Abstract
Two indices of answer copying, one using only identical wrong responses and the other using both right and wrong, were compared using naturally occurring test responses (no simulation of copying). Each index pointed to the existence of about the same number of exannnee pairs who had copied. However, index B implicated more higher scoring pairs that he g2 index, which, in turn, implicated more lower scoring pairs. An adaptation Of g2 using only wrong-answer correspondences was found to identify even more pairs of highly likely copiers than B among higher scoring examinees. A spliced index, the regular g2 for lower scoring pairs and g2 based on identical wrong responses for higher scoring pairs, was found to identify substantially more pairs of likely copiers than any single index. The methodological approach adopted made it possible to evaluate the indices without knowing which pairs of examinees actually copied.
Get full access to this article
View all access options for this article.