Coding scheme: REFERENCE
5.6 Results and discussion
5.6.2 Cohesive characteristics of abstracts. The frequency of the types of cohesive reference we found in the abstracts is shown in Table 15, where the
percentages indicate average numbers for the four analyses (2 analyses by 2 coders) out of the total number of cohesive ties of reference.
Type of cohesion Referring item Coding %
I. Pronominals 1 T: 6.9
singular, neuter functioning as non-possessive, as Head
it 13.6 6.5
plural functioning as possessive, as Deictic their 14.8 0.4
2. Demonstratives and definite article 2 T: 89.2
(1) demonstrative, near funct. as nominal, Deictic or Head
this/these 21.6 34.6
demonstrative, near funct. as place adverbial
here 21.7 3.5
(2) demonstrative, far funct. as nominal, Deictic or Head
that/those 22.6 0.4
demonstrative, far funct. as place adverbial
there 22.7 0.4
demonstrative, far funct. as time adverbial then 22.8 2.7
(3) definite article the 23 47.6
3. Comparatives (not complete list) 3 T: 3.9
(2) similarity funct. as deictic e.g., similar(ly), such 32.6 0.8 (3) difference funct. as deictic e.g., different, other, else 33.6 0.4 (4) comparison, quantity funct. as numerative e.g., more, less, as many;
ordinals
34.7 2.7
Table 15. Frequencies of referential ties 10 RA abstracts11
The most frequent types of ties are demonstratives (89.2%), with the definite article establishing almost half of the cohesive ties. Only two types of pronouns are present in the text, it contributing 6.5% to the number of ties, their to only 0.4%. This lack of
11 In Table 15: T = the total number for each main type, details under each; Coding is according to Halliday and Hasan’s (1976), summarized in Table 12.
115
pronouns is a characteristic of academic genres; therefore, this result is not surprising at all. Comparatives were mainly represented by quantifiers in the abstracts.
The 10 abstracts were, on average, 5.9 sentences long. The distances of ties are difficult to count as the method only allows us to list cohesive ties. Longer chains in this way are extremely difficult to follow; for it we need to count distances separately for each presupposed item, which is a messy and confusing procedure. In the abstracts there were on average 21 mediated (2.5/text) and 23 non-mediated (2.6/text) ties, but such analysis is obviously inapplicable in longer texts to reveal patterns of longer chains of reference.
5.6.3 Analytical problems concerning the types of reference. The numbers in the previous Section (5.6.2) will make more sense if we observe each type of tie where coders’ opinions differed. In this part of the paper I intend to summarize for the problematic items the reasons why they might be problematic and also the analytical problems we had during the analyses. The problems are collected from the think-aloud procedure and the intra- and inter-coder differences.
5.6.3.1 Titles. The analysis of cohesion starts with numbering the sentences successively, so the first question in our analysis is whether or not the title is part of the text. In our analysis we decided to number it as if it was a sentence, because later on in analyzing research articles we will have to handle section and subsection titles, which also form an inherent part of the text and need to be treated as such.
5.6.3.2 The definite article. As it is usually the case, the majority of the referring items were nominals with a definite article. Differences between the two coders
116
sometimes occurred as a result of different interpretations of grammatical and textual cohesion. In Abstract 1 below, for example, in sentence 6, the essays obviously refers to the three groups of essays in sentence 5. But about the other two items?
s. 5. The present study replicated the previous ESL study of two groups of essays written for the TOEFL Test of Written English with three groups of essays.
s. 6. Findings indicate that two topical structure variables, proportions of sequential and parallel topics in the essays, differentiate the highest group from the two lower rated groups.
(Abstract 1)
Strictly following the instructions in the taxonomy, the original presupposed item (here:
three groups of essays) is the presupposed item for all three definite NPs. Still, it is questionable, whether they point via the preceding item in the same sentence, or independently. The arrows in the two examples from Abstract 1 show these two interpretations. In the first interpretation, we only have one cohesive tie, whereas in the second, we have three ties.
Sample analysis A Sample analysis B
s. 5. The present study replicated the previous ESL study of two groups of essays written for the TOEFL Test of Written English with three groups of essays.
s. 6. Findings indicate that two topical structure variables, proportions of sequential and parallel topics in the essays, differentiate the highest group from the two lower rated groups.
s. 5. The present study replicated the previous ESL study of two groups of essays written for the TOEFL Test of Written English with three groups of essays.
s. 6. Findings indicate that two topical structure variables, proportions of sequential and parallel topics in the essays, differentiate the highest group from the two lower rated groups.
