The purpose of this research is to deepen our understanding of information flow in Taiwanese English learners’ compositions. Intact samples of compositions written for the General English Proficiency Test (GEPT) were obtained from The Language Training & Testing Center (LTTC) of Taiwan. Both quantitative and qualitative methods were conducted to investigate the proportion of different TP types in higher-rated and lower-higher-rated essays and the quality of TP.
The Corpus
The corpus of the present study was comprised of essays from the Guided Writing section of high-intermediate level General English Proficiency Test (GEPT), which was designed by The Language Training and Testing Center (LTTC) in Taiwan. GEPT came into existence in 1997, sponsored and supported by our Ministry of Education. It has five levels of tests, including (1) basic, (2) intermediate, (3) high intermediate, (4) high, and (5) superior level. In each level, all four skills are examined in the test. The results of the test can serve as convincing reference of examinees’ English proficiency, and it is widely accepted by schools, universities, and government agencies. The data of GEPT writing tests is worth exploring since the test takers are not limited to students but a wider range of people, such as businesspersons, engineers, researchers, customs officers, tour guides and so on. In this light, the general problems of Taiwanese learners of different backgrounds can be targeted. Unlike classroom environments or self-made tests, GEPT writing test is administered in a strictly controlled environment, and each tests is designed through a rigorous progress. The writings from the tests, therefore, are more reliable and can be more representative of students’ actual performance. GEPT is also a leveled test, which means that the examinees who take the same level of test might be more homogeneous in English proficiency. This feature can in turn narrows
33
the gap between all test takers at one certain level. In the present study, the corpus was made up of essays obtained from the Guided Writing section of high-intermediate level.
It was claimed by LTTC that examinees who passed this test may have a generally effective command of English and could deal with a wide range of topics; however, this study only included the expository essays from the test, so the genres of other types were not addressed.
The essays from the Guided Writing section of high-intermediate GEPT were rated holistically, and the rating scheme ranged from 1 (poor) to 5 (excellent).
According to LTTC, the essays rated above level 3 generally answer to the rubrics and were written with proper content, fluent structure, and fine command over grammar and vocabulary; those below level 2 could not conform to the requirements of the rubric and the content could barely be comprehensible to readers; level 1 essays particularly failed to reach the requirement of the rubrics and manifested a large amount of errors in organization and grammar. Therefore, level 1 was excluded since those essays were informed to be too short in length and manifested too many basic or local errors, making it difficult to study cohesion at discourse level. Based on the rating scheme, the researcher thus only applied for essays whose score ranged from 2 to 5, and 240 writings authorized by LTTC were provided.
After receiving the data, a pilot test was conducted on a 20 essays, 10 of which were from level 2 and the other 10 from level 4. The results revealed that level 2 essays were not proper for the analysis of this study because many of them did not reach the length requirement of the rubrics (150 words), and the writers’ command over the language was not effective. Some of the writings were off-topic, and their contents were mostly incomprehensible. Therefore, the 39 essays from level 2 were excluded.
34
To compare between higher-rated and lower-rated essays, 50 essays from level 3 were selected as the lower-rated group, and another 50 essays from level 4 composed the higher-rated group to exemplify difference, if present. The criteria for selecting these 100 essays was whether they conform to the word length requirement of the rubrics (at least 150 words); namely, essays with inadequate number of words were excluded. For the second research question, the essays were discussed from a
qualitative perspective, so all 106 essays from level 3 (50 essays), level 4 (50 essays), and level 5 (6 essays) were included for text analysis.
The total word count of higher-rated essays was 12924 words, and the total word count of lower-rated was 12352 words, which indicated that two groups of essays were similar and comparable. The total number of clauses in higher-rated essays was 839, and that of lower- rated essays is 813, which is slightly fewer than the higher- rated essays. Table 2 is the overview of the corpus:
Table 2
Note. Higher-rated essays included both level 4 and level 5, but only those from level 4 were quantitatively compared with lower-level essays.
The Writing Task
In the test, each essay was written on the same topic. Examinees would read the guidelines first, which included the topic and two main requirements. They were
35
required to compose an expository essay on their view points toward “talent show,”
and their essays should involve the two aspects: (1) specific reasons why young people participate in the shows and (2) possible consequences for attending the shows. The required length of the writing is 150-180 words; the genre of this writing was expository in nature that students were expected to elaborate their viewpoints on a topic with their arguments.
