ABSTRACT
Spatial abilities have long been relegated to a secondary status in accounts of human intelligence. Tests of spatial abilities are often viewed as measures of practical and mechanical abilities that are useful in predicting success in technical occupations, but not as measures of abstract reasoning abilities (Smith, 1964). This conflicts with the important role afforded to spatial imagery in accounts of creative thinking (Shepard, 1978), and with the observed correlations between spatial tests and other measures of intelligence. In fact, Spearman (see Spearman & Wynn Jones, 1950) considered spatial tests merely as unreliable measures of g. Hierarchical factor analyses generally support Spearman’s conclusion, especially for complex spatial tests. Such tests are primarily measures of g, secondarily measures of something task-specific, and thirdly measures of something that covaries uniquely with performance on other spatial tasks (Lohman, 1988). Simpler, speeded spatial tasks show lower g loadings, higher task-specific loadings, and higher spatial factor loadings. In this chapter, I first summarize and then attempt to explain these findings. The relationship between spatial task performance and g may reflect both statistical artifacts and psychological factors. Psychological factors include the attentional demands of maintaining and transforming images in working memory (Kyllonen & Christal, 1990) and the importance of mental models in reasoning (Johnson-Laird, 1983). Indeed, one can turn Spearman’s conclusion around and with equal conviction conclude that measures of g are, by and large, unreliable measures of the ability to generate and coordinate different types of mental models in working memory. Evidence that supports and challenges such a conclusion is reviewed.
