13 Research in Rural Education, Volume 5, Number 1, 1988 Gender, Urbanicity, and Ability! THEODORE COLADARCI, PH.D.2 AND WALTER G. McINTIRE, PH.D.2 This paper presents the results of an analysis of the relative contributions of gender and urbanicity in explaining variability among high school students on six measures of academic achievement and cognitive functioning. When adjustments for·SES were made, gender and urbanicity independently accounted for little variance on these measures. The High School and Beyond data base (1980-1982) was utilized. Implications for rural educators and future research are suggested. Psychologists have long investigated differences be- tween males and females on such variables as aptitude, academic achievement, and personality (8, 9,11). While gender differenceshave been declining over recent decades (3, 7), their magnitude has been found to vary by site and context (4). Interestingly, urbanicity is a context variable that remains to be examined in this regard. That is, does the magnitude of gender differences on various ability measures depend on whether subjects are from urban, suburban, or rural contexts? The present study was designed to address this question. Specifically, we examined the relative contributions of gender and context in explaining variability among high school students on measures of vocabulary, reading comprehension, mathematics, perceptual discrimination, paired-associate memory, and spatial reasoning. Method We employed the High School and Beyond (HSB) data base, a nationally representative sample of high school seniors in 1980 (10), HSB subjects were drawn from the 11,995 students who composed the 1980 senior cohort and participated in the 1982 follow-up study. The N for these analyses varied from 9,849 to 10,064, depending on the variables involved. We conducted all analyses with a modified HSB sampling weight in effect. For each subject, the HSB weight for the 1980 senior cohort, BYWT, was divided by the mean BYWT for these subjects to create the modified weight. This modified weight corrects for oversampling while preserving the sample size. For these weighted data, 48.4% of the students were male and 51.6% were female. Regarding context, 20.0%, 49.4%, and 30.6% of these students were attending urban, suburban, and rural schools, respectively. The ability measures are briefly described here. For more detailed discussion of these measures and their psychometric properties, see Heyns and Hilton (6). Vocabulary, which had 27 items, measured vocabulary through a synonym format. Reading was a 20-item test of reading comprehension. Mathematics had 33 items that called for quantitative comparisons. Mosaic Comparisons, an 89-item test, assessed the speed and accuracy with which one makes perceptual discriminations. Picture- Number was a paired-associate memory test containing 15 items. Finally, Visualization comprised 16 items asking one to visualize the shape a flat surface would assume if folded in a specified manner. Results Our discussion below focused on the magnitude of the results, rather than their statistical significance. Indeed, with Ns ranging from 9,849 to 10,064, even trivial correlations or mean differences can be statistically significant. Means, standard deviations, and intercorrelations are presented in Table 1. As would be expected, strong positive correlations (rs=.62 to .71) were found among Vocabulary, Reading, and Mathematics. Positive, if smaller, correlations (rs=.18 to .34) also were observed among Mosaic Comparisons, Picture-Number, and Visualization. Given that these three measures represent considerably different constructs, the smaller intercor- relations were not surprising. Similar correlations were found between the first and second sets of measures (rs=.21 to .49). To assess the relative contributions of gender and context for explaining variability in each of the six ability measures, we performed on each measure a two (gender) by three (context) analysis of covariance where socio- economic status (SES) served as the covariate. Instead of merely examining the statistical significance of each variance component, we determined the proportion of the total sum of squares in each measure that was 1A version of this paper was presented at the 1988 meeting of the American Educational Research Association in New Orleans. We wish to thank Scott F. Marion for his assistance in carrying out the statistical analyses and Sara Sheppard for preparing the tables. 2College of Education, University of Maine.