Abstract
The present study explores the knowledge and use of English lexical collocations and their relation to the speaking proficiency (assessed by two speaking tests) of Taiwanese EFL university learners. Data for the study were collected from 56 junior English majors at a national university of science and technology in southern Taiwan. Each student was asked to take three tests: (1) one lexical collocation test, measuring the subjects’ knowledge of lexical collocations; (2) one English speaking test, administered to collect the subjects’ use of lexical collocations and measure their speaking proficiency; and (3) PhonePass spoken English test, a standardized oral test, which was combined with the speaking test to measure the students’ speaking proficiency. Test results were examined for correlations (1) between the subjects’ knowledge of lexical collocations and their speaking proficiency, (2) between their use of lexical collocations and their speaking proficiency, and (3) between their knowledge of lexical collocations and their use of lexical collocations. The study findings showed that there was a significant correlation between Taiwanese EFL learners’ knowledge of lexical collocations and their speaking proficiency. However, no significant correlation existed between the subjects’ use of lexical collocations and their speaking proficiency. There was also no statistically significant correlation between the subjects’ knowledge and use of lexical collocations. The current study concluded that knowledge of lexical collocations seemed to be a more significant indicator of speaking proficiency than the subjects’ ability to use lexical collocations.

Keywords: lexical collocations, speaking proficiency, EFL education

1. Introduction
Vocabulary and grammar are two basic elements of a language, but traditionally most EFL teachers often emphasize learning grammar more than vocabulary in their teaching. The role of vocabulary has long been underestimated in EFL education. With the approaching new millennium, pioneer scholars and researchers have started to direct people’s attention to EFL vocabulary acquisition and began to emphasize the instruction of vocabulary in classroom practices (Channell, 1981; McCarthy, 1984; Nation, 1990; Nattinger, 1980). Among them, perhaps the most important one, Lewis (1993), proposed one groundbreaking teaching method, the Lexical Approach, and brought the field to systematically examine the nature of lexis in second language acquisition. He strongly argued that “language consists of grammaticalised lexis, not lexicalised grammar” (Lewis, 1993, p. vi). In Lewis’ view, learning collocations, the key component of grammaticalised lexis, is equal to language learning.

A collocation, in its simplest definition, consists of two words which are linked together in the memory of native speakers and occur together with some frequency in both written and oral discourse (Aghbar, 1990). For example, catch a cold and severe cold are two commonly used word combinations that qualify as collocations. The verb catch and the adjective severe recurrently co-occur with the noun cold. In addition, many also believed that knowing a word includes knowing its collocations (Lewis, 2000; Nation, 1990, 2001).
Many scholars have maintained that collocational knowledge is one important factor that contributes to the differences between native speakers and foreign language learners (Aston, 1995; Fillmore, 1979; Kjellmer, 1991; Pawley & Syder, 1983). Particularly, failure to use collocations accurately for EFL learners is a major indicator of foreignness (McArthur, 1992; McCarthy, 1990; Nattinger, 1980; Wu, 1996). The strongest position held so far is that collocational competence is an indispensable component in the process of second/foreign language acquisition (Lewis, 1997, 2000; Nattinger & DeCarrico, 1992; Richards & Rogers, 2001).

1.1 Purpose of the Study
University students in Taiwan are currently required to pass an English proficiency test as one of the graduation requirements. Among various English proficiency tests available, GEPT (General English Proficiency Test), TOEFL (Test of English as a Foreign Language) and IELTS (International English Language Testing System) are the most well-known among students. As a matter of fact, a speaking test is a basic component of these standardized exams. Nevertheless, acquiring speaking skills has been a challenge for most Asian students, and Taiwanese university students are no exception.

While searching for effective ways to improve EFL learners’ speaking proficiency, many researchers have proposed that a good control of collocations can help language learners to speak more fluently (Brown, 1974; Ellis & Schmidt, 1997; Nation, 2001; Nattinger & DeCarrico, 1992; Pawley & Syder, 1983; Schmitt, 2000; Sung, 2003). Collocational knowledge could be a key element in enhancing the EFL learners’ speaking ability. In Taiwan, a few researchers have carried out studies on collocational proficiency (Huang, 2001; Liu, 1999a; Tseng, 2002; Wu, 2005). However, most collocation studies in Taiwan investigate only the students’ written production; few have looked into their speaking performance. Hence, the present study examines whether Taiwanese EFL learners’ knowledge of collocations1 is related to their speaking proficiency. The purpose of this study is to examine and describe the knowledge and use of collocations and their relation with English speaking proficiency among Taiwanese EFL university learners. To fulfill the purpose of the study, the experiment carried out in this study aimed to answer the following research questions:
(1) To what extent is subjects’ knowledge of lexical collocations (KLC) correlated with their general speaking proficiency (GSP)?
(2) To what extent is subjects’ use of lexical collocations (ULC) correlated with their general speaking proficiency (GSP)?
(3) To what extent is subjects’ knowledge of lexical collocations (KLC) correlated with their use of lexical collocations (ULC)?

1.2 Significance of the Study
The study is quantitative in nature and examines Taiwanese EFL learners’ knowledge and use of collocations and further explores the relationship between collocations and their speaking proficiency. Since no study has been conducted to report such performance and the relationship between collocations and speaking in Taiwan, this study contributes directly to our understanding of the nature of collocations in speaking.
         
