Gender differences in memory test performance among children and adolescents

Archives of Clinical Neuropsychology 18 (2003) 865 878 Gender differences in memory test performance among children and adolescents Patricia A. Lowe a, Joan W. Mayfield b, Cecil R. Reynolds c, a University of Kansas, Lawrence, KS 66044, USA b Baylor Pediatric Specialty Services, Baylor, TX, USA c Department of Educational Psychology, Texas A&M University, College Station, TX 77843-4225, USA Accepted 27 May 2002 Abstract Gender differences among children and adolescents were examined on 14 separate measures of short-term memory. A nationally stratified sample of 1,279 children and adolescents, 637 males and 642 females, ranging in age between 5 and 19 years, were assessed on the 14 subtests of the Test of Memory and Learning (TOMAL). Factor structure of the TOMAL was determined to be invariant as a function of gender. Using age-corrected deviation scaled scores calculated at 1-year intervals, results of a one-way multivariate analysis of variance (MANOVA) revealed only two significant differences in absolute scores across gender on the 14 memory subtests. A profile of normal variations in patterns of memory test performance across gender revealing relative strengths for females on verbal tasks and males on spatial tasks is presented for clinical use and future normative comparisons. 2002 National Academy of Neuropsychology. Published by Elsevier Ltd. All rights reserved. Keywords: Gender differences; Memory test; Children; Adults 1. Introduction Gender differences in memory performance have seldom been examined explicitly (Herlitz, Nilsson, & Backman, 1997). However, studies which have examined the relationship between gender and memory test performance in adults have produced equivocal findings with some Corresponding author. Present address: 909 Pecan Street, Bastrop, TX 78602, USA. Tel.: +1-512-321-4320; fax: +1-512-321-4320. E-mail address: crrh@bluebon.net (C.R. Reynolds). 0887-6177/$ see front matter 2002 National Academy of Neuropsychology. PII: S0887-6177(02)00162-2

866 P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 studies reporting a significant relationship (e.g., Albus et al., 1997; Bolla-Wilson & Bleecker, 1986; Geffen, Moar, O Hanlon, & Clark, 1990; McGivern et al., 1998; Ruff, Light, & Quayhagen, 1989) and other studies finding no relationship (e.g., Freides & Avery, 1991; McCarty, Siegler, & Logue, 1982) between gender and memory test scores. Similarly, research investigating the association between memory test scores and gender in children and adolescents has produced mixed results (e.g., Aliotti & Rajabiun, 1991; Boivin, 1991; Forrester & Geffen, 1991; Huang, 1993; Robinson, Abbott, Berninger, & Busse, 1996; Temple & Cornish, 1993; Ullman, McKee, Campbell, Larrabee, & Trahan, 1997). Temple and Cornish (1993) assessed gender differences among a nonclinical sample of 64 females and 64 males, ranging in age from 9 to 21 years on a verbal memory recognition task. Temple and Cornish found that females outperformed males on this verbal memory task. In another study examining gender differences, Robinson et al. (1996) reported significantly higher scores for male than female preschoolers and kindergartners on a visual-spatial working memory task. Robinson et al s. sample consisted of 778 children who were considered advanced in mathematical reasoning ability. Likewise, Huang (1993) found Chinese adolescent males outscored adolescent females on a visual-spatial memory task, whereas Chinese adolescent girls outperformed adolescent boys on a verbal memory task. In contrast, other studies such as Aliotti and Rajabiun (1991), Forrester and Geffen (1991), and Ullman et al. (1997) have found no relationship between gender and memory test performance. Ullman et al. (1997) assessed 138 children, 67 females and 71 males, in grades 1 5. The authors found no significant gender differences on a task measuring visual recognition memory. Likewise, Forrester and Geffen (1991) found no gender differences on an auditory learning task for 40 girls and 40 boys between the ages of 7 and 15. Based on these results, it is unclear whether gender differences in memory test performance in children and adolescents really do exist. Moreover, these studies, which produced equivocal findings, examined gender differences in memory test performance using instruments that assessed very narrow-band aspects of memory, each with a different normative base. Whether a clearer picture of the relationship between gender and memory test performance would emerge if a comprehensive memory battery wherein all tests are normed on a common sample and using tasks designed specifically to assess children and adolescents memory was used instead of a narrow-band measure is not known. Four memory batteries are now available for children (Goldstein & Incagnoli, 1997), with some batteries being more comprehensive than others. The four memory batteries include the Wide Range Assessment of Memory and Learning (WRAML; Sheslow & Adams, 1990), the California Verbal Learning Test for Children (CVLT; Delis, Kramer, Kaplan, & Ober, 1994), the Children s Memory Scale (CMS; Wechsler, 1995), and the Test of Memory and Learning (TOMAL; Reynolds & Bigler, 1994). The TOMAL is the most comprehensive of these memory batteries and is designed to emphasize the broad-based as well as the narrow-band aspects of memory (Reynolds & Bigler, 1994). The TOMAL consists of 14 subtests (see Fig. 1). Of these 14 subtests, 10 subtests make up the core battery. The core battery consists of five verbal and five nonverbal subtests. Besides the core battery, four supplementary subtests (three verbal and one nonverbal) are included. Delayed recall tasks are also available in this comprehensive memory measure. The TOMAL has solid psychometric properties with very high reliability coefficients at the subtest-level making it particularly useful in the study of individual differences (see Reynolds

P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 867 Fig. 1. TOMAL subtests.

