Key Words: Data-driven Learning, concordancing, collocation, prepositions, concordancer, KWIC, Corpus Linguistics

Introduction
Prepositions are generally troublesome to the learners for whom English is a foreign/second language (Celce-Murcia, and Larsen-Freeman 1999, p.401). Boers and Demecheleer (1998, p.197) argue that prepositions are difficult for ESL/EFL learners because they have literal as well as figurative meanings. Jimenez Catalan (1996, p.174) claims that Spanish students have difficulty with mastering English prepositions. Jabbour-Lagocki (1990, p.162) believes that English prepositions are notoriously difficult for ESL/EFL learners to master because of L1 interference. For native speakers, prepositions present little difficulty, but for a foreign/second language learner they are confusing and largely problematic. For instance, we say, we are at the hospital; or we visit a friend who is in the hospital. We lie in bed but on the couch. We watch a film at the theater but on television. All these indicating that prepositions have strong collocational relations with other elements of language, and thus they are problematic for the EFL learners.   
   Although it is generally accepted that collocations are both indispensable and the same time problematic for foreign language learners and they therefore should play an important role in second language acquisition (SLA), especially for adult learners, learners’ difficulties with collocations have not been investigated in detail by EFL practitioners so far (Nesselhauf, 2003). Both the conventional approaches (such as Grammar Translation Method) and the modern approaches (such as Communicative approach) to SLA have in different ways underplayed the role of collocations. Shei and Helen (2000) believe that collocations have been largely neglected by researchers, course designers and EFL practitioners.

    The majority of Iranian EFL learners have a good knowledge of English grammar and vocabulary; however, they seem to have serious problems with the production of collocational patterns, specifically collocations of prepositions. This inefficiency seems to be, to some extent, due to the lack of collocational knowledge among Iranian EFL students, and to a large extent, the inadequate emphasis given to collocational patterns in their textbooks, and the type of instructions they receive. Moreover, such multi-word lexemes, i.e., prepositions and their collocational patterns have not usually been a major focus of teaching and research in our country.

   The importance of prepositions and their collocational properties on the one hand for the non-native learners of English, and the problems that Iranian EFL students have with collocation of prepositions on the other hand, highlight the significance of the present study. Furthermore, there is an abundant stock of phrasal and prepositional combinations in English that represent innumerable collocations, and the mastery over them and their collocational power can affect Iranian EFL learners' fluency as well as accuracy in both speaking and writing.

   One possible solution is to use the more recent Data-driven (DDL) as it emphasizes the collocational properties of language through concordancing lines. The present paper aims at investigating the role of DDL in the teaching collocation of English prepositions to Iranian EFL adult learners.
 
2. Literature review
One problematic question that remains unresolved concerns whether or not having a large store of vocabulary and a basic knowledge of grammar are enough for fluent and successful communication in second language acquisition (Rudzka , Channell, Ostyn, and Putseys, 1985). But it is usually the case that the majority of EFL learners, even advanced ones, have various problems in their oral or written productions (Bahns and Eldaw 1993; Rudzka , et al , ibid; Taiwo, 2004). This is in spite of the fact that they apparently seem to have sufficient lexical or grammatical knowledge. Such erroneous utterances like 'the manager of the university', 'heavy tea'; ' to take fish' and 'to be bad in something' are not due to poor lexical or grammatical knowledge. These problems arise partly from lack of knowledge about the companies that words keep, i.e., collocation.

   Hill (1999, p.4) states that lack of collocational competence can be a cause of EFL students' problems in learning English collocations. It may also be suggested that one reason for the EFL students' problems in learning English prepositions is that they usually try to learn the meaning and use of prepositions individually without paying sufficient attention to their collocational properties (Flowerdew, 1999). Zarei (2002) found that Iranian EFL learners have problems with English collocations. He classified English collocational patterns into ten categories of which the collocations of prepositions are among the most problematic patterns, while 'adjective+adverbs' and 'fixed expressions' rank among the least problematic collocational patterns for them. Zarei (ibid) further concluded that knowledge of collocations is an essential part of achieving native-like competence in English. Delshad (1980), too, found that Iranian EFL/ESL students have difficulty in the use of English prepositions. According to Delshad (ibid), Iranian EFL students usually misuse or omit English prepositions.

    With the introduction of digital computers and corpus linguistics, new trends are introduced into the teaching of collocations. One of these trends that is called concordancing has begun to find its way in language teaching. Concordancing is a method of analyzing language by studying structures and lexical patterns found in digital databases. This program allows for the study of large bodies of texts called corpora with a computer program, i.e., a concordancer. The concordancer can find a selected word and list sentences or portions of sentences containing that word, called the Key-Word-In Context (KWIC). In this format, the lexical or grammatical items that collocate with the key word are sorted to the left and right side of the key word. It can also identify collocations or words most often found together with the key word. This information can provide students with information on lexical or grammatical patterns in sample sentences of real language.