2. Review of Related Studies
The review of literature introduces empirical studies particularly targeting EFL learners (both non-Taiwanese and Taiwanese) and examines the possible connection between lexical collocations and English proficiency.

2.1 Studies of the Relation between Collocations and Language Proficiency of non-Taiwanese EFL Learners
Zhang (1993) was the first scholar to explore the possible correlation between the knowledge and use of English collocations and the quality of college freshmen’s writing. At a mid-size public university in Pennsylvania, in the United States, the 60 college freshmen in his study were categorized into two groups: 30 native and 30 non-native speakers of English. Within each group, two subgroups, i.e., Good writers and Poor writers, were established based on a writing test. Each subject completed one fill-in-the-blank collocation test and one writing task. The collocation test was used to measure the subjects’ collocational knowledge; the writing task was used to elicit the subjects’ use of collocations and writing proficiency. In this experiment, Zhang (1993) found that (1) native English writers performed significantly better than non-native writers on the collocation test, and Good writers within either group performed significantly better than Poor writers; (2) as for the use of collocations in their writing, native writers surpassed the non-native writers, and Good writers within either group surpassed Poor writers; (3) in terms of writing performance, a significant difference was found between non-native Good and Poor writers.

Zhang drew two conclusions based on the observed correlations. Collocational knowledge is a source of proficiency in writing among college freshmen. Besides, quantity, but more important, quality for use of collocations distinguish between native and non-native college freshmen writing as well as between Good and Poor college freshmen writing.

After Zhang, Al-Zahrani (1998) investigated the knowledge of English lexical collocations among four academic levels of Saudi EFL university students and the relationship between the participants’ collocational knowledge and their general language proficiency. In his study, the collocational knowledge of 81 Saudi male university English majors was measured by a cloze test, comprised of 50 “verb+noun” lexical collocations. Furthermore, the participants’ general English proficiency was assessed by a writing test and a paper-and-pencil TOEFL test. Al-Zahrani found that there was a significant difference in his subjects’ knowledge of lexical collocations among the different academic years. The knowledge of lexical collocations increased with the subjects’ academic years. Besides, he reported that there was a strong correlation between the subjects’ knowledge of collocations and their overall language proficiency. 

Sung (2003) examined the knowledge and use of English lexical collocations in relation to speaking proficiency of international students enrolled in a university in Pittsburgh area. A total of 72 non-native English speakers participated in her study. Each subject completed two tests: one collocation test and one speaking test. The first test was used to measure the subjects’ knowledge of lexical collocations while the latter was used to elicit the subjects’ use of lexical collocations and measure their speaking proficiency. Her results showed that there was a significant correlation between the knowledge of lexical collocations and the subjects’ speaking proficiency.
Up to this point, three major studies have investigated the relation between EFL learners’ collocational knowledge and their language proficiency (Al-Zahrani, 1998; Sung, 2003; Zhang, 1993). The results of the studies all showed similar positive correlations between these learners’ collocational knowledge and their language proficiency.

2.2 Studies of the Effect of Explicit Collocation Instruction on Language Skills of Taiwanese EFL Learners
In Taiwan, while the majority of studies have looked at error analysis of collocations (Chen, 2002; Hsueh, 2003; Liu 1999a, 1999b; Tang, 2004), only six researchers have applied collocation instruction in classroom to observe the effects on students’ language skills.

In one of her collocation studies, Liu (2000) led the field in investigating the effects of collocation instruction on students’ writing performance. Forty-nine freshmen English majors at a Taiwanese university participated in the study. During an 18-week semester, in a three-hour weekly class, Liu (2000) gave her students a series of twenty-minute mini-lessons on collocations (e.g., the introduction of six major lexical collocation patterns, collocations without direct L1 equivalents, and de-lexicalized verbs as collocates of nouns). To assess their writing ability and use of collocations, the subjects were asked to write a composition in class at the beginning and the end of the semester without using a dictionary. The two compositions were analyzed and compared for the patterns of acceptable and unacceptable lexical collocations. It was found that the students in the second composition generated a greater number and variety of acceptable lexical collocations although they did not improve much in their writing of the second composition. After Liu, Lin (2002) in a two-week course introduced 8 collocation activities (such as brainstorming for collocates of a word, underlining verb-noun collocations, and matching game)to her 89 senior high school students to explore the effects of collocation instruction on Taiwanese EFL learners’ vocabulary development.

Subjects also took a pre- and post-course collocation test (evaluating “productive collocation competence” (p. 84)) and also a pre- and post-course multiple-choice test (assessing just “recognitive collocation competence” (p. 85)). Lin reported that the students were found to have improved in both their receptive and productive competence after receiving the systematic instruction on collocations. In contrast with Liu (2000) and Lin (2002), Tseng (2002) examined the effects of direct collocation instruction on both vocabulary improvement and writing performance. In a 12-week period, Tseng taught ninety-four senior high school students in two groups, the experimental group (receiving collocation instruction) and the control group (no collocation instruction). She used some of Liu’s (2000) activities on collocation, including introduction of collocations, collocations without L1 equivalents, and added two more teaching activities in her study, i.e., using collocation dictionaries and identifying collocations in the textbook.

