论文部分内容阅读
摘要:本研究旨在考察总体等级评分(holistic rating)和句法成熟程度 (syntactic maturity)在评估学生作文之间的关系。 本研究通过计算句构成熟的五个指标(TNW, MTUL, MEFTUL, %EFTU and SR)来研究学生作文的句构成熟度。研究数据表明,根据总体等级评分标准所评估的作文分数同 MTUL 和MEFTUL这两个指标具有有意义的相关性,但是同TNW和%EFTU却存在无有意义的相关。根据这些发现,本研究提出了一些关于作文和课程设置方面的建议以供英语教师参考。
Abstract:
The study examines the relationship between holistic and objective measures in assessing ESL students' written composition. Five objective measures (TNW, MTUL, MEFTUL, %EFTU and SR) are calculated to measure the syntactic maturity of students' compositions. The data of the study indicate that significant correlations are found between holistically assigned scores and MTUL and MEFTUL. The holistic rating, however, is not significantly related to TNW, %EFTU and SR in the present study. According to these findings, implications are drawn for ESL composition and curriculum.
Hunt (1965) and O'Donnel(1970) developed objective measures of syntactic maturity, which was defined as the ability to manipulate the syntax of the language can be measured by objective measures such as mean T-unit length (MTUL), mean error-free T-unit length (MEFTUL), percent of error-free T-units(%EFTU), subordination ratio(SR), average clauses length, average sentence length and coordination ratio and the rest. In terms of Hunt, T-unit is “minimal terminal syntactic maturity” i.e., any main clause with all of its modifiers (Nystrand, 1982). Many researchers favor the use of syntactic maturity, because it is easy to count, provides more objective quantifiable measures of significant structural characteristics(Kerek; 1979; Stewart, 1978; Kameen, 1983; Tomlinson & Straehley, 1978). On the other hand, other researchers points out the problems of syntactic maturity including the unreliability of T-unit(Witte, 1983; Reed, 1983; Sharma, 1979; Witte and Sodowsky, 1978; Chew, 1978) Thus, more studies need to be done concerning the objective measures of syntactic maturity. The present study attempts to provide more empirical data in the study of written compositions by examining the relationship between holistic and objective measures in assessing the written proficiency of college-level ESL students.
Two key research questions are examined here: The first one is how holistic ratings as defined by high-rated students and low-rated students relates to syntactic maturity as measured by the total number of words(TNW), mean T-unit length (MTUL), mean error-free T-unit length(MEFTUL) and percent of error-free T-units(%EFTU) and subordination ratio(SR)? The second research question is how the holistically-assigned scores (N=20) relates to five objective measures of syntactic maturity? In the present study, four indices plus total number of words(TNW) were used to measure syntactic maturity of the compositions written by college-level ESL students.
1.Method
The analysis of the study involved the identification and tallying of six measures (TNW, MEDRUL,MTUL, MEFTUL, %EFUL, SR) used to characterize the students' written performance and functioning as dependant variables of this study, that is, the variables that were investigated in the study. The holistic rating of the students' composition (ten best compositions and ten worst compositions) was used as the independent variable of this study. The study attempted to investigate the relationship between these two kinds of variables to see whether they have any correlation with each other.
1.1 Sample
The subjects for this study were drawn from students enrolled in a college-level ESL composition course at an American university. These subjects came from different countries 20 students' compositions were examined in this research, which were 10 high-rated ones and 10 low-rated ones out of a total of 40 compositions. The essay prompt required them to describe three underlying rules shaping student behavior in American classrooms. They needed to show how American students behaved based on those rules by providing specific examples. In this case, the subject matter was controlled in the study.
1.2 Independent Variables
As noted earlier, the independent variables consisted of total number of words(TNW), mean T-unit length (MTUL), mean error-free T-unit length (MEFTUL) and percent of error-free T-units (%EFTU) and subordination ratio(SR). The TNW and MTUL were intended to measure the quantity of written performance. I hypothesized that the high-rated students would write more words and produce longer T-units than low-rated ones.
