This article reports the performance of 34 non-native speakers on four testing methods developed on the basis of reduced redundancy testing, namely a C-Test, schema-based cloze multiple choice item test (MCIT), text-driven cloze test and traditional cloze MCIT. A disclosed Test of English as a Foreign Language (TOEFL) was used as a criterion for estimating the empirical validity of the four methods and whether they tap into the same competence. The results indicated that among the four methods, schema-based and traditional cloze MCITs had the highest empirical validity. The principal component analyses of results revealed two factors: the ability to deal with cloze-based reduced redundancy and reduced redundancy in general. When the factors were rotated, only the C-Test and text-driven cloze test loaded on the first factor and thus validated themselves as measures of cloze-based reduced redundancy. The four methods, however, loaded on the second factor which was concluded to represent reduced redundancy in general. The addition of the TOEFL and rotation of the results revealed the same pattern. It also showed that the TOEFL had the highest loading on the reduced redundancy in general and thus validated reduced redundancy as a measure of language proficiency. Since the schema-based cloze MCIT had the second highest loading on the second factor, it provides the best measure of reduced redundancy and language proficiency.

Keywords