1 Referring in a second language
Introduction to the volume
Peter Crosthwaite and Jonathon Ryan
During the production of any discourse, one must use language to refer to people, places, objects or ideas under discussion. While there are few genuine typological universals that hold across all languages, the ability to introduce and maintain reference is a feature common to every language, and, as with the ability to discuss concepts such as space, causation and time, reference is a feature relevant to the coherence of any discourse produced, resolving the central question of “who did what (to whom)?”.
While, for reasons of space, we must necessarily eschew discussion of the various conceptual definitions of reference in the literature (e.g. Frege, 1892), in this volume we define reference as realised in the use of referring expressions, including zero, pronominal and nominal forms, to refer to entities under discussion. These expressions are then organised through such means as information structure, substitution and word order. The appropriate selection, position and marking of these referring expressions allow speakers and listeners (and writers/readers) to know when new information is being introduced into discourse, and when continued reference to old information is updated. This process of referential “movement” affects “about every third word of discourse (sometimes even more than that)” (Kibrik, 2001, p. 1124). Moreover, the complexity involved in managing reference across extended discourse and multiple referential targets is astounding, and in the face of such complexity, “the fact that people actually manage to understand one another most of the time seems almost magical” (Fretheim & Gundel, 1996, p. 7). Yet, the ability to introduce and maintain reference is one that is acquired by most children from a very young age (Serratrice & Allen, 2015).
The focus of the present volume is on how second language (L2) learners refer in their target language. Despite the apparent ease with which children acquire the ability to produce and maintain coherent reference in their first language as noted previously, research into second language acquisition has repeatedly shown that L2 learners struggle to acquire the means to make consistently accurate and felicitous reference in the target language. Rather, they are prone to producing reference that is ambiguous, erroneous, under-informative or overexplicit (or all of these), with serious implications for the coherence of any discourse that L2 learners are attempting to produce.
As pointed out by Hendriks in the foreword, while studies on L2 reference have revealed a range of findings explaining the difficulties involved in producing coherent reference in the L2, the time is now right to expand upon previous findings through new methodological advances, data from previously underexplored L1/L2 pairs and alternative theories of L2 reference acquisition and use. In the rest of this introduction, we explore how L1 reference has been characterised in the (applied) linguistics literature, before providing a brief overview of previous L2 treatments of reference in the field. We then discuss in more detail the need for this volume before introducing the individual contributions.
Approaches to L1 reference
There have been various treatments of reference in the (applied) linguistics literature over the past 40 years, each of which has sought to account for the conditions governing the speakers’ (or writers’) selection and positioning of referring expressions. Given considerations for space, we cannot hope to account for all of these, but have attempted to select treatments from those existing in only one domain (e.g. syntax) to those that take a more integrated (e.g. discourse-pragmatic) account of this phenomenon.
One of the most influential early syntactic treatments of reference is Chomsky’s (1981/1993) government and binding theory, developed to account for the positioning and coreference of pronouns to their antecedents (i.e. John saw his mother), although the syntactic binding component of this theory was modified in the later Minimalist approach to include semantic and phonological elements (Lasnik & Lohndal, 2010). Similarly, early typological accounts of coreference have focused on relative clause formation, such as the Noun Phrase Accessibility Hierarchy (Keenan & Comrie, 1977). These approaches took the view that there were universal grammatical principles dictating the conditions for reference in natural languages. However, there has been a great wealth of research in linguistics to challenge this view (which we do not attempt to cover here), and, ultimately, these theories did not seek to address reference as performed across greater units than the sentence.
Early functional treatments went beyond the sentence level to categorise reference as an underlying semantic property of the complete text, under the notions of cohesion and coherence (Halliday & Hasan, 1976; Hasan, 1984). Halliday and Hasan suggested that the cohesion of a given discourse text was formed through the presence of cohesive “ties” between presupposed referential, spatial, temporal or causal elements. These included a general notion of “reference” in the form of antecedent-anaphor relations (e.g. definite articles and demonstrative noun phrases (NPs)), alongside “substitution” (including pronouns and ellipsis). Hasan (1984) claimed that the coherence of a text could then be determined by the frequency of the cohesive “ties” within it, although this theory was subsequently determined to be overly superficial and did not hold up to empirical scrutiny (Carrell, 1982).
More recent accounts of reference have (quite rightfully) determined that reference cannot be placed within the sole domain of syntax or semantics, but that reference is subject to a range of conditions and constraints at the intersection of syntax, semantic, pragmatic and discourse interfaces. Reference cannot be treated simply a feature of the text, but is now thought of as (minimally) a two-way exchange between speaker/hearer (or writer/reader) realised in the “common ground” (following Chafe, 1974; see also Clark, 2015), with reference to entities within a given discourse held in “in two collaborating minds” (Gernsbacher & Givon, 1995, p. viii) in real time, as shown in examples (1) and (2) below:
(1) A: I just got back from Paris last week.
