论文部分内容阅读
Abstract：This study aims to investigate whether the reliability of lexical density as a reliable measurement to investigate the relationship between English speaking and writing. A randomly?鄄selected 40 college?鄄level ESL students of one American university participated in the study. In the study, the measures of lexical density were used to examine the lexical development of speaking and writing of a group of 10 students [out of the 40 students] enrolled in a college?鄄level ESL composition course. The rank-order of the subjects?蒺 spoken and written data showed that the measures of lexical density can indicate the differences between two modes but cannot differentiate between proficiency levels.
Keywords: lexical density; English as second language; the relationship between English speaking and writing
中文摘要：本研究主要探讨了词汇密度（lexical density）在研究英语说和写，即英语说和写的相关性方面是否有效可信。本研究随机选取了40名英语作为第二语言的大学生作为被试者，根据其在写作课中的一篇作文的成绩，最终选取了写作水平最好的5名学生和最差的5名学生参与到本研究中，运用词汇密度来考察这些学生英语说和写的关系。研究表明词汇密度似乎能够有效地区别英语说与写这两种表达方式，但是却不能很好地区别学生的英语水平。
关键词：词汇密度(lexical density）；英语作为第二语言（English as a second language）；英语说和写的关系(the relationship between English speaking and writing)
I. Introduction
Since Ure (1971) first coined the term, lexical density(LD), it has been recognized as a reliable indicator for distinguishing between oral and written language. Determining LD depends on distinguishing lexical and grammatical items in a text. Halliday?蒺s(1985) defines lexical words as content words and grammatical words as function words. L1 researchers using LD have reached the agreement that lexical density can sensitively and reliably distinguish between spoken and written data （DeVito，1965；Halliday，1979). In addition, a few studies have been conducted in L2 acquisition using the indices of lexical density(Linnarud, 1986; Hyltenstam, 1988;Lauren, 2002). The LD research in L2 acquisition has indicated that it can distinguish between spoken and written data as agreed in research in L1 acquisition, but it can not sensitively indicate L2 learners?蒺 proficiency levels. The purpose of the present study is to examine the reliability of lexical density as the measurement of the relationship between speaking and writing. Specifically, this study used the measures of lexical density to investigate the lexical development of speaking and writing. The researcher was interested in finding answers to the following research question. Is lexical density a reliable measurement to examine the relationship between English speaking and writing?
II. Materials and methods
1. Subjects
The 40 subjects of the study from different countries were enrolled in a college?鄄level ESL composition course, at one American university with different majors. The length of their stay in the United States ranged from 5 days to 1.5 years with a mean of 4.3 months. The average number of years spent studying English was 8.1 years with a range of 2 to 15 years.
2. Instrument
The instrument, the Video Oral Communication Instrument (VOCI) for ESL/EFL was used to assess the oral proficiency of the subjects. The measures of lexical density were employed to examine the lexical development of both the subjects?蒺 oral and written language. VOCI is a semi?鄄direct and tape?鄄mediated speaking test, which is used as an alternative for the OPI(Oral Proficiency Interview) to determine level of oral proficiency. This study particularly used the English version, the VOCI for ESL/EFL, which was developed at San Diego State University?蒺s Language Acquisition Resource Center (LARC) by Halleck and Young (1995). The VOCI for ESL/EFL used in this study consists of a total of 23 questions assessing the four proficiency levels: novice, intermediate, advanced, and superior levels defined by the ACTFL (American Council on the Teaching of Foreign Languages) guidelines (1986).
