Noun Collocations

0 downloads 0 Views 709KB Size Report
language teaching; the second language (L2) vocabulary knowledge includes chunks, ... Previous studies have indicated that advanced English as a foreign.
Computer Assisted Language Learning Vol. 18, No. 3, July 2005, pp. 231 – 250

Effects of Web-based Concordancing Instruction on EFL Students’ Learning of Verb – Noun Collocations Tun-pei Chan and Hsien-Chin Liou* National Tsing Hua University, Taiwan

This study investigates the influence of using five web-based practice units on English verb-noun collocations with the design of a web-based Chinese-English bilingual concordancer (keyword retrieval program) on collocation learning. Thirty-two college EFL students participated by taking a pre-test and two post-tests, and responding to a background questionnaire and an evaluation questionnaire. Results indicated that learners made significant collocation improvement immediately after the online practice but regressed later. Yet, the final performance was still better than students’ entry level. Different verb-noun collocation types and learners with different prior collocation knowledge were found to be not equally receptive to the practice effects. Both the online instructional units and the concordancer were acceptable to most participants.

Introduction Recently an interest in vocabulary learning has emerged in the field of foreign language teaching; the second language (L2) vocabulary knowledge includes chunks, such as collocations or idioms (see Nation, 2001). Collocation is the natural cooccurrence of a string of words whose meaning is inferable from literal concepts. Collocations are conventional (frequent, psychologically salient), semantically analysable (unlike genuine idioms), and unpredictable sub-sets of larger sets of logically and semantically possible collocations (Benson, Benson, & Ilson, 1986a; Nattinger & DeCarrico, 1992). L2 learners might not easily obtain the correct collocation if their inferencing process is based on their first language (L1) patterns. Their miscollocated chunks can become less acceptable when one of the components is substituted by another similar word close to the L1 equivalent conceptually *Corresponding author. 101 Sec 2 Kuang Fu Rd, foreign languages, Tsing Hua University, Hsinchu, Taiwan ROC 30043. Email: [email protected] ISSN 0958-8221 (print)/ISSN 1744-3210 (online)/05/030231–20 ª 2005 Taylor & Francis Group Ltd DOI: 10.1080/09588220500185769

232

T. Chan and H. Liou

(Benson, Benson, & Ilson, 1986b; Nattinger & DeCarrico, 1992). For instance, ‘‘take medicine’’ is a common mistake for English learners with a Chinese background because their first language word sense tends to lead to a corresponding equivalent ‘eat medicine’. Previous studies have indicated that advanced English as a foreign language (EFL) learners had common problems with collocations (see Bahns & Eldaw, 1993; Farghal & Obiedat, 1995; Nesselhauf, 2003). Since advanced language production relies heavily on a great amount of such pre-made items in memory (Howarth, 1998; Schmitt, 2004; Wray, 2002), language learning teachers have started to exert efforts on this component of foreign language knowledge (Johns, 1991; Lewis, 2000; Nation, 2001). Collocation Competence for EFL Learning To probe into learners’ collocation competence, Bahns and Eldaw (1993) have indicated that German advanced EFL college students had problems with productive knowledge of English collocations. Similarly, Nesselhauf (2003) analysed the writing of German advanced learners of English and found that the most frequent miscollocation type is wrong choices of verbs. He explained that the verb in a collocation has a restricted sense, which makes its correct use more difficult if learners cannot fully distinguish subtle differences among verb collocates. The collocations not congruent in learners’ first language (L1) and second language (L2) were far more difficult for learners to acquire. Liu (1999) investigated collocational errors in 127 Chinese college students’ final examination papers and their 94 compositions. The error analysis revealed that verb – noun collocation errors (verb + noun/pronoun, verb + propositional phrase) appeared most frequently. With the help of computer technology, Liu (2002) examined V – N miscollocations in Chinese learners’ essays through lexical semantic investigation. She indicated that 87% of the lexical miscollocations (233/265) were attributed to V – N miscollocations and 93% of them were due to the misuse of verb collocates. As for reasons behind miscollocations, 56% of missed collocations were semantically related such as synonyms (e.g. *carry out my goal instead of achieve my goal), hypernyms (e.g. *create songs instead of compose songs), and troponyms (e.g. *break the foundation instead of damage the foundation). Another 38% of the V – N miscollocations were traceable to L1 interference: split category (e.g. add and increase) and direct translation (e.g. *write homework instead of do homework). Among various types of collocations, the verb – noun lexical collocation was found to be particularly difficult for learners to acquire; further, V – N miscollocations can be attributed to three main reasons: (a) L1 interference; (b) misuse of de-lexicalised verbs; and (c) lack of knowledge of collocational restrictions in semantically related lexemes such as synonyms, hypernyms, and troponyms. Recognising the importance of collocation for foreign language learning, some researchers provided explicit collocation instruction on Chinese EFL learners in classroom settings, and found positive effects (Lin, 2002; Liu, 2002; Tseng, 2002). Liu (2000) has found that after collocation instruction, learners produced more

