Factor Structure of the Toefl Internet‐based Test Across Subgroups

Abstract

This study assessed the invariance in the factor structure of the Test of English as a Foreign Language™ Internet‐based test (TOEFL® iBT) across subgroups of test takers who differed in native language and exposure to the English language. The subgroups were defined by (a) Indo‐European and Non‐Indo‐European language family, (b) Kachru's classification of outer and expanding circles of countries (based on prevalence of English use in educational and business contexts), and (c) years of classroom instruction in the English language. The same factor structure (four first‐order factors corresponding to the test sections and a single higher‐order factor encompassing these factors) was identified in each subgroup. The results support the present scoring scheme for the TOEFL iBT assessment and suggest that the test functions the same way for diverse subgroups of test takers.

INVESTIGATING THE VALUE OF SECTION SCORES FOR THE TOEFL iBT® TEST

“…Meanwhile, the results supporting the distinctiveness of the constructs assessed in the reading, listening, and writing sections are mixed. Many issues might have contributed to the difference of the findings between the Stricker et al study versus Sawaki et al (2008) and Stricker and Rock (2008) on this issue. One potential reason has to do with the differences in the test design.…”

“…In this study a higher order factor model with a general factor for English as a second language or English as a foreign language (ESL/EFL) ability and four first-order factors corresponding to the four modalities were identified. Stricker and Rock (2008) conducted another factor analysis of the same field study test form, this time by modeling item parcels. They identified the higher order factor structure as well.…”

“…This is primarily because the present study was the first factor analysis of operational TOEFL iBT data. The previous factor analyses were conducted on a TOEFL iBT prototype by Stricker et al (2005) and on field study data by Sawaki et al (2008) and Stricker and Rock (2008). Following the suggestions in the literature to use different data sets for exploring the factor structure and confirming the findings of the exploratory analyses (e.g., Jöreskog, 2007, p. 58), EFAs were conducted on two randomly selected forms (the April and December forms), and CFAs were conducted on the other two forms (the July and September forms).…”

“…Relative goodness of fit of four CFA models was examined to choose the best among four proposed models for each sample. The models were constructed based on those tested by Sawaki et al (2008), Stricker and Rock (2008), and Stricker et al (2005).…”

“…A higher order factor model was identified as the final model by a recent study of TOEFL iBT field study data by Sawaki et al (2008). Stricker and Rock (2008) adopted a higher order factor model as well, although their final model did not include the cross-loadings. The specification of the first-order factor structure was the same as that for the correlated four-factor model, except that the cross-loading of the reading/listening/writing task on the listening factor was dropped from this model for model identification.…”