The measures of lexical density were used to investigate the relationship between speaking and writing in college?鄄level ESL students. Three indices of lexical density were used in this study, i.e. the number of lexical words per the total number of words(LD/TNW), the number of lexical words per clause (LD/C) and the number of lexical words per T?鄄unit (LD/TU) (T?鄄unit is any independent clause with all its dependent clauses.). As far as the calculation of these indices was concerned, the total number of words included in each language sample, and the total number of T?鄄units were counted, as well as the total number of lexical words, and the total number of dependent clauses. On this basis, three indices were then determined through the following formulas:
LD/TNW=number of lexical words/number of words
LD/C=number of lexical words/ number of T?鄄units+ number of dependent clauses
LD/TU=number of lexical words/number of T?鄄units
3. Data Collection
The data of the study includes both the written and spoken data. The written samples were taken from one essay of the subjects. The essay prompt is to describe three underlying rules shaping student behavior in American classrooms. Compared to other essays written for this course, the essay is written in class, so it has less planning time than other essays that are written after class by the students. Therefore, the use of the diagnostic essays increases the comparability between the spoken and written data.
40 essays were originally collected from the second written assignment of the class. Then, according to the rating of the instructor, 20 students took part in the study: 10 high?鄄rated and 10 low?鄄rated students. Then the diagnostic essays of these 20 students were collected and graded holistically on a scale of 100 by three raters. A higher interrater reliability was found between raters one and two (r=0.91; p=0.0002) than between rater three and the other two raters, respectively. As Davies, et al. (1999) suggests, a correlation coefficient of more than 0.8 indicates a good interrater reliability. So, the rating between raters one and two was used as the evaluation of the written samples. The ratings of these two raters were averaged, which were then transferred to the rank?鄄order of the written samples from one to 10. Accordingly, five high?鄄rated and five low?鄄rated students were finally chosen to take the spoken test in this study in order to obtain the corresponding spoken sample. The average of five high?鄄rated students?蒺 written score was 90.1 and that of the five low?鄄rated students?蒺 written score was 77.1. So, there were 13 points apart between the written samples of five high?鄄rated and five low?鄄rated students.
After the written data of these 10 students were collected, the spoken data were gathered through the students?蒺 participation in the VOCI. The subjects took the VOCI in a testing room alone in order for them to feel less nervous. Of the 23 questions in the VOCI, the subjects?蒺 answers to 17 questions were used for analysis mainly according to their familiarity of the topics with these questions. These 17 questions are questions 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 15, 16, 17, 18, 21, 22 and 23. In a specific manner, questions 1 and 2 function as the practice questions in order for the students to familiarize themselves with the use of the machines. Questions 13, 14, 19 and 20 were not used for data analysis because the researcher felt that the subjects might not be familiar with the topics such as “lasting peace” (Q 13), “abstract painting” (Q 14), “free trade” (Q 19), “televising trials” (Q 20). After the collection of the spoken samples, they were rated and rank?鄄ordered from one to 10 by a certified OPI tester according to the ACTFL Guidelines (1986). After the spoken samples were gathered, they were transcribed. In regard to the spoken sample, each subject?蒺s answers to all the above?鄄mentioned 17 VOCI questions were taken as a whole to be analyzed in comparison with the subjects?蒺 written samples.
III. Research Hypotheses
As mentioned earlier, this relationship is mostly concerned with two aspects: one is about the differences between speaking and writing and the other about the correlation. Consequently, this study was conducted according to the following hypotheses.
Hypothesis one: The written samples will positively correlate with the spoken samples in relation to the measures of lexical density for all subjects.
Hypothesis two: The written samples differ greatly from the spoken sample in terms of the lexical density.
IV. Data Analysis
Three procedures were carried out to analyze statistical data in regard to the purpose of this study. First, the objective measures of lexical density were marked and tallied in both the spoken and written samples. With respect to lexical density, means of each measure were computed separately concerning high?鄄and low?鄄rated samples and also were calculated separately regarding the spoken and written samples. Second, the data were analyzed using version 3.03 of the statistical software, GraphPad Prism. Pearson product?鄄moment correlation was carried out to demonstrate how each measure in the spoken and written data correlated with each other. Finally, the level of significance was computed with version 8 of SAS software to examine whether the results obtained were statistically significant or not. Results were considered significant at the p<.05 level.
V. Results and Discussion
This section first demonstrates the rank?鄄order of the subjects in light of their speaking and writing performance. And then, results of the measures of lexical density are analyzed to examine the correlation and differences between the spoken and written samples.