All the subjects in both groups did three kinds of tasks: (1) a questionnaire, (2) a pre- and post-instruction fill-in-the-blank collocation test, and (3) a pre- and post-instruction composition. Although the participants did not exhibit obvious improvement in their performance on the composition after receiving the collocation instruction, Tseng reported that the collocation instruction somewhat improved the subjects’ vocabulary development. This finding is somehow different from Lin’s (2002) study in which Lin reported her students’ progress in “collocation competences” (p. 85). A possible explanation might be the instruments they used in assessing their subjects’ improvement. In fact, unlike Tseng who analyzed an actual writing test, Lin relied on her fill-in-the-bank and multiple-choice tests only.

A teacher in a private Taiwanese university, Hsu (2002), also conducted an intensive business English workshop to investigate whether collocation instruction would strengthen Taiwanese college EFL learners’ development of the knowledge and use of collocations, as well as promote language proficiency. Seven English and two Banking & Finance majors at a private Taiwanese university took part in this workshop. The workshop ran for a total of 48 hours in one month (i.e., three hours a day, four days a week). As a teacher-researcher, Hsu collected and analyzed students’ writings, results of pre- and post-workshop collocation tests, teacher’s class notes, student interviews, and videotapes of the workshop sessions. After examining the data, he found that (1) collocation instruction appeared to have helped students acquire new collocations in both written and spoken discourse; and (2) there appeared to be a positive correlation between the students’ use of lexical collocations and their language proficiency. Based on the findings, he concluded that explicit collocation instruction is highly valuable and recommended that it should be incorporated into EFL classrooms.

Investigating a specific language skill, Lien (2003) addressed the effects of collocation instruction especially on reading comprehension. A total of 85 Taiwanese university students at three academic levels participated in her study. In the pre-experimental stage, each student took one fill-in-the-blank collocation test and reading comprehension pre-test. The collocation test and reading comprehension pre-test were administered in order to investigate the relationship between knowledge of collocations and reading comprehension. In the experimental stage, all subjects at the three academic levels received three types of instruction (collocation instruction, vocabulary instruction, and no instruction); that is, subjects served as their own control group as well as participating in the experimental conditions.

The subjects then took a reading comprehension post-test immediately after each type of instruction. The three reading comprehension post-tests were used to evaluate the effectiveness of the three types of instruction. In the post-experimental stage, a questionnaire was administered to collect the subjects’ opinions and attitudes toward the three types of instruction. The major findings of the study included: (1) collocational knowledge was associated with reading comprehension; (2) the subjects performed better after receiving collocation instruction compared with the other types of instruction; and (3) the subjects also reported positive attitudes toward the direct teaching of collocations.

Looking at another specific language skill, L. Hsu (2005)2 investigated the effects of explicit collocation instruction on EFL learners’ listening comprehension. Thirty-four Taiwanese university students participated in her study. Over a three-week period, each subject received three different types of instruction (single-item vocabulary instruction, lexical collocation instruction, and no instruction), and took one listening comprehension test right after each type of instruction. In the fourth week, the subjects completed a questionnaire. The results indicated that the subjects performed best after receiving lexical collocation instruction. Furthermore, the students indicated in the questionnaire that: (1) collocation instruction was their preferred instruction type; (2) they were willing to learn more about lexical collocations; and (3) they believed “their listening fluency can be improved as a result of instruction of lexical collocations” (p. 63).
In sum, empirical studies concerning the effects of explicit collocation instruction on Taiwanese EFL learners have been carried out to explore many aspects of language skills, including writing (Liu, 2000; Tseng, 2002), vocabulary development (Lin, 2002; Tseng, 2002), reading (Lien, 2003), listening (L. Hsu, 2005) and overall proficiency (Hsu, 2002). Among these studies, the results generally showed that direct collocation instruction was positively correlated with and possibly improved Taiwanese EFL learners’ language performance. In addition, the subjects in the majority of the studies displayed a positive attitude toward collocation instruction.

3. Research Method
3.1 Research Design
In this study, the data were gathered by three tests, including one written and two spoken tests. They are a fill-in-the-blank lexical collocation test, an English speaking test, and a PhonePass spoken English test. This study is also a quasi-experimental design; it is similar to true experimental designs, except that the participants are not randomly selected for the study (Mertens, 1997).

Three variables elicited from the three tests were used in the study. First, knowledge of lexical collocations (KLC) was measured by the subjects’ scores on the fill-in-the-blank lexical collocation test. Second, use of lexical collocations (ULC) was calculated by the quantity of lexical collocations collected from the subjects’ speaking test. Last, general speaking proficiency (GSP) was measured and combined from the average scores of the two speaking tests (i.e., English speaking test and PhonePass spoken English test). In this study, descriptive data were collected from the three tests. The Pearson correlation coefficient was adopted to determine the degree of a relationship between two quantifiable variables.

3.2 Study Setting and Subjects
The subjects in this study were 56 EFL junior English majors at a university of science and technology in southern Taiwan. These subjects were chosen for the following reasons. Firstly, the purpose of the study is to examine the relation between knowledge of lexical collocations and general speaking proficiency of the EFL learners. Furthermore, with at least 8 years formal English training, the junior English majors should have learned a certain number of collocations, and also would have had many opportunities to practice speaking English. Besides, no previous study has ever looked into this in Taiwanese university settings. Thus, the researchers chose only Taiwanese EFL learners as the target subjects. An investigation of these students’ performance would provide the researchers with sufficient data to explore their collocational performance and speaking proficiency.