868 P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 & Bigler, 1994). In a factor analytic study with the TOMAL standardization sample, Reynolds and Bigler (1995) examined two-, three-, and four-factor structures of the memory tasks of the TOMAL and came to the conclusion that a four-factor structure of memory represented the best fit to the data. The TOMAL s four factors include a Complex Memory Factor, Sequential Recall Factor, Backwards Recall Factor, and Spatial Memory Factor. In addition to these four factors, a general memory factor (g memory ), similar to g in cognitive measures (but not as strong, leaving more specific variance for each task), emerged. Whether these factors would be comparable across gender for a child and adolescent sample is not unknown. The purpose of the present study was to examine the factor structure of the TOMAL for boys and for girls to determine the comparability of the test s factors across gender and to determine whether any male female differences exist in mean levels of performance. A profile of normal variations in patterns of memory test performance across gender was also produced for clinical and future normative comparison. 2. Method 2.1. Subjects Participants for the current study consisted of 1279 individuals, 637 males and 642 females, from the TOMAL standardization sample. The subjects ranged in age from 5 to 19 years with a mean age of 10.65 years (S.D. = 3.62 years) for the total sample. The mean age of females was 10.84 years (S.D. = 3.68 years) and the mean age of males was 10.58 years (S.D. = 3.61 years). Racial composition of the total sample included Anglo European (82%), African American (11%) and Others (3%). The total sample included individuals who resided in the different regions of the United States: Northeast (15%), North Central (24%), South (34%), and West (23%). Four percent of the sample s residency status was not available. In addition to the nonreported residency status, data were not available for 45 individuals, representing 3% of the current sample, on the gender variable (i.e., gender was not coded during the data collection process) and thus, these individuals were excluded from the analyses. Data analyses revealed, however, a nonsignificant difference in age between those individuals who were excluded and those individuals in the current sample, t(1322) =.36, P>.05, as well as a nonsignificant difference in race (majority vs minority) between those individuals who were excluded and the current sample, z = 1.67, P >.05. However, a larger number of individuals proportion-wise from the southern as opposed to the nonsouthern region of the United States were included in the group of individuals excluded from the current sample, χ 2 (3) = 21.6, P<.05. The TOMAL standardization sample initially was stratified based on estimates of the 1990 U.S. Census and later corrected based on updated reports through 1992. The sample was stratified on the basis of age, gender, ethnicity, socioeconomic status, geographic region, and community size. Population-proportionate sampling was used to ensure the representativeness of the sample relative to the normal 1990 U.S. child and adolescent population. Data were collected from 17 states including California, Colorado, Florida, Georgia, Maine, Maryland, Massachusetts, Minnesota, New Hampshire, New York, North Carolina, Ohio, Pennsylvania,

P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 869 Texas, Utah, and Washington. A more detailed description of the TOMAL s standardization sample characteristics are provided by Reynolds and Bigler (1994). 2.2. Instrument The core and supplementary subtests (sans the four delayed recall tasks) of the TOMAL were used. The core and supplementary subtests are designed to give information on specific and general aspects of memory and are used to derive the Core Indexes, Supplementary Indexes, and the Composite Memory Index. The core subtests include Memory for Stories, Word Selective Reminding, Object Recall, Digits Forward, Paired Recall, Facial Memory, Visual Selective Reminding, Abstract Visual Memory, Visual Sequential Memory, and Memory for Location. The ten core subtests are used to derive the Composite Memory Index. In addition to the core subtests, there are four supplementary subtests, including Letters Forward, Letters Backward, Digits Backward, and Manual Imitation. Internal consistencies of the subtests and composites, represented by Cronbach s coefficient alphas and reported at one-year age intervals for the standardization sample, are uniformly high, with all composites and 63% of the subtests reliability coefficients at.90 and above; 31% of the subtests reliability coefficients are between.80 and.89; and only 6% of the subtests reliability coefficients fall below.80, with none below.74, excluding the delayed recall tasks. Test-retest reliability coefficients of the subtests and composites reported over an average 6-week testing interval ranging from.71 to.92, with all composite index reliability coefficient values exceeding.80. Extensive evidence of the validity of the TOMAL has also been noted. Evidence for content-descriptive, criterion-predictive, and construct validity of the TOMAL test scores has been reported (see Reynolds & Bigler, 1994, (Table 4)). 