The Journal of Social Psychology, 1984, 123, 3-8.
THE CATTELLIAN METHOD OF PREDICTING CHILD PERSONALITY*
University of New South Wales, Australia; and Guru Nanak Dev University, India
JOHN J. RAY AND SATVIR SINGH
A close reading of papers by Barton, Dielman, and Cattell suggests that supposed relationships reported between parental attitudes and child personality may have been less significant than at first appears. They may have been nothing more than the occasional large correlation that is to be expected in any large matrix of correlations by chance alone. A partial replication of the Barton, Dielman, and Cattell work was carried out by using two samples of 328 girls and their mothers and 327 boys and their mothers reached through schools of rural Punjab. Results very similar to those reported earlier were found -- confirming suspicions that non-significant results had previously been reported as significant.
As part of his monumental work with factor-analytically-produced personality inventories, Cattell has in recent years collaborated with several other workers to predict factors of child personality from the personalities of their parents (3, 4, 5). Children's motives were found to be predictable on the basis of parental attitudes towards the family, and their personality on the basis of reported child-rearing practices.
A closer inspection of the supposedly significant findings, however, discloses a number of difficulties. The first concerns the invariance of the factors used. The central claim of factor analysis, of course, is that it enables one to "discover" underlying variables behind the correlations between ostensibly different variables. While this is an admirable aspiration, the reality of its achievement has been called into question (1, 2, 8). For the variables discovered by factor analysis to be real in any sense, an essential requirement would seem to be that they be replicable, invariant from occasion to occasion and from sample to sample (6). Whether this is generally so is too large a question to be discussed here. What is pertinent, however, are the factors used by Cattell and his associates in their study of child personality. Dielman, Barton, and Cattell (5, p. 295) say: "Additional description of the Family Attitude Measure (FAM) and cross-validational evidence are available in an article by Barton et al. (1)." This gives the impression that at least the FAM inventory has been shown to have invariant factor structure. Factor invariance or "reality" can only be established by cross-validation and this has been provided.
If one examines the two articles on the FAM referred to by Cattell and his associates, however, one finds that the factor structure of the inventory observed on the two occasions was completely different. "The results of the analyses on two of the subtests were found to be interpretable and are described in detail. Few of the factors identified were as interpretable in the sense of the Delhees et al. factors. . . ." (3, p. 67). In the supposed cross-validation, half of the test was not interpretable at all even by Cattell's neologistic standards!
Perhaps the procedures of Cattell and his associates suggest less confidence in the invariance of his measures than even their claims would suggest: In the 1974 application of the FAM (5), they factor analyze it yet a third time and use as their predictors only those factors found on that occasion. How similar this third lot of factors is to either of the previous two we are not told. The reader must wonder whether child personality had been predicted by anything other than a mathematical artifact. Given the large number of correlation coefficients processed in the course of the Cattellian analytic method, one must also wonder whether the small number of weak correlations reported were in fact anything other than chance occurrences.
Dielman, Barton, and Cattell are singularly terse in their discussion of the statistical significance of their results. The only information they provide is an F ratio tabulated beside each R (multiple correlation coefficient). To the casual observer, the tabulated ratios would seem to confirm the significance of the multiple regressions. To someone with only a little experience with multiple regression, however, the F ratios seem not to match the Rs given. How can a multiple R of .17 give rise to an F of 6.50 with N of 250? Only, it would seem, when the number of predictors making up the multiple correlation is small.
At first sight, then, it appears that Cattell and his associates were aware of the perils of searching for high correlations out of a large matrix ("data snooping") and properly did a multiple correlation first to check that there was an overall significant prediction of the criterion. Only when this had been established did they proceed to look at the individual correlations. In fact, however, this procedure was not adopted. It would appear that the tabulated F refers not to the overall prediction of the criterion by all variables but only to the last "step" of the stepwise regression; it refers to the prediction provided by the single variable showing the highest correlation with the criterion. In other words, it may be inferred that (a) the Cattell group's multiple regressions told them that their factors of parental attitudes did not significantly predict any child personality traits but (b) they still went ahead and picked out those correlations that would be significant if they were the only ones examined.
