Measurement error: implications for diagnosis and discrepancy models of developmental dyslexia

Abstract

The diagnosis of developmental dyslexia (DD) is reliant on a discrepancy between intellectual functioning and reading achievement. Discrepancy-based formulae have frequently been employed to establish the significance of the difference between 'intelligence' and 'actual' reading achievement. These formulae, however, often fail to take into consideration test reliability and the error associated with a single test score. This paper provides an illustration of the potential effects that test reliability and measurement error can have on the diagnosis of dyslexia, with particular reference to discrepancy models. The roles of reliability and standard error of measurement (SEM) in classic test theory are also briefly reviewed. This is followed by illustrations of how SEM and test reliability can aid with the interpretation of a simple discrepancy-based formula of DD. It is proposed that a lack of consideration of test theory in the use of discrepancy-based models of DD can lead to misdiagnosis (both false positives and false negatives). Further, misdiagnosis in research samples affects reproducibility and generalizability of findings. This in turn, may explain current inconsistencies in research on the perceptual, sensory, and motor correlates of dyslexia.

Cognitive Profile of Students Who Enter Higher Education with an Indication of Dyslexia

“…Researchers have disagreed about whether this has theoretical consequences for the relationship between reading/writing skills and other abilities, or whether it is simply a consequence of the notoriously low correlations one is bound to find for difference scores of highly correlated variables (e.g. [63]). Our data leave little doubt that, as a group, dyslexics entering higher education show exactly the profile predicted by the traditional definition of dyslexia, even though at an individual level the difference scores may show large variability.…”

The mental health, emotional literacy, cognitive ability, literacy attainment and ‘resilience’ of ‘looked after children’: A multidimensional, multiple‐rater population based study

“…A discrepancy of ≥20 points occurred among 5% ( p > 0.01) of the norm sample (Elliot, ). Discrepancy analysis is contentious (BPS (British Psychological Society), ), but widespread (Cotton, Crewther, & Crewther, ). Diagnostic thresholds are also subjective, but an occurrence rate of less than 5% of the norm sample is typical (Kavale & Flanagan, ; Meyer, ).…”

Can Intelligence Testing Inform Educational Intervention for Children with Reading Disability?

“…While the discrepancy model can be challenged on the grounds that it introduces measurement error [2], its use has been primarily discredited because supposed differences between high and low IQ poor readers have not proven to be meaningful in educational terms. More recently, findings from neurophysiological studies have indicated no support for the argument that there are different aetiologies for IQ discrepant and nondiscrepant groups in relation to phonological awareness, a very influential aspect of reading disability [3,4].…”