Although the distinction between verbal ability and spatial ability is fundamental in all models of human abilities, differences in the relative strengths of verbal and spatial abilities have failed to show consistent interactions with instructional treatments. This study investigated the hypothesis that spatial tests measure different abilities in different subjects and so replicable Aptitude X Treatment interactions (ATI) with spatial ability are unlikely. Forty-three subjects attempted 144 items from a spatial synthesis task previously shown to be a good measure of spatial ability. Componential models suggested that subjects differed in their solution strategies and that most subjects shifted strategies predictably over items. Correlations between task variables and reference tests showed that some solution methods relied more on verbal-analytic skills whereas other solution strategies relied more on spatial-analogy skills. Sex differences and implications for the measurement of spatial abilities are also discussed. (Author)