| The current study investigated the content and construct validities of second language proficiency as measured by the Test of English as a Foreign Language (TOEFL)---i.e., the types of language (test content) and the types of skills (constructs) involved in a second language test score when derived from narrowly-defined groups of second language learners. Learner characteristics were controlled to elaborate on quantitative and qualitative relationships between examinee demographics and test characteristics. In addition, tables of items representing relative strengths and weaknesses were created for several descriptive profiles of second language learners. Such information may prove extremely helpful to the ESL classroom practitioner (e.g., knowing what weaknesses should be expected for remediation) as well as in efforts to prepare curriculum and materials that are suited for the developmental needs of specific language learners (e.g., the staging and sequencing of syllabi and curriculum). With statistical and descriptive methodologies, the present study addressed essential facets of interlanguage, including: ethnolinguistic background, developmental stages, test features, and the dimensionality of measurement. Group comparisons were made using Multivariate Analysis of Variance (MANOVA) of the section and parcel scores of the TOEFL as regressed on proficiency (quartile) and ethnolinguistic features (factor). Follow-up ANOVA tests were used to determine specific group differences. Finally, item-level analyses, with the use of a chisquare technique, were designed to probe group differences as found between items. Groups were as follows: three native language groups (Japanese, Portuguese, Turkish); five nationalities (Brazil, Cyprus, Japan, Portugal, Turkey); and test sites (domestic and overseas). Groups were further sub-divided by overall proficiency level into quartiles. The comparisons between these profile-groups and a generic ESL group constitute the essential feature of this study. This study thus provides information about: (1) the diagnostic potential of section subscores and/or parcels of test items; (2) the dimensionality of test sections as an artifact of the population or sample being studied; (3) distinctive developmental, interlanguage characteristics of certain ethnolinguistic groups; and (4) the diagnostic utility of a "scale of proficiency levels" per the content areas (sections), to be used for instructional washback. |
