Page 90

in reading by ignoring those who are just on the other side of an arbitrary categorical boundary (Shaywitz et al., 1992).

In prospective epidemiological studies, the rates of specific reading retardation in Yule et al.'s sample was 3.5 percent of 10-year-olds and 4.5 percent of 14-year-olds on the Isle of Wight and 6 percent of the 10-year-olds in London; the criterion was scores that were two standard deviations (SD) from the mean (Yule et al., 1974). In contrast, Silva et al. (1985) found only 1.2 percent of a sample of New Zealand schoolchildren met the same SD criterion. Similarly, Rodgers (1983), examining populations in Great Britain and Northern Ireland, reported that 2.29 percent of children had scores falling two SD below the mean for reading achievement.

The Connecticut longitudinal study, using a less stringent discrepancy criterion of a 1.5 SD discrepancy between predicted and actual reading achievement based on a regression equation or on a criterion of low achievement in reading, found 17.5 percent of the population of schoolchildren in primary and middle school to have reading difficulties (Shaywitz and Shaywitz, 1996). Other available prevalence data are limited either by the population base or by definitional concerns. For example, in Canada, a privately appointed multidisciplinary committee, the Commission on Emotional and Learning Disorders in Children (1970), estimated that between 10 and 16 percent of school-age children required diagnostic and remedial help in learning. This finding is consistent with findings in U.S. studies that 14.8 percent of students in grades 3 and 4 (Mykelbust and Boshes, 1969) and 14 percent of students in grades 7-11 (Meier, 1971) met criteria for underachievement.

Reviewing both population-based studies and numbers of school-age children receiving special education services, the Interagency Committee on Learning Disabilities (1987), in a report to Congress, estimated the prevalence of learning disabilities as ranging from 5 to 10 percent. The vast majority of children identified as having learning disabilities, and therefore reading difficulties, are identified by grade 4. Standard measures of reading are often inappropriate for identifying reading difficulties in older individuals, particularly those who can identify words accurately but not automatically. Preva-