Coding weight C Cognitive statuses and coding of referents

underlies the scale is a complex “degree of accessibility of the mental entity for the addressee” as the speaker assesses it Ariel 1996:15. In her Accessibility Theory, degree of accessibility is based on “many variables,” prominent among which are discourse topicality, recent mention, and lack of other possibilities Ariel 2004:99. 51 Topical and recent active concepts with no competitors are easier to access than their opposites, though for different reasons. Sometimes, in place of actual discourse topicality status in a given context, Ariel cites semantic properties of prototypical topics, particularly humanness. Since, however, there are different dimensions to “givenness”“accessibility,” it is not surprising if they at times make different use of linguistic expressions. At the left end of the scale, as noted above, center of attention combines criteria from both discourse topicality and activation, and in English it is commonly used both for discourse topics and very recent referents. But under certain conditions, languages may have referents that are even more accessible than “center of attention.” Mambila Northern Bantoid, Nigeria uses zero reference for a special case of discourse topicality: the global VIP main participant of a narrative, once established, is generally referred to by zero, whereas “participants other than the main one are re-identified by a noun every time they are mentioned” Perrin 1978:110f.; see Dooley and Levinsohn 2001:60 for an example. Thus, the Givenness hierarchy could be extended to the left and possibly branched to include categories of “ultra high accessibility,” which would have to do with discourse topicality for Mambila. 52 So although the Givenness hierarchy may be generally correct, it is in fact a composite, it requires extensions to cover other statuses, and its details need to be adapted to particular languages.

