Читать книгу Special Indefinites in Sentence and Discourse - Sofiana Lindemann - Страница 23

More recent approaches to anaphora resolution emphasize the crucial role played by intersentential semantic coherence relations in correct pronoun assignment. Following the work of Hobbs (1979), Kehler (2002), Rohde (2008), Kehler et al. (2008), Kaiser (2010) argued that coherence relations, which hold between two adjacent clauses, can be of various types, as for example causal relations (e.g. as a result), temporal relations (e.g. next), to name just a few. Moreover, discourse participants are said to build forward-looking expectations about the ensuing discourse based on the coherence relations they assume to hold between sentences. To illustrate how the coherence relations contribute to pronoun interpretation, consider examples (18a) and (18b), adapted from Kehler et al. (2008).

(18)	(a)	Bush narrowly defeated Kerry, and as a result heKERRY took some days off.
	(b)	Bush narrowly defeated Kerry, and then heBUSH took some days off.

Given the result-interpretation in (18a), the pronoun he in the subordinate clause is preferentially interpreted to refer back to the direct object referent, Kerry, than to the subject of the sentence (i.e. Bush). Under a reading in which the events described in the matrix and subordinate clause represent an enumeration of the events that happened, as in (18b), the subject pronoun he in the second sentence is preferentially interpreted to have Bush as an antecedent. An analysis of pronoun resolution in terms of coherence relations was shown to mediate between the conflicting subject and thematic roles biases, which often yielded contradictory results (see the discussion in Kehler et al. 2008). Under carefully controlled conditions, coherence-based analyses were shown to override other semantic or structural approaches to pronoun resolution, and pronoun use was interpreted as an epiphenomenon of a more general preference to establish particular coherence relations in local contexts, between two adjacent sentences (Hobbs 1979, Kehler et al. 2008, Kaiser 2010).

The features discussed above, namely recency, givenness, syntactic and semantic prominence, implicit causality and coherence relations were shown to impact the accessibility of a referent in terms of pronominalization. These characteristics, together with some of the predictions or biases usually associated with them, are summarized in Table 2.4.

Bias	Predictions
Recency	recently mentioned> not recently mentioned
Givenness	old information> new information
Syntactic prominence	subject> object> other
Parallelism	parallel position> non-parallel position
Semantic prominence	e.g. Goal>Source in a transitive event representing these two thematic roles
Implicit Causality	different types of verbs show different NP-biases
Coherence relations	e.g. Elaboration relation: preference for subject continuation

Table 2.4:

Factors contributing to pronoun use and interpretation

The factors presented in Table 2.4 predict that a referent which represents given information, which is mentioned as a subject, bearing the semantic Goal role (vs. Source), which is mentioned in the immediately following clause, tends to be picked up by a less explicit type of referring expression. On the contrary, a referent mentioned in object position, which conveys new information, and which is not mentioned in the immediately following sentence is less accessible and will need a more explicit type of referring expression to be mentioned in the subsequent discourse. The biases presented above have been shown to impact the accessibility of referents, however, as Arnold (1998, 1999) and Kaiser (2010) have pointed out, a single-factor system cannot explain accessibility as a whole. Rather, the factors differ with respect to how influential they are relative to one another at a particular point in the discourse.

It is important to note that these factors have been singled out in comprehension or interpretation studies, by employing more often than not a backward-looking perspective. In other words, these studies were mostly concerned with predictions pertaining to the most likely antecedent of an already given pronoun. Moreover, coreference relations were computed between adjacent sentences. The bundle of characteristics discussed so far make predictions about the likelihood of subsequent pronominalization, but not about the likelihood of (more or less frequent) subsequent mention or about the type of referring expression used next besides pronoun use. In this book I present cross-linguistic evidence to show that the type of referring expression used (i.e. marked vs. unmarked indefinite noun phrases) can be used as a heuristic to predict the discourse structuring potential of referents introduced in this way. For example, the two referents, Paul and a girl, are introduced in the first sentence conjunct in (19). The first-mentioned referent, Paul is the grammatical subject and the semantic Agent of the sentence. This referent is mentioned in the immediately following sentence by a zero anaphor in a parallel subject position. These characteristics taken together predict that Paul is the most accessible referent in the two-sentence discourses presented in (19a) and (19b).

(19)	(a)	Paul	a	întâlnit	o fată.	A	fost	impresionat	de	părul ei.
		Paul	Aux	met	a girl	Aux	was	impressed	by	hair her
		‘Paul met a girl. He was impressed by her hair.’
	(b)	Paul	a	întâlnit	-o	PE	o fată.	A	fost	impresionat […].
		Paul	Aux	met	CL	PE	a girl	Aux	was	impressed […]
		‘Paul met a girl. He was impressed […].’

The second referent is introduced in the first sentence of (19a) as a direct object realized as the indefinite noun phrase o fată (‘a girl’) and by means of the indefinite noun phrase pe o fată (‘PE a girl’) in (19b). The referents of the direct objects are taken up in the next sentence by a pronoun and in syntactically parallel positions (i.e. as direct objects). Thus, the second referent, pe a girl, in (19b) is less accessible than the first-mentioned referent, Paul, as it conveys new information and is mentioned in direct object position. By limiting our analysesto to two-sentence contexts as the ones presented above, we miss an important point about the discourse contribution of the use of indefinite noun phrases realized by direct objects in Romanian. First, speakers can choose between two possible constructions to refer to the same referent in direct object position, namely one in which the indefinite noun phrase remains unmarked, as illustrated in (19a) and one in which the indefinite noun phrases receives the marker pe, as in (19b). Second, referents associated with pe-marked indefinite noun phrases are more prone to be mentioned next and will shift the focus. These effects will become visible, however, only in larger discourse units. Results of the multi-sentence completion studies presented in Chapter 5 illustrate this observation.

The approaches discussed so far cosider accessibility a graded and dynamic phenomenon that is subject to constant change as the discourse unfolds. While various factors have been shown to affect accessibility, a constant indicator of a referent’s accessibility status at a particular point in the discourse is the type of referring expression used to refer back to the antecedent of this referent. Highly accessible or topical referents tend to be coded by means of attenuated types of referring expressions (e.g. zero pronouns, personal pronouns). On the contrary, less accessible referents tend to be picked up by more elaborated types of referring expressions (e.g. definite (unmodified or modified) noun phrases). Moreover, the preceding discourse (i.e. the discourse in which the referent was introduced for the first time) is considered of utmost importance in determining the status of a referent in a given discourse. In other words, the way in which the antecedent was mentioned in the previous sentence(s) decisively impacts the prominence of its anaphor. Note, again, that accessibility was computed locally, between adjacent sentences. Thus, establishing the prominence of a referent is a circular method, as it requires looking back at the properties of the antecedent in order to define the accessibility of the anaphoric expression (see Arnold 2001 for a similar remark). Nevertheless, this analysis has at least some advantages, as it provides us with a means to calculate the prominence of referents and make predictions about an antecedent and the correlation between this antecedent and its subsequent coreferential expressions.

The theories on accessibility presented so far are backward-looking as they focus on the preceding discourse to determine the accessibility of a given referent. Few studies focused on the behaviour of referents that are realized in structural or semantic non-prominent positions. Such non-prominent referents were considered at best ‘secondary topics’ (Givón 1983) or, non-preferred backward-looking centers in Centering Theory’s terms. In what follows, I discuss a more recent expectancy-based account that reconciliates between the two discrepant views on accessibility as activation and topicality. This expectancy-motivated approach will be used to explain the contribution of the indefinite noun phrases discussed in this book.