Effects of Concordancing on Collocation

233

varieties of collocations in their writing, particularly in de-lexicalised verbs. In a traditional classroom context, Lin (2002) examined the effects of collocation teaching on receptive and productive collocation competence of high-achievers and lowachievers in a group of EFL high school students regarding the V – N lexical collocation structures. The results indicated that all students made more progress in receptive collocation tests than productive ones, but low-achievers performed better in productive tests after collocation teaching. Both groups held positive attitudes toward collocation teaching activities. Tseng (2002) divided 94 high school participants into an experimental group, who received 12 weeks of explicit collocation instruction, and a control group, who did not receive any training. Before the instruction, the students took a pre-test on collocation, wrote a composition, and filled in a background questionnaire about vocabulary learning behaviors. The results in questionnaires indicated that students knew little about the concept of collocation. After collocation instruction, the experimental group far exceeded the control group in the post-test regardless of their prior collocation levels. The experimental group’s performance was found to have no significant difference across the six collocation types they investigated, while at the beginning of the collocation teaching they found de-lexicalised verb collocations easier to produce. From studies discussed above, it is clear that collocations as one important type of knowledge in foreign language have posed learning difficulties to EFL learners; specifically for those from a Chinese background. Moreover, types of collocations seem to behave differently under the influence of collocation instructions. Concordancing for Learning of Collocations Traditionally, collocation instruction was mostly conducted in a classroom-based context where teachers took full responsibility. Although inductive learning of collocation patterns with the help of a great number of examples in order to draw regularities has been recommended, without a computer concordance, such teaching methods are time-consuming and thus very demanding for teachers, if not impossible. A concordancer is a sophisticated computer retrieval program with a large amount of information in the form of computer language corpora accessible to encourage data-based inductive learning. Such a program can display many examples of a key word or phrase and thus enable learners to examine usable concentrated data from a great deal of authentic language. Concordancers can be made as monolingual English programs, bilingual ones, or even multilingual ones. With bilingual concordancers, contrastive contexts of language use examples are more easily accessible (Wang, 2001). Specifically for the learning of collocation items, few studies have explored the possibility of utilising concordancers to help EFL college students except the one by Sun and Wang (2003). Sun and Wang examined the effectiveness of inductive and deductive teaching approaches when EFL learners were learning collocations of different difficulty levels with the help of an online monolingual concordancer. Four collocation patterns were selected based on two experienced EFL experts’

234

T. Chan and H. Liou

difficulty judgments. Eighty-one senior high students were randomly divided into two groups. The two groups used two corresponding online exercise versions designed with the deductive or the inductive approach respectively. The deductive group was given the rule explanations and example sentences while the inductive group had to induce the patterns independently from the concordances. A pretest, online exercises, and a post-test were given with the design of error correction items. Both groups took the pre-test, a one-hour instruction section, and the post-test. Post-test results indicated that overall the inductive group improved significantly more than the deductive group. Easier collocations were prone to be more suitable for the inductive approach with the help of concordancing, whereas there were no significant differences between students’ performance influenced by inductive or deductive methods with respect to difficult collocations. The way Sun and Wang distinguished their four collocation phrases into easy or difficult categories was arbitrary—whether the difficulty level was determined by students, teachers, or even linguists was unknown. With a sample of four collocations under their investigation, the scope of linguistic phenomena was extremely limited. Besides, like other studies, they did not pay enough attention to the dominant EFL weakness, V – N miscollocations. Whether the instructional effects fade out for a certain period of time after the collocation teaching ends was not examined. Further, few studies have examined the surplus value which the L1 translation in a bilingual concordancer may bring. Lastly, learners’ inherent preference for the deductive or inductive learning style was not considered. In sum, few studies on collocation learning have made full potential of new computer bilingual or multilingual concordancers. The current study therefore aims at investigating Chinese EFL college students’ learning of well sampled verb – noun (V – N) collocations with the help of computer-assisted language learning (CALL) that has the design of online practice and the incorporation of a web-based bilingual concordancer as an online reference tool. Such a learning process is named as concordancing. The study probes into the factors of individual variations of treatment (with concordancing or not), different collocation types, and learner prior knowledge, in additional to the main teaching effects and retention, in order to bridge the gaps pointed out from the survey of previous research above. Three main research questions are addressed in the study: 1.

Can CALL collocation practice with a bilingual concordancer improve students’ verb – noun collocation knowledge and performance? If so, do the effects retain after a certain period of time? 2. Do different collocation types, teaching methods, and students at different collocation levels have an impact on the learning effects? 3. What is students’ feedback on the online collocation practice units and the bilingual concordancer? To which extent do the bilingual examples in the concordancer help their learning? Does students’ appreciation of the practice units have relations with the learning effects?

Effects of Concordancing on Collocation

235

Method A one-group pre-test post-test experiment was designed to examine whether and in what way a Chinese – English bilingual concordancer can assist EFL students’ learning of verb – noun collocations. Instructional Design of Five Online Units on V-N Collocations Previous literature has indicated that V – N lexical collocations are the dominant EFL learners’ weakness and that the verb collocate is often misused by EFL learners (Liu, 2002; Nesselhauf, 2003). The design of the online materials was based on these two considerations. In the format of textual explanation, semantic grid analysis, bilingual concordancer, and interactive exercises with voice reading of online information (to motivate learners in the learning process), five web-based units were designed. Three units were taught with the incorporation of a bilingual concordancer; two units were taught without the use of the concordancer. In the three concordancing units, a web-based bilingual concordancer, TOTALrecall, (developed in Liou, Chang, & Yeh et al., 2003, http://candle.cs.nthu.edu. tw/totalRecall/totalRecall/totalRecall.aspx) was used as the referential tool with which students were asked to consult appropriate verb collocates while doing the online units (see Figure 1). Learners could read a great number of examples in concordances and learn how to induce language patterns from them. V – N collocations such as English de-lexicalised verbs which tend to be semantically