3.3 Instruments
3.3.1 Lexical collocation test
The fill-in-the-blank lexical collocation test consisted of 50 items. The subjects were expected to provide the best answer for the target collocate in the blank. In this specific test, the first letter/phoneme was provided because it could help subjects trigger the appropriate target collocate, and reduce the possibility of guessing by test takers. A sample question for L2 type is “If you want to have a b_____ future, you need to set your goals in advance, and turn your plan into actions.” The anticipated collocate for the question is “bright.”
The test included five major types of lexical collocations as categorized by Benson, Benson, & Ilson (1997). Each type had 10 test items. Table 1 summarizes the distribution of the number for lexical collocations in the test. The underlined part of speech in Table 1 indicates the missing part in the test which the subjects needed to fill in. 

Table 1 Number of Items Used in the Test of Lexical Collocations

3.3.2 Pear Film Speaking test
Subjects took the Pear Film speaking test to elicit collocations, which would provide the quantity of their use of lexical collocations (ULC). The test results were later considered with those of the PhonPass speaking test as a measurement of the subjects’ speaking proficiency. The Pear Film was produced under the direction of Chafe (1980) at the University of California, Berkeley. Chafe noted that the Pear Film was designed for the purpose of analyzing how people talk about what they have experienced and later how they recall these experiences. This 6-minute film covering a series of natural events can be easily understood. After watching, viewers should be able to recall what they saw and describe how events in the film related to their daily life.
When attending the speaking test of Pear Film, the subjects were invited to watch the film individually and answered the following three questions orally:
(1) After watching the Pear Film, please tell me what happened in the film?
(2) Please explain what you consider to be the meaning behind the film.
(3) Please relate an aspect of the film (any aspect) to something in your own life or something you have observed in real life.

The subjects were encouraged to speak as much as possible as there was no time limit. Later, the subjects’ oral descriptions were transcribed, and the quantity of the lexical collocations was counted and recorded.

3.3.3 PhonePass spoken English test
The PhonePass spoken English test is a standardized test developed by Townshend and Bernstain in 1996 (Validation summary for PhonePass, 2005). It is the only English spoken testing system which combines computerized design, telephone, Internet and professional test contents to measure the test taker’s speaking proficiency over the telephone. There are three major characteristics of PhonePass spoken English test. This test is continuously and conveniently available from any telephone on a 24-hour basis. Also, the test is scored automatically by a computer-based system. Lastly, the test scores can be efficiently displayed on the PhonePass website (http://www.phonepass.com.tw/) one hour after taking the test. Currently three versions (i.e., 10-minute, 7-minute, and 5-minute) of the PhonePass spoken English tests are available.

The format used in this study was a 7-minute version for each subject. The test consists of 44 items which are divided into four sections (Parts A through D). Each of the four parts presented the test takers with a different task: (A) reading aloud, (B) repeating, (C) answering short questions, and (D) building sentence. In Part A, the subjects read the sentences shown in the test sheet. For example, they read the sentence, There are advantages and disadvantages to eating alone. In Part B, the subjects had to repeat what they heard on the telephone. For instance, a voice says, Leave town on the next train and the subjects would repeat. In Part C, the subjects must give a simple answer to the questions. For example, a voice says, Would you get water from a bottle or a newspaper? and the subjects had to give short answers of a bottle or from a bottle. In Part D, the subjects had to arrange the scrambled words or phrases into a complete sentence. For example, a voice says, was reading.. my mother.. her favorite magazine.. and the subjects had to unscramble them and make one complete sentence of My mother was reading her favorite magazine.

ORDINATE Corporation, the publisher, has conducted series of studies to examine the validity and reliability of PhonePass spoken English test. In terms of inter-rater reliability, it reported a positive correlation of 0.73 between PhonePass spoken English test results and paper-pencil TOEFL scores, and a high correlation of 0.94 with human scorers (Validation summary for PhonePass, 2005).

3.4 Data Collection Procedures
This study consisted of two experimental stages. In the first stage, each subject was required to take the lexical collocation test and then English speaking test of Pear Film individually. In the second stage, the researchers delivered the PhonePass spoken English test sheet to the subjects and explained how to operate the test. All the subjects were asked to take the PhonePass spoken English test one week after completing the first stage of experiment. Some subjects were excluded because they did not complete the PhonePass spoken English test before January 13, 2006 (as this was last class of the academic semester before winter vacation). Altogether, there were 56 subjects at the end of the study.

3.5 Scoring
3.5.1 Scoring for lexical collocation test
The responses on the lexical collocation test were rated by using a 3-point partial-credit scoring method, suggested by Aghbar and Tang (1991). Each item was assigned 0-2 points, totaling 100 points for the 50-item test. The researchers scored the lexical collocation test with the help of the BBI dictionary of English word combinations (Benson, Benson, & Ilson, 1997), two online corpora: Simple Search of British National Corpus (http://sara.natcorp.ox.ac.uk/lookup.html) and VLC Web Concordancer (http://vlc.polyu.edu.hk/concordance/WWWConcappE.htm) and two native speakers who are both experienced EFL teachers. Online corpora were utilized in this study because evidence showed that a corpus could demonstrate how words collocate (Schmitt, 2000; Scholfield, 1997). The two online corpora were used because they were both free of charge. In addition, they were designed to include a wide range of samples from different registers and genres. Although these two corpora are much closely related to British English, the researchers also asked help of two native speakers from the United States, making efforts to balance the judgment of lexical collocations between British English and American English.

