Examining Psychometric Qualities of Reading Rate Data Collected under Two Different Conditions

抄録

Reading rate data produced by 207 Japanese EFL college students taking one of two versions of a computer-administered sentence reading test was examined for reliability and validity as a reading proficiency measure. The test had 40 items, each consisting of one sentence followed by a multiple-choice question. The two versions presented sentences in different manners: in one version, they appeared under the moving-window condition and in the other, under the stationary condition. In either version, presentation was subject-paced and reading time was recorded, based on which reading rate was computed for each item in three different units: words per minute, syllables per minute, and characters per minute. Main findings included: (1) the moving-window condition produced faster reading rates than the whole-sentence condition; (2) the moving-window condition did not allow anomalous subject behavior, from which the stationary condition suffered severely; (3) the moving-window condition produced more reliable data than the stationary condition; (4) the mean reading rates recorded under the moving-window condition were more valid indicators of the readers' proficiency than those elicited under the stationary condition; (5) no clear differences were observed between the mean reading rates based on all the data points and those derived from the correctly responded items only; and (6) no one unit had a clear advantage over another.