Figure 1. The display of bilingual concordances

236

T. Chan and H. Liou

unmotivated, and non-congruent collocations between two languages which are prone to native speakers’ intuitions were taught in these units, since concordances can provide a richer amount of repetitive authentic data on a particular use and help learners develop their own intuition (see Figure 2 for a sample of online units). In the other two units taught without the concordancer, pattern explanations were provided, followed by several examples and exercises; for example, the grid of semantic field analyses was employed to give a systematical analysis on semantic discrimination of synonymous verbs (see Figure 3). V – N collocations associated with subtle semantic discriminations were taught in this manner such as synonyms, hypernyms, and troponyms. Since their fine differences might not be easy to distinguish, the explanation of patterns may facilitate learners’ understanding of collocation usage. The conconcordancing and non-concordancing methods are called treatment effects in this study. An online record-keeping computer program was used for tracking students’ learning interactions such as the words learners looked up in the concordancer, the number of times each word was looked up, and learners’ answers to the five online units The interactive exercise types in the five online units included multiple choice, Chinese – English sentence translation, and gap-filling sentences. Each unit contained about twenty questions and took about one hour for students to complete. The incidents in the five units were over 100 common V – N miscollocations chosen from previous research on Chinese college students’ production and manual analysis of local students’ compositions. The five online units covered the content of four common error types: (a) synonymous verbs (e.g. construct/build/establish); (b) hypernymy (e.g. create/compose) and troponymy (e.g. break/damage) verbs; (c) de-

Figure 2. The cloze task asking learners to find appropriate verbs from the concordancer

Effects of Concordancing on Collocation

237

Figure 3. The semantic field grid for synonymous verbs

lexicalised verbs (e.g. make, do); and (d) English V – N collocations in lack of translation equivalents in Chinese (e.g. brew tea, ‘‘pao cha’’ in Chinese). Participants and Research Instruments Thirty-two non-English major college freshman students in Taiwan participated in the study. They were taking a required course, Freshmen English, four hours per week about one month after they were admitted into the university from various high schools. Two types of instruments were designed for the study: three tests and two questionnaires. First, three sets of tests were used for the pre-test, immediate posttest, and delayed post-test respectively with 36 items designed to measure students’ V – N lexical collocation knowledge at different time points. The items in the three collocation tests were identical except that the sequence of presentation was rearranged. The tested items were in the fill-in-the-blank format with the first letter of the verb provided as illustrated below. He a_________ great success and became the leading landscape architect of the day. [answer: achieved]

Among the 36 items in the three collocation tests, 26 items were sampled from the collocations taught in the five online units so as to assess whether students can acquire the target V – N collocations after being taught. Another 10 items consisted of collocations which were not included in the five units so as to evaluate whether effects of the online units can be transferred to other V – N incidents through awareness-raising by means of exposure. In terms of the 26 items designed in the three tests, 16 items were sampled from collocations taught in the three concordancing units and 10 items were chosen from the two nonconcordancing units. Besides, the content of the 36 items was sampled from the four types of collocations: synonymous verbs (VN1, 10 items), hypernymy and troponymy verbs (VN2, 8 items), de-lexicalised verbs (VN3, 8 items), and V – N collocations that are non-congruent between Chinese and English (VN4, 10 items).

238

T. Chan and H. Liou

Secondly, two questionnaires in Chinese were designed to survey students’ perception before and after the online practice: a background questionnaire and an evaluation questionnaire. The Background Questionnaire with 14 items was designed to probe into learners’ vocabulary learning behaviors, and their preferences for the deductive or inductive teaching methods. The Evaluation Questionnaire with 33 items was used to investigate students’ perceptions in using online practice sections (see an example below), including their attitudes toward online collocation units, the web-based bilingual concordancer, and the concordancing and non-concordancing methods. Most items were designed with a 5-point Likert preference scale (ranging from ‘‘like it a lot’’, 5 point, to ‘‘dislike it very much’’, 1 point, see an example below). Example of an item from the Evaluation Questionnaire: My feelings about learning English V – N collocations after using the units which incorporate the use of the bilingual concordancer. 5—I like it a lot. 4—I like it. 3—I have no preference. 2—I dislike it. 1—I dislike it strongly.

The procedures were as follows. First, students were given the background questionnaire and took the pre-test in order for researchers to determine their entry collocation knowledge. In the instructional period, students were required to do various online tasks designed in the five units either in-class or after class. The five units were used across five weeks infused with regular classroom teaching. Last, students filled in the evaluation questionnaire and took the immediate post-test. About two and a half months later, students took the delayed post-test. Results The results of the current study are presented in three parts: (a) learners’ performance of collocation knowledge and collocation awareness at three time points (operationalised by scores of the pre-test, the post-test, and the delayed post-test); (b) individual linguistic, treatment, and learner differences versus variations of learning effects; and (c) learners’ appreciation of various aspects of the online learning process as related to the effects. Overall Learning Effects of Online Instruction at Three Time Points The paired t-test comparison of the post-test scores and pre-test scores indicated students’ significant gains in collocation knowledge (p = .000 5 .05, see Table 1).

Effects of Concordancing on Collocation

239

Table 1. Comparison of pre-test and post-test scores for instructional effects

The pre-test overall score The post-test overall score

N

Mean

S.D.

T

Df

Sig.

