turkishenglish.com - Comparison of relative clauses in ESL students




            This study was undertaken in order to further explore the phenomenon of avoidance.  In particular, the study sought to determine whether or not the supposed avoidance of RCs by L2 learners with prenominal RCs in their NL could be accounted for in terms of the function and distribution of RCs in discourse.  A comparison of the IL production of Spanish and Korean students revealed that the Korean students produced significantly fewer RCs than the Spanish students.  This finding supported Hypothesis 1, which posited that because of the prenominal position of RCs in their NL, Korean students would produce fewer RCs in English.  It was suggested that a large part of the difference in total production was due to three primary factors related to the distribution of RCs in discourse and to discourse level strategies involving RCs. 

            The first difference was the distribution of the GR of the NPRel and the Head Noun between the two groups.  The Korean students demonstrated a preference for S relatives over O and X relatives.  The Spanish students used more O relatives than S relatives.  The Korean pattern of distribution is best explained as transfer from their NL, and not as avoidance of O relatives (cf. Gass 1980).  This study is the first I know of to demonstrate transfer on this level, and as such may be of interest to applied linguists still trying to pin down exactly what aspects of an NL are amenable to transfer in SLA. 

            The different distribution of the NPRel was indicative of some differences with regard to how the RCs functioned in discourse.  The Spanish students used more RCs which performed a grounding function and/or which linked a referent to a first or second person pronominal form.  Spanish students and probably native English speakers as well (cf.  Fox and Thompson 1990:304) have a strong tendency to link referents to human NPs in the form of pronouns in order to make them relevant.  The Korean students rarely used RCs in this way and this contributes to their underproduction with respect to the NL frequency.

            The second difference was the type of Head Noun used by the two groups.  The Spanish data had a much higher incidence of indefinite pronouns and generic Head Nouns that lend themselves to modification by RCs.  The Korean students used very few of these, though these forms are acceptable and common in their NL. 

            The third difference was the discourse function of the RCs.  Spanish students used more RCs to identify known referents with old information than the Korean students.  They also used more RCs for what I termed “topical cohesion.”  

            Taken together, these three differences in usage of RCs point toward an explanation for the underproduction of RCs by the Korean students in terms of discourse processes.  In other words, these students organize and arrange text in such a way that their production of RCs is lower than that of the Spanish students.  There was, however, some evidence that also suggested students might be consciously avoiding the structure.

            The Korean students produced more functionally non-native RCs than the Spanish students, and over half of these were clear cases of NL transfer.  They commented on the difficulty they had using RCs correctly, and though they frequently used RCs on class compositions, on a midterm placement exam they used few or no RCs.  This could be interpreted as an awareness on the part of the Korean students of their own difficulty with the construction which consequently leads to avoidance of the structure.  The source of avoidance seems to be, at least on a conscious level, a problem not with the morphosyntax of English RCs but rather with how to use them when encoding particular propositions - a question of function. 

            In summary, this study supports a growing consensus among many researchers in SLA who feel that the influence of the mother tongue is felt most keenly at the level of discourse and not syntax.  Rutherford (1983) noticed that Chinese students produced more sentential subjects than students from other countries.  This was interpreted as the influence of the topic-comment structure of Chinese.  Rutherford summarized the attitudes of many researchers saying, “I take these observations  as evidence that it is therefore discourse and not syntax that gives gross overall shape to interlanguage.”  Schacter (1974) thoroughly exposed the inadequacies of Error Analysis since EA could not detect areas of difficulty faced by the learners which manifested themselves as avoidance or underproduction.  She supported a return to the strong version of the CA Hypothesis.  This study suggests that, in conjunction with CA, interlanguage data be subjected to more rigorous discourse analysis in order to determine more precisely the influence of the NL on the SLA process.  This is advocated because a simple contrastive analysis, though it does reveal the differences between languages, is still not able to predict how these differences will affect the learners' IL competence.  Current research into the effect of NL discourse structure on SLA is promising and should be continued.     



            A clear implication of this study is that students with prenominal RCs in their NL might benefit from special attention to this structure in the classroom.  This would entail not only a careful analysis of the functions of RCs in native texts but also the development of pedagogical materials which focus on the function of RCs in the organization of discourse.  Until more detailed analyses of native English text have been done, it is difficult to offer concrete advice as to how these students should be instructed, but I will hazard a few suggestions.  First, Korean students could be given practice using RCs in conjunction with Predicate Nominal Heads to frame new sub-topics, to describe new terms, and to focus information.  Second, Korean students could be given sentence combining exercises which include presupposed information used for grounding, especially with first and second person pronouns.  Sentence combining exercises are frequently used in the classroom, but the sentences in these combining exercises are not usually logically presupposed in relation to one another.  In addition to this, writing teachers might discuss the explicitness of English writing and demonstrate how RCs can facilitate this.  These meager proposals indicate the need for more research into exactly how it is that English is used by native speakers, and this in turn will affect the teaching of English to second language learners.