When a response could be found from the BBI dictionary of English word combinations, it would receive 2 points. Alternatively, if the response was not found in the BBI dictionary of English word combinations but was found in either one of the two online corpora or confirmed by either one of the two native speakers, then the subjects would receive 1 point. This decision was made based on two considerations. First, online corpora may include many possible collocations which are used loosely, representing a wider range of possibility. Second, the native speakers’ judgment may merely be limited to the regional feature of their English. Besides, as the purpose of the test was to elicit the subjects’ collocational knowledge, rather than judge their grammar or spelling errors, errors of these types were overlooked.

3.5.2 Scoring for Pear Film speaking test
The scoring of Pear Film speaking test was conducted holistically by two raters who are both experienced English teachers. The IELTS holistic scoring system was adopted to evaluate the subjects’ performance based on the two reasons: its popularity in Taiwan and speaking test component.
To ensure inter-rater reliability, a practice session was held. All the materials including IELTS scoring system were introduced, so that the two raters could fully understand the rating scale. After watching the Pear Film and listening to a sample of spoken data, the two raters assigned scores and made recommendations. Their ability to assign scores consistently was practiced before they were exposed to the current research rating task. When starting to score the data of the speaking test of Pear Film, each spoken data was scored by the two raters separately. When the difference in scoring was greater than 2 points, the raters were asked to re-rate the speaking data. Then the average of their ratings was used as the final score for the subject’s speaking proficiency in the test of Pear Film.

3.5.3 Tallying the frequency of lexical collocations from Peal Film speaking test
The spoken data obtained from Pear Film speaking test were transcribed and all the lexical collocations were manually counted. An acceptable lexical collocation was recorded based on the same criteria (i.e., the verification from 2 online corpora or 2 native speakers) adopted in evaluating the lexical collocation test. Since all the subjects’ total length of speeches differed, the total words uttered were different. In order to make comparison at an equal scale, the total words in each subject’ spoken data were converted to 400 words3.

3.5.4 Scoring for PhonePass spoken English test

Figure 1. A Sample Score of PhonePass Spoken English Test
           
After obtaining the scores of the two major speaking tests, the researchers combined them to produce general speaking proficiency (GSP). Because the scoring systems between the two speaking tests were totally different, the scores of the Pear Film speaking test were converted into the measurement of the PhonePass spoken test.

3.6 Data Analysis and Statistical Procedures
All the data were computed by Statistical Package for the Social Science (SPSS) version 12. To understand the general picture of the subjects’ knowledge of lexical collocations (KLC), use of lexical collocations (ULC) and their general speaking proficiency (GSP), a preliminary analysis was first conducted by computing the descriptive statistics of these three variables for the study. In the second stage, Pearson product-moment correlation4 (Bachman, 2004; Haslam & McGarty, 2003; Schweigert, 1994; Weiss, 2005) was adopted to measure possible correlations of KLC and GSP, ULC and GSP, KLC and ULC respectively and to answer the 3 research questions of the study.

4. Results & Discussion
4.1 Analysis of Research Question One
Research Question One asked: To what extent is the knowledge of lexical collocations (KLC) correlated with the general speaking proficiency (GSP) among the subjects? The Pearson correlation was used to measure correlations between the scores of KLC and GSP among all subjects; in addition, the L1-L5 subtypes in the KLC scores were also further investigated to analyze their correlations with GSP. Table 2 presents the correlations between KLC and GSP.

Table 2 Pearson Correlation between Knowledge of Lexical Collocations and General Speaking Proficiency (N=56)

Note. The asterisks ** indicate correlation is significant at the 0.01 level; the asterisk * indicates correlation is significant at the 0.05 level.
           
Table 2 shows that the relationship between KLC scores and GSP scores is positively correlated. Statistically significant correlation exists between GSP scores and total KLC scores for the subjects (r = 0.561 at p < 0.01 level). Specifically, the L1-L5 subtypes of KLC scores are also moderately correlated with GSP scores. In brief, there is a moderately positive correlation between GSP scores and KLC scores among the 56 subjects. This could suggest that KLC could be a predictor to measure the subjects’ general English proficiency.

In the same manner, the results of previous studies also showed positive relationship between the knowledge of collocations and other language skills, such as writing (Zhang, 1993); language proficiency (Al-Zahrani, 1998; Bonk, 2000), reading (Lien, 2003) and speaking (Sung, 2003). Thus, it could be concluded that knowledge of collocations is indicative of non-native speakers’ language proficiency.

4.2 Analysis of Research Question Two
Research Question Two asked: To what extent is the use of lexical collocations (ULC) correlated with the general speaking proficiency (GSP) among the subjects? The Pearson correlation was employed to measure correlation between the frequency of ULC and the scores of GSP among all subjects; moreover, the L1-L5 subtypes in the frequency of ULC would be also further investigated to determine their correlations with GSP. Table 3 presents the correlations between ULC and GSP.

Table 3 Pearson Correlation between Use of Lexical Collocations and General Speaking Proficiency (N=56)

Note. The symbol “(a)” indicates that the correlation cannot be computed because one variable is constant5.
According to Table 3, most of the variables are weak but positively correlated. The correlation between the GSP scores and the total frequency of ULC is not strong enough to be significant among all subjects. The case between the GSP scores and the individual L1-L5 subtypes is the same. In short, the results show that only a positively weak correlation occurs between the GSP scores and the frequency of ULC. This might suggest that the ULC does not seem to be a strong predictor for measuring the subjects’ general English proficiency.