1. The Correlation between Spoken and Written Samples with Measures of
Lexical Density of All Subjects
Table 1 indicates the correlation between spoken and written data. From the table, the spoken and written data correlated positively with regard to the three indices with the correlation coefficients of LD/TNW being 0.35, that of LD/C being 0.15 and that of LD/TU being 0.75. Of these three measures, only LD/TU demonstrated a relatively strong and significant correlation between the two samples.
Table 1: The Correlation between Spoken and Written Data with Regard to the Measures of Lexical Density for All Subjects (*significant at p<.05)
From the table, we see that a positive correlation was found between the spoken and written data with respect to one of measures of lexical density. Only LD/TU indicated a relatively strong and significant correlation between two samples. That is to say, students with high proficiency in an L2 produced more lexical words per T-unit when they speak and write than those with low proficiency. Therefore, this finding proves that LD/TU can sensitively demonstrate the relationship between speaking and writing concerning the levels of proficiency. This result will be supported even further in the later section.
To conclude, significant correlation between the two samples was not found concerning all the measures of lexical density except LD/TU. Therefore, it can be concluded that measures of lexical density are not good indicators of levels of proficiency. The findings of this section agree with those of L2 studies (e.g, Linnarud, 1986; Lauren, 2002) with the measures of lexical density in that the measures of lexical density, specifically, LD/TNW and LD/C cannot differentiate between L2 learners?蒺 proficiency levels. Therefore, the measures of lexical density cannot differentiate between proficiency levels, so they cannot serve to measure whether there is a positive relationship between the spoken and written data.
2. The Differences between Speaking and Writing with Measures of Lexical
Density
This section is particularly devoted to the analysis of the differences between speaking and writing in relation to the measures of lexical density. Table 2 reflects the mean differences between these two modes with respect to three measures of lexical density. As noted in the table, all the measures demonstrated a higher density in the written discourse than in the speaking discourse. In a specific manner, LD/TNW was 0.36 lexical words/number of words in the spoken data and 0.43 lexical words/number of words in the written data. With respect to LD/C, the spoken samples showed a lower index (LD/C=2.53 lexical words/clause) than the written samples (LD/C=3.43 lexical words/clause). Concerning LD/TU, it was lower in the spoken samples (LD/TU=3.93 words/T-unit) than in the written samples (LD/TU=6.26). Moreover, the differences between these two modes were found very significant in relation to all the measures of lexical density.
Table 2: Average of All the Measures of Lexical Density of Spoken and Written Data (*Significant at p<.05)
The results obtained from Table 2 show that written samples overall are more lexically dense than the spoken samples. This finding is in line with L2 studies (e.g. Hyltenstam, 1988) using the measures of lexical density in that written production has a higher density than spoken production. As far as these three measures are concerned, it seems that LD/TU is more sensitive than the other two measures, since it indicated the largest difference between the two modes as shown by the percentage difference as indicated in the table. The percentage difference is computed by subtracting the means of each measure of the written data from that of the spoken data and then dividing the number by the mean of the spoken data. This figure is used to show how much difference is between the two modes. As shown in the table, the percentage difference of LD/TU is 59% followed by LD/C being 36% and LD/TNW being 19%. So, it can be said that LD/TU is the most sensitive indicator of the difference between the two modes with LD/C being the second best indicator and LD/TNW the third best indicator.
In sum, the spoken samples are shown in the present study to have more lexical density than the written samples. And, of all the measures of lexical density, LD/TU seems to be the best indicator of the differences between the spoken and written data.