2.4.3 Coding weight C

ODING WEIGHT is a useful way of describing referential and other expressions. The following simplified scale, involving both prosodic and morphosyntactic criteria, is adapted from Givón 1983:18: Coding weight of referring expressions: Heavy: full NPs stressed or independent pronouns unstressed free pronouns bound pronouns including “agreement” Light: zero anaphora 53 Besides the form of referring expressions, there are at least two other other dimensions of coding weight: • Besides the coding of the referring expressions themselves, it appears that certain syntactic positions in the sentence are less prominent, hence less “weighty,” than others. Cross-linguistically, it appears that right-detached position is less “weighty” than the left-detached position Givón 1983:19, Lambrecht 1994:203f.; see example from Mithun 1996 at end of this section. 54 • Intonational criteria should be considered. Referring expressions which have a sentence accent or which are bounded by pause, or at least by an intonation boundary, are more weighty than those which occur in other prosodic contexts Rauh 1983:37, Mithun 1996, Ariel 2004:100. 51 There are other explanations, such as “degree of awareness” Emmott 1997:216–221 and Almor’s 1999 “information load hypothesis” cited by Zwaan and Singer 2003:100: “anaphoric phenomena are regulated by the processing cost of identifying the referent and computing the new information signalled by the anaphor.” Cf. also Tomlin 1987:458: “The speaker uses a pronoun to maintain reference as long as attention is sustained on that referent. Whenever attention focus is disrupted, the speaker reinstates reference with a full noun….” A similar formulation in terms of relevance theory is found in Sperber and Wilson 1986:47 and Wilson 1992. 52 Mambila also has zero reference for same-subject deletion in third person within sentences Perrin 1978:105. 53 In other places in this treatment, as is common, “zero” can also designate an elided argument, with no free expression but possibly with verb agreement. 54 Other “surface level characteristics of discourse” which affect coding weight include order of mention, syntactic role, depth of embedding, and competing candidates or referential distance Maes et al. 2004:118. Coding weight thus has at least three distinct aspects: the form of the referring expression, its position in the sentence, and its intonation. In general, coding weight increases as accessibility decreases. In the Givenness hierarchy Figure 11, both coding weight and processing cost generally increase to the right, due to lower activation, lesser identifiability or less probable topicality. Other accessibility factors come to light as we consider basic functions of reference in discourse Lambrecht 1994:183. The following list is adapted from Dooley and Levinsohn 2001:55: Basic functions of referential expressions in discourse: Semantic: to identify referents unambiguously, distinguishing them from other possible ones that occur in the context or that can be reasonably inferred: cost increases with danger of ambiguity and disruption Conceptual: to signal the activation, identifiability or discourse topicality status of referents: cost decreases towards the left in the Givenness hierarchy Contrastive: to signal constrast with another referent: cost increases contrastiveness Hierarchical processing: to signal the onset and possibly the end of a discourse unit: cost increases at the beginning and sometimes slightly at the end of a discourse unit The composite processing cost of an act of reference therefore depends largely on its basic discourse functions. 55 Since most or all of these are a matter of degree, accessibility is complex and scalar. At different points we have used the notion of MINIMAL CODING for a language. Intuitively, it designates the lowest coding weight that the language typically uses for referring expressions for the subject of the independent clause in simple sentences. 56 Minimal coding in English is a pronoun, whereas in “pro-drop” languages it is zero. Non-subject arguments may require more coding: for example, standard Portuguese has zero minimal coding for subjects, but objects must have at least an clitic pronoun: não sei explicá-lo NEG know.1SG explain-3SG.O ‘I don’t know how to explain it’. Adapting terminology from Levy 1982:298–300, we can say that a referent in a particular context has MINIMAL IDENTIFICATION if its referring expression has just the coding weight that its accessibility requires, thus reflecting the “referential heuristic” of Maes et al. 2004:118: “not to expect to use more referential material than necessary to uniquely identify a discourse referent.” The referent is OVERCODED if it has more than minimal identification and UNDERCODED if it has less. Minimal identification, which is context-sensitive, is not to be confused with typical MINIMAL CODING . The English expressions in the Givenness hierarchy Figure 11 represent minimal identification at each position; only on the left end is minimal coding found. Undercoded referring expressions are generally “infelicitous,” hence do not generally occur except in special morphosyntatic, discourse, or pragmatic conditions, such as zero reference for Mambila global VIPs see above in §2.4.2; Gundel et al. 1993:292. 57 55 Some coding choices are made for what might be called secondary discourse functions which have little to do with accessibility. For example, the following example is found in Appendix B, lines 07–08: She considered staying in the car, but the garage door opened and the husky figure of her husband, Jim, appeared. Jim, who’d been stowing away patio furniture, liked to call his slightly built wife Toughie, because of her energy and determination. The NP his slightly built wife is overcoded; in this context the pronoun her would be sufficient. As a descriptive NP, it goes far beyond reference per se: it provides information about the referent which is relevant to other discourse goals. As with nonrestrictive relative clauses such as who’d been stowing away patio furniture and nonrestrictive modifiers such as husky in this same material, this kind of reference is “syntactic opportunism”: it goes beyond the characteristic function of the element and simply adds new information in an economical way. 56 Coding can vary with syntactic situation; in English, a subject that is correferential with that of the independent clause in the same sentence is commonly zero. In standard Portuguese, minimal coding for subjects is zero but for objects is an unstressed pronoun. 57 Text producers occasionally commit “infelicities”—discourse-pragmatic performance errors, somewhat like ungrammaticalities in formal grammar. When text producers becomes aware of these, such as through verbal Since the Givenness hierarchy is cumulative and implicational, however, overcoding is permitted op. cit., p. 290. However, the cognitive principle of relevance Sperber and Wilson 1995:261 predicts that a reference which is overcoded for one function is probably performing another function which compensates for the increased coding. In English, for example, “many full NPs which occur in narratives where one could have expected pronouns are functioning to signal the hierarchical structure of the text … to demarcate new narrative units” Fox 1987:168. An example is provided in the following excerpt from Appendix E: Example text 13: Oral text 4, excerpt Labov and Waletzky 1967:16 08 An’ I remember real well what happened. 09 Bunch of us kids was out there playin’; an’ no one meanin’ any harm about it. 10 But anyway, Mrs. Hatfield came down an’ took away her money from Mr. Hatfield, you know, for the peaches, ’cause she know he was gonna buy drinks with it. 11 ’Nd Mr. Cassidy was laying out there in the yard. 12 ¶And Mr. Cassidy just looked up, and he said to Bill, just—just jokin’, just in a kiddin’ way, he said “Uh huh,” he says, “that’s—another dollar bill you won’t get to spend for drink, hunh.” Lines 08–11 are a paragraph of setting information containing a secondary event line 10. Line 12 begins an action paragraph which turns out to be the story’s peak. The full form of And as a sentence- initial conjunction appears to confirm this We note that Mr. Cassidy in line 12 is overcoding, since the subject is the same as in line 11 and there would have been no ambiguity with the pronoun he. The NP appears to be overcoded with respect to the semantic and conceptual functions but not for the hierarchical textual function initiation of a new discourse unit. If overcoding occurs gratuitously, without any intention of compensating payoff, then it can result in difficulties and even errors of interpretation: “If a character is referred to by a proper name after a run of pronominal references, then the name itself serves to indicate that a shift in topic is occurring” Sanders and Gernsbacher 204:84, citing Vonk et al. 1992. An intermediate level of overcoding is found in Central Pomo, an SOV native American language of Northern California: active NPs often occur right-detached, with monotone intonation: “Only what you can eat, take only that much,” it seems she said, the old lady to her children Mithun 1996:231. According to Mithun, right-detached NPs occur “for several purposes. They most often confirm the identity of a given referent when several entities are active at the same time. They may signal discourse structure, providing a summary at the end of an episode or scene” loc. cit.. That is, right-detached NPs can occur because of the semantic function for disambiguation or the hierarchical function for signalling the closure of a discourse unit. This is an intermediate degree of overcoding because, although NPs are used, they have no sentence accent and the right-detached position is not as discourse-prominent as the left-detached position Givón 1983;19, Lambrecht 1994:203f.. Another type of “light” NPs is found in Koiné Greek when active nouns occur without articles §3.5.2.

2.5 Conceptual and formal evidence of discourse organization