This finding corresponded with Hsu’s (2002) study in which he found that there is no strong relationship between the subjects’ quantity of collocations and their speaking performance. However, the finding was somewhat different from Sung’s (2003) study. Sung reported that there was a moderate correlation between the speaking proficiency and the frequency of collocations used in the speaking test among the non-native speakers residing in the United States. One possible explanation might be due to the different focuses on the categories of lexical collocations. Sung (2003) included more subtypes of lexical collocations (L1-L7 based on Benson et al., 1997) in her recording of students’ spoken data. In the current study, the researchers looked at the subjects’ L1-L5 subtypes of lexical collocations which are fewer than Sung’s. Another possible explanation might be that the subjects in Sung’s study were residents in the United States; they may be more proficient in English.

Still, it is beyond the scope of the current study to draw a concrete conclusion on the relationship between use of collocations and speaking proficiency. Therefore, further larger scale research could investigate the correlation between these two variables.

4.3 Analysis of Research Question Three
Research Question Three asked: To what extent is the knowledge of lexical collocations (KLC) correlated with the use of lexical collocations (ULC) among the subjects? Table 12 presents the correlations between KLC and ULC and the 5 subtypes found in KLC and those from ULC for all the subjects. According to Table 4, only a weak but positive correlation occurs between the KLC scores and the frequency of ULC among the subjects. In other words, the linear correlations between KLC scores and frequency of ULC are not significant. Such similar finding was formerly reported by Zhang (1993) who also found that a weak correlation existed between the subjects’ knowledge of collocations and use of collocation from their writing.

Table 4 Pearson Correlation between Knowledge of Lexical Collocations and Use of Lexical Collocations (N=56)

Note. The symbol “(a)” indicates that the correlation cannot be computed because one variable is constant6.

However, the findings of Zhang (1993) and the current study are somewhat different from Sung’s study in 2003. In her results, there was a moderately significant correlation between the knowledge of lexical collocations and frequency of lexical collocations in the speaking performance of the non-native speakers in the United States. The relationship between knowledge of collocations and their use of collocations is not obvious and we are far from drawing any convincing conclusion. Therefore, further investigation could look into this particular relation.

4.4 Summary of Research Findings
The present study intended to examine the possible connections among the three variables: KLC, ULC, and GSP. Table 5 clearly shows the Pearson correlation among the three variables.

Table 5 Pearson Correlation among KLC, ULC, and GSP (N=56)

Note. The asterisks ** indicate correlation is significant at the 0.01 level.

Based on Table 5, three major findings can be summarized:
(1) The scores of knowledge of lexical collocations (KLC) showed significant correlations with the subjects’ general speaking proficiency (GSP).
(2) The total frequency of use of lexical collocations (ULC) did not show any significant correlations with the subjects’ general speaking proficiency (GSP).
(3) The subjects’ scores of knowledge of lexical collocations (KLC) did not show significant correlations with the subjects’ use of lexical collocations (ULC).
Thus, the current study concludes that knowledge of lexical collocations seemed to be a more suitable indicator for determining Taiwanese college EFL learners’ speaking proficiency than their Use of lexical collocations. In fact, ULC may not be a good indicator for measuring either the subjects’ KLC or their general speaking proficiency. Findings of the current study are summarized and visually illustrated in Figure 2.

5. Concluding Remarks
5.1 Pedagogical Implications
The study has demonstrated that knowledge of lexical collocations is significantly associated with Taiwanese university EFL learners’ spoken English. Future bigger scale studies might still need to be carried out in order to examine whether or not direct instruction on lexical collocations will benefit EFL learners’ speaking proficiency. Nevertheless, a cautious way to deal with lexical collocations can be the introduction of a lexis-focused syllabus (Willis, 1990; Lewis, 1993, 1997) as a supplementary component to any current existing syllabus. In other words, teachers can at least try to provide EFL learners sources of lexis, such as corpus, concordance, and dictionaries of collocations, ensuring these learners to have direct exposure to chunks of English language and learn to extract and use patterns of lexical collocations effortlessly.

The next pedagogical implication concerns the subtype of lexical collocations for Taiwanese EFL learners. The study showed that the subjects scored lowest in the L4 subtype (adv + adj) in their lexical collocation test. None of the subjects ever produced the L4 subtype of lexical collocations (adv + adj) in their speaking test of Pear Film. In other words, Taiwanese university EFL students might not have gained much knowledge of the L4 subtype of lexical collocations and still have not mastered well the adverb usage in their oral production. Based on such a finding, EFL teachers should especially concentrate on the practice of this certain subtype (adv + adj) in class.

5.2 Limitations of the Study
The fact that the subjects of the study were not randomly selected is perhaps one obvious limitation. The subjects were limited to the English-major junior students at a national university of science and technology from southern Taiwan. Therefore, while the study reveals interesting findings about the relationship between the Taiwanese EFL university learners’ lexical collocational knowledge and their speaking proficiency, these findings may not be easily generalized beyond the subjects of the study.

The reliability of self-design lexical collocation test is another possible concern as well. Since the subjects were only tested once on their knowledge of lexical collocations, the results of their performance may not be reliable enough to truly indicate their real scores. In order to minimize the degree of performance errors, future researchers may also need to consider enhancing instrument reliability.

Another limitation of this study is related to the scope of collocations investigated. This study only examined lexical collocations and did not include grammatical collocations. As a result, the performance of the subjects on the collocational test only reveals their knowledge of lexical collocations, but does not represent their overall collocational knowledge.

