| This study is an investigation of perceived word difficulty, i.e., the sense of certainty that a word is known, measured through the Yes/No lexical decision task. The primary purpose is to estimate the Item Response Theory difficulties of 6,000 frequently used vocabulary items under the assumption that words with lower item difficulties will be better known than words with higher difficulties. Given the large number of items, the only practical testing format is the Yes/No test, therefore the study was conducted in two parts. The first part was an investigation of the internal validity of the Yes/No format for measuring word difficulty. The second part was the utilization of Yes/No tests and IRT to estimate difficulty indices for more than 6,000 words. The initial investigation compared the Yes/No test, Vocabulary Knowledge Scale, and Vocabulary Levels Test (N = 165). All statistical procedures, including item and score factor analysis, and unidimensionality analysis, indicate the Yes/No test is a reliable, valid, and unidimensional measure of word difficulty. In the main study, 1,106 predominately female, Japanese junior college and four-year university EFL students responded to one or more of 133 different test forms (a total of 5,100 tests) testing 6,029 words. An IRT analysis returned difficulty estimates for 5,413 words. Subsequent analysis with a repeated measures t-test showed a significant difference between base words and derived forms using Level 3 Word Family affixes (t = -14.981, df = 303, p < .01), and a correlation between word difficulty and transformed frequency count of -.59 (p < .01). For any given ability estimate, lexical size, expressed in number of words known, is estimated from the sum of the probabilities of each word being known. Finally, for any ability estimate, both high frequency unknown words, and low frequency known words can be identified using the twin criteria of difficulty and frequency. By eliminating words that have a high probability of being known from a target word list, the efficacy of a pedagogical program is increased. Additionally, this process identifies familiar low-frequency words which can be used for material development. |