   Later on, the idea of DDL was proposed by Johns (1991) as an innovative approach to the implementation of concordancing materials in the realm of second language acquisition (SLA). In DDL, a concordancer searches huge amounts of linguistic data (called Corpora). The concordancer is electronic software that depicts frequent lexical/grammatical patterns of language within authentic contexts. According to Johns (ibid) the language learner using this approach is essentially a research worker whose learning is driven by access to authentic linguistic data.

   In the same vein, Batstone (1995) claims that DDL is a pedagogic continuum from product to process. It has the advantage of product approach since the specific aspects of language are presented to the learners by multiple exposures within contexts. At the same time, it has a process approach towards learning because DDL promotes creativity and self-discovery learning among learners. While still very much a new methodology, DDL appears to utilize the strengths of both product and process approaches to teaching grammar successfully ( Hadley, 2004).

   DDL approach suggests that grammar learning should consist largely of consciousness-raising activities rather than the teaching of rules. Consciousness-raising is defined as the deliberate attempt to draw the learner's attention specifically to the formal properties of the target language. In DDL approach, learners are not seen simply as recipients of knowledge, but as researchers studying the regularity of the language. Teachers encourage learners' search without knowing in advance what patterns they will discover.

   One of the important factors that shape learners’ inter-language (IL) is transfer, and it has been a controversial issue in SLA research. Ellis (1986) remarks that SLA is strongly affected by the learners' first language, and the best support for such a claim is the foreign accent in the second language learners’ speech. Towel and Hawkins (1994) enumerate five observable phenomena about SLA of which transfer of L1 patterns into L2 is of prime significance. They further write that, “Transfer seems to affect all linguistic levels: pronunciation, syntax, morphology, lexicon and discourse, (p. 7).” Gass and Selinker (1983) believe that two interrelated processes are affecting SLA: first, the learner’s built-up body of knowledge formed based on the available second language data, and second, the learner’s utilization of the knowledge of the first language (L1) and other languages known to him or her.

   The importance of transfer as a property of second language has been evaluated differently throughout the history of SLA. Although its origin goes back to behaviorism, it remains one of the most fundamental aspects of second language acquisition. Oldin (1989) asserts that transfer has been documented to occur at all the levels of linguistic analysis such as phonology, syntax, lexis, and grammar. Although phonological transfer is most common in SLA, lexical and collocational transfer seems to be the major cause of poor proficiency.

   EFL learners have trouble where collocations are language specific. Thus, in such a case, they tend to carry over the collocational patterns of their L1 into L2 settings. Gabrys-Biskup (1992) contends that lexical transfer occurs in the learners’ use of collocations.

   Consequently, the question of transferability of collocational patterns from L1 into L2 setting is an indication of cross-linguistic effect in the context of inter-language acquisition. There is now abundant empirical evidence that first language transfer is a real and central phenomenon that must be fully taken into account in SLA (Ellis, 1986). And one of the areas of SLA that is strongly influenced by L1 is the transfer of collocational patterns (Gabrys-Biskup, ibid).

3. Research methodology
3.1. Research questions
The introduction of the DLL approach to the teaching of collocations is rather new and a number of questions and controversial issues are awaiting answers and clarification. Based on our review of the literature and the reported results of the pre-studies done in this area, this study sets itself the goal to shed some lights on the following research questions:
1. What is the role of the DDL approach in the development of collocational knowledge of prepositions among Iranian EFL students?
2. To what extent is the collocation of prepositions affected by Iranian EFL learners' L1?
3. Does collocation of prepositions exert the same degree of difficulty for different levels of language proficiency among Iranian EFL students?

3.2. Research  hypotheses
For doing this research, the afore-mentioned questions were then formulated into the following null hypotheses.
Ho.1. DDL approach has no role in the development of collocational knowledge of prepositions among the Iranian EFL students.
Ho.2. Iranian EFL learners' L1 has no significant effect on their knowledge of collocation of prepositions.
Ho.3.There is no significant difference in the collocation of prepositions in terms of difficulty for different levels of language proficiency among the Iranian EFL learners.

3.3. Participants
Two hundred English majors served as the participants of this study. They were selected through cluster random sampling from among 450 EFL majors studying at three universities in Shahrekord. At first, the Michigan Test of English Language proficiency (MTELP) was administered to determine their level of proficiency. The possible scores range from zero to one hundred. The mean score and the standard deviation of the MTELP were 54 and 21 respectively. Then according to the mean performance and the standard deviation of the test, subjects were assigned into high, mid and low groups. Those students whose scores fell one standard deviation below and above the mean were assigned as the mid group. Those subjects whose scores were two standard deviations below and above the mean were classified as the low and high groups respectively. In the next step, the participants in each low, mid and high group were randomly assigned to experimental and control groups. The schematic representation of the participants in different levels in this study was shown in Table 1. 
   