5.3 Recommendations for Future Research
Collocations, an important focus in EFL research, have started to gain increasing attention in the past two decades. As the current study investigated the subjects’ knowledge and use of collocations as well as their speaking proficiency, several recommendations may be made for future research.
First, a strong relationship was found between the knowledge of lexical collocations and speaking proficiency among the Taiwanese university EFL learners in the study. However, the subjects’ use of lexical collocations did not show significant correlation with either knowledge of lexical collocations or their speaking proficiency in the study. Thus, Taiwanese EFL learners’ use of collocations could be further examined.

In the current study, the researchers used only one type of speaking test in the study to elicit the subjects’ use of lexical collocations. Future studies could consider using different discourse tasks for eliciting the subjects’ spoken data, such as face-to-face conversations, interviews, and speeches. Second, the current study aimed at the subjects’ knowledge and use of collocations in relation to their speaking proficiency. Future research could be extended to explore the relationship between knowledge of collocations and other language skills, including listening, reading, and writing. It could provide a better understanding of the connection between Taiwanese EFL learners’ collocational knowledge and their general English proficiency. Lastly, the study also found Taiwanese EFL students performed differently on the five major categories of lexical collocations. The students performed the worst on the L4 subtype (adv + adj) both in their knowledge and use of lexical collocations. Therefore, it could be worthwhile to further investigate the patterns of development in the EFL learners’ knowledge and use of lexical collocations.

The above are some potential directions this current study can recommend for future research on collocations. In fact, this study has stirred up more questions than it can possibly answer. The role of collocations in second language acquisition and teaching is not yet being fully understood. Much still needs to be done, particularly on the possible relationship between the acquisition of lexical collocations and EFL learners’ language skills. It would take a lot of joint efforts from the fields of first and language acquisition, EFL/ESL, and applied linguistics to uncover.

Endnotes:
1. The present study focuses exclusively on lexical collocations, rather than grammatical collocations. Henceforth, throughout this study lexical collocations and collocations will be used interchangeably.
2. L. Hsu (2005) refers to Li-chen Hsu; Hsu (2002) is used for Jeng-yih Hsu.
3. Both Zhang (1993) and Hsu (2002) convert their subjects’ length of written/spoken data for a possible comparison even though this could be a compromise for accurate data collection.
4. Henceforth, the term Pearson correlation will be used in the study.
5. In the spoken data obtained from Per Film test, no L4 subtype lexical collocation was used.
6. The spoken data obtained from Per Film test for ULC on L4 subtype was zero.

References
Aghbar, A. A. (1990 October). Fixed expressions in written texts: Implications for assessing writing sophistication. Paper presented at a meeting of the English Association of Pennsylvania State System Universities (ERIC Document Reproduction Service No. 352 808). 

Aghbar, A. A. & Tang, H. (1991). Partial credit scoring of cloze type items. Paper presented at 1991 Language Testing Research Colloquium, Education Testing Service, Princeton, NJ, 1-24.

Al-Zahrani, M. S. (1998). Knowledge of English lexical collocations among male Saudi college students majoring in English at a Saudi university. Unpublished doctoral dissertation, Indiana University of Pennsylvania, Pennsylvania.

Aston, G. (1995). Corpora in language pedagogy: Matching theory and practice. In G. Cook & B. Seidlhofer (Eds.), Principle and practice in applied linguistics: Studies in honour of H. G. Widdowson (pp. 257-270). Oxford: Oxford University Press.

Bachman, L. F. (2004). Statistical analysis for language assessment. Cambridge: Cambridge University Press.

Benson, M., Benson, E., & Ilson, R. (1997). The BBI dictionary of English word combinations. Amsterdam: John Benjamins.

Brown, D. (1974). Advanced vocabulary teaching: The problem of collocation. RELC Journal, 5(2), 1-11.

Chafe, W. (Ed.). (1980). The pear stories: Cognitive, cultural, and linguistic aspects of narrative production. Volume III in the series. Norwood, NJ: Ablex Publishing Co.

Channell, J. (1981). Applying semantic theory to vocabulary teaching. English Language Teaching Journal, 35, 115-122.

Chen, P. C. (2002). A corpus-based study of the collocational errors in the writings of the EFL learners in Taiwan. Unpublished master thesis, National Taiwan Normal University, Taiwan.

Ellis, N. C. & Schmidt, R. (1997). Morphology and longer distance dependencies: Laboratory research illuminating the A in SLA. Studies in Second Language Acquisition, 19, 145-171.

Fillmore, J. C. (1979). On fluency. In C. J. Fillmore, D. Kempler, & W. Wang (Eds.), Individual differences in language ability and behavior (pp. 85-101). New York: Academic Press.

Gay, L. R. (1996). Educational research: Competencies for analysis and application. Englewood Cliffs, NJ: Merrill.

Haslam, S. A. & McGarty, C. (2003). Research methods and statistics in psychology. London: Sage Publications Ltd.

Huang, L. S. (2001). Knowledge of English collocations: An analysis of Taiwanese EFL learners. Texas papers in foreign language education: Selected proceedings from the Texas foreign language education conference 2001, 6(1), 113-129. (ERIC Document Reproduction Service No. 465 288).

Hsu, J. Y. (2002). Development of collocational proficiency in a workshop on English for general business purposes for Taiwanese college students. Unpublished doctoral dissertation, Indiana University of Pennsylvania, Pennsylvania.