In addition, the two variables, MEFTUL and %EFTU were evaluated in the study to measure the quality of written performance. Concerning these two variables, the hypothesis was that high-rated compositions might contain more error-free T-units than low-rated ones. Correspondingly, the percentage of error-free T-units of high-rated compositions would be higher than low-rated ones.
As far as the subordination ratio was concerned, this variable was used as an additional measure of syntactic maturity of students'written performance, which was the number of subordinate clauses for every main clause or clauses per T-unit. I noted that some students' written proficiency was very low, even though they used a lot of subordinate clauses in their compositions according to my previous teaching experience and my study of students' compositions in the research. Moreover, the subordinate clauses they used in their compositions were not appropriate and full of mistakes. Therefore, I had my hypothesis that there was no significant difference between holistic rating and the number of subordinate clauses students used in their compositions.
1.3 Dependant Variable
The dependant variable of this study was the holistic rating of the instructor. The scores of 10 high-rated students ranged from 94 to 100 while those of 10 low-rated students varied from 73 to 81. Since these 20 compositions were graded by the same instructor both holistically and analytically, the intra-rater reliability was obtained in this study.
1.4 Data Analysis
Three procedures were carried out in analyzing statistical data related to holistic judgments of written performance and objective measures of syntactic maturity. First, the objective measures of TNW, MTUL, MEFTUL, %EFTU and SR were marked and tallied. Means of each objective measures were computed separately concerning high-rated compositions and low-rated ones. Next, bar graphs were drawn to vividly demonstrate the difference of TNW, MTUL, MEFTU, %EFTU and SR between high-rated students and low rated students in terms of mean and the percentage difference of each variable between high and low rated students. Finally, a Pearson product-moment was carried out to show how holistically assigned score correlated with each variable of objective measures.
2. Results
The first key research question in the study concerns whether there is any difference between high-rated students and low-rated students in terms of TNW,MTUL, MEFTUL,%EFTU and SR respectively. Thus, the analyses were illustrated in the following sections to reveal these relationships separately.
Table I and II provided detailed information of the raw scores. TNW, MTUL, MEFTUL, %EFTU and SR of ten high-rated students and ten low-rated students respectively, both of which showed five objective measures of each student separately.
Next, as revealed in Table III, there was a positive correlation between high-rated and low-rated students concerning five dependant variables (r=0.9999, p<0.0001), which was both high and significant. This finding indicated a high positive relationship between high-rated and low-rated students regarding five objective measures of syntactic maturity. That is to say, compared with low-rated students, high-rated students as a whole tended to produce more words, longer T-units, more error-free T-units and use more subordinate clauses. In addition, the Pearson correlation coefficient and probability value revealed that five variables inter-correlated with each other significantly. Besides, Table III also showed the percentage difference of each variable between high and low-rated students, which was computed by subtracting the mean of each variable of low-rated students from that of high-rated students and then dividing the number by the mean of the low-rated students. As noted in Table II, the biggest difference was found in MTUL(%44) followed by MEFTUL (36%), TNW(18%), SR(15%)and %EFTU (6%).
Finally, concerning the second research question, Table IV revealed the results of correlations computed between the holistic ratings of all 20 students as a whole and five dependant variables. As noted in the Table, five objective measures indicated a positive correlation with the holistic rating of students' written performance; i.e., as students scored higher, the five objective measures tended to increase. The positive correlation confirmed that students with higher written proficiency had a tendency to produce more words, longer T-units and more error-free T-units and would like to incorporate more subordinate clauses. However, Table IV also demonstrated that only two variables, MTUL and MEFTUL correlated significantly with the holistic measure. In other words, there were significant positive relationships between the students' scores and their ability to perform on the written exam in terms of MTUL and MEFTUL. Interestingly, the percentage difference of each variable between high and low-rated students as mentioned in Table III and Bar Graph II also showed that high and low-rated students could be better differentiated as far as MTUL and MEFTUL were concerned, because their percentages were obviously bigger than those of other variables.
