Poster #12 - Which dimensions of numerical cognition are most strongly related to math performance?

• In Event: Poster Session 08

Abstract

Early math performance is predictive of long-term educational and occupational outcomes. As such, identifying the core cognitive skills underlying math performance is a critical step towards supporting children’s academic development. There are a variety of tasks designed to assess core domain-specific numerical cognition skills. Extant studies suggest that symbolic number skills may be most strongly related to math (e.g., Schneider et al., 2017), though relatively few studies have directly compared the various types of numerical tasks within the same analysis (Lyons et al., 2014). To build upon these findings, the current study proposes a novel organizational structure for characterizing numerical cognition tasks which delineates two key dimensions: (1) the manner in which stimuli are represented (e.g., symbolically using Arabic digits versus non-symbolically using dots), and (2) the precision of quantification that is required (e.g., determining the exact magnitude versus approximating the relative magnitude). Tasks can be categorized into four quadrants within a two-by-two matrix using these dimensions (i.e., symbolic-exact, non-symbolic-exact, symbolic-approximate, and non-symbolic-approximate tasks). Prior research has generally not separated the two dimensions, instead considering symbolic tasks as tapping the exact number system and non-symbolic tasks as tapping the approximate number system, which fails to address symbolic-approximate and non-symbolic-exact number tasks. Consequently, it remains unclear which dimensions of numerical cognition are most strongly related to math skills. The current study addresses this gap by including tasks that fulfill the full range of the two-by-two matrix (symbolic/non-symbolic, exact/approximate) within the same analyses. We hypothesized that tasks using symbolic stimuli and requiring exact responses would be most strongly related to mathematics skills, given their closest resemblance to calculation.
Analyses included participants from the Colorado Learning Disabilities Research Center twin sample (N=192; ages 8-16). Numerical cognition was measured using the Number Sets (symbolic- and non-symbolic-exact; Geary et al., 2009) and Numeracy Screener (symbolic- and non-symbolic-approximate; Nosworthy et al., 2013). Mathematics performance was assessed using a composite of WRAT-R Math, WJ-III Math Fluency, and WJ-III Applied Problems. We used a multilevel mixed-effects linear regression to account for twin relationships. Mathematics performance was regressed on the four types of numerical cognition measures along with demographic covariates (age and sex). A Bonferroni-adjusted p-value of .0125 was used to account for multiple testing effects. Contrary to our hypothesis, non-symbolic-exact skills were found to be the most strongly related to math performance (β=.52, p<.001), followed by symbolic-exact (β=.34, p=.006). Non-symbolic-approximate and symbolic-approximate skills were not significantly associated with math skills.
In conclusion, when assessing tasks varying in terms of precision of quantification, we found that exact number tasks were significantly related to math while approximate tasks were not, which was consistent with our hypothesis. Further, when considering variations in stimulus representation across the exact number tasks, it was surprising that a non-symbolic task outperformed a symbolic task in statistically predicting math scores, given that math problems utilize symbolic values (i.e., Arabic digits). Such a result suggests that building children’s understanding of exact quantities using non-symbolic stimuli (e.g., manipulatives) may be an important instructional technique to support math development.

Authors

• Nina Jean Anderson, University of Denver
  Presenting Author

• Eliza L Kramer, University of Denver

• Bruce F. Pennington, University of Denver

• Erik G. Willcutt, University of Colorado Boulder

• Lauren M. McGrath, University of Denver