VII. Conclusion
The conclusion of the study is presented by a discussion of outcomes of the hypotheses and implications of the results. Hypothesis 1 investigates the correlation between speaking and writing with the measures of lexical density. The study demonstrated that statistical difference was only found with regard to LD/TU. Concerning the other two measures of lexical density, they indicated a low and non?鄄significant correlation. Hypothesis 2 is concerned with the differences between two samples. The study showed that all the measures of lexical density were higher in the written samples than in the spoken samples. Besides, such difference was found statistically significant in relation to all the measures. Therefore, this hypothesis is sustained. The findings of the present study seemed to indicate that there was a positive relationship between speaking and writing according to the rank?鄄order of the subjects. However, this implication cannot be generalized, because other aspects of this relationship have to be considered to get an overall picture of this relationship. The present study found that the measures of lexical density are good indicators of the differences between the two modes but cannot differentiate between proficiency levels. Therefore, the indices of lexical density can be adopted to carry out L2 teaching and research concerned with the difference between the two modes.
References
[1] Beaman, K. (1984). Coordination and subordination revised: Syntactic complexity in spoken and written narrative discourse. In D.Tannen (Ed.), Coherence in spoken and written discourse. Norwood, NJ: Ablex.
[2] DeVito, J.A.1965: A linguistic analysis of spoken and written language. Central States Speech Journal 18, 81-85.
[3] Halleck, G.B. & Young, R.F. (1995). Video oral communication instrument (VOCI) for ESL/EFL. Language Acquisition Research Center, San Diego, CA.
[4] Halliday, M.A.K. (1979). Differences between spoken and written language: Some implications for literacy teaching. In Page, G. Elkins, J. & O?蒺Connor, B. (Eds). Communication through reading proceedings of the fourth Australian reading conference (pp. 37-52) Adelaide, SA: Aust. Read. Assoc.
[5] Halliday, M.A.K. (1985). Spoken and written language. Deakin University Press.
[6] Hyltenstam, K.(1988). Lexical characteristics of near?鄄native second?鄄language learners of Swedish. Journal of Multilingual and Multicultural Development 9 (1&2), 67-84.
[7] Lauren, U.(2002). Some lexical features of immersion pupils?蒺 oral and written narration. Working Papers 50, 63-78.
[8] Linnarud, M. (1986). Lexis in composition: A performance analysis of Swedish learners?蒺 written English. CWL Gleerup.
[9] Ure, J. (1971). Lexical density and register differentiation. In Perren, G. E. & Trim, J.L.M (Eds). Applications of linguistics (pp.443-452). Cambridge: Cambridge University Press,443-452.
Keywords: lexical density; English as second language; the relationship between English speaking and writing
中文摘要：本研究主要探讨了词汇密度（lexical density）在研究英语说和写，即英语说和写的相关性方面是否有效可信。本研究随机选取了40名英语作为第二语言的大学生作为被试者，根据其在写作课中的一篇作文的成绩，最终选取了写作水平最好的5名学生和最差的5名学生参与到本研究中，运用词汇密度来考察这些学生英语说和写的关系。研究表明词汇密度似乎能够有效地区别英语说与写这两种表达方式，但是却不能很好地区别学生的英语水平。
关键词：词汇密度(lexical density）；英语作为第二语言（English as a second language）；英语说和写的关系(the relationship between English speaking and writing)
I. Introduction
Since Ure (1971) first coined the term, lexical density(LD), it has been recognized as a reliable indicator for distinguishing between oral and written language. Determining LD depends on distinguishing lexical and grammatical items in a text. Halliday?蒺s(1985) defines lexical words as content words and grammatical words as function words. L1 researchers using LD have reached the agreement that lexical density can sensitively and reliably distinguish between spoken and written data （DeVito，1965；Halliday，1979). In addition, a few studies have been conducted in L2 acquisition using the indices of lexical density(Linnarud, 1986; Hyltenstam, 1988;Lauren, 2002). The LD research in L2 acquisition has indicated that it can distinguish between spoken and written data as agreed in research in L1 acquisition, but it can not sensitively indicate L2 learners?蒺 proficiency levels. The purpose of the present study is to examine the reliability of lexical density as the measurement of the relationship between speaking and writing. Specifically, this study used the measures of lexical density to investigate the lexical development of speaking and writing. The researcher was interested in finding answers to the following research question. Is lexical density a reliable measurement to examine the relationship between English speaking and writing?
