Reading and math are related generally, and comorbidly at the level of disability. Language, working memory (WM), processing speed (PS), and attention are four domain-general processes important for reading and math separately (Floyd et al., 2003; Fuchs et al., 2010; McDougal et al., 2022). Research of shared cognitive predictors is rarer (e.g., Cirino et al. 2018; Peterson et al., 2017). Therefore, the present study aims to evaluate these factors' shared contribution to both reading and math (both timed and untimed) among middle school students. We hypothesized that each of the four cognitive domains would relate significantly to all academic outcomes, and that together, they would account for the relationship between math and reading performance. We also expected that language and attention would be more relevant for reading than for math; that WM would be more relevant for math than for reading, and that PS would be more relevant for timed than for untimed measures.

Two-hundred-eighteen Hispanic middle school students completed cognitive assessments on visual attention, visual search, objective attention, behavioral attention, phonological awareness, rapid automatized naming, vocabulary, WM, and PS. Timed and untimed reading and math were measured using the KTEA-3 (Kaufman & Kaufman, 2014). Latent variables were formed, and analyses were conducted via path analyses.

A measurement model delineated language, attention, WM, and PS with nine latent variables with 20 indicator variables, with good model fit. Variables from each of the four domains significantly correlated with both reading and math outcomes. However, for untimed (overall R2 = 47.8%) and timed reading (overall R2 = 56.8%), language and behavioral attention were the only unique predictors. For untimed math (overall R2 = 51.8%), WM, PS, and behavioral attention were unique predictors. Finally, for timed math (overall R2 = 26.1%), WM was the only unique predictor. Reading and math were correlated with one another, whether untimed (r = .43) and timed (r = .40). Although the set of predictors reduced these correlations, the residual relation between reading and math remained significant, for both untimed (p = .002) and timed (p = .037) outcomes. When specific paths were constrained, language was found to be more important for untimed reading than untimed math but was similarly important for timed outcomes. WM was more important for math than reading, whether timed or untimed. Attention and PS were similarly important for achievement outcomes.

The present work supported prior work documenting the relation of reading and math, and the relation of language, attention, WM, and PS to both types of achievement. However, unique contributions were much more sporadic, and some, but not all, cognitive domains showed differential prediction. These results highlight the role of shared variance among predictors (Cirino et al., 2018; Cirino et al., 2019), and raise questions as to other sources of the overlap between reading and math, whether timed or untimed. The nature of the sample also raises interesting replicability and generalizability issues but advances our understanding of the relation between cognitive and achievement skills.