32 32

10.59 19.53

3.26 3.95

7 14.880

31

.000*

Note. * p 5 .05.

Students’ overall V – N collocation knowledge increased and reached the statistical significance level after learning from the five online units. Students’ awareness of collocation in the three tests (operationalised as performance on the 10 non-taught items) given at different time points were compared. The results showed that students’ awareness was significantly raised in the immediate post-test (p = .000 5 . 05) and their collocation awareness retained in the delayed post-test after two and a half months (p = .309, ns, comparison of the post-test and delayed post-test scores). Items in the tests seem very challenging as in the pre-test learners obtained an average of 10 out of a total of 36. Though they did not push up their performance to a great extent after the online instruction, such a significant difference (about 9-point increase) indicated learners have benefited from the instruction. To understand the ‘‘retention’’ (defined as the score difference between the posttest and the delayed post-test) and ‘‘residual’’ effects (the score difference between the pre-test and the delayed post-test) of CALL collocation instruction after almost four months from the beginning of the project, the results of comparisons with the delayed post-test were analysed. These closer examinations are to investigate how much learners still remembered after the instruction ends for two and a half months (the retention effects). If they do forget, do they regress to the extent that their collocation knowledge level is lower than that when they start in the beginning of the project (the residual effects of the learning)? It was found that students’ overall degree of retention of collocations between the immediate post-test and the delayed post-test decreased significantly (p = .000 5 .05). Yet, residual effects of the overall collocation knowledge stayed: students’ delayed post-test scores were significantly higher than those of their pre-test scores (p = .000 5 .05, see Table 2). Learners regressed two and a half months after the online instruction ended, but the performance was still better than the entry knowledge when they first started to learn about collocations from the online units. Individual Linguistic, Treatment, and Learner Differences Versus Learning Effects Thirty-six verb – noun collocations in the three tests were carefully chosen and categorised into four types according to their linguistic differences: (1) synonymous verb pairs; (2) hypernymy and troponymy verbs pairs; (3) de-lexicalised verb pairs; and (4) verb pairs that are non-congruent in Chinese and English. These are called individual linguistic differences in the online content sampled for testing on learners.

240

T. Chan and H. Liou

Table 2. The paired t-test comparison for retention of overall collocation knowledge between the pre-test and the delayed post-test

The pre-test overall score The delayed post-test overall score

N

Mean

S.D.

T

Df

Sig.

32 32

10.59 15.47

3.26 4.09

7 9.119

31

.000*

Note. * p 5 .05; Total score = 36.

Students’ sub-scores of the four different collocation types were divided by the total item numbers involved in each category respectively to examine whether they behaved differently. The repeated measure test results showed that there were significant differences among the mean scores of the four types in the pre-test (F [3, 36] = 11. 393, p = .000 5 .05). To factor out the beginning differences of the four types, gain scores were used for comparison of their learning effects. The differences of their gain scores (the post-test scores subtracted by their pre-test scores) were also found to be significantly different (F [3, 93] = 17.722, p = .000 5 .05, see Table 3). The post hoc comparisons using the Least Significant Difference test discovered that the highest gain scores of the VN3 (de-lexicalised verbs) were significantly different from those of VN1 and VN2 (VN1 stands for synonymous verb pairs; VN2, for hypernymy and troponymy verbs pairs). The second highest gain scores of VN4 (collocations that are non-congruent in Chinese and English) were also significantly different from those of VN1 and VN2 (all p values were smaller than 0.05). These showed that students made more progress in de-lexicalised verbs and non-congruent collocations between Chinese and English after learning from the online collocation units. Further comparisons of students’ retention of the four different VN types showed that mean changes of collocations in lack of direct translations were lowest and significantly different from the other three VN types (VN4 v. VN1, p = .000 5 .05, VN4 v.VN2, p = .002 5 .05, VN4 v. VN3, p = .025 5 .05). This indicates that learning effects of VN4 can maintain better, if the criterion is to see the different performance between the measurement taken immediately after the online instruction and that two and a half months later. Yet, if the criterion is to see the different performance between the measurement taken in the very beginning and that four months later, the findings are slightly different. Comparisons of students’ residual effects revealed that the mean changes of collocations with de-lexicalised verbs (VN3) were the highest followed by VN4, while the mean changes of collocations involving hypernyms and troponyms (VN2) were the lowest and VN1 the second lowest (i.e. VN3 4 VN4 4 VN1 4 VN2 and ‘‘ 4 ’’ means ‘‘smaller than’’). Collocations with de-lexicalised verbs (VN3) and collocations in lack of direct translations in two languages (VN4) may be more receptive to CALL collocation instruction; students made more progress in VN3 and VN4, and students’ residual effects of VN3 and VN4 were also better.

Effects of Concordancing on Collocation

241

Table 3. The repeated measure statistics for the gain scores of the 4VN types between the post-test and the pre-test The mean score N

VN1

VN2

VN3

VN4

Sum of squares

Df

Mean square

F

Sig.

32

.1469

.1055

.3828

.3563

1.934

3

.645

17.722

.000*

Note. * p 5 .05; VN1 stands for synonymous verb pairs, VN2 stands for hypernymy and troponymy verbs pairs,VN3 stands for de-lexicalized verbs, and VN4 stands for V – N collocations that are noncongruent in Chinese and English.

