Identification of Stable Asymptotic Performance on Computer-Based Cognitive Tests

Abstract

Examining whether human cognitive performance is affected by environmental conditions requires stable performance measures prior to stressor exposure. This study evaluated the stability and reliability of six computer-based cognitive performance tasks. A Microsoft Excel Visual Basic for Applications (VBA) macro program, the Stability and Reliability Analysis System (SRAS), was developed to evaluate performance of the cognitive tests using three approaches for identifying stability: Graphical Analysis, Learning Curve Fitting, and Statistical Analysis. The results of the comparative evaluation indicated that the SRAS macro program was effective in determining differential stability for the various tasks and measures. Across all tasks, the use of a compound graphical analysis approach was better than a single graph method in terms of providing a more reliable estimation of task stability. Learning curves were fit to each performance measure. For most tasks, the best-fit models were power and logarithmic models. The statistical analysis methods provided conservative estimates of task stability.