Halliday and Hasan’s (1976) taxonomy is frequently applied for cohesion analysis. It can be used for almost any text type of any length. Experience shows, however, that with longer texts, the analysis easily becomes very complicated, as cohesive chains become more and more difficult to follow with each sentence. Besides the difficulty of applying the taxonomy, it has a number of weaknesses as regards its validity and reliability.

 

 

Lexical cohesion entails a mutual connection between lexical items. Conjunctions also stand out, as they might have an organizing function between larger segments of text. It is also emphasized that cohesion is basically a semantic relation expressed through either the lexical or the grammatical system (Halliday & Hasan, 1976). As the two systems are not easily divided, the guiding principle is that “the more general meanings are expressed through the grammar and the more specific meanings through the vocabulary” (Halliday & Hasan, 1976, 5). This statement is particularly true for reference, where the existence of a referential relationship is expressed by a grammatical element (e.g., a demonstrative), and the identity of the presupposed item is specified by the meaning of the noun that follows. Consequently, a valid description of a referential chain needs to include a straightforward description of such lexical relationships. As a number of revised and current approaches have been developed for lexical analysis, the original 1976 taxonomy of lexical relationships is no longer valid and needs to be replaced by one of the current taxonomies for the present paper, which is Károly’s (2002) taxonomy summarized in Table 16 above. Besides describing the use of cohesive devices in a text, identifying the non-cohesive use of the observed elements might indeed prove as relevant to the interpretability of the text as their cohesive use. Therefore, a full description of cohesion realized by a particular cohesive device will also include the ratio of cohesive and non-cohesive use of those items. Another methodological weakness of Halliday and Hasan’s (1976) approach is its failure to account for cohesive chains that display patterns other than a linear referent-referring item sequence. Cohesive chains often divide into two or more chains or might unite into one chain, which considerably affects the complexity and accessibility of texts. A further methodological problem is the treatment of cohesive errors. Inappropriately used or ambiguous cohesive ties may occur both in NS and NNS writing. For this reason, Ting (2003) suggests that any application of Halliday and Hasan’s (1976) taxonomy needs to be complemented with a qualitative analysis to deal with cohesive errors in order to be valid. In this study, besides a qualitative analysis, a visual representation of cohesive ties will be used. This representation not only allows the inclusion of cohesive errors in the analysis, but also defines them, by indicating a missing item for referring items that need a presupposed element for their decoding. A further methodological problem is the treatment of cohesive errors. Inappropriately used or ambiguous cohesive ties may occur both in NS and NNS writing. Consequently, Ting (2003) suggests that any application of Halliday and Hasan’s (1976) taxonomy needs to be complemented with a qualitative analysis to deal with cohesive errors in order to be valid. In this study, besides a qualitative analysis, a visual representation of cohesive ties will be used. This representation not only allows the inclusion of cohesive errors in the analysis but also defines them by indicating a missing item for referring items that need a presupposed element for their decoding. As the revision of the original taxonomy seems inevitable, Chapter 6, 7 and 8 describe the steps to its revision to create a more reliable instrument for analyzing reference in texts for a more valid representation of cohesive chains in the present corpora.