It is likely that reading ample concrete examples of de-lexicalised verbs in concordances and mutual translations in the bilingual concordancer have helped students master de-lexicalised verbs and collocations in lack of direct translations in two languages. With regard to the different effectiveness of the two teaching methods (inductive and deductive methods perhaps drawing from their past English learning experiences in high school), students did not show significant preference for either one of the methods, based on the analyses of the background questionnaire items. From the tracker program’s records, the 32 students looked up words using the bilingual concordancer about 2900 times for the five units. Gain scores comparisons (difference between the pre-test and the post-test scores) indicate that the scores of items taught via concordancing were significantly higher than those which were not taught using a concordancer (p = .000 5 .05). That is, students improved more on collocations taught with the help of the concordancer. The retention effects associated with different teaching methods revealed that learners’ retention of collocations taught with the help of a concordancer was worse (p = .002 5 .05, see Table 4), but the residual effects were significantly better than those taught without the use of a concordancer (p = .004 5 .05). It is likely that the concordancing effects take a longer time for learners to assimilate and demonstrate in their performance of collocation knowledge; thus, its instant effects were not shown in the immediate post-test but in the delayed post-test after two and a half months. Students’ performance was also compared among those with different prior collocation levels. Thirty-two participants were divided into the high-level group (17 of them were classified as the HG with their mean higher than 10.59) and low-level group (15 of them were classified as the LG with their mean lower than 10.59) according to their pre-test scores. The comparison of their pre-test scores verified the validity of the grouping: the HG significantly outperformed the LG. The gain scores comparisons (between the pre-test and the post-test) pointed out that the gain scores of the LG were significantly higher than those of the HG (p = .046 5 .05, see Table 5). But no significant differences were observed between two groups’ degrees of retention between the immediate post-test and the delayed post-test (p = .088 4 .05),

242

T. Chan and H. Liou

Table 4. The paired t-test results for mean changes of two teaching methods between the immediate post-test and the delayed post-test N T1 mean changes T2 mean changes

32 32

Mean

S.D.

7 6.2500E-02 .1737 7 .2109 .1723

T

Df

Sig.

3.393

31

.002*

Note. * p 5 .05; T1: items drawn from the three concordancing units; T2: items drawn from the other two non-concordancing units.

Table 5 Comparison of HG’s and LG’s gain scores between the pretest and the immediate posttest N High 17

Mean

S. D.

Low

High

Low

High

Low

t-value

Df

Sig.

15

7.8235

10.2000

3.4321

2.9809

2.0777

30

.046*

Note: *P 5 .05

and none for the two groups’ degrees of residual effects of collocation between the pre-test and the delayed post-test (p = .657 4 .05). The two groups’ retention and residual effects of collocation were nearly similar. Students’ Appreciation of the Online Instruction Related to the Effects The results of the Background Questionnaire pointed out that few students (21.9%) had ever noticed accompanying words of the target vocabulary as they recalled from their past high school English learning experiences. Around 90% of the total students (87.5%) responded they did not know the term ‘‘collocation’’. Few learning strategies associated with collocations were introduced in their high school instruction. The results of the Evaluation Questionnaire showed that students held positive attitudes toward learning English V – N collocations via the online instruction module (mean = 3.53 out of 5, meaning ‘‘like it a lot’’) and their future use of similar design (about 90% of the total students). Specifically, they liked the bilingual concordanceraided sessions less as compared with their preference for the overall online practice (mean = 3.03) (see Table 6). Besides, students were satisfied with the design of the online units (mean = 3.84). To see whether students’ appreciation of the concordancer and its Chinese translation have correlations with the immediate effects, the regression, or the residual effects of collocation learning, responses drawn from several related items in the Evaluation Questionnaire and the score differences were computed. None of the correlation coefficients were found to achieve the statistically significant level; but

Effects of Concordancing on Collocation

243

Table 6. Participants’ general attitudes toward online explicit collocation instruction Item

Mean

Percentage

1. My feelings about learning English V – N collocations after using six online collocation units.

3.53*

Like it a lot (6.3)

2. If I can access the Internet, will I use similar designs to learn other English collocation types such as Adj. + N, Adj. + Adv.?

1.91*

Like it (40.6) No preference (53.1) Dislike it (0) Dislike it very much (0) Will (90.6)

3. My feelings about learning English V – N collocations after using the units which incorporate the use of the bilingual concordacner.

3.03

Will not (9.4) Like it a lot (3.1)

1.47

Like it (15.6) No preference (62.5) Dislike it (18.8) Dislike it very much (0) Will (46.9)

4. If I can access the Internet, will I use the bilingual concordancer, TotalRecall, to learn English collocations?

Will not (53.1) Note. For items with 5-point preference scale, the highest score is 5; for items with binary choices, the highest score is 2.

three of them show slightly higher correlations. It was found that the more learners agreed that the Chinese translation in the bilingual example sentences of the concordancer had helped students with the online tasks and that they liked the tool, the higher their retention scores of L1 – L2 non-congruent verb pairs were (correlation coefficient, R = 0.213). Unsurprisingly, learners had less difficulty locating collocations of L1 – L2 non-congruent pairs, and thus the residual effects were better (R = 7 0.341). In contrast, learners had more difficulty locating collocations with de-lexicalised verbs, and their improvement scores (immediate learning effects) were higher (R = 0.319). It is likely that collocations of different association strengths among their respective words may play a role. The Mutual Information (Church & Patrick, 1989) as a criterion of their strengths was chosen to compute using the British National Corpus. It was found that the Mutual Information scores of L1 – L2 non-congruent verb pairs were higher than those of de-lexicalised verb pairs as shown in Table 7.