This adoption of a contrast-wise error-rate approach may of course be quite defensible, but in the circumstances it surely should have been at least remarked. Dielman, Barton, and Cattell (5) not only failed to defend their practice in this respect but presented their data in such a way that the difficulty was not apparent. They do enter a caveat: "These results must necessarily stand as tentative pending cross-validation; and they are regarded by the authors as hypothesis generating rather than hypothesis testing" (p. 295). This statement is surely faint recompense for presenting non-significant results as worthy of consideration. Many researchers make modest disclaimers about the generalizability of their findings but the reader is surely not expected to find that such disclaimers stand in lieu of reporting non-significance.
If the above reconstruction is correct, we are left wondering about the multiple regressions in a later article by the same authors (4). No significance data at all are reported for them. The individual correlations between single variables and criteria that are reported, however, seem generally very similar in magnitude to those reported in 1974. It is therefore curious to note how much more positive is the summary of findings on the later occasion: "It was found that the majority of the personality variables could be significantly predicted from the child-rearing practices measures" (p. 75).
Why are effects of extremely similar magnitudes merely hypotheses in 1974 but findings in 1977? It is true that the 1977 study had an initial claim to being a replication of an earlier study by Dielman, Cattell and Rhoades (which in 1977 was still being cited as "In press, 1971") (1) but nowhere in the article do we have any comment on whether the 1977 study did in fact replicate any finding of the "In press" study. Since the purpose of the 1977 study was described as "to compare such relationships (i.e. the relationships of the 1977 study), should they exist, to those found in younger children in the Dielman, Cattell, and Rhoades study" (p. 76), may we infer that non-report indicates that no similarities were found?
Clearly, the reporting of results by Cattell and his associates leaves many questions unanswered and the only way to remove speculation is to attempt an independent replication of their work. The present paper is an attempt to do this.
Because of indications from Cattell's own work of the instability in the factor structure of the FAM, it seemed pointless to make further use of it. Instead, the apparently better-established Schaefer and Bell Parental Attitude Research Instrument (PARI) was used (10). For the measure of child personality, however, the Cattell SMAT (School Motivational Analysis Test) was used. As some supplement to the SMAT, however, projectively expressed need for achievement and fear of success were also measured, as were intelligence (9) and demographic variables.
Because of Cattell's claims for cross-cultural invariance in his factors, it was thought not inappropriate that a replication of his work be done in India. The Ss were schoolchildren reached in rural Punjab villages. Data were analyzed separately for boys and girls and each child took home a questionnaire for his or her mother to answer. The research was limited to mothers as it was thought that they had the greatest opportunity to influence children.
Inasmuch as there appears to be some difficulties with Cattell's procedure of entering factor scores in multiple regression equations (7), all scores on all variables were obtained simply by scoring the various inventories as per their respective manuals.
Data were received from 328 girls and their mothers and 327 boys and their mothers. With 23 predictor variables (mothers' attitudes) in each case, an R with a corresponding F ratio of 1.55 or more would be significant using a contrast-wise error-rate approach. As an experiment-wise error-rate approach would have been more rigorous, this does, then, constitute a rather generous test of Cattell's hypothesized relationship between parental attitudes and child personality.
Out of the 30 child characteristics tested, only two were significantly predicted by mothers' attitudes among the girls and only one among the boys. Among the girls these were factors Q2 and Pr (Rs of .326 and .324; Fs of 1.58 and 1.56) and among the boys it was socioeconomic status (R of .37; F of 2.15). As even the significant relationships were extremely weak (borderline significance even with N large), any attempt at interpretation seemed pointless. For all intents and purposes, child characteristics were not predicted by parental attitudes.
It would appear that the impression created by Cattell and his associates to the effect that child personality can be predicted by parental attitudes is misleading. Independent research in the Cattellian mold here did produce findings very similar to those reported by Cattell, but fuller reporting of those findings reveals them as non-significant. The appearance of significance given by the writings of Cattell and his associates is shown-to be obtainable only by "data snooping," by the exploitation of those chance oscillations in levels of correlation which must occur when large matrices are handled.
