A study of the relationship between noncompensatory processing and the working memory load of matrix completion items was conducted. Data were taken from the British Cohort Study of 1970, First Follow-up (N = 14,875). To assess compensation, the GMIRT Rasch model (Spray & Ackerman, 1986), variable compensation model, was used with MCMC estimation via WINBUGS. In support of these analyses, a simulation study assessing parameter recovery for the GMIRT model was conducted. Sample size, item pool size, and interability correlation were manipulated. Adequate parameter recovery was observed when difficulty parameters were constrained equal across dimensions. In the application study, there was some evidence to support the relationship between working memory load and compensation. |