Areas for Further Study

            This study has raised several important questions regarding transfer and second language acquisition.  The structure of the IL of the Korean students has obviously been influenced by the NL, particularly with respect to the GR of the NPRel.  One puzzling question in this study, however, was why Korean students strongly tend not to transfer their NL use of generic Heads with RCs in their acquisition of English.  To help in answering this question, further study should be done to determine to what extent Korean students and Spanish students differ with respect to the use of complement clauses, which are closely related to RCs with generic Heads both in form and meaning.

            Another question raised by this study is “What role does the Accessibility Hierarchy play, if any, in SLA?”  Previous studies have claimed that the AH, as a language universal, played an important role in the acquisition of language (cf. Gass 1980).  The distribution of the NPRel in the Korean data, especially the early stages, is very much like the NL distribution, which happens to reflect the AH very well.  But, as we see, there are two possible explanations for the distribution - the AH and mother tongue influence. The Spanish students, on the other hand, use more O relatives (33%) than S relatives (31.2%), and while the percentages are very close, there is certainly no preference for S relatives as would be expected if the AH were playing a significant role in their  acquisition of RCs.  I have found no studies as to whether the production of the Spanish students is like that of their native language, but since the AH has not exerted an overwhelming influence on their production we have reason to believe that the distribution of the IL has been influenced by the NL.  Still more convincing evidence would be obtained by looking at the IL of English speakers learning either Korean or Japanese.  These languages have a predilection for S relatives while English has no preference for these relatives.  If an English speaker learning either of these languages produced an equal number of S and O relatives, we would have very strong evidence that it is the distribution of the NL and not the AH which affects SLA most heavily.     

            Another area for further research is the extent to which the position of the RC with respect to the Head Noun affects the distribution and function of RCs in discourse.  Relative clauses in Korean and Japanese have very similar distributions and functions, but they are also closely related languages.  What about other unrelated languages with prenominal RCs?  Are they similar to Korean and Japanese?  The same question holds for languages with postnominal RCs.  As a matter of fact, the issue may be even more crucial for these languages since very few studies of any kind have been done on these languages because these speakers seem to have very little trouble with RCs in English.  Is the ease with which they acquire RCs in English due to similarities beyond the position of the RC with respect to the Head Noun?  Can it be linked to the distribution of the GR of the NPRel and to the functions of RCs in discourse?

            In addition, languages which are a sort of "mixed bag" typologically speaking should also be researched.  For example, Chinese is an SVO language, but has prenominal RCs.  According to Zhao (1989), RCs in Chinese do not encode new asserted information.  Could this be due to the tendency of SVO languages to put new asserted information in clause final position?  As a matter of fact, RCs seem to  perform a very limited number of functions in Chinese.  If typologies are more than the creation of imaginative linguists (and I think they are), then this suggests that VO languages, for instance, may be better suited for postnominal RCs and when this pattern is broken, the RC is not as efficient or versatile in encoding information. 

            If it could be demonstrated that typological parameters, such as word order, have ramifications cross-linguistically for discourse level phenomena like those discussed, then applied linguists might gain new insight into how language distance, specifically typological distance, affects SLA.  Though some are skeptical about whether or not a better understanding of language acquisition processes will be pedagogically useful (Corder 1983), this is definitely a question of interest to theoretical linguists, particularly those interested in language typology and universals.  We should also be careful about pre-judging the usefulness of any valid linguistic discovery. 

            Finally, I would like to make a suggestion about the design of future study in this area.  This study compared the IL production of RCs in free compositions.  This was done so that the production of RCs would be as unconstrained as possible thereby allowing the researcher to get a more accurate picture of how RCs function in the IL of the learner.  This freedom, however, allowed students to develop their essays in a number of different ways.  To better control for variables related to content and organization, it might be profitable to design a study similar to the Pear Stories project (Chafe 1980).  In this study, subjects were shown a short silent movie and then asked to recount the events in their own words.  This would constrain the macro-structure and content of the data somewhat, while still allowing the subject considerable freedom as to how the information should be encoded.  

Next: Appendix A