Two experiments investigated children's strategies for solving geometric matrices that were correctly or incorrectly completed and that varied in number of elements and number of transformations. Examining the relationship between working memory and item complexity, the first experiment tested 90 boys and girls of 7, 10, and 13 years of age for auditory and visual memory and for performance on Raven's and geometric matrices. Results revealed that 7-year-old subjects were less accurate than either 10- or 13-year-olds; 10-year-olds were slower than 13-year-olds. Developmental differences in strategies included the finding that older children increased latency to solution with increased item complexity, while the youngest group did so only for up to three "feature" matrices. The second experiment, conducted with 30 children, 10 years of age, focused on such matrix characteristics as kind of transformation and nature of falseness. Two questions were raised: Does strategy use differ across addition, expansion, and circular motion transformations? and, Will strategy use vary depending on where the falseness lies in the matrix? Strategic solution seemed to be perceptually driven in addition processes and in circular motion transformations. It was suggested that consideration of what cues are relevant to problem solution appears to be triggered in large part by conceptual/perceptual characteristics of these matrices. Results were considered in terms of the interaction of individual abilities with quantitative and qualitative characteristics of matrix problems. (Author/RH)