A correlation structure modeling method for comparison of mediated effects is outlined. The procedure permits point and interval estimation of differences in mediator effects, and is useful with models postulating 1 or more predictor, intervening, or response variables that may also be latent constructs. The approach allows scale-free evaluation of differences in effects of any explanatory upon criterion variables transmitted via studied mediators, and is applied on data from a study of older adults with recent vision loss. (Contains 1 figure, 3 tables and 1 footnote.)