Approaches to Analysis
In this section, the procedure to analyze TP in the corpus was explained. The framework for identifying Theme, Rheme, and the markedness of Theme was presented first; then, the way of identifying TP was explained.
Before quantitative analysis, each essay was broken down into “clauses,” which was the basic unit of meaning as defined in Halliday and Husan’s (1976) model. In the next step, the Theme and Rheme of each clause were specified based on Halliday and Matthiessen’s (2013) approach. They suggested that each clause can be divided into a Theme and a Rheme, and the former always precedes the latter. In declarative clauses, for example, Theme was identified as the element(s) appearing in front of the main verb, which may be the Subject and sometimes adjuncts or conjunctive phrases.
However, although Theme might sometimes be manifested by an adverbial group or a prepositional phrase, some researchers argued that the Subject of the clause should be included in Theme when conducting Theme-Rheme analysis (e.g., Green,
Christopher, & Mei, 2000; Lin, 2011). This study also adopted this approach, so the boundary of Theme and Rheme is identified as the following:
(1) In declaratives, any element before the main verb of the main clause is the Theme; the remainder of the clause including the main verb is the Rheme (Green, Christopher, & Mei, 2000).
36
(2) In Yes/No interrogative, the finite operator and the Subject belong to the Theme; in Wh- interrogatives, the WH- element is the Themes. The remainder of the clause it the Rheme (Halliday and Matthiessen, 2013).
(3) In imperatives, the predicator (sometimes “let’s”) is the Theme; in negative conditions, the negation is also included in the Theme. The remainder of the clause is the Rheme (Halliday and Matthiessen, 2013).
The above three rules were used to divide a single clause into Theme and Rheme, but for clause complexes, their thematic structure was analyzed differently based on the type of clause complexes. The following explained how Theme and Rheme were identified in two types of clause complexes in the present study:
(1) Paratactic clause complexes: in this kind of clause complexes, there were two independent clauses which were linked by conjunctions (e.g., for, and, nor, but, or, yet, so). This study adopted Eggins’ (2004) model that each of the clauses should be treated separately, which meant that the Theme and Rheme of each independent clause were identified. In example (a) below, the Theme of the first independent clause is Teenagers, and the Theme of the following independent clause is but their parents:
(a) Teenagers love talent shows, but their parents do not.
Theme1 Rheme1 Theme2 Rheme2
(2) Hypotactic clause complexes: in this kind of clause complexes, there were one dependent clause and one independent (main) clause, and these two clauses can appear in two orders: the main clause preceding the dependent clause or the dependent clause preceding the main clause. The present study treated any element before the main verb of the main clause as the Theme of the clause complexes (Green, Christopher, & Mei, 2000), and the remainder
37
of the clause would be the Rheme. Therefore, the two clauses were not analyzed independently as in paratactic clause complexes, but treated as a whole. In this way, writers’ choice to place dependent clause as marked Theme and how this tendency contributed to cohesive development could be exemplified. In example (b) below, the dependent clause preceded the main clause, and the Theme of this clause complex would involve the dependent clause and the Subject of the main clause (If teenagers have adequate time, they); the Rheme of this clause complex is can attend the shows:
(b) If teenagers have adequate time, they can attend the shows.
Theme Rheme
However, if the main clause preceded the dependent clause as shown in example (c), the Theme would only include Teenagers, which appeared before the main verb of the main clause. The remainder of the clause complex, including the dependant clause, would be the Rheme:
(c) Teenagers can attend the shows if they have adequate time.
Theme Rheme
After the Theme of each clause or clause complexes were identified, the markedness of each Theme were analyzed because marked Theme were used strategically by writers for certain reasons, and they were related to the cohesive development of succeeding clauses (Eggins, 2004; Martin & Rose, 2007). The identification of the markedness of Theme followed the framework of Halliday and Matthiessen (2013), and it is described as the following:
(1) In declaratives, if the Theme does not play the role of the Subject of the main clause, it is marked. For example, time adverbs and adjunctive adverbs are marked Themes.
38
(2) In interrogatives, if the Theme is not the finite operator and the Subject (as in Yes/No interrogative) or the WH- element (as in Wh- interrogatives), it is marked. For example, in sentence “on the left is it?” the Theme “on the left”
is marked.
(3) In imperatives, if the Theme is not the predicator (or “let’s”), it is marked.