244

T. Chan and H. Liou Table 7. Mutual Information scores for the collocation types

V

N

Quench Milk Convene Fulfil Brew Practice Try Perform Charge Evoke File Compose Give Achieve Process Convey Obtain Take Maintain Have Take Reduce Give Take Make

Thirst Cow Meeting Ambition Tea Medicine Luck Task Battery Admiration Complaint Poem Indication Success Datum Information Information Initiative Confidence Impact Measurement Sauce Party Call Circle

co_occur

v_no

n_no

mi_score

VN type

14 27 15 27 8 2 36 263 20 2 12 10 377 195 35 65 368 261 22 919 64 2 127 79 26

26 107 49 788 86 112 2511 3682 1888 493 574 832 44880 7322 776 819 6887 72670 6308 191840 72670 9755 44880 72670 90312

53 286 3576 496 1374 354 361 2184 392 165 991 570 643 2078 3855 9418 9418 1139 1628 2295 552 145 3966 1797 634

10.28212 7.838465 5.505678 5.291211 5.271061 4.9768 4.737653 4.543478 4.352646 4.258141 4.105011 4.104565 3.625801 3.606648 3.515518 3.187374 2.791762 2.204379 1.817809 1.791829 1.5231 1.402327 0.718375 0.553338 0.26647

4 4 4 1 4 4 4 1 4 2 4 2 3 1 4 1 2 3 1 3 3 2 3 3 3

Note. VN4: L1 – L2 non-congruent pairs; VN3: pairs with de-lexicalized verbs

As a brand new e-referencing tool, the bilingual concordancer was unfamiliar to most learners in Taiwan in spite of training handouts and demonstration sessions given before this experiment. Most learners were not familiar enough with the concordancing to impact learners’ perception of the e-referencing tool and learning at the same time. Surveying students’ problems with learning with concordancing, it was found that they had most difficulty in locating appropriate answers in the corpus and had to spend much time searching for collocations. In the open-ended Item 33 of the Evaluation Questionnaire (pointing to further improvement of the bilingual concordancer), students emphasised the virtue of two enhanced features: the partof-speech marking of the key word and the addition of specialised collocation aids (output ‘‘make a presentation’’ or ‘‘give a presentation’’ instead of a jungle of example sentences with ‘‘presentation’’).

Effects of Concordancing on Collocation

245

Discussion In the beginning of the project when students were admitted into a university from high schools, students barely knew the concept of collocation, and they did not pay much attention to collocation knowledge in English vocabulary learning, according to the results of the Background Questionnaire. These explained why students performed poorly in the collocation pre-test. After receiving explicit online collocation instruction with the web-based bilingual concordancer, students’ overall V – N lexical collocation knowledge has been significantly enhanced in light of their progress in the immediate post-test. The findings were consistent with those of Lin (2002), Sun and Wang (2003), and Tseng (2002): explicit collocation instruction was effective in promoting EFL learners’ collocation knowledge, including the online format with a bilingual concordancer. Previous studies were not designed with a delayed post-test and their retention effects were thus unknown. It seems that in this study, students became more aware of general collocation cases after being exposed to the five online units. Both students’ retention and residual effects were evident as we can see in their better performance of the untaught items immediately after the online project and after a time lapse. This may also imply that students can transfer the instructional effects (taught collocations plus patterns and examples encountered during concordancing) to novel contexts (untaught collocations). Although students’ degree of retention of collocations decreased after two and a half months, the residual effects of explicit collocation instruction plus practice sustained after about four months since students’ performance in the delayed post-test was still better than their entry level in the pre-test. In other words, lack of the online instruction or input (across the winter break between two semesters) during the time lapse may cause learners to forget but some amount of effects still stayed compared with their knowledge level of the very beginning. Explicit collocation instruction was also found to differently affect students’ performance of the four V – N collocation types. Using different effect measures results in different findings for the four collocation types. Overall, de-lexicalised verbs seemed to be more effectively learned under the current instructional context in this study, followed by L1 – L2 non-congruent verb collocations. One explanation is to attribute the findings to research artifacts. Yet, after a careful examination, some clues are suggested. L1 – L2 non-congruent verb pairs were stronger than de-lexical verb pairs, based on the Mutual Information scores computed, and seemed easier to locate using the bilingual concordancer. The concordancer’s Chinese translations also helped students learn better. In contrast, collocations with de-lexicalised verbs were weaker and harder to locate; yet its learning, retention, and even residual effects, were better than L1 – L2 non-congruent verb pairs. More efforts on concordancing in order to locate de-lexicalised verb collocations may result in deeper level of processing and thus enhance learning and retention. In spite of the possible interpretation, no significant and thus strong enough correlations were found in the current study. Future research is called for to address the puzzle.