Table IV
Pearson Correlation for Written Performance Criteria and Holistic scores (N=20)
3. Discussion
The purpose of this study was to examine how objective measures of syntactic maturity related to the holistically-graded scores assigned to the compositions written by college-level ESL students. As noted earlier, three hypotheses were made on the relationship between five dependent variables and the holistic assessment of students' written work. Accordingly, the findings of the study were summarized in the following discussion in relation to three hypotheses followed by the implications of the study.
3.1 Hypothesis 1
The first hypothesis was concerned with two dependant variables, TNW and MTUL, which indicated the quantity of written performance. According to this hypothesis, the researcher assumed that students with high written proficiency would produce more words and longer T-units than those with low written proficiency. The data revealed a significant difference between MTUL and the scores, but not between TNW and the scores. Meanwhile, the highest percentage was discovered in MTUL among five variables in Table III. Specifically, ‘good’ students can be differentiated from ‘poor’ students with MTUL, but not with TNW. ‘Good’ students usually write longer T-units and but not necessarily produce more words. This finding correlates with that of other researchers. Stewart (1978) investigated the written syntactic maturity of students from grade ten through sixth year university and pointed out that words per T-unit appeared to be the best of the three measures of syntactic maturity (MTUL, Clause length and SR).
3.2 Hypothesis 2
The second hypothesis was about the quality of written performance as measured by MEFTUL and %EFTU. The research hypothesized that a significant difference could be found between these two variables and the holistic grading. The data revealed that MEFTUL correlated significantly with the scores while no significant correlation was found between %EFTU and the holistic scoring. Therefore, like MTUL, MEFTUL could serve as a good index to differentiate “good” students from “poor” students. The finding of the study was consistent with Sharma (1979) given that MEFTUL was found to be significantly correlated with the holistic rating. Nevertheless, as revealed in Pearson correlation for written performance criteria and holistic rating, MEFTUL didn't correlate with holistic rating as high as MTUL did(see Table IV), which was also indicated as far as the probability value is concerned. So it would be safe to say that MEFTUL was the second most reliable index of objective measures in differentiating students' written performance.
3.3 Hypothesis 3
The third hypothesis related to the use of subordinate clauses in students' composition, by which the researcher assumed that no significant difference existed between the variable of subordinate ratio and the holistic assessment based on the researcher's previous teaching experience and the observation on the compositions involved in the study. The finding of the study confirmed this hypothesis and revealed that there was no significant difference between subordination ratio and holistic rating. In this sense, SR is not a reliable index in determining written performance. This finding was in line with Chew(1978), who found that there was no direct correlation between a paper's quality and the ratio of clauses per T-unit. On the whole, the findings of the study correlate with many studies in that T-unit length is proved to be the most reliable index of syntactic maturity to differentiate between “good” and “poor” students with regard to written performance (Hunt,1965; O'Hare,1973; Tomlinson & Staehley,1978; Stewart,1978; Kameen,1983).
4. Conclusions
To conclude, the study revealed that significant correlations existed between holistic assessment and objective measures of syntactic maturity in terms of Mean T-unit length and Mean Error-Free T-unit length. Meanwhile, the study indicated no significant correlation between the use of subordinate clauses and students' holistically-assigned scores. Based on these findings, the researcher reinforced that sentence-combining exercises could be applied to “poor” students to increase their syntactic complexity. However, it is recommended that the technique of embedding subordinate clauses into a main clause may not benefit “poor” students. Therefore, this kind of practice should be avoided in ESL student composition class. In addition, the study indicated that counting objective measures of syntactic maturity could be used in ESL composition and curriculum.
References
Chew, C.(1978). A study to determine the relationship between indexes of syntactic maturity and a quality grade in a written examination. Retrieved November 7.