For example, in sentence “don’t you tell me the truth,” the Theme “don’t you” is marked.
Quantitative Analysis of the Proportion of TP Types
To answer research question one, after all Themes and Rhemes were identified in each clause, the TPs across clauses were then specified and analyzed. Five types of TP were located in each writing, which were based on Danes’ (1976) model and other studies (Chao, 2002; Crompton, 2004; Hawes & Thomas, 1996, 1997). These five TPs included: (1) Simple linear TP, (2) TP with a constant Theme, (3) TP with a derived Theme, (4) Exposition of split Rheme, and (5) Rhematic Progression. In some TPs, however, there might be a “gap” in the middle of TP. In the present study, the analysis followed principles in previous studies (Crompton, 2004; Ebrahimi & Ebrahimi, 2012a; McCabe, 1999) that the referent of a Theme cannot be “two clauses” further from the Theme. That is, the referent of the Theme must be in the previous clause or the “previous but one” clause (Crompton, 2004); otherwise, the connection would not be counted as a link but a “break” (Ebrahimi & Ebrahimi, 2012a; Hawes & Thomas, 1997; Jalilifar 2010).
The procedure of identifying TP was adapted from Crompton’s (2004), including two rounds. In the first round, the four major types of TP proposed by Danes (1976) were identified first, including Constant TP, Linear TP, Derived TP, and Split-rheme TP. In the second round, Constant TP and Linear TP which include one irrelevant
39
clause in the middle of the progression were identified. Rhematic Progressions were also identified in the second round instead of the first because it links information from Rheme to Rheme, which was not a conventional type of progression. After all these TPs were specified, the clauses that did not fit into our TP typology or TP with a gap of more than two or more clauses between the Theme and its referent were counted as “breaks,” and these “breaks” were calculated apart from the TPs.
The definition to each of these TPs and the procedure in analyzing TPs were provided in Table 3 below:
40 Table 3
Procedure in Analyzing TP (adapted from Crompton, 2004, p. 235) Step TP types Criteria
First Round
1 Constant TP Theme refers to an item also referred to in the Theme of the previous clause
2 Linear TP (a) Theme refers to an item also referred to in the Rheme of the previous clause
(b) Theme refers to the whole of the previous clause
3 Derived TP Theme is derived from a hyperTheme 4 Split-rheme
TP
Rheme of the first clause is split into two or more ideas, and these ideas are then developed in the Themes of subsequent clauses
Second Round
5 Constant TP Theme refers to an item also referred to in the Theme of the “previous but one” clause
6 Linear TP (c) Theme refers to an item also referred to in the Rheme of the “previous but one” clause (d) Theme refers to the whole of “the previous
clause but one”
7 Rhematic Progression
Rheme refers to an item also referred to in the Rheme of the previous clause or the previous but one clause
8 “Break” (a) The gap between the Theme and its referent is larger than “two clauses or more”
(b) TPs which cannot fit the TP typology
41 Qualitative Analysis of the Deployment of TP
To answer research question three, which focuses on the features of learners’
essays in terms of the deployment of TP, marked Themes, and breaks from a qualitative perspective, a data-driven analysis were conducted to reveal distinctive characteristics. A total of 106 essays from level 3 to level 5 were compiled for iterative rounds of qualitative analysis.
The procedure of the qualitative analysis is as follows: The original text of both levels were compiled in two text files. The TPs in each essay were presented in tables for analysis and observation so that how writers organized their information and connected their ideas with TP can be revealed. In the first round, how TP manifested in lower-rated essays were examined, and some specific features of the patterns were seen. These emergent patterns were recorded and listed. Then, in the second round, the TP in higher-rated essays were explored, and the features of TP these writings were also recorded to allow for comparison between the two levels. By comparing these features from the two levels, it could be speculated that some patterns were exclusive to lower-rated essays, and these may be consider problematic features of lower-rated essays. For example, after viewing the TPs in lower-rated essays, it was found than constant TPs appeared repetitively in some of these essays, while this problem was not seen in higher-rated essays. The comparison specified that a sequence of constant TPs were distinctive in lower-rated essays. Through this recursive process, a few features the quality of TP were indicated and compiled.
Since this study adopted a data-driven approach, the procedure includes iterative rounds of analysis. In the end, all the articles were reviewed for three to four times in case some detailed were overlooked during the previous observation.
42