IQ/achievement discrepancy certainly indicates the existence of a problem (it is a discrepancy after all) but the measure says nothing useful about its aetiology (its origin or cause). Most frontline research, even recent research, and most consequent theorising on the nature or remediation of ‘dyslexia’, however, rests on this definition of the sample population as distinct, different and dyslexic by application of this discrepancy criterion, and by this criterion alone. (eg Brooks & Weeks 1998, Fisher et al 1999, Hanley 1997, Hogben 1997, Hynd et al 1995, Nicolson & Fawcett 1999 and many, many, many more. Some are coy – Shaywitz’s widely read book (2005) states clearly that the discrepancy criterion is no longer valid as a diagnostic tool, but lets slip that the research the book is largely based upon is the Connecticut study, from a few years back. The sample population for that study was identified as ‘dyslexic’ by exactly that discrepancy criterion.)

Such studies, and such conclusions, may thus be fundamentally, even fatally, flawed. Many subsequent studies rest their arguments entirely upon, or take as their starting point, the apparent results and conclusions from these earlier studies, almost all of them based upon the discrepancy criterion. Where sample selection is invalid, then so are results and so are conclusions and so is almost all we think we know about ‘dyslexia’. This is fundamental, so let us look carefully and sceptically at the evidence and the debate. How valid is the discrepancy criterion as a selection tool for ‘dyslexia’?

Samuelsson et al (1999 p. 94-95) state that ‘… the use of IQ achievement discrepancy definitions does not constitute a valid approach to define sub-groups of poor readers.’ Stanovich (1991 p. 22) says that ‘Defining dyslexia by reference to discrepancies from IQ is an untenable procedure.’ Siegel & Himel say (1998 p.91) that ‘… the use of an IQ-based discrepancy definition of dyslexia is invalid on variables closely related to the reading process.’ and on p. 102 that ‘…the use of IQ to define dyslexia (or reading disabilities) seems fatally flawed because of the confounding with SES [socio-economic status] and age.’ Adams (1990 p. 59) says that ‘Whereas IQ & general cognitive skills seem not to have much bearing on early reading achievement, early reading failures seem to result in a progressive diminution in IQ scores and general cognitive skills.’ Stanovich & Stanovich (1997 p.6) write that ‘In the largest longitudinal growth-curve investigation ever conducted, Francis et al (1996) reported that IQ was completely useless for predicting future reading growth among poor readers.’ Miles & Miles (1999 p. 114) say that ‘It is perhaps strange that this notion of a discrepancy definition survived as long as it did.’

A single issue of the journal Dyslexia considers this subject. In it various authors dispute the validity of the IQ/achievement discrepancy criterion as diagnostic tool. (Miles 1996, Share 1996, Stanovich 1996, Tunmer & Chapman 1996, and see Cotton et al 2005.) Stuebing et al (2002) have produced a highly regarded meta-analysis of research in this area, clearly showing that the discrepancy criterion should no longer be regarded as valid, and it should no longer be used as a tool for the diagnosis of dyslexia. If this is true today then it was also true yesterday, of course.

Keith Stanovich seems frustrated. He writes that

The persistence of the discrepancy concept in LD [learning difficulties] signals that the field is not yet ready to put itself on a scientific footing and that it will continue to operate on the borders of pseudoscience… The field suffers greatly from its tendency to base practice on concepts and psychometric technologies that have been superseded by subsequent scientific advance. I am referring here to the field’s persistence in linking the definition of learning disability to the concept of aptitude-achievement discrepancy and identifying aptitude with intelligence test performance. (Stanovich 2005 p. 103)

Fletcher (2004) observes that ‘Recent empirical synthesis and consensus reports share the common finding that IQ is ineffective in the identification of LD [learning disability] …’ and that ‘It is widely recognised that the presence of IQ-discrepancy … does not mean that the student has a neurobiological disorder.’ He says that ‘The clear consensus … is to abandon the ability-achievement discrepancy model.’ (And see Fletcher, Denton & Francis 2005 & Fletcher, Francis, Morris & Lyon 2005 reporting ‘serious psychometric problems’ causing low reliability and validity of aptitude-achievement discrepancies as indicative of learning disability - it is not possible accurately, well, or in some cases at all to measure that which we have claimed to measure).