Hunt, K.W. (1965a). Grammatical structures written at three grades levels. Research Report N.3, Urbana, Illinois: NCTE.
Hunt, K.W. (1965b). Sentence structures used by superior students in grades four and twelve and superior adults. Cooperative Research Project, No.5-0312 ED 010 047.Tallahassee, Florida: Florida State University.
Kameen, P. (1983). Syntactic skill and ESL writing quality. In A.Freedman, I. Pringle & J.Yalden. (Eds.) Learning to write: first language/second language (pp.139-158). London: Longman.(162-170)
Kerek, A. (979). Measuring and evaluating syntactic maturity: some questions of methodology. Retrieved November 7, 2002 from ERIC database. ED 173783.
O'Donnell, R.C; Griffin, W.J. and Norris, R.C. (1967). Syntax of kindergarten and elementary school children: a transformational analysis, NCTE Research Report No.8.Urbana, Illinois: National Council of Teachers of English.
O'Hare, F. (1971). Sentence combining: improving student writing without formal grammar instruction. (NCTE Research Report No.150) Urbana, IL: National Council of Teachers of English.
Reed, W.M.(1983). Daly and Miller's writing apprehension test and Hunt's T-unit analysis: two measurement precautions in writing research. Retrieved November 7, 2002 from ERIC database. ED 248539.
Sharma, A.(1979).Syntactic maturity: assessing writing proficiency in a second language. Retrieved November 7, 2002 from ERIC database ED185105. Paper presented at the international conference on frontiers in language proficiency and dominance testing (3rd, Carhondate.IL, September 26-28, 1979).
Stewart, N. (1978). Syntactic maturity from high school to university: a first look. Research in the Teaching of English 12.1, 37-46.
Tomlinson, B & Straehley, M(1978). Developing and Measuring Mature Syntax. Retrieved November 7, 2002 from ERIC database. ED158244.
Witte, S & Sodowsky, R (1978). Syntactic maturity in the writing of college freshman. Retrieved November 7, 2002 from ERIC database. ED163460.
Witte, S.(1983). The reliability of mean T-unit length: some questions for research in written composition. In A. Freedman, I. Pringle, & J.Yalden. (Eds.) Learning to Write: First language/second language (PP. 172-177). London: Longman.
Abstract:
The study examines the relationship between holistic and objective measures in assessing ESL students' written composition. Five objective measures (TNW, MTUL, MEFTUL, %EFTU and SR) are calculated to measure the syntactic maturity of students' compositions. The data of the study indicate that significant correlations are found between holistically assigned scores and MTUL and MEFTUL. The holistic rating, however, is not significantly related to TNW, %EFTU and SR in the present study. According to these findings, implications are drawn for ESL composition and curriculum.
Hunt (1965) and O'Donnel(1970) developed objective measures of syntactic maturity, which was defined as the ability to manipulate the syntax of the language can be measured by objective measures such as mean T-unit length (MTUL), mean error-free T-unit length (MEFTUL), percent of error-free T-units(%EFTU), subordination ratio(SR), average clauses length, average sentence length and coordination ratio and the rest. In terms of Hunt, T-unit is “minimal terminal syntactic maturity” i.e., any main clause with all of its modifiers (Nystrand, 1982). Many researchers favor the use of syntactic maturity, because it is easy to count, provides more objective quantifiable measures of significant structural characteristics(Kerek; 1979; Stewart, 1978; Kameen, 1983; Tomlinson & Straehley, 1978). On the other hand, other researchers points out the problems of syntactic maturity including the unreliability of T-unit(Witte, 1983; Reed, 1983; Sharma, 1979; Witte and Sodowsky, 1978; Chew, 1978) Thus, more studies need to be done concerning the objective measures of syntactic maturity. The present study attempts to provide more empirical data in the study of written compositions by examining the relationship between holistic and objective measures in assessing the written proficiency of college-level ESL students.
