Investigating item response times in computerized adaptive testing

It is often argued that computerized testing offers more differential information than traditional test applications. However, this has not been demonstrated very often. The following study presents results from more than 100 subjects who worked on a computerized adaptive test based on 273 matrix items. Time on items is the additional information used here. Indirectly, the problem of the diagnostic value of achievement attained and time used is raised. As far as item times are concerned it surprising that improper solutions took more than 20 seconds longer on average. However, this does not indicate anything about different underlying cognitive processes. Also, no systematic decline of item time for a longer testing was observed, but a far bigger variance of gradients for correct responses emerged. Time gradients and achievement are not correlated substantially The correlation observed between achievement and speed of response suggests a bivariate test score interpretation.