Snow, Freedman, and Ford (1986) entered into the debate over gender differences, concluding that such differences tended to characterize Wechsler-Bellevue but not WAIS studies. They reported a nonsignificant .37 correlation between the percentage of males and the size of V P IQ discrepancy for 16 samples tested on the WAIS, compared to a significant (.05 level) correlation of .58 for 12 samples tested on the WechslerBellevue. However, their conclusion is unwarranted. Appropriate methodology does not depend on one coefficient falling just short of statistical significance and the other one emerging as significant. The coefficients must be shown to be significantly different from each other to support their conclusion. The values do not differ sufficiently to justify Snow et al. s interpretation of the data. Nonetheless, Snow et al. are correct in trying to understand variables that might interact with gender to account for the male female differences observed in the literature. These authors also showed that the percentage of males correlated about .90 with age, education, and chronicity in Wechsler-Bellevue studies, although these findings were not replicated in the WAIS investigations. The ultimate impact of Snow et al. s results is to remind researchers and clinicians that the best way to study the relationship of any variable to the impact of unilateral brain damage is to hold as many other variables constant as is reasonably possible, given the practical constraints on neuropsychological research.

Turkheimer and Farace (1992) more recently conducted a meta-analysis of gender-difference studies to follow up on the data presented by Inglis and Lawson (1982), Bornstein and Matarazzo (1982), Snow et al. (1986), and Kaufman (1990). Turkheimer and Farace examined the Verbal and Performance IQs in 12 studies by conducting a repeated-measures analysis of variance with three dichotomous independent variables: gender, hemisphere of lesions, and type of Wechsler scale (Verbal or Performance). They conducted their analyses with both weighted and unweighted means and found the results to be highly similar for both. The results of their repeated-measures analysis revealed a significant main effect for laterality, with left-hemisphere lesions producing greater deficits (M = 91.9) than right-hemisphere lesions (M = 96.1; p < .008), across gender and Verbal and Performance scales. However, when examining the effects of lesion across genders in the separate Verbal and Performance scales, the V-IQ versus P-IQ difference produced by the right lesions was substantially larger than the V P dif-

ference produced by left lesions (1.7 point discrepancy vs. an 11.2 point discrepancy). The main effect for gender was also significant, as female patients had lower IQs than did males across type of scale and laterality. The final main effect of type of scale (V or P) showed that the mean V-IQ of 96.3 was significantly higher than the mean P-IQ of 91.6. Most important was the significant three-way interaction found between gender, type of scale, and laterality. Male and female patients showed roughly equivalent scores on V-IQ following left hemisphere lesions, but female patients with left-hemisphere lesions had lower P-IQs (note that P-IQ is not supposed to be associated with left hemisphere functions). Male and female patients with right hemisphere lesions had roughly equal P-IQs, but female patients had lower V-IQs; again, V-IQ is not ordinarily associated with the right hemisphere. Turkheimer and Farace conclude that the primary effect appears to be that female patients are more sensitive to lesions in the hemisphere opposite to that thought to be dominant for a function. The sets of means presented in Table 8.13 reveal that males and females with unilateral lesions did not differ in their means on the IQ (either V or P) that is supposed to be associated with a specific hemisphere (shown in bold print),