2.3. Procedure To assess accurately differences in mean levels of performance of two or more groups across any set of variables (such as a profile of cognitive test scores such as on the TOMAL or a Wechsler scale for example), one must first determine whether the latent structure of the variables is common across the groups being compared. If the latent constructs assessed are different or lack common dimensionality, artifactual differences in performance may appear when mean scores are compared (e.g., see Jensen, 1980, for a detailed explanation). Essentially this can occur because the groups may not have been measured on common dimensions or constructs and thus one is comparing performance on different constructs even though the stimuli may be constant across groups. If the groups differ on these unknown dimensions, differential regression effects may occur as well and produce artifactual differences in profiles across groups. The best approach to evaluate the comparability of the latent structure of the variables across groups is to make use of some form of comparative factor analysis (Jensen, 1980; Reynolds, 2000), and that is the approach taken herein (see also Table 3). The responses of the 1279 children and adolescents were converted to age-corrected deviation scaled scores (Mean = 10, S.D. = 3) based on 1-year intervals using the standard conversion tables in the TOMAL Manual. The children s responses were then factor analyzed for the total sample through the method of principal axis factoring with R 2 in the diagonal of

870 P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 the item correlation matrix as the initial communality (h 2 ) estimates. Principal axis factoring was selected as the method of extracting factors because it is a conservative method and it does not assume perfect reliability of the variables input to the analysis. Once extracted, the factors were rotated through oblique (Promax) and orthogonal (Varimax) procedures. Exploratory approaches to the data seemed more appropriate since the choice of factor extraction methods among psychometricians remains largely subjective. The Promax rotation procedure, an oblique solution, was chosen because it simplifies the pattern or factor structure and allows the natural intercorrelation of the underlying factors to be seen and to influence the analyses. On the other hand, the Varimax rotation procedure, an orthogonal solution, was selected because it tends to add clarity to the factor structure, approximates simple structure, and assumes that the underlying factors are independent. Both of these rotation procedures also tend to produce clinically useful solutions that have the greatest applications to practice questions. Both sets of results are reported here as an aide to future researchers as well as to assess any variations in outcome that may have occurred by virtue of forcing the results to an orthogonal or an oblique solution. Four-factor Promax and Varimax solutions were obtained for the current sample, as had occurred in the original analyses of the TOMAL s standardization sample. The four-factor Promax and Varimax solutions were the clearest and most interpretable solutions for the current sample. Four-factors were then extracted through principal axis factoring and rotated via the Promax and Varimax procedures for the male and female subsamples. To determine the degree of similarity of the factors across gender, two indices of factorial similarity were calculated between pairs of corresponding factors as recommended in several sources (e.g., Reynolds, 2000), the coefficient of congruence (r c ; Gorsuch, 1988; Harman, 1976) and the salient variable similarity index (s; Cattell, 1978). The coefficient of congruence is a measure that evaluates the shared variance between factors. A value of.90 or above indicates a high degree of similarity between factors or factorial invariance. Since the coefficient of congruence statistic is sensitive to violations of the assumptions of equivalent variance covariance matrices, normal distributions, and to factor size, the salient variable similarity index was also computed along with r c. The salient variable similarity index is a nonparametric statistic that is not a measure of shared variance between factors. The s statistic assesses factorial similarity by distinguishing categorically those