Paper Summary
Share...

Direct link:

The Latent Structure of Spatial Skill and Mathematics

Tue, April 17, 2:15 to 3:45pm, Millennium Broadway New York Times Square, Floor: Seventh Floor, Room 7.01

Abstract

The link between spatial skill and mathematics is well established; however, little is known about the particular processing links involved or how these may change over development (Mix & Cheng, 2012). Because both domains are comprised of multiple skills and concepts, they might interact on a general level, one of many specific levels, or both. Further, although a general relation is likely to hold at all age points, the specific interconnections may change over time.

To find out, we completed a factor analysis with data from 1592 children tested in two waves. Children were drawn from three grades: kindergarten (n = 526, 263 boys, mean age =6.04), third (n = 537, 238 boys, mean age =9.05 years) and sixth (n =529, 252 boys, mean age =11.78 years). All children completed a battery of 6 spatial and 5 mathematics tasks, including Block Design (WISC-4), Mental Rotation, Visual-Spatial Working Memory, Map Reading, Perspective-Taking, and the Test of Visual-Motor Integration (VMI), place value concepts, calculation, word problems, fraction concepts, number line estimation (e.g., Siegler et al., 2006), and, in the 6th grade, algebra and geometry. Children in Wave 2 also completed a spatial scaling task (Boyer & Levine’s, 2012, proportion matching).

We have previously presented our findings from Wave 1. Using an exploratory factor analysis, we found evidence for two distinct but highly correlated factors—one comprised of spatial tasks and the other comprised of mathematics tasks. There also were a few significant cross-loadings—variables that loaded significantly onto both factors—but the particular pattern of cross-loadings changed across age. In the present study, we sought to replicate this pattern using the data from Wave 2. First, we simultaneously fit the same model to both datasets (Wave 1 and Wave 2) using multi-sample confirmatory factor analysis (see Table 1).

We also carried out two exploratory factor analyses on Wave 2 alone, one with only the Wave 1 tasks included (see Table 2) and one with the spatial scaling task added (see Table 3). We consistently obtained the same structure: (1) the two-factor model fit both datasets best, (2) the two factors are comprised of spatial and mathematics tasks, respectively, and (3) the two factors are highly correlated. However, the patterns involving cross-loadings were less clear-cut. Although there were significant cross-loadings, and many of these were the same as we found in Wave 1, they were not all the same. Also, the significance of the cross-loadings was affected by the type of analysis and the tasks included. Thus, we suggest interpreting these cross-loadings with caution.

The combined dataset also provided enough power for us to examine the factor structure of spatial skill and mathematics for boys versus girls. In kindergarten, a parallel CFA revealed that mental rotation loaded onto both the spatial factor more heavily for males than it did for females. Separate EFAs also showed that mental rotation significantly cross-loaded onto the mathematics factor for boys, but not girls. However, this gender difference was not apparent after kindergarten, suggesting it fades with development.

Authors