DEVELOPMENTAL REVIEW 7, 142- 144 (1987)

Confusing Developmental and Individual Differences: A Reply to Anderson

TAMARGLOBERSON School of Education, Tel-Aviv University

Anderson, criticizing certain methodologies in the psychological literature, has taken my article (T. Globerson, 1985, Developmental Review, 5, 261-273) as an example. In his criticism Anderson claims that the findings of this research are artifact of the erroneous methodology (ANOVA) I used. In my reply I explain the rationale for my methodology. I also report the results of the data analyses sug- gested by Anderson, and show that my results are still valid, and are not an artifact of my methodology. D 1987 Academic PESS, hc.

Anderson (1987, Developmental Review, 7, 131-141) summarized my 1985 article as follows:

“Globerson (1985) compared the relative effects of an individual differ- ence variable field dependency (FD/I) with a developmental variable (age) on a measure of mental (M) capacity (Pascual-Leone), in an analysis of variance (ANOVA) design. Age is a highly significant effect and the sum of squares associated with this factor accounts for 66% of the vari- ance in M-capacity. FD/I is not significant and accounts for less than 1% of the variance in M-capacity. This is interpreted as indicating that M-ca- pacity is a purely developmental phenomenon showing negligible indi- vidual differences (p. 131).”

Taking my study as an example of inappropriate data analysis (i.e., ANOVA), Anderson made the following claims about my research:

(1) The procedure of dichotomizing a continuous variable (FD/I) and treating it as a factorial variable in an analysis of variance underestimates the size of the relationship between field dependence and M-capacity. As a consequence, he claims, the above result is an inevitable artifact of the experimental design. Anderson suggests abandoning ANOVA on behalf of other methods, such as multiple regression, to assess the true relation- ship between three variables: age, field-dependence/independence, and mental capacity.

(2) The fact that field dependence is a less reliable measure than age reduces its contribution, relative to age, to the variance in M-capacity.

(3) Rigidly controlling the distribution of one of the independent vari- ables (choosing restricted age groups) while letting the other variables

Requests for reprints should be sent to Tamar Globerson, School of Education, Tel-Aviv University, Ramat Aviv 69978, Tel-Aviv, Israel.

142 0273-2297187 $3.00 Copyright 0 1987 by Academic Press, Inc. All rights of reproduction in any form reserved.

CONFUSING DIFFERENCES 143

vary freely means that the ANOVA reports that the controlled variable is the most “important.”

In my response to Anderson’s claims I explain why I chose to carry out the two-way ANOVA; then I show that despite Anderson’s reasonable critique, my conclusions are still valid and are not an artifact of the data analysis. I demonstrate this by doing the kind of data analyses that An- derson has suggested.

The design of my study was such that only age was controlled by se- lecting certain ranges of age. The selection of certain age groups, based on a developmental theory, is a common practice in developmental re- search. The reason 1 chose to categorize both field dependence into two groups of field independent and field dependent subjects, and age into three normative age groups, was that I wanted to examine the contribu- tion of prototypical ages and prototypical cognitive styles to M-capacity, rather than the contribution of individual differences. Hence, analyses were carried out on groups of field independent and field dependent chil- dren of the same age.

What is “prototypical” of the whole distribution of FD/I? Surely not extreme scores. Rather, given a normal distribution, one would reason- ably group all subjects above the overall mean, separate them from all subjects below the mean, and create a dummy, two-value variable. The values on this dummy variable come now to represent, respectively, the higher and lower portions of the distribution.

Such a procedure obviously eliminates the within-group variance as all subjects within each one of the FI and FD groups receive the same score of FD/I. However, the same was done with the age variable where all subjects were similarly assigned either one of three values on the now dummy variable of age.

Still, Anderson’s critique cannot be totally dismissed. The possibility of underestimating the contribution of FD/I to M-capacity due to low reliability or to the restriction of variance has not been ruled out entirely. As Anderson shows, there might be cases where such factors may indeed yield artifactual results. To rule out such a possibility, I reanalyzed the original data, following Anderson’s suggestions, using multivariate anal- yses rather than two-way ANOVA. The analyses were carried out on the entire sample (n = 335), using the following measures: two measures of M-capacity-Compound Stimulus Visual Information (CSVI) and Serial Stimulus Visual Information (SSVI) (for detailed descriptions of the various tests see Globerson (1985)); a measure of FD/I-WISC-R Block Design (raw scores); and age (measured in months).

Zero-order Pearson correlations between the three variables yielded the following results (coefficients in parentheses are those corrected for

144 TAMAR GLOBERSON

attenuation, based on reliability scores of .95 for the M-capacity mea- sures, and .84 for the FD/I measure; see Case & Globerson, 1974). The correlation between M-capacity (CSVI) and age was .80 (.82), and be- tween FD/I and age, .35 (.38). The correlation between M-capacity (SSVI) and age was .74 (.76). The correlations between FD/I and the M-capacity measures (CSVI, SSVI) were .43 (.48) and .36 (.40), respec- tively.

Since tests of FD/I include a developmental component, they correlate with M-capacity (for a detailed discussion see Case & Globerson (1974)). A multiple regression analysis was carried out to demonstrate the relative “importance” of age and FD/I to the variance of M-capacity, taking into consideration the common variance between age and FD/I. The results of the R2 change for M-capacity (CSVI) when age was entered before FD/I were 64 (age) and .02 (FD/I). The reverse order, where FD/I was entered before age yielded an R2 change of .18 (FD/I) and .48 (age). The results pertaining to the other measure of M-capacity (SSVI) were as follows: When age was entered first, R2 change was .55 (age) and .Ol (FD/I); for the reverse order, .13 (FD/I) and .43 (age).

To summarize, FD/I accounts for no more than 2% of the M-capacity variance (regardless of measure) when entered after age. Age, entered first, accounts for up to 64% of the M-capacity variance. When FD/I is entered first, it accounts for up to 18% of the M-capacity variance, while age (entered second) continues to account for close to 50% of the M-ca- pacity variance. These findings are very similar to the ones obtained by the two-way ANOVA as reported in my 1985 article.

It thus appears that age accounts for far more variance on the M-ca- pacity measures than the individual variability of field dependencejinde- pendence, suggesting that M-capacity is indeed a developmental rather than stylistic capability.

REFERENCES Case, R., & Globerson, T. (1974). Field independence and central computing space. Child

Development, 45, 772-778. Globerson, T. (1985). Field dependence/independence and mental capacity: A develop-

mental approach. Developmental Review, 5, 261-273.

RECEIVED: November 22, 1986.