variables that load or do not load on both factors of a matched pair from those variables that load on only one factor of a matched pair or fail to meet the factor loading criterion cutoff of ±.20 on one or both factors of a matched pair. The probability of this overlap of the variables on factors of a matched pair that is categorically defined is known as the s statistic. A conservative value ±.20 to denote saliency, instead of ±.10, was chosen to take a conservative stance and not over interpret the results. An s value may range from 1.00 to 1.00 with a 0.00 value indicating no relationship between the corresponding factors. Significant tests are used to assess s, whereas the magnitude of r c is more crucial; r c is a measure of shared variance but s is not. A statistically significant s-value, which suggests factorial invariance, is determined by the magnitude of the s-value, the number of variables (e.g., subtests) present, and the number of variables located in the hyperplane (see Cattell, 1978, for tables). In addition to the factor analytic procedures, means and standard deviations were computed separately for males and females on the 14 subtests. A multivariate analysis of variance

P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 871 (MANOVA) was computed to compare the mean levels of performance on the 14 subtests across gender. The MANOVA was followed by univariate F tests with Bonferroni type adjustments. The General Linear Model (GLM) Multivariate Analysis procedure was used in these computations. The GLM model was selected over more conventional multivariate analysis procedures because of its correction for unequal cell sizes. Fourteen, one-way analyses of variance (ANOVAs) were then conducted across gender with the 14 subtest scaled scores serving as the dependent variables in separate analyses. The GLM Factorial Analysis of Variance procedure was used in computing the ANOVAs to correct for unequal cell sizes. A profile of normal variation in patterns of memory-test performance across gender was also determined based upon patterns of relative, as opposed to absolute, strengths and weaknesses. Partial correlations between gender and subtest scaled score for each subtest holding the gender Composite Memory Index scaled score constant were calculated. All analyses were performed using the 1996 version of SPSS, release 7.5. 3. Results and discussion Prior analyses of the entire sample (Reynolds & Bigler, 1995) revealed that a four-factor solution to the intercorrelation matrix was the best fit to the data. Following Reynolds (2000) recommendations, this same solution was extracted separately for males and females following a determination that more than one latent construct was in evidence. The four-factor solutions for the sample in the present study yielded Kaiser Meyer Olkin (KMO) measures of sampling adequacy of.85. These values are considered to be good and suggest the correlations between pairs of variables can be explained by the other variables, and factor analysis is an appropriate method used to analyze these relationships. Bartlett s test of sphericity was also computed and was found to be statistically significant, χ 2 (91) = 3458.20, P =.001 for the four-factor solutions for the total sample. Therefore, the hypothesis that the population correlation matrix is an identity matrix was rejected. The four-factor Promax and Varimax solutions for the male and female subsamples are presented in Tables 1 and 2. Visual inspection of Tables 1 and 2 reveal similarities as well as a few minor differences in outcome across gender. One subtest that loaded highest on the Sequential Recall for males, Visual Sequential Memory, loaded highest on the Spatial Memory Factor for females in both the Promax and Varimax solutions. Facial Memory moved from the Spatial Memory Factor for females to the Complex Memory Factor for males for both solutions. On the other hand, Memory for Stories loaded highest on the Complex Memory Factor for females but loaded on both the Complex Memory Factor and Spatial Memory Factor for males in both the Promax and Varimax solutions. These changes, however, are relatively inconsequential. No consistent pattern of alteration is evident, and the difference in the magnitude of the loadings is small. The coefficients of congruence and the salient variable similarity indices between corresponding factors are shown in Table 3. The coefficients of congruence for the male female comparisons range from.80 on the Spatial Memory Factor to.97 on the Sequential Memory Factor for the Promax solution and from.92 on the Backward Recall Factor to.99 on the Sequential Recall Factor for the Varimax solution. In addition, all of the salient variable similarity indices were statistically significant (P <.05) for both the Promax and Varimax solutions.