Hsu, L. C. (2005). The effect of lexical collocation instruction on Taiwanese college EFL learners’ listening comprehension. Unpublished master thesis, National Kaohsiung First University of Science and Technology, Taiwan.

Hsueh, S. C. (2003). An analysis of lexical collocational errors in the English compositions of junior high school EFL students. Unpublished master thesis, National Kaohsiung Normal University, Taiwan.

Kjellmer, G. (1991). A mint of phrases. In K. Aijmer & B. Altenberg (Eds.), English corpus linguistics: Studies in honour of Jan Svartvik (pp. 111-127).  Harlow, Essex: Longman.

Lewis, M. (1993). The lexical approach: The state of ELT and a way forward. London: Language Teaching Publications.

Lewis, M. (1997). Implementing the lexical approach: Putting theory into practice. London: Language Teaching Publications.

Lewis, M. (2000). Teaching collocation: Further developments in the lexical approach. London: Language Teaching Publications.

Lien, H. Y. (2003). The effect of collocation instruction on the reading comprehension of Taiwanese college students. Unpublished doctoral dissertation, Indiana University of Pennsylvania, Pennsylvania.

Lin, Y. P. (2002). The effects of collocation instruction on English vocabulary development of junior high school students in Taiwan. Unpublished master thesis, National Kaohsiung Normal University, Kaohsiung, Taiwan.

Liu, C. P. (1999a). A study of Chinese culture university freshmen’s collocational competence: “Knowledge” as an example. Hwa Kang Journal of English Language & Literature, 5, 81-99.

Liu, C. P. (1999b). An analysis of collocational errors in EFL writings. The Proceedings of the 8th International Symposium on English Teaching, (pp. 483-494), Taipei: Crane.

Liu, C. P. (2000). An empirical study of collocation teaching. The Proceedings of the 17th International Symposium on English Teaching, (pp. 165-178), Taipei: Crane.

McArthur, T. (1992). The Oxford companion to the English language. Oxford: Oxford University Press.

McCarthy, M. J. (1984). A new look at vocabulary in EFL. Applied Linguistics, 5(1), 12-22.

McCarthy, M. (1990). Vocabulary. Oxford: Oxford University Press.

Mertens, D. M. (1997). Research methods in education and psychology: Integrating diversity with quantitative & qualitative approaches. California: Sage Publications, Inc.

Nation, I. S. P. (1990). Teaching and learning vocabulary. New York: Newbury House.

Nation, I. S. P. (2001). Learning vocabulary in another language. Cambridge: Cambridge University Press.

Nattinger, J. (1980). A lexical phrase grammar for ESL. TESOL Quarterly, 14(3), 337-344.

Nattinger, J. & DeCarrico, J. (1992). Lexical phrases and language teaching. Oxford: Oxford University Press.

Pawley, A. & Syder, F. H. (1983). Two puzzles for linguistic theory: Nativelike selection and nativelike fluency. In J. C. Richards & R. W. Schmidt (Eds.), Language and communication (pp. 191-226). New York: Longman.

Richards, J. C. & Rogers, T. S. (2001). Approaches and methods in language teaching (2nd ed.). Cambridge: Cambridge University Press.

Schmitt, N. (2000). Vocabulary in language teaching. Cambridge: Cambridge University Press.

Scholfield, P. (1997). Vocabulary reference works in foreign language learning. In N. Schmitt & M. McCarthy (Eds.), Vocabulary: Description, acquisition, and pedagogy (pp. 279-302). Cambridge: Cambridge University Press.

Schweigert, W. A. (1994). Research methods & statistics for psychology. Pacific Grove, California: Cole Publishing Company.

Sung, J. (2003). English lexical collocations and their relation to spoken fluency of adult non-native speakers. Unpublished doctoral dissertation, Indiana University of Pennsylvania, Pennsylvania.

Tang, Y. T. (2004). A study of the collocation errors in the oral and written production of the college students in Taiwan. Unpublished master thesis, National Taiwan Normal University, Taiwan.

Tseng, F. P. (2002). A study of the effects of collocation instruction on the collocational competence of junior high school students in Taiwan. Unpublished master thesis, National Taiwan Normal University, Taiwan.

Validation summary for PhonePass (2005). ORDINATE Corporation Technical Report.

Weiss, N. A. (2005). Introductory statistics. Boston: Pearson Education, Inc.

Willis, D. (1990). The lexical syllabus: A new approach to language teaching. London: Harper Collins Publishers.

Wu, L. H. (2005). A study of English verb-noun collocational knowledge of technological university English majors in Taiwan. Unpublished master thesis, National Kaohsiung First University of Science and Technology, Taiwan.

Wu, W. S. (1996). Lexical collocations: One way to make passive vocabulary active. The Proceedings of the 11th Conference on English Teaching and Learning in the Republic of China, (pp. 461-480). Taipei: Crane.

Zhang, X. (1993). English collocations and their effect on the writing of native and non-native college freshmen. Unpublished doctoral dissertation, Indiana University of Pennsylvania, Pennsylvania.

Websites
The Pear Film. Retrieved October 20, 2005, from http://www.linguistics.ucsb.edu/faculty/chafe/index.htm
PhonePass score search. Retrieved January 14, 2006, from http://www.phonepass.com.tw/

Simple Search of British National Corpus. Retrieved from January 20, 2006 to
 February 17, 2006, from http://sara.natcorp.ox.ac.uk/lookup.html