246

T. Chan and H. Liou

Another way to interpret the findings is to rely on the frequency criterion. Delexicalised verb collocations occur often in the corpus. Learners might take advantage of abundant input to strengthen the associations of their verbs and nouns. On the other hand, Chinese – English mutual translations in the parallel corpus may have facilitated the noticing, reading, and perhaps assimilation of examples in the two languages for learners to note the L1 – L2 non-congruent collocation usage in two contrastive language examples. The L1-based approach strongly recommended in Nesselhauf’s study (2003) was supported in the current study to have assisted learners to acquire some collocation patterns. These patterns could hardly be acquired without being instructed with reference to L1. With regard to the treatment effects of the two collocation teaching approaches, the concordancing and non-concordancing approaches had different effects on collocation learning. Students made more progress in items learned with the help of the concordancer in the immediate post-test. The findings coincided with Sun and Wang’s study (2003): students benefited more from the concordancing method for either difficult or easy collocation patterns. It seemed that the concordancer can empower learners to see concrete collocation examples or to self-induce patterns that eventually facilitated collocation acquisition. Though students’ retention of these items was weaker in the delayed post-test, their residual effects were superior. It seemed that although students might fail to comprehensively internalise induced collocation patterns in time, students’ residual of collocations might sustain, given more time. This suggests that in the long term, the concordancing method can assist students’ collocation learning more, though such an induction-based approach may take longer time for the learning effects to surface. The assumption requires further research to verify. Learners’ entry-level collocation knowledge also has an impact on their later performance. The low-level group (LG) were found to have made significantly more improvement than the high-level group (HG) after the online instruction. The results corresponded to studies by Lin (2002) and Lee and Liou (2003): low-achievement students made significantly more progress in the productive collocation task, partly because the low-achievers had more room to push up their performance. The HG’s and LG’s degrees of retention and residual effects were similar. Overall, the low-level students benefited more from collocation instruction because of better gains and the same degrees of retention and residual effects compared with those of the high level students. Students’ responses in the Evaluation Questionnaire indicated that they viewed the collocation learning via the CALL approaches as helpful. Students were satisfied with the program design of the online collocation units which could give instant answers and explicit feedback. Most of them wanted to use similar online design to learn other types of collocation. Together with the performance data, it can be seen that students liked explicit collocation teaching and indeed benefited from it, with better learning outcomes. Although students did not like the bilingual concordancer-aided collocation teaching so much, the collocation tests’ results pointed out that students made more progress with the help of the concordancer and students’ residual effects

Effects of Concordancing on Collocation

247

were also significantly better than those of the non-concordancing method. It can be claimed then, that the bilingual concordancer has great potential to assist collocation learning. In spite of the observation that students felt difficulty in locating some of the appropriate collocates in concordancing, most of them agreed that concordancing inspired them to notice collocation use in context. It might be that students were not used to concordancing-based inductive learning as their past English experiences were dominated by the deductive ways of learning. Conclusion and Implications In light of the results found in this study, EFL college students’ overall verb – noun lexical collocation knowledge was improved via the explicit CALL collocation instruction with the help of the bilingual concordancer. Although students regressed after two and a half months, their knowledge of verb – noun collocation remained to the degree that it was superior to their entry levels. In terms of different collocation types, learners improved and remembered more regarding collocations with delexicalised verbs and collocations in lack of direct translations between L1 and L2. Moreover, the concordancing method seemed to effectively scaffold learners’ collocation learning. Low-level students benefited more from collocation instruction (see Table 8 for a summary of details) and regarding learners’ perception: they held a positive attitude towards the online collocation instruction module, but their appreciation of the online units did not tally with their learning effects. Previous research has indicated that collocations are largely neglected in English vocabulary teaching and, consequently, EFL students are particularly weak in collocation use. The current study has supported the positive effects of incorporation of a web-based bilingual concordancer into CALL practice units for learners to acquire collocation knowledge. Concordancing not only scaffolds EFL learners’ collocation learning but also raises their awareness of collocations that have not been taught through CALL. Our study has also found that students like to learn collocations with the CALL design and the web-based bilingual concordancer as designed in the study. The CALL approach can be used by other EFL or ESL (English as a second language, taught in the U. S. or U.K.) instructors with a similar curriculum infusion approach to help students’ vocabulary or collocation learning. As students have produced four collocation types with different learning effects, it might be necessary for instructors to teach collocations with different degrees of intervention, time investment, or material coverage. For instance, collocation teaching incorporating the use of the concordancer provides rich input for learning of the de-lexicalised verbs. The high occurrence of de-lexicalised verbs in the corpus can help reinforce learners’ impression. In addition, the bilingual concordancer with mutual translations in Chinese and English seems to ease learning of the L1 – L2 noncongruent verb collocations which can sometimes be negatively affected by learners’ L1. More guiding aids directly from the English language system are necessary for semantic-involved collocations as they are challenging for learners but the concordancer does not seem to provide immediate usefulness.

248

T. Chan and H. Liou Table 8. A summary of major findings of the study Three time intervals

4 VN types

Progress

Retention

Residual

Students made more progress in VN3 and VN4

Students regressed most in VN4

Students’ residual of VN3 was best and that of VN2 was worst. Their residual of VN4 was slightly better than that of VN1 and VN2 though not reaching statistical significance Students’ residual of collocations taught in the concordancing method was better The HG’s and LG’s residual of collocation was similar

Two teaching methods Students made more progress with the help of the concordancing method Learners’ prior The LG progressed collocation levels more than the HG did

Students regressed more in collocations taught with the concordancer The HG’s and LG’s retention was similar

