This research studied the psychometric properties of a reading comprehension test using the Cloze's technique. The participants were 612 freshmen undergraduate students whose task was to discover the omitted words. It was observed a good fit to a two parameter model of Item Response Theory. The total score composed by the sum of the correct completions achieved high level of reliability (KR-20 = .84). Considering the item complexities, two groups were formed: one composed of very easy items, and the other with very complex ones. A parcel of item complexity was explained by the grammatical category of the word to be discovered (r= 0.62, p < 0.001). From the highest to the lowest difficulty level: adjectives, verbs, adverbs, nouns, pronouns, prepositions, articles were observed. The cognitive processes that could explain the differences in the difficulties correlated with grammatical categories were analized. Trough this analysis and trough Item Response Theory (IRT) the advances of item referenced interpretation in describing reading comprehension proficiency were discussed.
Reading comprehension; Cloze's Technique; Item Response Theory