The present work was not at all intended as a challenge to the notion that child personality is influenced by parental attitudes and personality. It was intended merely as an examination of the usefulness of the Cattellian methods and instruments in depicting that influence. The fact that both the present work and the Cattellian work can now be seen to lead to the conclusion that parental attitudes do not influence child personality may however be worthy of some consideration as information about the world. It seems that we do now have at least some grounds for concluding that the well-known attitude-behavior discrepancy may extend to this field also. It may be only parental behavior that influences child personality, not parental attitudes.
1. ARMSTRONG, J. S: Derivation of theory by means of factor analysis or Tom Swift and his magic factor analysis machine. Amer. Stat., 1967, 21, 17-21.
2. ARMSTRONG, J. S., & SOELBERG, P. On the interpretation of factor analysis. Psychol. Bull., 1968, 70, 361-364.
3. BARTON, K., DIELMAN, T. E., & CATTELL, R. B. An item factor analysis of intrafamilial attitudes of parents. J. Soc. Psychol., 1973, 90,67-72.
4. BARTON, K., DIELMAN, T. E., & CATTELL, R. B. Child-rearing practices related to child personality. J. Soc. Psychol., 1977, 101, 75-85.
5. DIELMAN, T. E., BARTON, K., & CATTELL, R. B. Relationships among family attitude dimensions and child motivation. J. Genet. Psychol., 1974, 124, 295-302.
6. DIELMAN, T. E., CATTELL, R. B., & WAGNER, A. Evidence on the simple structure and factor invariance achieved by five rotational methods on four types of data. Multivariate Behav. Res., 1972, 7, 223-242.
7. KUKUK, C. R., & BATY, C. F. The misuse of multiple regression with composite scales obtained from factor scores. Educ. & Psychol. Meas., 1979, 39, 277-289.
8. LYKKEN, D. T. Multiple factor analysis and personality research. J. Exper. Res. in Personal., 1971, 5, 161-170.
9. RAVEN, J. C. Guide to the Standard Progressive Matrices. London: Lewis, 1960.
10. SCHAEFER, E. S., & BELL, R. Q. Development of a parental attitude research instrument. Child Devel., 1958, 29, 339-361.
The University of New South Wales P.O. Box 1, Kensington, N.S.W., Australia 2033
* In the article, "Child-rearing practices related to child personality," by K. Barton, T. E. Dielman, and R. B. Cattell, J. Soc. Psychol., 1977, 101, 75-85, there were, in fact, two references cited as "in press." Their full citations follow: Dielman, T. E., Barton, K., & Cattell, R. B. Cross-validation evidence on the structure of parental reports of child-rearing practices. J. Soc. Psychol., 1973, 90, 243-250. Dielman, T.E., Cattell, R. B., & Rhoades, P. A. Child-rearing antecedents of early school personality factors. J. Mar. & Fam., 1972, 34, 431-436.
Go to Index page for this site
Go to John Ray's "Tongue Tied" blog (Backup here)
Go to John Ray's "Dissecting Leftism" blog (Backup here)
Go to John Ray's "Australian Politics" blog (Backup here)
Go to John Ray's "Gun Watch" blog (Backup here)
Go to John Ray's "Education Watch" blog (Backup here)
Go to John Ray's "Socialized Medicine" blog (Backup here)
Go to John Ray's "Political Correctness Watch" blog (Backup here)
Go to John Ray's "Greenie Watch" blog (Backup here)
Go to John Ray's "Leftists as Elitists" blog (Not now regularly updated)
Go to John Ray's "Marx & Engels in their own words" blog (Not now regularly updated)
Go to John Ray's "A scripture blog" (Not now regularly updated)
Go to John Ray's recipe blog (Not now regularly updated -- Backup here)
Go to John Ray's Main academic menu
Go to Menu of recent writings
Go to John Ray's basic home page
Go to John Ray's pictorial Home Page (Backup here)
Go to Selected pictures from John Ray's blogs (Backup here)
Go to Another picture page (Best with broadband)