Can Response Speed Be Fixed Experimentally, and Does This Lead to Unconfounded Measurement of Ability?

Abstract

Goldhammer (this issue) proposes an interesting approach to dealing with the speededness of item responses. Rather than modeling speed as a latent variable that varies from person to person, he proposes to use experimental conditions that are expected to fix the speed, thereby eliminating individual differences on this dimension in order to make unconfounded comparisons of a person’s ability possible. We applaud his efforts for considering the gains that can be obtained by changing the test conditions to better match the measurement aims of ability tests, rather than just considering altering the measurement model. We agree that the model provides an interesting theoretical exploration into possible conditions under which the measurement of speed and ability would not be confounded by the speed ability compromise. However, the model is only able to achieve this unconfounded measurement of ability by imposing a number of restrictive assumptions. We believe that the merit of the approach will depend on the extent to which these assumptions are likely to be met in practice. We will discuss two main concerns: issues with the practical realizability of fixing effective speed and consequences of fixing speed for the measurement of ability.