There has been chronic and widespread dissatisfaction with procedures used to identify students as learning disabled. Critics have especially questioned the use of ability-achievement discrepancy as a criterion for eligibility for special education services (Donovan & Cross, 2002; Fletcher et al., 2002; Lyon et al., 2001; Siegel, 1989, 1992; Stanovich, 1991, 1993; U.S. Department of Education Office of Special Education and Rehabilitative Services, OSERS, 2002; Vellutino et al., 1996). Because of such extensive discontent with discrepancy as a method of identification, the Individuals With Disabilities Education Improvement Act (IDEA, 2004) now allows an alternative method of identification: response to intervention (RTI). Without debating the merits of RTI, it is fair to say that the field of learning disabilities is about to enter a state of flux with respect to how students are identified; some would go even further and assert that the very construct of learning disabilities is being questioned (Kavale, 2005; Mastropieri & Scruggs, 2005). They see RTI as essentially redefining learning disabilities as being low achievement rather than unexpected low achievement relative to ability.
One of the most frequently died pieces of evidence used to argue that learning disabilities identification practices, including the use of discrepancy, are flawed has been the wide variability across states in the prevalence rates of learning disabilities (Algozzine & Korinek, 1985; Batsche et al., 2005; Kavale & Forness, 1998; MacMillan & Siperstein, 2002; Reschly, 2002; Reschly & Hosp, 2004; Ysseldyke & Algozzine, 1984). For example, Reschly and Hosp have stated:
The original purpose of the IQ-achievement
discrepancy in the federal regulations was to
provide guidance to states on SLD [specific
learning disabilities] identification and to exercise
control over SLD prevalence.
The goal ... has not been achieved. Prevalence
continues to vary significantly across
the states for reasons that are not simply related
to the stringency of the SLD IQ-achievement
criteria or to other obvious
features of SEA requirements. (p. 210)
In a similar vein, Reschly, in a discussion of high incidence versus low incidence categories, wrote:
SLD varies from a low of about 3% in Kentucky
to a high of over 9.5% in Rhode Island.
ED [emotional disturbance] is prevalence
varies from an unrealistically low 0.1% in
Arkansas to a more realistic 2% in Minnesota.
There are no ready explanations for
variations by factors of 3 and 20 in SLD and
ED prevalence, respectively. It stretches
credulity to posit that variations of that magnitude
actually exist in the respective populations
of those and the other states. (p. 120)
In the 1980s, when researchers and policy makers first began to point to interstate prevalence variability as indicative of the vague or arbitrary criteria used to determine eligibility for learning disabilities services, Hallahan et al. (1986) conducted what was (and to this point has remained) the only direct comparison of state prevalence rates for all categories of special education. Although there was then and still continues to be undeniable variation in the prevaleace rates of learning disabilities from state to state, the categories of special education are also widely discrepant with respect to mean prevalence. For example, for 6- to 17-year-olds in the 2001 to 2002 school year, the national prevalence rates for learning disabilities, mental retardation, and visual impairments were 5.49%, 1.07%, and 0.05%; respectively (U.S. Department of Education, 2003).
Hallahan et al. (1986) applied a statistic, the coefficient of variation (CV), which statisticians recommend for comparing variability when the means of the conditions being compared are radically different (Friedman, 1972; Lang & Secic, 1997). …