Faster, smarter? Working memory capacity and perceptual speed in relation to fluid intelligence

Abstract

Numerous studies have found that working memory capacity and perceptual speed predict variation in fluid intelligence. Within the cognitive ageing literature, perceptual speed accounts for substantial ageing variance in working memory capacity and fluid intelligence. However, within young adults, the interrelationships among these three abilities are less clear. The current work investigated these relationships via confirmatory factor analyses and structural equation modelling using tasks with verbal, spatial, and numerical content. The results indicate that working memory capacity and perceptual speed were not related in a large, cognitively diverse sample of young adults. However, both working memory capacity and perceptual speed accounted for unique variance in fluid intelligence. The results are discussed in relation to previous research with young and older adults.

Keywords:

• Fluid intelligence
• Individual differences
• Perceptual speed
• Working memory capacity

Acknowledgements

We thank Zach Hambrick for providing the Letter, Pattern, and Number Comparison task materials. David McCabe provided Figure 1A. We thank Phil Ackerman, Zach Hambrick, Mike Kane, and Paul Verhaeghen for helpful comments on a previous version of the manuscript and the members of the Attention & Working Memory Lab for assistance with data collection.

Notes

¹Unless stated otherwise, for the rest of the paper, we will use PS to refer to the PS-Scanning factor in the taxonomy of PS abilities identified by Ackerman and Cianciolo (2000).

²The complete results of this reanalysis can be obtained by contacting the first author. The fit was excellent: χ²(24) = 24.03, p =.46; χ²/df=1.00; NNFI =1.00; CFI = 1.00; RMSEA<.01; SRMR=.05.

³Before providing the results of the confirmatory factor analyses (CFAs) and structural equation models (SEMs), we note the criteria used to assess model fit provided by LISREL. A nonsignificant (p>.05) χ²-value is desirable, although with sufficiently large sample sizes, a significant χ²-value will be obtained and not necessarily be indicative of poor model fit. We also report a ratio of the χ²-value and the degrees of freedom in the model, with a ratio value of two or less indicating acceptable fit. Values of the nonnormed fit index and the comparative fit index greater than .90 indicate acceptable model fit (Kline, 1998). Root mean square error of approximation values and standardised root mean square residual values less than .08 indicate acceptable model fit (Kline, 1998). In order to statistically compare models, χ²-tests of the difference (Δχ²) between the two models were used, with p<.05 indicating better statistical fit. In addition, the Akaike information criterion was used to compare models, with the model associated with the smallest value representing the best statistical fit.