Two key research questions are examined here: The first one is how holistic ratings as defined by high-rated students and low-rated students relates to syntactic maturity as measured by the total number of words(TNW), mean T-unit length (MTUL), mean error-free T-unit length(MEFTUL) and percent of error-free T-units(%EFTU) and subordination ratio(SR)? The second research question is how the holistically-assigned scores (N=20) relates to five objective measures of syntactic maturity? In the present study, four indices plus total number of words(TNW) were used to measure syntactic maturity of the compositions written by college-level ESL students.
1.Method
The analysis of the study involved the identification and tallying of six measures (TNW, MEDRUL,MTUL, MEFTUL, %EFUL, SR) used to characterize the students' written performance and functioning as dependant variables of this study, that is, the variables that were investigated in the study. The holistic rating of the students' composition (ten best compositions and ten worst compositions) was used as the independent variable of this study. The study attempted to investigate the relationship between these two kinds of variables to see whether they have any correlation with each other.
1.1 Sample
The subjects for this study were drawn from students enrolled in a college-level ESL composition course at an American university. These subjects came from different countries 20 students' compositions were examined in this research, which were 10 high-rated ones and 10 low-rated ones out of a total of 40 compositions. The essay prompt required them to describe three underlying rules shaping student behavior in American classrooms. They needed to show how American students behaved based on those rules by providing specific examples. In this case, the subject matter was controlled in the study.
1.2 Independent Variables
As noted earlier, the independent variables consisted of total number of words(TNW), mean T-unit length (MTUL), mean error-free T-unit length (MEFTUL) and percent of error-free T-units (%EFTU) and subordination ratio(SR). The TNW and MTUL were intended to measure the quantity of written performance. I hypothesized that the high-rated students would write more words and produce longer T-units than low-rated ones.
In addition, the two variables, MEFTUL and %EFTU were evaluated in the study to measure the quality of written performance. Concerning these two variables, the hypothesis was that high-rated compositions might contain more error-free T-units than low-rated ones. Correspondingly, the percentage of error-free T-units of high-rated compositions would be higher than low-rated ones.
As far as the subordination ratio was concerned, this variable was used as an additional measure of syntactic maturity of students'written performance, which was the number of subordinate clauses for every main clause or clauses per T-unit. I noted that some students' written proficiency was very low, even though they used a lot of subordinate clauses in their compositions according to my previous teaching experience and my study of students' compositions in the research. Moreover, the subordinate clauses they used in their compositions were not appropriate and full of mistakes. Therefore, I had my hypothesis that there was no significant difference between holistic rating and the number of subordinate clauses students used in their compositions.
1.3 Dependant Variable
The dependant variable of this study was the holistic rating of the instructor. The scores of 10 high-rated students ranged from 94 to 100 while those of 10 low-rated students varied from 73 to 81. Since these 20 compositions were graded by the same instructor both holistically and analytically, the intra-rater reliability was obtained in this study.
1.4 Data Analysis
Three procedures were carried out in analyzing statistical data related to holistic judgments of written performance and objective measures of syntactic maturity. First, the objective measures of TNW, MTUL, MEFTUL, %EFTU and SR were marked and tallied. Means of each objective measures were computed separately concerning high-rated compositions and low-rated ones. Next, bar graphs were drawn to vividly demonstrate the difference of TNW, MTUL, MEFTU, %EFTU and SR between high-rated students and low rated students in terms of mean and the percentage difference of each variable between high and low rated students. Finally, a Pearson product-moment was carried out to show how holistically assigned score correlated with each variable of objective measures.
2. Results
The first key research question in the study concerns whether there is any difference between high-rated students and low-rated students in terms of TNW,MTUL, MEFTUL,%EFTU and SR respectively. Thus, the analyses were illustrated in the following sections to reveal these relationships separately.
Table I and II provided detailed information of the raw scores. TNW, MTUL, MEFTUL, %EFTU and SR of ten high-rated students and ten low-rated students respectively, both of which showed five objective measures of each student separately.