 Table 1. Schematic representation of the participants in control and experimental groups

Group	Level	No	Level	No	Level	No	Total
Control (conventional)	high	30	mid	40	low	30	100
Experimental (DDL)	high	30	mid	40	low	30	100
						total	200

3. 4. Instrumentation
In order to test the research hypotheses, four testing instruments and two types of teaching materials are developed and used by the researchers in this study. They are as follows:
1. As it was already mentioned, in the first phase of this study, the Michigan Test of English Language Proficiency (MTELP) was given to the participants to determine their levels of proficiency.

2. To determine the collocational knowledge of the participants, a completion test on collocations of prepositions was given to the students as the pre-test. This test comprised 60 items, and its reliability was estimated to be 0.86 using Kr-21 formula. This test was given to 200 participants who had been selected through cluster random sampling.

3. Having received fifteen two-hour sessions instructions on collocation of prepositions with specific treatment for each group, a completion test on collocation of prepositions was administered as the post-test in order to determine the impact of specific instructions the participants received. This test also contains 60 items, and its reliability was estimated to be .82.

4. A translation task that deliberately included problems of Iranian students with prepositions and their collocations was assigned to the students in order to determine the possible effects that students' first language might create in their development of collocational knowledge of prepositions. Two experts in TEFL (Ph.D. colleagues) were consulted for the accuracy and appropriateness of the translated texts.

5. Teaching materials on collocations of prepositions were taken from Brown Corpus (2005). These texts were concordanced by the Web Concordancer Online.

6. Teaching materials on the prepositions were selected from different grammar books such as ‘The Comprehensive English Grammar by Quirk et. al (1998), Modern English Part (I) written by Marcella Frank (1993) and English Idioms by Seidl and  McMordie ((1987). All the teaching materials in the conventional and DDL groups were at the same level of difficulty for the participants.

 3.5. Procedures
As it was already mentioned, at first the participants were randomly divided into control and experimental groups. In the second stage, the Michigan Test of English Language proficiency (MTELP) was used to determine the overall proficiency of the participants. Then the mean score and the standard deviation of MTELP were used as criteria for dividing the participants into low, mid and high levels of proficiency. Altogether six groups participated in this study, i.e. three experimental (DDL approach) and three control groups (conventional approach). In the third stage, they were pre-tested by a completion test on collocation of prepositions. The reliability coefficient of this test was estimated to be 0.86 using the Kr-21 formula. In the fourth stage, the participants attended English classes one hour per week in a fifteen-week semester. The structures taught to the students were prepositions and their collocational properties. The rational behind such a decision was that most of the EFL students showed to have problems with the selection and use of English prepositions and their collocational properties.

   To conduct the present study, six general patterns of collocation of prepositions are recognized and investigated by the researchers. These patterns are as follows:
1. adjective + preposition collocation: good at ,bored with, superior to, tired of
2. preposition + noun collocation: in astonishment, with embezzlement, on credit, with 
   distress, by car
3. noun + preposition collocation:  motivation in, admiration for, argument about  
4. verb + preposition collocation:  consist of, insist on, indulge in, resist in
5. preposition + preposition collocation:  out of, next to, in front of, in lieu of
6. idiomatic expressions:  to be at best, to be in the air

     Then the participants went through a fifteen-session treatment, i.e., the control groups underwent a conventional instruction in which prepositions and their collocational patterns were explicitly taught to the participants in English or Farsi. The materials used for the conventional-based instruction were taken from conventional grammar books such as Practical English Grammar, Comprehensive English Grammar and English Idioms. On the other hand, the experimental groups received a data driven-based instruction that was based on concordancing lines presented in KWIC format. The concordances were taken from the Brown Corpus Online (2005) which were searched by the Web Concordancer. These lines were then given to the participants in printouts. The materials which were taught to the two groups were at the same level of difficulty. Finally, after the specific treatments given to each group, a post-test was administered to check the effects of the instructions. 

For the second research hypothesis, errors of collocations were extracted and classified. Since there was not a systematic approach to the classification of collocational errors, the researchers used Jack C. Richards' taxonomy of errors. In his taxonomy, Richards (1974) classifies L2 errors into two major categories, i.e., inter-lingual and intra-lingual. According to Richards, while inter-lingual errors are due to the interference of L1 collocational patterns into L2 settings, intra-lingual errors reflect the arbitrary and unpredictable nature of the English collocations, as well as the learners' lack of adequate knowledge of L2 collocations. Blank items were excluded in this study. Finally, the post-test was given in order to compare the mean performance of the groups.

4. Data analysis and results
Concerning the first research question, the researchers intended to investigate whether the effect of DDL instruction on the development of collocational knowledge of prepositions among Iranian EFL students was significant. To answer this question, the researchers formulated the first null hypothesis:
Ho.1. DDL approach has no role in the development of collocational knowledge of prepositions among Iranian EFL students.
     
   To statistically verify the first research hypothesis, the collected data were subjected to the statistical analysis of variance (one-way ANOVA). The results of this analysis showed that there was a significant difference between the performance of the participants in the DDL group and conventional group. Table 2 displays the results of the ANOVA for the first research hypothesis.

Table 2. The results of ANOVA (one-way) for the first research hypothesis(The comparison between conventional approach and DDL approach)