Confirmatory factor analysis was used to examine the appropriateness of using with elderly persons Levenson's multidimensional locus of control (LoC) scale that measures beliefs in Internal Control (I), Control by Powerful Others (P), and Chance (C). Data were obtained from elderly individuals in Florida and Texas and reformulated to allow cross-validation at each step in a series of model evaluations. Results indicated that Levenson's three-factor specification of control was not a valid representation of the samples' responses. A model that specified the elimination of eleven unreliable items and the formation of a new external control factor that was based on the remaining C and P items provided an adequate fit to the data for both samples. The failure to confirm Levenson's three-factor structure poses a construct-validity and measurement-equivalence problem when making age-based comparisons of LoC scores.