II. Materials and methods
1. Subjects
The 40 subjects of the study from different countries were enrolled in a college?鄄level ESL composition course, at one American university with different majors. The length of their stay in the United States ranged from 5 days to 1.5 years with a mean of 4.3 months. The average number of years spent studying English was 8.1 years with a range of 2 to 15 years.
2. Instrument
The instrument, the Video Oral Communication Instrument (VOCI) for ESL/EFL was used to assess the oral proficiency of the subjects. The measures of lexical density were employed to examine the lexical development of both the subjects?蒺 oral and written language. VOCI is a semi?鄄direct and tape?鄄mediated speaking test, which is used as an alternative for the OPI(Oral Proficiency Interview) to determine level of oral proficiency. This study particularly used the English version, the VOCI for ESL/EFL, which was developed at San Diego State University?蒺s Language Acquisition Resource Center (LARC) by Halleck and Young (1995). The VOCI for ESL/EFL used in this study consists of a total of 23 questions assessing the four proficiency levels: novice, intermediate, advanced, and superior levels defined by the ACTFL (American Council on the Teaching of Foreign Languages) guidelines (1986).
The measures of lexical density were used to investigate the relationship between speaking and writing in college?鄄level ESL students. Three indices of lexical density were used in this study, i.e. the number of lexical words per the total number of words(LD/TNW), the number of lexical words per clause (LD/C) and the number of lexical words per T?鄄unit (LD/TU) (T?鄄unit is any independent clause with all its dependent clauses.). As far as the calculation of these indices was concerned, the total number of words included in each language sample, and the total number of T?鄄units were counted, as well as the total number of lexical words, and the total number of dependent clauses. On this basis, three indices were then determined through the following formulas:
LD/TNW=number of lexical words/number of words
LD/C=number of lexical words/ number of T?鄄units+ number of dependent clauses
LD/TU=number of lexical words/number of T?鄄units
3. Data Collection
The data of the study includes both the written and spoken data. The written samples were taken from one essay of the subjects. The essay prompt is to describe three underlying rules shaping student behavior in American classrooms. Compared to other essays written for this course, the essay is written in class, so it has less planning time than other essays that are written after class by the students. Therefore, the use of the diagnostic essays increases the comparability between the spoken and written data.
40 essays were originally collected from the second written assignment of the class. Then, according to the rating of the instructor, 20 students took part in the study: 10 high?鄄rated and 10 low?鄄rated students. Then the diagnostic essays of these 20 students were collected and graded holistically on a scale of 100 by three raters. A higher interrater reliability was found between raters one and two (r=0.91; p=0.0002) than between rater three and the other two raters, respectively. As Davies, et al. (1999) suggests, a correlation coefficient of more than 0.8 indicates a good interrater reliability. So, the rating between raters one and two was used as the evaluation of the written samples. The ratings of these two raters were averaged, which were then transferred to the rank?鄄order of the written samples from one to 10. Accordingly, five high?鄄rated and five low?鄄rated students were finally chosen to take the spoken test in this study in order to obtain the corresponding spoken sample. The average of five high?鄄rated students?蒺 written score was 90.1 and that of the five low?鄄rated students?蒺 written score was 77.1. So, there were 13 points apart between the written samples of five high?鄄rated and five low?鄄rated students.
After the written data of these 10 students were collected, the spoken data were gathered through the students?蒺 participation in the VOCI. The subjects took the VOCI in a testing room alone in order for them to feel less nervous. Of the 23 questions in the VOCI, the subjects?蒺 answers to 17 questions were used for analysis mainly according to their familiarity of the topics with these questions. These 17 questions are questions 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 15, 16, 17, 18, 21, 22 and 23. In a specific manner, questions 1 and 2 function as the practice questions in order for the students to familiarize themselves with the use of the machines. Questions 13, 14, 19 and 20 were not used for data analysis because the researcher felt that the subjects might not be familiar with the topics such as “lasting peace” (Q 13), “abstract painting” (Q 14), “free trade” (Q 19), “televising trials” (Q 20). After the collection of the spoken samples, they were rated and rank?鄄ordered from one to 10 by a certified OPI tester according to the ACTFL Guidelines (1986). After the spoken samples were gathered, they were transcribed. In regard to the spoken sample, each subject?蒺s answers to all the above?鄄mentioned 17 VOCI questions were taken as a whole to be analyzed in comparison with the subjects?蒺 written samples.
