Abstract Entity Anaphora In Argumentative Texts-pragmatic Features And Referent Interpretation

This dissertation first defines abstract entity anaphora and differentiates it fromother kinds of anaphora, especially concrete individual anaphora (i.e. nominalanaphora). Thus abstract entity anaphora is established as a special kind of anaphora,which is distinguished by its antecedent (a phrase, a clause, a sentence or a sentencecluster). The anaphor refers back to the proposition or concept expressed by theantecedent instead of the linguistic form of the antecedent.To date classical and modern novels have been used as the data source in most ofthe anaphora study and other genres have scarcely been analyzed. However, this studyfocuses on argumentative texts since by comparison among argumentative, expositoryand narrative texts the frequency of abstract entity anaphora in argumentative textsranks the highest. That motivates us to select argumentative texts as our data sourceand accordingly expands the scope of anaphora study. More than160,000-wordargumentative texts (80,000-word English texts and80,000-word Chinese ones) areselected from newspapers, journals and books. In data analyses both qualitative andquantitative analyses are employed. Qualitative analysis produces the pragmaticfeatures of abstract entity anaphors—recapitulation (recapitulating the content,recapitulating the property, recapitulating and commenting), obscuration anddemarcation. Quantitative analysis consists of referential distance, the frequencies ofantecedents and anaphors, comparisons with concrete individual anaphora, and theprobabilities of the concurrences of each kind of anaphor with its typical antecedentsand hence the pragmatic tendencies. The results show that abstract entity anaphoradisplays much difference from concrete individual anaphora in each aspect mentionedabove. And it is indicated that the anaphors are not fully constrained by Ariel’s (1990)Accessibility Hierarchy. What’s more, in all those respects English abstract entityanaphora manifests congruity with Chinese one despite the linguistic difference.At present the theories of anaphora resolution are postulated especially forconcrete individual anaphora and thus fail to solve the problems in abstract entityanaphora. In order to deal with abstract entity anaphora Asher (1993) puts forwardSegmented Discourse Representation Theory (SDRT). This dissertation introducesSDRT and applies it to Chinese abstract entity anaphora so as to test its universality.Apart from that, it is also found in this study that SDRT can be applied to solvebackwards anaphora which Asher never touches upon. The major difference between backwards anaphora and forwards anaphora lies in the constituent graph.Although SDRT can account for most of the instances of abstract entity anaphora,yet it appears incompetent in dealing with some phenomena in the data of this study.Moreover, the inadequacies of this theory lead to its incompetence as well. Thisdissertation points out seven problems and proposes corresponding solutions.First of all, in order to simplify the complicated discourse relations involved inthe SDRS construction, new simplified categorization is suggested based on Sanderset al.’s (1993) theory and Asher’s practice of topic construction. Discourse relationsare simplified into three categories: Coordination, Subordination and BinaryStructures.As for the topic construction of Binary Structures, Asher reveals his vaguenessand confusion. This dissertation advances several solutions according to the differentways of information distribution. In addition, we establish the dominating role oftopic about which Asher shows his hesitation.The two discourse constraints in SDRT—AVAILABILITY andWELL-FOUNDEDNESS may only provide validity for given referents rather thanhelp determine correct referents since they allow more than one referent. In view ofthat, this dissertation avers that it is the semantic information in the discourse thatrealizes referent interpretation. Hobbs’(1985) coherence theory and Lambrecht’s(1994) Delayed Construal may support the view. However, for abstract entityanaphora, the semantic information may not be restricted to the verbs or adjectives inthe anaphoric sentences as is postulated in Hobbs’ theory. It is found in our data thatcorrect referents may be suggested by the semantic information in the anaphors(demonstrative NPs or full NPs), in the anaphoric sentences and in the sentence(s)between the antecedents and the anaphors (if there is any).The data presents another phenomenon—implicit referent, which SDRT fails totouch upon. And this can never be solved directly by topic construction of SDRS.Only after proposition abstraction the implicit referent can be identified. Our datadisplay the following antecedent types of this phenomenon: binary structures,rhetorical questions and negative sentences. Partial abstraction can be made in a seriesof declarative sentences. Generally speaking, the semantic information in theantecedent and the anaphor should be consonant and coherent. Otherwise, theincoherence may entail proposition abstraction to identify the referent. Therefore, theinformation in the anaphoric sentence may serve as the trigger for proposition abstraction.Currently many theories about anaphora aim at ambiguity resolution. But Ashergoes against that. He exaggerates the cognitive difference of people and thus inprocessing certain types of discourse he assigns two attachment sites—“highattachment” and “low attachment”. This is bound to bring about referent ambiguity.This study makes an investigation into the comprehension of these types of discoursefrom pragmatic perspective. Twelve discourse segments are devised according to thefeatures of these types of discourse (including Asher’s original discourse segments).250freshmen participate in this investigation. The statistical significance by X2testsmanifests that the readers have strong tendencies in processing these types ofdiscourse, i.e. they tend to identify the same referent. This result makes it reasonableto reject Asher’s implausible practice.Since argumentative texts are used as the data of this study, long-distanceanaphora particular to such texts occur in the data collection. If we use SDRT toconstruct SDRSs sentence by sentence to handle such phenomenon, it will lead toconfusion. Moreover, SDRT constraints for abstract entity anaphora can not beapplied to such long-distance referent resolution. In this dissertation the“superstructure for argumentative texts” is designed and a hypothesis to solvelong-distance anaphora is postulated. Under these circumstances, AVAILABILITYand Provisional, Descriptive Constraint (A) can be applied to the long-distanceanaphora.In general, this dissertation has four aspects of significance. Firstly, it establishesabstract entity anaphora as an independent and a special kind of anaphora. Secondly,the research methods are the combination of empirical study and theoreticalhypotheses as well as the coalescent of dynamic study and static study. Thirdly, thisdissertation is not limited to the application of SDRT to Mandarin Chinese andbackwards anaphora. Instead, based on the linguistic phenomena in the data, itchallenges and improves the theory, and it even negates some aspects and meanwhilebrings forward new solutions. Finally, the use of argumentative texts as dataconstitutes another characteristic of this dissertation.