The study suggests that an eclectic approach that combines concordancing and other traditional instructional methods may help learners better than the approach that relies on a single dominant teaching method. From the students’ responses, they have been found to be accustomed to the traditional method which usually dominated English teaching in high school education of Taiwan. However, performance of collocation tests points out such a method was not superior to the concordancing method as learners seem to better remember those collocations over a longer lapse of time. Thus, the concordancing method as designed in the current study can be encouraged in an educational setting where only rule- or explanationbased instruction is the norm. For learners educated in such a setting, students’ concordancing skills need to be strengthened; they might then be motivated and feel more confident about the concordancing method in spite of the findings that learners felt frustrated when not being unable to locate some incidents and they did not have enough appreciation of a concordancer. Besides, teachers can play the role of facilitators by posing leading questions to direct students’ attention to important examples and to help students probe into the contextual differences of collocation use presented in the web-based concordances. Also teachers can design proper materials with clear goals for students to search for certain linguistic features by making use of the concordancer. It is believed that with careful planning, the concordancing method can contribute to English learning. Technology infused into a traditional curriculum takes time for both learners and teachers to become accustomed to. The study does not include a control group for experimental comparison and the evidence may weaken its claimed effectiveness. Further, it has investigated only

Effects of Concordancing on Collocation

249

verb – noun collocations; the results may not apply to other types of collocations in English. Yet, the study has identified important factors that may have an impact on EFL learners’ learning of collocations: learners’ prior knowledge level, teaching methods, and collocation types. In the future, a larger-scale study may explore into those factors in more depth. Acknowledgements The study was funded by the National Science Council in Taiwan under the project number NSC92-2524-S007-2. The authors would like to express their gratitude to Dr Yuli Yeh, Dr Chih-Cheng Lin, Dr Jason S. Chang, Hung-Tzu Huang, and the anonymous reviewers of Computer Assisted Language Learning for their comments on this paper. Research assistants and participating students involved in the study are also acknowledged. References Bahns, J., & Eldaw, M. (1993). Should we teach EFL students collocations? System, 21, 101 – 114. Benson, M., Benson, E., & Ilson, R. (1986a). Lexicographical description of English. Amsterdam: John Benjamin. Benson, M., Benson, E., & Ilson, R. (1986b). The BBI combinatory dictionary of English: A guide to word combination. Amsterdam: John Benjamin. Church, K. W., & Patrick, H. (1989). Word association norms, mutual information and lexicography. Proceedings of American Computational Linguistics, 27, 76 – 83. Farghal, M., & Obiedat, H. (1995). Collocations: A neglected variables in EFL. International Review of Applied Linguistics, 33, 313 – 331. Howarth, P. (1998). Phraseology and second language proficiency. Applied Linguistics, 19, 22 – 24. Johns, T. (1991). From printout to handout: Grammar and vocabulary teaching in the context of data-driven learning. ELT Journal, 4, 27 – 45. Lee, C. Y., & Liou, H. C. (2003). A study of using web concordancing for English vocabulary learning in a Taiwan high school context. English Teaching & Learning, 27, 35 – 56. Lewis, M. (2000). Teaching collocation: Further development in lexical approach. London: Language Teaching Publications. Lin, Y. P. (2002). The effects of collocation instruction on English vocabulary developments of senior high students in Taiwan. Unpublished Master’s thesis, National Kaohsiung Normal University, Taiwan.> Liou, H. C., Chang, J., Yeh, Y., Liaw, M., Lin, C., Chen, H., You, G., Chuang, C., & Gao, Z. (2003). Using corpora and computational scaffolding to construct an advanced digital English learning environment: The CANDLE project. In the Proceedings of APAMALL 2003 and ROCMELIA 2003 (pp. 62 – 77). Taipei, Taiwan: The Crane Publishing Co. Liu, C. P. (1999). An analysis of collocational errors in EFL writings. In The proceedings of the Eighth International Symposium on English Teaching (pp. 483 – 494). Taipei, Taiwan: The Crane Publishing Co. Liu, C. P. (2000). An empirical study of collocation teaching. In the Proceedings of the seventeenth conference on English teaching and learning in the Republic of China (pp. 165 – 178). Taipei, Taiwan: The Crane Publishing Co. Liu, L. E. (2002). A corpus-based lexical semantic investigation of verb-noun miscollocations in Taiwan learners’ English. Unpublished master’s thesis, Tamkang University, Taipei. Nation, I. S. P. (2001). Learning vocabulary in another language. Cambridge: Cambridge Press.

250

T. Chan and H. Liou

Nattinger, J. R., & DeCarrico, J. D. (1992). Lexical phrase and language teaching. Oxford: Oxford University Press. Nesselhauf, N (2003). The use of collocations by advanced learners of English and some implications for teaching. Applied Linguistics, 24, 223 – 242. Reber, A. S., Kassin, S., Lewis, S., & Cantor, G. (1980). On the relationship between implicit and explicit modes in the learning of a complex rule structure. Journal of Experimental Psychology: Human Learning and Memory, 6, 492 – 502. Schmitt, N. (2004). Formulaic sequences: Acquisition, processing, and use. Philadelphia: John Benjamins. Sun, Y. C., & Wang, L. Y. (2003). Concordancers in the EFL classroom: Cognitive approaches and collocation difficulty. Computer Assisted Language Learning, 16, 83 – 94. Tseng. F. P (2002). A study of the effects of collocation instruction on the collocational competence of senior high school students in Taiwan. Unpublished master’s thesis, National Taiwan Normal University, Taipei. Wang, L. (2001). Exploring parallel concordancing in English and Chinese. Language Learning & Technology, 5, 174 – 184. Retrieved from http://llt.msu.edu/vol5num3/ wang/default.html, downloaded 3/4/2004. Wray, A. (2002). Formulaic language and the lexicon. Cambridge: Cambridge Press.