B: That must have been an interesting trip.
(2) A: Did you talk to Kate yesterday?
B: Yes, I told her about the arrangements
(Clark, 2015, p. 330).
The speakers’ (or writers’) specific use of a shorter NP such as the demonstrative pronoun (1) or personal pronoun (2) at that particular moment (rather than the use of a fuller NP form) is designed to allow the listener (or reader) to resolve the reference with as minimal effort and maximal efficiency as possible (Hawkins, 2004). Under such accounts, either an incorrect (i.e. an error at the syntax/semantics level) or – for the first time – an infelicitous NP (at the pragmatics/discourse level) are both likely to result in the listeners/readers’ inability to appropriately resolve the reference, potentially leading to miscommunication with an accompanying breakdown of overall coherence (Ryan, 2012).
Once researchers had realised the central importance of pragmatics and discourse to the realisation of reference, a range of syntax-pragmatic scales/hierarchies of referring expressions were then proposed in the literature to account for the felicitous selection of referential NP forms according to particular discourse-pragmatic contexts. A particularly influential account under this paradigm is that of Gundel, Hedberg, and Zacharski’s (1993) Givenness Hierarchy, where the speakers’/writers’ selection of referring expressions depends on the relative “cognitive status” of the referential target, depending on their relative “givenness” in the discourse as it unfolds (a more detailed account of this framework is provided in Jennifer Killam’s chapter within this volume). An alternative account is that of Givon’s (1995) Topicality Scale, where the selection of referring expressions signals the “topicality” of a referent according to its level of “activation,” as calculated by such measures as the relative distance between references to said referent. In other words, zero or anaphoric pronouns may be used for “continued [coreferential] activation,” while indefinite nouns, demonstrative NPs, definite NPs (including relative clause modifiers), L-dislocation and grammatical role/voice change are used for “discontinued [non-coreferential] activation,” with these forms listed in order of their relative level of activation. While not without their flaws, the strength of these accounts was their ability to account for reference beyond the sentence level and to account for variation in referential forms used according to contextual/discourse-related factors shared between both speaker and listener (or reader/writer) during production.
Probably the most influential account of reference along discourse/pragmatic lines is that of Ariel’s (1991, 2008, 2010) Accessibility Theory (AT), which is used in Jonathon Ryan’s; Jo Lumley’s; Peter Crosthwaite and Min Jung Jee’s; and Ewa Lenart’s chapters in this volume. Ariel’s scale of referring expressions as organised according to the relative level of “accessibility” they denote are shown in this Accessibility Scale (Ariel, 2008, p. 44):
Full name > long definite description > short definite description > last name > first name > distal demonstrative > proximate demonstrative > stressed pronoun > unstressed pronoun > cliticised pronoun > verbal person inflections > zero
According to Ariel, referring expressions are used to “encode” (Ariel, 2008) the level of inference required to resolve any reference at a given moment, taking into consideration the target referent(s)’ relative distance between discourse-old and repeated mentions, competition between referents of similar types, degree of salience of the referent, and the unity (breaks in continuity) of reference within a given discourse sequence. Interestingly, in a way, Ariel’s hierarchy of referring expressions hearkens back to the universalist position taken by Chomsky et al. in that the configuration of forms along the hierarchy is believed to hold cross-linguistically.
Most recently, corpus-based and computational accounts of reference (mostly in the form of algorithms for coreference resolution) have become increasingly prominent, given the potential benefits of successful coreference resolution for machine translation and automated essay scoring, among other applications. While computational linguistics-based accounts of reference lie outside the scope of this volume, the history of these accounts has also followed a trajectory from mainly syntax-led approaches to those incorporating semantic and discourse information as key components of successful coreference resolution across a variety of languages (e.g. Poesio, Stevenson, Eugenio, & Hitzmann, 2004; Lee, He, Lewis, & Zettlemoyer, 2017).
Approaches to L2 reference
Like L1 reference, work on L2 reference has been conducted within the domains of syntax, semantics, discourse-pragmatics and the intersection of each. If we are to take the distinction between these domains at face value, the second language learner will have already internalised the kind of language-universal syntax-semantic-pragmatic tendencies of linguistic form and referential function (e.g. the NP hierarchies such as Ariel’s AT) during the course of acquiring their first language. We should also consider that if typological universals [of the type predicted by AT] are universal to natural human languages, “then they should also hold for interlanguages” (Callies, 2009, p. 107). The challenge for L2 learners, therefore, is to first acquire a new set of referential forms in the target language, ...