| This study examined the extent to which a computer analysis of adult English as a second language students' Test of Written English (TWE) compositions could yield data reflecting their English proficiency as measured by the Test of English as a Foreign Language (TOEFL) and the TWE. Using multiple regression, the study examined the relationship of TOEFL, three TOEFL subscores, and TWE scores to numerical indices representing vocabulary size, vocabulary diversity, vocabulary complexity, deviation from standard English, and syntactic complexity computed from TWE compositions.;Predictor and criterion data were generated by institutional administrations of the Test of English as a Foreign Language (TOEFL) and the test of Written English (TWE). Predictor data were derived by analyzing 73 essays using a text analysis program developed by the researcher. The values for the five criterion variables were the examinees' scores on TOEFL, TOEFL subsections, and the TWE. The usefulness of the predictors, as potential indicators of English proficiency, was then assessed using five multiple regression analyses.;Coefficients of determination were obtained of.24,.32,.45, and.42, and.68, respectively, significant at the.001 level or better. Vocabulary size was the best predictor in all five models. Vocabulary complexity also contributed substantially in all cases except the Listening Comprehension subscores. Deviation from standard English made modest contributions to predicting the Listening Comprehension and Structure and Written Expression subscores. Syntactic complexity also made modest contributions to predicting the Structure and written Expression and total TOEFL scores. Vocabulary diversity did not contribute even 1% to the predicted variance in any case.;Computer assessment of English proficiency is not possible based solely on the results of this study, but the results support the feasibility of such a technique. The combined strength of active vocabulary size, and vocabulary complexity in predicting TWE scores (R... |