872 P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 Table 1 Factor pattern coefficients for the female and male subsamples for the four-factor Promax solution Factor I Factor II Factor III Factor IV Subtest F M F M F M F M Memory for Stories.03.09.40.15.15.06.07.15 Word Selective Reminding.06.05.45.68.08.09.17.15 Object Recall.12.00.36.69.06.03.21.01 Digits Forward.89.93.02.00.04.08.07.10 Paired Recall.00.14.79.47.02.15.17.08 Letters Forward.76.61.02.03.01.02.05.04 Digits Backward.04.11.02.06.70.58.01.10 Letters Backward.02.01.04.07.86.89.02.04 Facial Memory.22.08.10.39.04.28.47.26 Visual Selective Reminding.05.03.04.12.01.06.44.28 Abstract Visual Memory.03.04.19.07.20.09.28.43 Visual Sequential Memory.11.23.11.12.09.06.54.10 Memory for Location.02.06.01.08.04.00.43.70 Manual Imitation.21.21.02.02.06.20.37.27 Although the r c values for the Complex Memory Factor and Spatial Memory Factor for the Promax solution are below.90, these r c values are very close to the.90 value. In addition, the s-value is significant in each case between the male and female corresponding factors. On the other hand, the Varimax solution produced r c values above.90 for all male female Table 2 Factor pattern structure coefficients for the female and male subsamples for the four-factor Varimax solution Factor I Factor II Factor III Factor IV g memory Subtest F M F M F M F M F M Memory for Stories.10.16.41.21.23.13.19.21.47.35 Word Selective Reminding.17.18.46.62.09.18.28.05.50.53 Object Recall.21.14.40.65.11.10.31.18.51.54 Digits Forward.82.84.17.17.22.18.16.09.69.66 Paired Recall.09.04.65.46.11.20.10.20.47.45 Letters Forward.72.60.18.13.21.22.24.17.68.58 Digits Backward.22.29.20.12.64.56.22.23.67.73 Letters Backward.22.26.16.23.75.81.21.17.67.73 Facial Memory.05.12.22.39.13.12.41.29.35.34 Visual Selective Reminding.15.08.13.19.11.12.38.30.39.33 Abstract Visual Memory.13.17.31.21.29.18.35.44.54.48 Visual Sequential Memory.21.27.09.23.07.12.44.18.41.40 Memory for Location.11.08.15.10.10.10.37.62.39.41 Manual Imitation.31.31.16.14.14.28.38.33.53.52

P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 873 Table 3 Coefficients of congruence and salient variable similarity indices between corresponding factors of the TOMAL for males and females Coefficient of congruence Salient variable similarity index Factor Promax Varimax Promax Varimax I.97.99.75.86 II.83.93.75.80 III.95.92.50.55 IV.80.93.86.63 P<.05. P<.01. corresponding factors as well as statistically significant salient variable similarity indexes for each pair of factors. Psychometricians are far from agreement on which solution to use when judging whether factors are similar across nominal groupings of subjects. The Varimax solution appears to be a better fit, in this case, to the data than the Promax solution. Thus, the better fit across groups for the Varimax solution and reasonably good fit across groups for the Promax solution, a potentially inferior solution, support the notion that these specific factors are highly similar if not actually invariant across gender. A large general memory factor, the first unrotated factor, also emerged. This finding suggests a strong tendency for the memory tasks to trend in a constant direction. The large value of the coefficient of congruence (r c =.99) and statistically significant salient variable similarity index (s =+1.00) show that this general memory factor is constant across gender. Since the factor structure of the TOMAL is so highly similar across gender, the difference in the mean scores between males and females were examined on all 14 subtests. The mean scores and standard deviations for males and females are depicted in Table 4. Prior to computing the MANOVA, a Box s M test of equality of covariance matrices, Bartlett s test of sphericity, and tests of multicollinearity and singularity were performed. Box s M test was significant, F(105, 3803951) = 1.53, P<.05. Although a violation of the homogeneity of variance covariance matrices assumption exists, it is unlikely that the MANOVA is affected due to a large sample size and the subsamples are relatively equal in size. Under these circumstances, the alpha level is conservative and the MANOVA statistic is robust to violations of the homogeneity variance covariance matrices (Tabachnick & Fidell, 1989). In contrast, Bartlett s test of sphericity was significant, χ 2 (104) = 3.727.19, P<.001, and lead to the rejection of the hypothesis that the correlation matrix was an identity matrix. Furthermore, no evidence of multicollinearity and singularity was present as regressions were performed with each dependent variable in turn serving as the dependent variable with all other dependent variables serving as the independent variables in the analyses. Based on these analyses, R 2 for the dependent variables ranged from.14 to.48 and tolerance statistics for the dependent variables ranged from.53 to.87. The MANOVA was then computed and results are reported in terms of Pillai s Trace. The MANOVA produced a significant main effect for gender, F(14, 1092) = 3.30, P<.001, but with a very small effect size (d =.04).

