The Binormal Hypothesis of Specific Learning Disabilities

Abstract

The concept of specific learning disabilities has its roots in the medical literature of the nineteenth century. According to the medical model the cause of specific learning disabilities are presumed to lie in specific cognitive dysfunctions. This hypothesis predicts two qualitatively distinct types of learner and a bimodal distribution of assessment scores. Evidence for bimodality has been sought in the distribution of residuals generated from the regression of standardised measures of attainment on IQ, however this technique has been widely criticised. Recent advances in computer adaptive assessment, coupled with Rasch interval level measurement, have opened up the possibility of seeking evidence for bimodality in the distribution of assessment scores directly.

In the present study the binormal distribution was developed as a model for describing bimodality. The binormal distribution is conceived as two superimposed normal distributions and is defined by five parameters. The algebraic relationship between the five parameters was first determined, and then a methodology was developed for deriving objective estimates of those parameters. The methodology was applied to a unique dataset of over 80,000 children aged between seven and eleven years of age, and across four assessment domains; picture vocabulary, reading, mathematics and arithmetic.

The methodology was found to be sensitive to factors that might influence the shape of the distribution of assessment scores such as gender, number of years of schooling, and ceiling effects, and this affected its utility. Nevertheless evidence was found for the existence two qualitatively distinct groups of reader. The pattern in these results was consistent with a developmental transition from beginning to fluent reader. Evidence was also found for a developmental lag between boys and girls, which would explain the higher prevalence of dyslexia reported for boys in many studies. The methodology produced inconsistent results when applied to the other assessments, and no evidence was found to either confirm or disprove the existence of specific dysfunctions as predicted by the medical model.