Text extract from Abstract 1
Another example is Abstract 5, where one of the coders coded the conversations as forming a cohesive tie with conversations in the previous sentence, while the other coder did not count it, because the presupposed item could also be a conversation in the same sentence, which would not be considered an instance of a cohesive tie.
117
s. 2. This paper reports on how speakers of Kiswahili, native and non-native, close conversations.
s. 3. In this paper I show that 1) closings in Kiswahili are quite elaborate and may extend to over five turns at talking, 2) an exchange of ’goodbyes’ does not usually signal the end of a conversation, 3) there is no strict ordering of features and 4) some closing features are linked to the opening part of the conversation.
(Abstract 5)
Sentence 3 is also a good example of a sentence where both markers had doubts about the sentence boundaries.
For the definite article, it is frequently difficult to decide whether it is cohesive or not, as its presence may be attributed to grammatical necessity. As in the following example, where the definite article was counted as a tie in only one of the analyses:
s. 1. Title: Multivariate approaches to contrastive rhetoric
s. 2. The purpose of this article is to describe applications of multivariate quantitative research approaches to contrastive rhetoric research.
...
s. 4. We will then present the rationale for the multivariate approach used here.
(Abstract 3)
While on the one hand, the noun is definite in sentence 4 because it is specified by the adjectival modifier that follows it, it is also definite because it points to the full form multivariate quantitative research approaches two sentences before it. This ambiguity has already been referred to in Section 3.2.2, where we mentioned an example of a fourfold reference. As regards the analysis, Halliday and Hasan (1976) do not give any practical guidance. This, and similar examples with modifiers caused many of the mismatches between the different analyses of abstracts.
5.6.3.3 Comparatives and conjunctions. As discussed in Section 3.2.3, comparatives are in a sense the odd one out, as the members of this category do not form a closed set; moreover, they are not likely to form chains of more than two items.
118
In addition, the analysis of abstracts showed that we need to be careful with the ambiguous interpretation of items such as different. If it has an indefinite meaning of various, it is not cohesive, but was still categorized as such in one instance:
s. 5. In this paper, we focus on one type of cohesive tie, conjunction, and compare its use in four different American English genres – fiction, journalism, religion, and science.
s. 6. Our results show that methods of conjunction in these genres vary in a statistically significant way and that conjunctions, although few in number of types and tokens, play a major role in structuring these different text types.
(Abstract 2)
Another problem was that in this analysis of abstracts we followed the original Cohesion Analysis where the description of the method does not specify whether in phrases with multiple referring items we should code it as one tie or two, and if the former, which type has precedence over the other and why. Consequently, many inter-coder mismatches were due to inconsistency in this respect as well.
Although they are not to be regarded referential, in many cases it is difficult to show a difference between conjunctions and reference. One such example in the present analysis is more recently (from Abstract 1, example below) which was marked as conjunction by one coder, and as comparative reference by the other who regarded as the presupposed item the time frame of the previous sentence (i.e.: recent) which is linguistically only realized by the present perfect tense of the sentence. Still, this disagreement again points to the need to clarify that it is essentially in nominal types of phrases where more establishes comparative chains.
s. 3. It has also been used to distinguish between essays written by groups of native English speakers with varying degrees of writing proficiency (Witte, 1983a, 1983b).
s. 4. More recently, TSA has distinguished between higher and lower rated ESL essays, but with different results from those found with native speakers of English (Connor & Schneider, 1988).
(Abstract 1)
119
The comparative items higher and lower compare two sets to each other, and are not cohesive. The link between sentences 3 and 4 above is exemplified by a lexical tie (essays – ESL essays) which we are not dealing with in this analysis. The conclusion from such mistakes is that when observing one phenomenon in connection with texts, it is not enough to give a description of what it is, but it has to be explicitly stated what it is not, by defining other uses a certain item may have and why it is not cohesive in those senses.
The taxonomy is not very clear about the difference between comparatives, demonstratives and conjunctions concerning items such as then or the second. These may establish referential links between sentences (demonstrative: then; comparative:
ordinal second), or may also be conjunctions that establish temporal, sequential relations (Halliday & Hasan, 1976, p. 266).
In the following example from our data the markers did not agree on the categories of the items in bold (in the example below) on the basis of the taxonomy.
s. 5. Two separate studies using a multivariate approach will then be described.
s. 6. One examines patterns of variation in English and Brazilian Portuguese newspaper editorials (Dantas-Whitney and Grabe to appear); the second examines the variations in writing among Ecuadorian Spanish and Anglo-American English university students (Lux 1991).
(Abstract 3)
While one is a pro-form here (substituting one study) and as such, it is an instance of ellipsis, one and the second together point back to two separate studies.