III. Research Hypotheses
As mentioned earlier, this relationship is mostly concerned with two aspects: one is about the differences between speaking and writing and the other about the correlation. Consequently, this study was conducted according to the following hypotheses.
Hypothesis one: The written samples will positively correlate with the spoken samples in relation to the measures of lexical density for all subjects.
Hypothesis two: The written samples differ greatly from the spoken sample in terms of the lexical density.
IV. Data Analysis
Three procedures were carried out to analyze statistical data in regard to the purpose of this study. First, the objective measures of lexical density were marked and tallied in both the spoken and written samples. With respect to lexical density, means of each measure were computed separately concerning high?鄄and low?鄄rated samples and also were calculated separately regarding the spoken and written samples. Second, the data were analyzed using version 3.03 of the statistical software, GraphPad Prism. Pearson product?鄄moment correlation was carried out to demonstrate how each measure in the spoken and written data correlated with each other. Finally, the level of significance was computed with version 8 of SAS software to examine whether the results obtained were statistically significant or not. Results were considered significant at the p<.05 level.
V. Results and Discussion
This section first demonstrates the rank?鄄order of the subjects in light of their speaking and writing performance. And then, results of the measures of lexical density are analyzed to examine the correlation and differences between the spoken and written samples.
1. The Correlation between Spoken and Written Samples with Measures of
Lexical Density of All Subjects
Table 1 indicates the correlation between spoken and written data. From the table, the spoken and written data correlated positively with regard to the three indices with the correlation coefficients of LD/TNW being 0.35, that of LD/C being 0.15 and that of LD/TU being 0.75. Of these three measures, only LD/TU demonstrated a relatively strong and significant correlation between the two samples.
Table 1: The Correlation between Spoken and Written Data with Regard to the Measures of Lexical Density for All Subjects (*significant at p<.05)
From the table, we see that a positive correlation was found between the spoken and written data with respect to one of measures of lexical density. Only LD/TU indicated a relatively strong and significant correlation between two samples. That is to say, students with high proficiency in an L2 produced more lexical words per T-unit when they speak and write than those with low proficiency. Therefore, this finding proves that LD/TU can sensitively demonstrate the relationship between speaking and writing concerning the levels of proficiency. This result will be supported even further in the later section.
To conclude, significant correlation between the two samples was not found concerning all the measures of lexical density except LD/TU. Therefore, it can be concluded that measures of lexical density are not good indicators of levels of proficiency. The findings of this section agree with those of L2 studies (e.g, Linnarud, 1986; Lauren, 2002) with the measures of lexical density in that the measures of lexical density, specifically, LD/TNW and LD/C cannot differentiate between L2 learners?蒺 proficiency levels. Therefore, the measures of lexical density cannot differentiate between proficiency levels, so they cannot serve to measure whether there is a positive relationship between the spoken and written data.
2. The Differences between Speaking and Writing with Measures of Lexical
Density
This section is particularly devoted to the analysis of the differences between speaking and writing in relation to the measures of lexical density. Table 2 reflects the mean differences between these two modes with respect to three measures of lexical density. As noted in the table, all the measures demonstrated a higher density in the written discourse than in the speaking discourse. In a specific manner, LD/TNW was 0.36 lexical words/number of words in the spoken data and 0.43 lexical words/number of words in the written data. With respect to LD/C, the spoken samples showed a lower index (LD/C=2.53 lexical words/clause) than the written samples (LD/C=3.43 lexical words/clause). Concerning LD/TU, it was lower in the spoken samples (LD/TU=3.93 words/T-unit) than in the written samples (LD/TU=6.26). Moreover, the differences between these two modes were found very significant in relation to all the measures of lexical density.
