Application of the Rule Space Model to the reading comprehension section of the Test of English as a Foreign Language (TOEFL)

The purpose of this study was to apply the Rule Space Model, a cognitive diagnostic test model, to a foreign language proficiency test. The analyses focused on the reading comprehension section of the Test of English as a Foreign Language (TOEFL). The study consisted of the following three parts: (a) identifying a set of attributes which would underlie the examinee's performance on the reading comprehension section of the TOEFL, (b) determining how consistently each test item is evaluated by different raters using the attributes, and (c) examining how well the set of attributes explains the examinee's performance on the test. Sixteen prime attributes were first identified, and an additional 11 interaction attributes were created based on the prime attributes. Attributes with strong objective criteria yielded fairly consistent ratings by the raters; whereas some attributes allowed the raters' subjective judgment, and as a result, received less consistent ratings. The results of the rule space analyses indicated that prime and interaction attributes together classified more than 80% of the examinees in one or more theoretically derived knowledge states and accounted for about 96% of the variance in the examinee's performance on the test. The relative difficulty of each attribute was determined by examining attribute probabilities, and developmental paths were also devised. The findings suggested that the Rule Space Model had potential for providing information regarding the examinee's present cognitive state. Some possible further studies were suggested.