Numerical and nonnumerical estimation in children with and without mathematical learning disabilities

Abstract

There are currently multiple explanations for mathematical learning disabilities (MLD). The present study focused on those assuming that MLD are due to a basic numerical deficit affecting the ability to represent and to manipulate number magnitude (Butterworth, 1999, 2005; A. J. Wilson & Dehaene, 2007) and/or to access that number magnitude representation from numerical symbols (Rousselle & Noël, 2007). The present study provides an original contribution to this issue by testing MLD children (carefully selected on the basis of preserved abilities in other domains) on numerical estimation tasks with contrasting symbolic (Arabic numerals) and nonsymbolic (collection of dots) numbers used as input or output. MLD children performed consistently less accurately than control children on all the estimation tasks. However, MLD children were even weaker when the task involved the mapping between symbolic and nonsymbolic numbers than when the task required a mapping between two nonsymbolic numerical formats. Moreover, in the estimation of nonsymbolic numerosities, MLD children relied more than control children on perceptual cues such as the cumulative area of the dots. Finally, the task requiring a mapping from a nonsymbolic format to a symbolic format was the best predictor of MLD. In order to explain these present results, as well as those reported in the literature, we propose that the impoverished number magnitude representation of MLD children may arise from an initial mapping deficit between number symbols and that magnitude representation.

Keywords:

• Mathematical learning disabilities
• Learning disabilities
• Dyscalculia
• Mathematical development
• Number magnitude
• Numerical estimation
• Number magnitude representation

Notes

¹The method section of Price et al.'s (2007) paper does not give a detailed description of the items used.

² The AES for these children were entered in a repeated measures ANCOVA with targets (8, 12, 16, 21, 26, 34, 64) and tasks (HM to AN, HT to AN, HT to HM, AN to HM) as within-subject factors, groups (control or MLD) as the between-subjects factor, and the IQ as covariate. The group effect was significant, F(1, 21) = 8.68, η² = .30, p = .008, with a statistical power of the test of .802.