Methods of the current analyses

Infancy

Data from the Composite Infant Scale were analyzed using the mental and motor scores obtained at each of the three ages: 6, 15, and 24 months. In contrast to previously published studies, the new analyses were restricted to an Atole/Fresco comparison of a cohort of subjects who were exposed to the treatment both pre- and postnatally. A general linear model was used to analyze the data, with treatment entered as a class variable. All analyses controlled for socioeconomic status (SES), sex, and attendance and included a treatment x SES interactive term (a complete description of the socioeconomic and attendance variables is given in Chaps. V and VI, respectively).

Preschool

To reduce the large numbers of tests included in the preschool battery, the battery was factor analyzed (with varimax rotation) before undertaking the current analyses. Only tests considered to have minimally acceptable test-retest stability coefficients (.6 or above for most age/sex groups) were included in this analysis. Several other tests were also eliminated from consideration because the scoring system was ambiguous (i.e., Reversal Discrimination Learning Test II), because the data file was unavailable (i.e., Animal House), or because the tests did not load with any other measures in exploratory factor analyses (Conservation of Matter, Incidental and Intentional Learning). Finally, two tests - Verbal Inferences and Knox Cubes - were not included at age 3 owing to very high rates of missing data. All tests that were used for the factor analysis are listed in Appendix B.

The results of this analysis are presented in Table 2. At all ages, the first factor accounted for about 35% of the variance and the second for between 22% and 9%, declining with the age of the children. A third factor, with eigenvalues below 1 at all ages, was dropped from further analyses.

The first factor represents a general perceptual-organizational (Embedded Figures Test, Incomplete Figures, Odd Figures, Block Design, and Memory for Designs) and a verbal (Verbal Analogies, Memory for Objects, and Vocabulary Recognition) factor. The second loads most heavily on digit and sentence memory at all ages. The two other tests labeled as memory measures (Memory for Objects and Memory for Designs) loaded primarily on the first factor, suggesting that these assess cognitive skills other than the verbal repetition required in the Memory for Sentences and Digit Memory tests. The Draw-a-Line Slowly test loaded on both factors.

TABLE 2: FACTOR LOADING FROM VARIMAX ROTATION FOR PRESCHOOL BATTERY TESTS

TEST		AGE IN YEARS
		3	4	5	6	7
General factor (Factor 1):
	Embedded Figures	.78	.40	.14	.58	.59
	Digit Memory	.06	.04	.04	.05	.17
	Sentence Memory	.28	.37	.26	.23	.22
	Recognition Vocabulary	.81	.72	.58	.54	.69
	Draw-a-Line	-.22	-.10	-.53	-.32	-.46
	Puzzle	-.09	-.03	.00	.09	.16
	Memory for Objects	.63	.71	.42.	19	.53
	Verbal Inferences		.61	.37	.52	.57
	Knox Cubes: Slow			.62	.61	.35
	Incomplete Figures			.62	.70	.74
	Odd Figure			.47	.69	.74
	Block Design			.60	.69	.58
	Memory for Designs			.57	.72	.69
	Eigenvalues	1.81	1.72	2.71	3.18	3.78
Memory factor (Factor 2):
	Embedded Figures	.02	-.03	.09	.10	.13
	Digit Memory	.87	.87	.87	.81	.83
	Sentence Memory	.81	.74	.82	.79	.85
	Recognition Vocabulary	.17	.17	.24	.44	.17
	Draw-a-Line	-.08	-.46	-.27	-.45	-.28
	Puzzle	.13	.00	.05	.09	.01
	Memory for Objects	.24	.05	.24	.45	.05
	Verbal Inferences		.28	.41	.45	.11
	Knox Cubes: Slow			.00	.19	-.28
	Incomplete Figures			.17	.37	.15
	Odd Figure			.11	.16	.11
	Block Design			.03	.16	.15
	Memory for Designs			-.03	.23	.07
	Eigenvalues	1.52	1.64	1.85	2.73	1.73

NOTE - Sample size is over 600 per age group.

Cross-age correlations within each factor suggest moderate stability ranging between .29 and .67 for Factor 1 and between .26 and .66 for Factor 2. In general, the magnitude of these coefficients increased with age.

As with the Composite Infant Scale, analyses were restricted to an Atole-Fresco comparison of a cohort of subjects who had been exposed to the treatment both pre- and postnatally. At least 100 subjects were available at ages 3, 4, 5, and 6 years, but their number was too small (N = 13) to permit analysis of the 7-year data. In addition to age-specific scores, two composite factor scores were calculated by summing across ages for individuals with data at multiple time points. All these data were analyzed using general linear model regression and entering treatment as a class variable. Analyses adjusted for sex, attendance, and SES and included a treatment x SES interactive term.

TABLE 3: RESULTS OF REGRESSION (General Linear Model) OF COMPOSITE INFANT SCORES ON TREATMENT FOR MAXIMUM EXPOSURE COHORT

Variable Name		N	R²	Treatment a
6 months:
	Mental	279	.02	-.007
	Motor	279	.03	-.035
15 months:
	Mental	274	.03	1.12 †
	Motor	274	.03	1.59 †
24 months:
	Mental	242	.05*	.157
	Motor	242	.12***	4.58***

a The value reflects the least square mean difference (Atole - Fresco), adjusted for SES (house quality, father's occupation, mother's schooling), sex, and attendance.

† p <.10.

*p <.05.

***p <.001.

TABLE 4: RESULTS OF REGRESSION (General Linear Model) OF TREATMENT ON PRESCHOOL BATTERY SCORES

Variable Name		N	R²	Treatment a
3 years:
	Factor 1	260	.05*	.090
	Factor 2	260	.04	.176
4 years:
	Factor 1	248	.06*	.269*
	Factor 2	248	.01	.064
5 years:
	Factor 1	178	.07 †	.489**
	Factor 2	178	.04	-.221
6 years:
	Factor 1	101	.09	.152
	Factor 2	101	.09	-.289

a The value reflects the least square mean difference (stole - Fresco), adjusted for SES (house quality, father's occupation, mother's schooling), sex, and attendance.

† p <.10.

*p <.05.

**p <.01.

FIG. 1. - SES x treatment interaction: General factor score at age 4 years