874 P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 Table 4 Means and (S.D.) on the TOMAL by gender Gender Subtest Male Female Memory for Stories 10.06 (3.23) 10.07 (3.16) Word Selective Reminding 9.87 (2.85) 10.56 (2.87) Object Recall 9.86 (3.05) 10.47 (3.09) Digits Forward 10.02 (3.32) 10.11 (3.50) Paired Recall 9.88 (2.77) 10.06 (2.84) Letters Forward 10.09 (3.16) 10.23 (3.23) Digits Backward 10.23 (3.14) 10.36 (3.33) Letters Backward 9.97 (3.39) 10.42 (3.51) Facial Memory 9.81 (2.65) 10.15 (2.96) Visual Selective Reminding 9.43 (3.31) 9.73 (3.32) Abstract Visual Memory 10.10 (2.97) 9.75 (3.06) Visual Sequential Memory 9.83 (3.02) 9.83 (2.98) Memory for Location 9.93 (4.08) 9.49 (3.93) Manual Imitation 9.85 (3.21) 9.72 (3.32) P<.003. Before computing the follow-up univariate F tests, Bonferroni type adjustments (which are notably conservative in nature) were made due to the possibility of an inflated Type I error rate due to multiple significance testing. The alpha level for these univariate F tests was set at.003. In addition, Levene s test of equality of error variances was calculated for each dependent variable. Levene s test of equality of error variances was found to be significant for the Facial Memory subtest only, F(1, 1277) = 8.02, P<.05. Although a violation of the homogeneity of variance assumption exists, it is unlikely that the ANOVA is affected due to the relatively large sample size and small actual difference between the variances. Under these conditions, ANOVA statistics are robust to violations of homogeneity of variance (McCall, 1975). Univariate F tests were then calculated. The univariate F tests revealed that females scored significantly higher than males on the Word Selective Reminding subtest, F(1, 1277) = 16.02, P<.001, but with a small effect size (d =.01) and Object Recall subtest, F(1, 1277) = 11.07, P<.001, but again with a small effect size (d =.01). The Object Recall subtest requires the examinee to pair verbal and nonverbal stimuli and to recall the association verbally, whereas the Word Selective Reminding subtest is a verbal free-recall task. Both Word Selective Reminding and Object Recall are considered to be verbal tasks. These results are consistent with studies of intelligence in which females have been found to outperform males on certain verbal tasks (Born, Bleichrodt, & van der Flier, 1987; Maccoby & Jacklin, 1974). However, these differences represent absolute differences in scores that may be related to male female differences on the overall composite scores. It is also of interest to ascertain whether there are any differences remaining when overall level of memory skill is controlled across groups. This can be calculated by use of the method of partial correlation wherein total test score (in the case of the TOMAL, the Composite Memory Index) is held constant or partialled from the analysis.

P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 875 Table 5 Partial correlations between gender and each subtest holding the gender composite memory index constant Subtest Memory for Stories.02 Word Selective Reminding.12 Object Recall.09 Digits Forward.02 Paired Recall.01 Letters Forward.00 Digits Backward.00 Letters Backward.04 Facial Memory.05 Visual Selective Reminding.03 Abstract Visual Memory.10 Visual Sequential Memory.03 Memory for Location.10 Manual Imitation.04 P<.01. r Partial correlations between gender and subtest scaled score for each subtest holding the gender Composite Memory Index scaled score constant ranged from.10 to.12 (see Table 5). Four partial correlations were significant: Object Recall, Memory for Location, Abstract Visual Memory, and Word Selective Reminding. Thus, a gender difference independent of any overall memory difference exists for these four subtests. Object Recall and Word Selective Reminding, as mentioned previously, are verbal tasks and females, as a whole, did better on these tasks than males. On the other hand, males outperformed females on the Memory for Location and Abstract Visual Memory subtests. Memory for Location and Abstract Visual Memory are spatial memory tasks. Again, these results are consistent with studies of intelligence in which males have been reported to outperform females on spatial tasks (Kaufman, 1990; Kaufman, Kaufman Packer, McLean, & Reynolds, 1991). The results from the current study provide substantial support for the presence of a general memory factor (g memory ), similar to g on cognitive measures, and four specific factors of memory: Complex Memory Factor, Sequential Recall Factor, Backwards Recall Factor, and Spatial Memory Factor on the TOMAL. These results are consistent with previous studies on the factor structure of the TOMAL (Mayfield & Reynolds, 1997; Reynolds & Bigler, 1995). The findings also suggest that the general memory factor and four specific memory factors are similar across gender. In other words, these factors are essentially invariant across the nominal grouping variable of gender. Since the latent structure is similar for males and females on the TOMAL, these results argue for the same score interpretation for both males and females. The mean level of performance for females was statistically significantly different from males on only two subtests, Word Selective Reminding and Object Recall. Word Selective Reminding and Object Recall are verbal tasks. These findings are consistent with the literature in which females have been reported to outperform males on tests of certain verbal abilities (e.g., Born et al., 1987; Grant & Adams, 1996; Maccoby & Jacklin, 1974, (Table 5)). The results also

876 P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 do not substantiate the presence of bias in the test scores or in the construct validity of the TOMAL across gender. After controlling for any overall memory effects, females scored higher on two verbal subtests: Word Selective Reminding and Object Recall, and males scored higher on the Memory for Location and Abstract Visual Memory subtests, the key spatial memory tasks on the battery. These findings are consistent with studies of intelligence with regards to pattern,, females performing higher on certain verbal tasks and males performing higher on certain spatial tasks (Born et al., 1987; Grant & Adams, 1996; Maccoby & Jacklin, 1974). However, as also happens with ethnic differences on memory tasks, the magnitude is reduced (Mayfield & Reynolds, 1997). Male female differences on the TOMAL are thus consistent with known outcomes in many other studies of gender differences in aptitude (e.g., Kaufman, 1990; Kaufman et al., 1991; Maccoby & Jacklin, 1974), but are of no real clinical consequence. Despite differences by gender in neurological organization of cognitive functions (Kolb & Whishaw, 1996), there is a common substrate to memory across gender. Acknowledgments The authors thank Arthur MacNeill Horton, Associate Editor, for handling all editorial matters including reviewing and publication decisions regarding this manuscript independently. References Albus, M., Hubmann, W., Mohr, F., Scherer, J., Sobizack, N., Franz, U., Hecht, S., Borrmann, M., & Wahlheim, C. (1997). Are there gender differences in neuropsychological performance in patients with first-episode schizophrenia? Schizophrenia Research, 28, 39 50. Aliotti, N. C., & Rajabiun, D. A. (1991). Visual memory development in preschool children. Perceptual and Motor Skills, 73, 792 794. Boivin, M. J. (1991). The effect of culture on a visual-spatial memory task. Journal of General Psychology, 118(4), 327 334. Bolla-Wilson, K., & Bleecker, M. L. (1986). Influence of verbal intelligence, sex, age, and education on the Rey Auditory Verbal Learning Test. Developmental Neuropsychology, 2(3), 203 211. Born, M. P., Bleichrodt, N., & van der Flier, H. (1987). Cross-cultural comparison of sex-related differences on intelligence tests. Journal of Cross-Cultural Psychology, 18, 283 314. Cattell, R. B. (1978). The scientific use of factor analysis in the behavioral and life sciences. New York: Plenum Press. Delis, D. C., Kramer, J., Kaplan, E., & Ober, B. A. (1994). California Verbal Learning Test Children s version. New York: The Psychological Corporation. Forrester, G., & Geffen, G. (1991). Performance measures of 7- to 15-year-old children on the Auditory Verbal Learning Test. The Clinical Neuropsychologist, 5(4), 345 359. Freides, D., & Avery, M. E. (1991). Narrative and visual spatial recall: Assessment incorporating learning and delayed retention. The Clinical Neuropsychologist, 5, 338 344. Geffen, G., Moar, K. J., O Hanlon, A. P., & Clark, C. (1990). The Auditory Verbal Learning Test: Performance of 16 86 year olds of average intelligence. The Clinical Neuropsychologist, 4, 45 63. Goldstein, G., & Incagnoli, T. (1997). Contemporary approaches to neuropsychological assessment. New York: Plenum Press.