Next, as revealed in Table III, there was a positive correlation between high-rated and low-rated students concerning five dependant variables (r=0.9999, p<0.0001), which was both high and significant. This finding indicated a high positive relationship between high-rated and low-rated students regarding five objective measures of syntactic maturity. That is to say, compared with low-rated students, high-rated students as a whole tended to produce more words, longer T-units, more error-free T-units and use more subordinate clauses. In addition, the Pearson correlation coefficient and probability value revealed that five variables inter-correlated with each other significantly. Besides, Table III also showed the percentage difference of each variable between high and low-rated students, which was computed by subtracting the mean of each variable of low-rated students from that of high-rated students and then dividing the number by the mean of the low-rated students. As noted in Table II, the biggest difference was found in MTUL(%44) followed by MEFTUL (36%), TNW(18%), SR(15%)and %EFTU (6%).
Finally, concerning the second research question, Table IV revealed the results of correlations computed between the holistic ratings of all 20 students as a whole and five dependant variables. As noted in the Table, five objective measures indicated a positive correlation with the holistic rating of students' written performance; i.e., as students scored higher, the five objective measures tended to increase. The positive correlation confirmed that students with higher written proficiency had a tendency to produce more words, longer T-units and more error-free T-units and would like to incorporate more subordinate clauses. However, Table IV also demonstrated that only two variables, MTUL and MEFTUL correlated significantly with the holistic measure. In other words, there were significant positive relationships between the students' scores and their ability to perform on the written exam in terms of MTUL and MEFTUL. Interestingly, the percentage difference of each variable between high and low-rated students as mentioned in Table III and Bar Graph II also showed that high and low-rated students could be better differentiated as far as MTUL and MEFTUL were concerned, because their percentages were obviously bigger than those of other variables.
Table IV
Pearson Correlation for Written Performance Criteria and Holistic scores (N=20)
3. Discussion
The purpose of this study was to examine how objective measures of syntactic maturity related to the holistically-graded scores assigned to the compositions written by college-level ESL students. As noted earlier, three hypotheses were made on the relationship between five dependent variables and the holistic assessment of students' written work. Accordingly, the findings of the study were summarized in the following discussion in relation to three hypotheses followed by the implications of the study.
3.1 Hypothesis 1
The first hypothesis was concerned with two dependant variables, TNW and MTUL, which indicated the quantity of written performance. According to this hypothesis, the researcher assumed that students with high written proficiency would produce more words and longer T-units than those with low written proficiency. The data revealed a significant difference between MTUL and the scores, but not between TNW and the scores. Meanwhile, the highest percentage was discovered in MTUL among five variables in Table III. Specifically, ‘good’ students can be differentiated from ‘poor’ students with MTUL, but not with TNW. ‘Good’ students usually write longer T-units and but not necessarily produce more words. This finding correlates with that of other researchers. Stewart (1978) investigated the written syntactic maturity of students from grade ten through sixth year university and pointed out that words per T-unit appeared to be the best of the three measures of syntactic maturity (MTUL, Clause length and SR).
3.2 Hypothesis 2
The second hypothesis was about the quality of written performance as measured by MEFTUL and %EFTU. The research hypothesized that a significant difference could be found between these two variables and the holistic grading. The data revealed that MEFTUL correlated significantly with the scores while no significant correlation was found between %EFTU and the holistic scoring. Therefore, like MTUL, MEFTUL could serve as a good index to differentiate “good” students from “poor” students. The finding of the study was consistent with Sharma (1979) given that MEFTUL was found to be significantly correlated with the holistic rating. Nevertheless, as revealed in Pearson correlation for written performance criteria and holistic rating, MEFTUL didn't correlate with holistic rating as high as MTUL did(see Table IV), which was also indicated as far as the probability value is concerned. So it would be safe to say that MEFTUL was the second most reliable index of objective measures in differentiating students' written performance.
