The category of comparison differs from the category of personals or demonstratives mainly in that its lexico-grammatical representation is not realized by a closed set of items. Again, as with other types of reference, the referent may be a passage of any extent. There are two main types of comparison, where one is termed general and expresses identity (e.g., same, identical), similarity (e.g., such, similarly, so) or difference (e.g., other, different(ly), else) “without respect to any particular property” (Halliday & Hasan, 1976, 77). The other type – particular comparison – is based on quantity (e.g., more, fewer, so many) or quality (e.g., comparative adjectives or adverbs). The latter contains an open set of lexical items, which means that it would be difficult to find all the items in a document by computer search. In addition, this open category would create an unnecessary source of error and unreliability in our analysis because such comparisons in academic writing are extremely rare (on average 2-3 items per RA).

 

 

In the discussion of each of the types of cohesion, Halliday and Hasan (1976) concede that their categories (lexical cohesion, substitution, reference and ellipsis) of items are in many ways arbitrary:

The intention of this study is to build an analytical tool for the analysis of reference working with a manageable number of possible referring items so as to be able to computerize some of the analytical procedure; therefore, narrowing down the category of comparatives to the most relevant items is unavoidable. Among the first nine most frequent items in our RA corpus (as, … less) five are also among Biber et al.’s (1991) analyzed reference items, as pre- or semi-determiners based on their syntactic positions, especially when they have a comparative function. These are more, less, same, other and such. Therefore, to a certain extent, Biber et al.’s (1991) approach is followed in setting up the framework for analysis. As for the remaining four forms of comparison, the items different(ly), similar(ly), as and so will be kept as comparative reference, based on their high frequency. In their Grammar of Spoken and Written English, Biber et al. (1991) deal with cohesive reference extensively in their overview of nominals in discourse. Their statements are based on a study of a large corpus of texts, among them an approximately five-million-word corpus of academic prose. They describe chains of co-reference, which they define as “sequences of noun phrases all referring to the same thing” (Biber et al., 1991, 234). Apparently, as their focus is primarily on nominals, they go further than describing reference only. Chains of reference for them also include “repeated noun or synonym” (ibid., 235); that is, they do not separate lexical cohesion from reference. While this approach is very close to the present one, here it is maintained that reference occurs only when there is a trigger item that cannot be interpreted on its own. The remaining paragraphs of this section attempt to combine Halliday and Hasan’s (1976) comparative reference and Biber et al.’s (1991) determiners and semi-determiners into clearly distinguishable categories of reference. This will finally lead to the final set of cohesive reference items that this study will work with. In their discussion of cohesive devices, Biber et al. (1991) use the term anaphoric expressions, implying that they are less concerned with the directionality of reference, and use anaphoric and cataphoric to describe both structural and textual cohesion interchangeably. Bearing this in mind, their corpus-based findings can still be used by adapting their categories for our present purposes.

 

Ellipsis is the type of reference that stands out, as it is not among the types of reference in Halliday and Hasan’s (1976) taxonomy; according to Biber et al. it “signals the closest type of connection in a referential chain” (Biber et al., 1991, 235). Although it is very tempting to include ellipsis in our analysis, as it is easy to accept it as a building block of a referential chain of cohesive ties, it would lead us to an unnecessary side-track and an infinite list of reference items. If ellipsis is included in the analysis, which is in fact substitution by zero, then substitution as a category should also be kept. This process is infinite and would lead us nowhere; nevertheless, this problem shows how central the notion of reference is in cohesion analysis. It also raises the question of whether it would be possible to cover the analysis of all the four cohesive categories – lexical cohesion, reference, substitution, and ellipsis – by one single analytical tool to gain a more comprehensive but perhaps less detailed and specific picture of the patterning of cohesion in texts. This, however, will remain a question for future research. The remaining three main types of co-reference above do not constitute a full list. Biber et al. (1991, 278-293) describe the range of uses, distribution and referential properties of certain determiner types with a referential property: quantifiers, numerals and semi-determiners.

 

 

B. Quantifiers specify the reference of nouns in terms of quantity, in that they “specify the number or the number of entities referred to” (Biber et al., 1991, 69), for example each person, all the students, some theories, many books, etc. (Biber et al., 1991, 275). While most of them are combined with one of the definite determiners, without quantifiers it would not always be possible to identify the exact reference of the NP. Besides, some of them are used on their own with the presupposed noun phrase partially or fully ellipted (which latter case it is not an instance of cohesive reference). As such, they form a strong tie with their presupposed expression, as in the example below:

 

Inclusive quantifiers (B.1) refer to groups in different ways. All is used to refer to a whole group and is frequently combined with a demonstrative determiner. Group members are referred to by each or every depending on whether the separate individuals are stressed (each) or the individual as a member of the group (every). Both refers to two entities with countable nominals (as in the example above). Large quantity (B.2) is mainly expressed by many or much and their comparative and superlative forms with an accompanying NP. The determiner some specifies moderate or small quantity (B.3). For the cohesive use of some with a specified noun, there is an example here from RA7: some in sentence 166 points back to participants in sentence 154.

 

Arbitrary/negative member or amount (B.4). Reference to fewer than two out of two entities is usually carried out by either or neither used anaphorically as determiner (examples from Biber et al. 1991, 258) as in Either extreme is possible. As for any and no, there were no cohesive instances in the RA corpus.

In sentence 75 in RA14 in our corpus, the author promises the reader a description of three possible explanations. In other words, the reader anticipates the three explanations, and on finding each of them, he or she will establish an anaphoric link to three above. One may argue that this is only anaphoric cohesion and there is no cataphoric reference taking place. However, when there are only two explanations provided and the third is missing (which is a typical mistake of novice writers, see Section 6.4), then this unfulfilled expectation proves that something went wrong in the text. If this anticipation is not satisfied, it will not be possible to recover all three explanations referred to. As discussed later in Chapter 6, an important aspect of the cohesive structure would be lost without discussing the cohesive aspect of numerals. Cardinals have a similar function in texts, but in the anaphoric direction, in fulfilling the expectations set by cataphoric numerals or in reference to items in lists or to sets of items. Cardinals always combine with the definite article, and in this sense, they further specify its reference, but do not form cohesive ties on their own. To restrict our analysis to the text-organizing function of numerals and avoid having too many reference items, the set of items will only include numerals from one to ten (and first to tenth).

Out of these four uses, only the first, classifying function can participate in cohesive ties.

 

 

The combination of these determiners is rather common. In the analysis of reference in this study, any combination of these items in one phrase will be counted as one instance of reference: e.g., all the, some of the, this same, the former, all four, etc. Finally, it seems we have attained our goal and arrived at a list of determiners that function cohesively. The main change that needs to be implemented to apply Halliday and Hasan’s (1976) original taxonomy is to narrow down the set of comparatives to a closed (or at least smaller) set of determiners with specific referential functions. This way, the taxonomy becomes more consistent, in that it uses the referential properties of noun specification as a criterion for distinguishing different types of reference, rather than semantic categories (such as: expressing identity, similarity, or particular comparison).