Table 2: Average of All the Measures of Lexical Density of Spoken and Written Data (*Significant at p<.05)
The results obtained from Table 2 show that written samples overall are more lexically dense than the spoken samples. This finding is in line with L2 studies (e.g. Hyltenstam, 1988) using the measures of lexical density in that written production has a higher density than spoken production. As far as these three measures are concerned, it seems that LD/TU is more sensitive than the other two measures, since it indicated the largest difference between the two modes as shown by the percentage difference as indicated in the table. The percentage difference is computed by subtracting the means of each measure of the written data from that of the spoken data and then dividing the number by the mean of the spoken data. This figure is used to show how much difference is between the two modes. As shown in the table, the percentage difference of LD/TU is 59% followed by LD/C being 36% and LD/TNW being 19%. So, it can be said that LD/TU is the most sensitive indicator of the difference between the two modes with LD/C being the second best indicator and LD/TNW the third best indicator.
In sum, the spoken samples are shown in the present study to have more lexical density than the written samples. And, of all the measures of lexical density, LD/TU seems to be the best indicator of the differences between the spoken and written data.
VII. Conclusion
The conclusion of the study is presented by a discussion of outcomes of the hypotheses and implications of the results. Hypothesis 1 investigates the correlation between speaking and writing with the measures of lexical density. The study demonstrated that statistical difference was only found with regard to LD/TU. Concerning the other two measures of lexical density, they indicated a low and non?鄄significant correlation. Hypothesis 2 is concerned with the differences between two samples. The study showed that all the measures of lexical density were higher in the written samples than in the spoken samples. Besides, such difference was found statistically significant in relation to all the measures. Therefore, this hypothesis is sustained. The findings of the present study seemed to indicate that there was a positive relationship between speaking and writing according to the rank?鄄order of the subjects. However, this implication cannot be generalized, because other aspects of this relationship have to be considered to get an overall picture of this relationship. The present study found that the measures of lexical density are good indicators of the differences between the two modes but cannot differentiate between proficiency levels. Therefore, the indices of lexical density can be adopted to carry out L2 teaching and research concerned with the difference between the two modes.
References
[1] Beaman, K. (1984). Coordination and subordination revised: Syntactic complexity in spoken and written narrative discourse. In D.Tannen (Ed.), Coherence in spoken and written discourse. Norwood, NJ: Ablex.
[2] DeVito, J.A.1965: A linguistic analysis of spoken and written language. Central States Speech Journal 18, 81-85.
[3] Halleck, G.B. & Young, R.F. (1995). Video oral communication instrument (VOCI) for ESL/EFL. Language Acquisition Research Center, San Diego, CA.
[4] Halliday, M.A.K. (1979). Differences between spoken and written language: Some implications for literacy teaching. In Page, G. Elkins, J. & O?蒺Connor, B. (Eds). Communication through reading proceedings of the fourth Australian reading conference (pp. 37-52) Adelaide, SA: Aust. Read. Assoc.
[5] Halliday, M.A.K. (1985). Spoken and written language. Deakin University Press.
[6] Hyltenstam, K.(1988). Lexical characteristics of near?鄄native second?鄄language learners of Swedish. Journal of Multilingual and Multicultural Development 9 (1&2), 67-84.
[7] Lauren, U.(2002). Some lexical features of immersion pupils?蒺 oral and written narration. Working Papers 50, 63-78.
[8] Linnarud, M. (1986). Lexis in composition: A performance analysis of Swedish learners?蒺 written English. CWL Gleerup.
[9] Ure, J. (1971). Lexical density and register differentiation. In Perren, G. E. & Trim, J.L.M (Eds). Applications of linguistics (pp.443-452). Cambridge: Cambridge University Press,443-452.