Does Answer-Changing Affect Test Quality?

Abstract
Test papers of 289 seventh-grade examinees were scored using students' initial responses, then rescored using their final responses on a subtest of the Metropolitan Achievement Test. Item analyses were conducted on both sets of data to determine item biserial correlations with total score and K-R 20 coefficients. Results indicate that reliability coefficients and item discriminations are not adversely affected by examinee response changes. For these seventh graders the proportion of examinees changing responses was much lower than has been recorded in previous studies of college examinees. In general, however, the seventh graders tended to profit from response changes, just as college examinees do. Implications of these results for counselors and directors of testing programs are presented.

This publication has 4 references indexed in Scilit: