Examining Within-Person Changes in Speed in a Verbal Reasoning Test

Hierarchical Models are the most commonly used statistical methods for describing the relationship between speed and accuracy on timed tests. These methods have the advantage of increasing measurement precision by borrowing information from response time (RT) models, but assume that both speed and accuracy are fixed across the test. Unfortunately, this assumption has been shown to be violated in practice. Also, researchers have recently sought additional research on the fit of the lognormal RT model and on modeling conditional dependencies among response times. This dissertation had four research objectives: 1) exploring modeling approaches for working speed across a verbal reasoning test, 2) determining how RT information impacts individual ability estimates, 3) identifying examinees who change their working speed using individual factor scores from the joint models for speed and accuracy, and 4) exploring the congruence between two modeling approaches capable of explaining test-taking behaviors with different assumptions. Findings from this study suggest that working speed is multiphasic and the relationship between speed and accuracy is not fixed. More specifically, there are three distinct phases that represent the beginning, middle, and end of the test, with the middle of the test providing the most information. The incorporation of response time information in the measurement model produces smaller standard errors and less measurement error. The implications of this study are twofold: 1) researchers should evaluate the assumption of conditional independence in the RT model and choose an appropriate model for RTs, and 2) the effects of time pressure can be seen as early as the middle of the test as rapid-guessing behavior is a gradual process.