P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 877 Gorsuch, R. L. (1988). Exploratory factor analysis. In J. Nesselroade & R. Cattell (Eds.), Handbook of multivariate experimental psychology (2nd ed., pp. 231 258). New York: Plenum Press. Grant, I., & Adams, K. (1996). Neuropsychological assessment of neuropsychiatric disorders. New York: Oxford Press. Harman, H. H. (1976). Modern factor analysis (2nd ed.). Chicago: University of Chicago Press. Herlitz, A., Nilsson, L. G., & Backman, L. (1997). Gender differences in episodic memory. Memory & Cognition, 25(6), 801 811. Huang, J. (1993). An investigation of gender differences in cognitive abilities among Chinese high school students. Personality and Individual Differences, 15(6), 717 719. Jensen, A. R. (1980). Bias in mental testing. New York: The Free Press. Kaufman, A. S. (1990). Assessing adolescent and adult intelligence. Boston: Allyn and Bacon. Kaufman, A. S., Kaufman-Packer, J. L., McLean, J. E., & Reynolds, C. R. (1991). Is the pattern of intellectual growth and decline across the adult life span different for men and women? Journal of Clinical Psychology, 47, 801 812. Kolb, B., & Whishaw, I. (1996). Fundamentals of human neuropsychology (4th ed.). New York: Freeman. Maccoby, E. E., & Jacklin, C. N. (1974). The psychology of sex differences. Stanford, CA: Stanford University Press. Mayfield, J. W., & Reynolds, C. R. (1997). Black White differences in memory test performance among children and adolescents. Archives of Clinical Neuropsychology, 12(2), 111 122. McCall, R. B. (1975). Fundamental statistics for psychology (2nd ed.). Atlanta: Harcourt Brace Jovanovich. McCarty, S. M., Siegler, I. C., & Logue, P. E. (1982). Cross-sectional and longitudinal patterns of three Wechsler Memory Scale subtests. Journal of Gerontology, 37, 169 175. McGivern, R. F., Mutter, K. L., Anderson, J., Wideman, G., Bodnar, M., & Huston, P. (1998). Gender differences in incidental learning and visual recognition memory: Support for a sex difference in unconscious environmental awareness. Personality and Individual Differences, 25, 223 232. Reynolds, C. R. (2000). Methods for detecting and evaluating cultural bias in neuropsychological tests. In E. Fletcher-Janzen, T. Strickland, & C. R. Reynolds (Eds.), Handbook of cross-cultural neuropsychology (pp. 249 286). New York: Plenum Press. Reynolds, C. R., & Bigler, E. D. (1994). Test of Memory and Learning (TOMAL). Austin, TX: PRO-ED. Reynolds, C. R., & Bigler, E. D. (1995). Factor structure, factor indexes, and other useful statistics for interpretation of the Test of Memory and Learning (TOMAL). Archives of Clinical Neuropsychology, 11(1), 29 43. Robinson, N. M., Abbott, R. D., Berninger, V. W., & Busse, J. (1996). The structure of abilities in math-precocious young children: Gender similarities and differences. Journal of Educational Psychology, 88(2), 341 352. Ruff, R. M., Light, R. H., & Quayhagen, M. (1989). Selective Reminding Tests: A normative study of verbal learning in adults. Journal of Clinical and Experimental Neuropsychology, 11(4), 539 550. Sheslow, D., & Adams, W. (1990). Wide range assessment of memory and learning. Wilmington, DE: Jastak. Tabachnick, B. G., & Fidell, L. (1989). Using multivariate statistics (2nd ed.). New York: Harper Collins. Temple, C. M., & Cornish, K. M. (1993). Recognition memory for words and faces in schoolchildren: A female advantage for words. British Journal of Developmental Psychology, 11(4), 421 426. Ullman, D. G., McKee, D. T., Campbell, K. E., Larrabee, G. J., & Trahan, D. E. (1997). Preliminary children s norms for the Continuous Visual Memory Test. Child Neuropsychology, 3(3), 171 175. Wechsler, D. (1995). Children s Memory Scale. San Antonio, TX: The Psychological Corporation. Further reading Kaufman, A. S., McLean, J. E., & Reynolds, C. R. (1988). Sex, race, residence, region, and education differences on the 11 WAIS R subtests. Journal of Clinical Psychology, 44, 231 248. Reynolds, C. R. (1995). Test bias and the assessment of personality and intelligence. In D. Saklofske & M. Zeidner (Eds.), International handbook of personality and intelligence (pp. 545 573). New York: Plenum Press.

878 P.A. Lowe et al. / Archives of Clinical Neuropsychology 18 (2003) 865 878 Reynolds, C. R. (1997). Measurement and statistical problems in neuropsychological assessment of children. In C.R. Reynolds & E. Fletcher-Janzen (Eds.), Handbook of clinical child neuropsychology (2nd ed., pp. 180 203). New York: Plenum Press. Reynolds, C. R., & Bigler, E. D. (1997). Clinical neuropsychological assessment of child and adolescent memory with the Test of Memory and Learning. In C. R. Reynolds & E. Fletcher-Janzen (Eds.), Handbook of clinical child neuropsychology (2nd ed., pp. 296 319). New York: Plenum Press.