3.3 Hypothesis 3
The third hypothesis related to the use of subordinate clauses in students' composition, by which the researcher assumed that no significant difference existed between the variable of subordinate ratio and the holistic assessment based on the researcher's previous teaching experience and the observation on the compositions involved in the study. The finding of the study confirmed this hypothesis and revealed that there was no significant difference between subordination ratio and holistic rating. In this sense, SR is not a reliable index in determining written performance. This finding was in line with Chew(1978), who found that there was no direct correlation between a paper's quality and the ratio of clauses per T-unit. On the whole, the findings of the study correlate with many studies in that T-unit length is proved to be the most reliable index of syntactic maturity to differentiate between “good” and “poor” students with regard to written performance (Hunt,1965; O'Hare,1973; Tomlinson & Staehley,1978; Stewart,1978; Kameen,1983).
4. Conclusions
To conclude, the study revealed that significant correlations existed between holistic assessment and objective measures of syntactic maturity in terms of Mean T-unit length and Mean Error-Free T-unit length. Meanwhile, the study indicated no significant correlation between the use of subordinate clauses and students' holistically-assigned scores. Based on these findings, the researcher reinforced that sentence-combining exercises could be applied to “poor” students to increase their syntactic complexity. However, it is recommended that the technique of embedding subordinate clauses into a main clause may not benefit “poor” students. Therefore, this kind of practice should be avoided in ESL student composition class. In addition, the study indicated that counting objective measures of syntactic maturity could be used in ESL composition and curriculum.
References
Chew, C.(1978). A study to determine the relationship between indexes of syntactic maturity and a quality grade in a written examination. Retrieved November 7.
Hunt, K.W. (1965a). Grammatical structures written at three grades levels. Research Report N.3, Urbana, Illinois: NCTE.
Hunt, K.W. (1965b). Sentence structures used by superior students in grades four and twelve and superior adults. Cooperative Research Project, No.5-0312 ED 010 047.Tallahassee, Florida: Florida State University.
Kameen, P. (1983). Syntactic skill and ESL writing quality. In A.Freedman, I. Pringle & J.Yalden. (Eds.) Learning to write: first language/second language (pp.139-158). London: Longman.(162-170)
Kerek, A. (979). Measuring and evaluating syntactic maturity: some questions of methodology. Retrieved November 7, 2002 from ERIC database. ED 173783.
O'Donnell, R.C; Griffin, W.J. and Norris, R.C. (1967). Syntax of kindergarten and elementary school children: a transformational analysis, NCTE Research Report No.8.Urbana, Illinois: National Council of Teachers of English.
O'Hare, F. (1971). Sentence combining: improving student writing without formal grammar instruction. (NCTE Research Report No.150) Urbana, IL: National Council of Teachers of English.
Reed, W.M.(1983). Daly and Miller's writing apprehension test and Hunt's T-unit analysis: two measurement precautions in writing research. Retrieved November 7, 2002 from ERIC database. ED 248539.
Sharma, A.(1979).Syntactic maturity: assessing writing proficiency in a second language. Retrieved November 7, 2002 from ERIC database ED185105. Paper presented at the international conference on frontiers in language proficiency and dominance testing (3rd, Carhondate.IL, September 26-28, 1979).
Stewart, N. (1978). Syntactic maturity from high school to university: a first look. Research in the Teaching of English 12.1, 37-46.
Tomlinson, B & Straehley, M(1978). Developing and Measuring Mature Syntax. Retrieved November 7, 2002 from ERIC database. ED158244.
Witte, S & Sodowsky, R (1978). Syntactic maturity in the writing of college freshman. Retrieved November 7, 2002 from ERIC database. ED163460.
Witte, S.(1983). The reliability of mean T-unit length: some questions for research in written composition. In A. Freedman, I. Pringle, & J.Yalden. (Eds.) Learning to Write: First language/second language (PP. 172-177). London: Longman.