While we have already settled the question (in Chapter 3) that ellipsis will not be counted as an instance of reference in this sentence, the analysis of second part of sentence 6 is quite confusing. Both the definite article and the ordinal suggest reference;
however, without one reference to the full NP in sentence 5 would not be established. In any case, this example shows on the one hand, that in the description of the analytical
120
method we need to emphasize the difference between ellipsis and reference, and on the other hand, that in academic writing, it is a constantly recurring question whether or not to include nominal ellipsis in our analysis.
5.6.3.4 Cataphoric reference. The direction of reference was mainly a problem with the items that made reference to the whole paper (the (present) study, this/the paper, this/the article, here, this work). One of the coders interprets them as anaphorically referring to the title – she considered the title as the representation of the whole paper. The other did not include them as they are in a sense exophoric, pointing to the actual printed text. At this point, we can see the importance of the treatment of the title as different interpretations may lead to substantial differences in the analyses.
Resolving this issue here, we decided that while we accept that the reference of items referring to the paper itself are partly exophoric, these items also co-refer with each other and can in this way establish chains of reference pointing to a single entity. For our present analysis it is reasonable to regard them as ties of reference. Additional arguments for this approach will come in Chapter 6 of this dissertation, where, in the analysis of longer texts with a number of subsections, the orientation of the reader to various text segments in the co-text (that is, the surrounding textual context) will turn out to be an important aspect of texture.
5.6.3.5 Other errors. As cohesion is largely dependent on reader interpretation and the amount of effort put into the interpretation process, some mistakes are inevitable. In many cases, the “depth” of our understanding of the text changes from the first reading to the second reading, as a result, our accuracy in noticing cohesive items may change.
121 One such example is the following:
s. 4. More recently, TSA has distinguished between higher and lower rated ESL essays, but with different results from those found with native speakers of English (Connor & Schneider, 1988).
s. 5. The present study replicated the previous ESL study of two groups of essays written for the TOEFL Test of Written English with three groups of essays.
(Abstract 1)
The definite article in bold in the second sentence was identified as referring to the study described in the previous sentence only in one analysis. It is also ambiguous (between being referential and exophoric) and is easily overlooked, as the definite article may also be necessary for grammatical reasons (because of the possessive construction).
Another source of ambiguity is that one item may point to various directions at the same time. A typical example from our data is the results (3 instances in the Abstracts) which points both anaphorically to the summarized data in the abstract, and the results detailed in the article.
5.6.3.6 Reference by determiners versus lexical cohesion. We have already pointed to the overlap between the categories of reference and lexical cohesion. In Halliday and Hasan’s (1976) definition the difference between reiteration (a form of lexical cohesion) and reference as a type of grammatical cohesion is not at all straightforward. Reiteration in their definition is
“…the fact that one lexical item refers back to another, to which it is related by having a common referent. [...] A reiterated item may be a repetition, a synonym or near-synonym, a superordinate, or a general word;
and in most cases it is accompanied by a reference item, typically the.”(ibid., p. 278).
122
They discuss this issue from the perspective of lexical cohesion, only mentioning the reference item as if it could establish a cohesive reference tie on its own.
Looking at it from the perspective of analyzing reference, we cannot assume that the definite article in itself (having the property of definiteness, but no meaning on its own) could point to another noun. Obviously, the reference item (for pointing elsewhere in the text), whether it is a determiner or a comparative and the lexical item (for specifying the item pointed to) are both necessary for us to be able to identify a cohesive tie. Hasan later also admits that “the separation of lexical and grammatical cohesive chains quite clearly did violence to certain aspects of the text’s semantic organization” (1984, p.
200). For example, “pronouns continue as subsequent mentions of an introducing noun phrase until there is a shift to a new discourse segment, when the full noun phrase surfaces again” (McCarthy, 1994, p. 270) in other words, that is when the full noun phrase re-enters the discourse. As reference to the same entity is maintained with this noun phrase, an analysis of reference will also have to take them into consideration.
Concerning the relationships between lexical items we will build on Károly’s (2002) and Hasan’s (1984) categories of lexical repetition, with the limitation that we will only refer to noun phrases, and not to other word classes. (The invented examples in Table 16 are slightly shortened sentences from our RA corpus). Károly’s (2002) study of lexical repetition in text (building on Halliday and Hasan’s (1976), Hoey’s (1991) and Biber et al.’s (1991, p. 265)) describes anaphoric referential relationships where the lexical item is not a synonym or an exact repetition as indirect anaphoric reference (e.g.:
the car ← the handbrake, the windscreen).
123