CONFIRMATORY FACTOR ANALYSIS OF THE TEST OF ENGLISH AS A FOREIGN LANGUAGE

Abstract

In a recent study, confirmatory factor analyses indicated that, for each of several language groups, TOEFL performance can be characterized by two factors, associated with (a) the Listening Comprehension section and (b) the other sections of the test (Hale, Stansfield, Rock, Hicks, Butler, & Oiler, 1988). This conclusion was inconsistent with that drawn in the most comprehensive previous factor‐analytic study of the TOEFL, which suggested a three‐factor solution for each of several languages (Swinton & Powers, 1980). The present study investigated the inconsistency in conclusions drawn from these two studies and provided further information about the factor structure of the TOEFL.It was hypothesized that the inconsistency between studies was related to the populations under investigation, as the earlier study used TOEFL examinees in both domestic and overseas test centers, whereas the more recent study used domestic examinees only. The present data did not support this hypothesis, however. Confirmatory factor analyses were conducted for each of several language groups, using data from the test form used in the more recent study (a 1984 TOEFL), and these analyses yielded essentially similar results for both domestic and overseas populations, as well as for the combined population. In all cases, the data supported a two‐factor interpretation, with the two factors related to the Listening Comprehension section and to the nonlistening sections.Other hypotheses were that the inconsistency between studies was due to differences in factor‐analytic methodologies used or to changes in the test over time. Confirmatory factor analyses were conducted using the data from the earlier study (taken from a 1976 TOEFL) as well as data from the more recent study. These analyses supported a two‐factor interpretation in both cases, for each of several language groups, with the two factors associated with Listening Comprehension and with the other sections of the test. Thus, the use of different factor‐analytic methodologies in the two previous studies undoubtedly contributed to the inconsistency, although further work would be needed to determine exactly what aspects of the methodology played a role. The data also provided tentative evidence that the basic factor structure of the test may not have changed substantially between 1976 and 1984.The role of examinee proficiency in determining the TOEFL's factor structure was also examined. For each of several language groups, low‐ and high‐proficiency groups were defined on the basis of approximately one third of the items in the TOEFL, drawn from all sections of the test, and factor analyses were then performed on an abbreviated TOEFL, which consisted of the remaining items. For both proficiency levels within each language group a two‐factor structure appeared to underlie performance, with the factors once again linked to Listening Comprehension and to the other, nonlistening sections of the test.