is to map the information in the current utterance onto the antecedent-trigger in the mental discourse representation of the interlocutors. The referring expressions should enable the addressee to identify the referent onto which the current information needs to be attached.
There are several linguistic means to do this: in English, these means typically involve zero anaphors, pronouns, proper nouns, and definite noun phrases, all of which are considered to be definite. These expressions differ in their lexical specificity, which consequently impacts on their level of identificational explicitness (Vonk et al., Reference Vonk, Hustinx and Simons1992: 302; Schiffrin, Reference Schiffrin1994: 199). Which device is chosen by the speaker seems to depend largely on how effective it is considered to be in terms of fulfilling its identificational role, which in turn is related to various factors, such as the interlocutors’ representation of the ongoing discourse event, the co-text and context, the assumed degree of accessibility, or the cognitive status of the entity in question and the intentions of the speaker. The pragmatic principle of the maxim of quantity can be applied here too; the choice of referring expression seems to be a matter of locating it on a continuum which balances providing sufficient information for identification with not providing too much.
However, perhaps identification of a referent is not the only function of referring expressions. The wordings of referring expressions are not intended to purely assist in the identification of the referent, but rather also to fulfil a different discourse function (cf. Hanks, Reference Hanks, Gundel and Abbott2019). Fox (Reference Fox1987) found that anaphora is also governed by rhetorical organisation. For example, the principal determinant of whether a pronoun is used rather than a repeated noun phrase is whether the referring expression is in the same structural unit as the initial mention. Further, a full NP is often used at the beginning of ‘a new rhetorical unit’ (Fox, Reference Fox1987: 136). Fox also suggests that non-structural factors play a role in the choice of referring expression, such as ‘categorisation of the referent, further information about the referent, and comparison and contrast of the referent with other people’ (ibid.).
In addition, Vonk et al. (Reference Vonk, Hustinx and Simons1992: 303) found that referring expressions that are more specific than necessary for identification of the antecedent indicate ‘an episode boundary’. They use the following text to illustrate this point (Vonk et al., Reference Vonk, Hustinx and Simons1992: 303):
1. Sally Jones got up early this morning.
2. She wanted to clean the house.
3. Her parents were coming to visit her.
4. She was looking forward to seeing them.
5. She weighs 80 kilograms.
6. She had to lose weight on her doctor’s advice.
7. So she planned to cook a nice but sober meal.
The use of the pronoun she in sentence 5 above may not cause any identificational problems, but the more specific Sally would ‘[make] the sentence sound better’ (Vonk et al., Reference Vonk, Hustinx and Simons1992: 304). Vonk et al. (Reference Vonk, Hustinx and Simons1992) showed empirically that when a device is used that is more specific than needed for the recovery of the intended entity, it also has a discourse structuring function. That is, it marks the beginning of a new theme concerning the same discourse referent. So, it seems that referring expressions do not function merely as identificational devices, but discourse restraints also play a role in the choice of expression.
It is generally accepted that once a speaker introduces a referent for the first time, subsequent mentions of the referent continue in a fairly predictable way. That is, an initial mention is typically introduced or presented using an indefinite noun phrase (but if the speaker assumes that the addressee is able to recover the identity of the referent, a definite NP may be used), and subsequent mentions (which are therefore recoverable) tend to be encoded as definite (as we discussed in Chapter 2). But as Fries (Reference Fries, Scott and Thompson2001: 89) notes, ‘it is not always the case that speakers match actual recoverability with presented recoverability’, and in such cases the referential identities of the referents are perhaps not what is most important in the discourse.
4.4.2 Atypical Reference as an Exploitation of the Norm
Earlier in this chapter, atypicality was defined as a low-frequency, unconventionalised occurrence of a pattern in a given register. Similar to Hanks’ (Reference Hanks2013) exploitations, such uses bring additional meaning, often making a further pragmatic implicature. In referring, atypical uses of expressions can be identified when there is a mismatch between the form of the expression and the form expected by the givenness status of the intended referent. If typical referring expressions follow the expectancies of the Givenness Hierarchy in order to allow the addressee to identify the referent, then atypical references will deviate from this hierarchy, either giving more or less information than is necessary to identify the intended referent. It is this under- and overspecification of the referent which then leads to pragmatic implicatures with functions that go beyond the simple identification of the referent. We will come back to this point in Chapter 8, where we examine formally indefinite expressions which are being used to refer to a highly accessible referent.
For now, let us consider the extract in example (89), which has been taken from the first page of the novel All the Pretty Horses by Cormac McCarthy (Reference Macarthy1992).
The candleflame and the image of the candleflame caught in the pierglass twisted and righted when he entered the hall and again when he shut the door. He took off his hat and came slowly forward. The floorboards creaked under his boots. In his black suit he stood in the dark glass where the lilies leaned so palely from their waisted cut-glass vase. Along the cold hallway behind him hung the portraits of forebears only dimly known to him all framed in glass and dimly lit above the narrow wainscotting. He looked down at the guttered candlestub. He pressed his thumbprint in the warm wax pooled on the oak veneer. Lastly he looked at the face so caved and drawn among the folds of funeral cloth, the yellowed moustache, the eyelids paper thin. That was not sleeping. That was not sleeping.
Several paragraphs later, the reader is introduced to another referent by means of a pronoun. Note also that the identity of he has still not been revealed.
She looked up from the stove when he came in and looked him up and down in his suit. Buenos días, guapo, she said
It is not until eleven paragraphs later that we learn the name of the protagonist, John Grady Cole. The she turns out to be a character only known as Abuela (grandmother in Spanish), an old Mexican woman who had lived on the ranch since the turn of the century and who helped raise John Grady. We never find out her name.
According to the Givenness Hierarchy, use of the personal pronouns he and she in the introductory paragraphs to All the Pretty Horses would mark these referents as ‘in focus’, that is, the highest givenness status. Due to the implicational nature of the hierarchy, this means that the two referents are also ‘activated’, ‘familiar’, ‘uniquely identifiable’, etc. However, at this point in the story, the reader is not yet able to identify the two referents at all. As the narrative unfolds, the discourse representation of the referent develops in that we learn about his actions upon arriving home, his clothing, and the time in which the scene was set. The reader, therefore, has a mental representation of the referent, but identification is still not possible in terms of referent resolution, and the referential content of the discourse representation of the individual is not ratified until we encounter a lexical expression. Hence, the reader is here presented with an atypical use of reference. Usually, when introducing a new entity into the discourse, one might expect the use of a proper name (e.g. John Grady Cole), or an indefinite noun phrase with added descriptive material (e.g. a man in a black suit and hat). Instead, the writer has chosen to refer to the referent simply by pronouns for the extent of several paragraphs. This is an atypical use for a pronoun, which should, in theory, signal identifiability (i.e. being able to pick out who or what we are talking about and not simply a type of entity with particular qualities).
As we will see in Chapter 7, this type of incongruency would make it difficult for the addressee to accept the referring expression. In a conversation, if a speaker uses a personal pronoun and the addressee does not share the level of activation assumed by the speaker, the addressee will request clarification and the conversation will not continue until the identity of the referent is resolved. However, if we consider this use of pronouns within the register of fictional writing, we find that this use is a literary device that allows the writer to ‘trigger the interpretation that a discourse entity is highly prominent’ (Epstein, Reference Epstein2002: 349). In using this literary strategy, the writer chooses ‘a definite description to introduce an important entity at the start of a narrative, for the purpose of calling the reader’s attention to that entity’ (ibid.). Because this choice of expression for a referent that is not in focus is deliberate and creative, we may want to think of it as an exploitation (Hanks, Reference Hanks2013), although perhaps it has become a norm, to some extent, as a conventional literary device. We consider this use to be atypical since the personal pronouns in example (89) are intended to indicate not that the referent should be viewed as in focus but rather that the entity will be the primary topic of concern in the immediately following discourse. Holding back on the identity of the individual forces the question of identity to be addressed by detail and inference, which contributes an element of suspense. Clearly, here, the function of definite descriptions is not a straightforward matter of identifiability.
4.5 Summary of the Chapter
The aim of this chapter was two-fold. First, we wanted the chapter to serve as an introduction to the current section on typical reference and the third section of the volume on atypical reference and extensions. Second, we explored and developed our understanding of the concept of typicality in relation to atypicality. We settled on a view of typicality as conventionalised language use which can be said to fall within the norms of a given register, and atypicality as an instance of language which can be said to be infrequent and unconventionalised for a given register. The following two chapters explore typical reference in two different modes: spontaneous discourse (Chapter 5) and written discourse (Chapter 6).
The remainder of this book, in addition to providing an empirical account of reference in the atypical uses of reference in Chapters 7 to 10, aims to show that these atypical uses serve the needs of the speaker, expressing functions which go beyond the central function of identifiability of referring expressions. As a result, we will argue that atypical reference, despite being atypical by our definition, is often actually quite normal and common. It is more a matter of perspective.
