Results

Psychoeducational tests

Vocabulary. Table 1 presents the results of the hierarchical regression analyses) for the vocabulary test. In general, the results are in the predicted direction with older adolescents, subjects from higher SES families and those who entered school earlier and stayed in longer performing better on the measure of vocabulary. After controlling for all of these variables, the treatment contributed an additional 5% of the variance (Ff = 22.35, P<0.001) in performance with Atole subjects performing significantly better (b = 3.93) than Fresco subjects.

TABLE 1 Results of hierarchical regression analyses for vocabulary ¹

Step	Variables	R2	F(eq.)	F-to-enter	b	Direction of effects favors
1	Sex	0.03	3.16*	0.01	-0.112
	Age			8.01**	3.290	Older subjects
	Attendance		1.48	-0.571
2	Socioeconomic status	0.11	10.66***	32.26***	1.168	Higher SES
3	Age at entry	0.21	14.25***	18.98***	-1.064	Younger subjects
	Maximum grade			19.17***	2.327	Higher grade
4	Treatment	0.26	16.20***	22.35***	3.930	Atole
5	Treatment by grade	0.30	15.:13***	6.13**
	socioeconomic status			11.28***

¹ Adapted from Pollitt et al. l 1993).
* P<0.05.
** P<0.01.
*** P<0.001.

When entered into the model, both interactive terms were significant and accounted for an additional 4% of the variance. In the case of the SES-by-treatment interaction, the slope for Atole subjects was nonsignificant whereas that for Fresco subjects was positive and significant (b = 1.373, P<0.001). Although there was no relation between SES and performance in Atole villages, performance improved in Fresco villages with increasing SES level (Fig. 1). At lower ends of the SES distribution, subjects who received Atole supplements performed significantly better than those who received Fresco supplements; whereas at higher SES levels there were no differences between them.

FIGURE 1 SES-by-treatment interactions for vocabulary.

The grade-by-treatment interaction showed a different pattern. The slope was positive and significant for Atole (b = 3.861, P<0.001) but not for Fresco subjects. Differences between treatment groups increased with grade attained, such that children from Atole villages scored significantly higher than Fresco children at the upper ends of the grade distribution (Fig. 2). For those at the lower end of grade attainment, there were no differences between Atole and Fresco subjects.

FIGURE 2 Maximum grade-by-treatment interaction for vocabulary.

Results of the hierarchical analyses for other outcome variables were similar (Table 2). After controlling for potentially confounding variables, there were significant effects of Atole on performance on tests of numeracy, knowledge, vocabulary and reading achievement. The percent of variance accounted for by inclusion of the treatment variable was generally small, yet statistically significant, ranging between 1 and 5%. Examination of significant interactive terms permits the identification of subgroups in whom effects were greatest. In almost all instances, effects of Atole were evident in children from families at the lowest levels of SES. In several cases (e.g., reading, vocabulary and reading achievement), effects were observed in children with the highest levels of education.

TABLE 2 Summary of results of hierarchical regression analyses for psychoeducational tests¹

	R2	F values
Dependent variable	Full model	Treatment†	SES by treatment	Treatment by grade
Literacy	56	0.44	0.14	0.54
Numberacy	48	7.75**	10.06***	0.01
Knowledge	27	8.57**	6.74**	0.79
Raven	15	0.22	8.39**	2.18
Reading	30	0.03	2.49	5.36*
Vocabulary	30	22.35***	11.28***	6.13**
Reading	30	20.05***	14.91***	13.14***
Achievement

¹ Adapted from Pollitt et al. (1993).

† After controlling for age at testing, gender, attendance, SES, age at school entry and maximum grade attained. SES = socioeconomic status.

*P<0.05.
**P<0.01.
*** P<0.001.

Information processing

Results of the regression analyses on reaction time of the memory task are presented in Table 3. In contrast to the results on the psychoeducational tests, none of the predictor variables were associated with performance, with the exception of grade attainment. After controlling for all potential confounders, treatment was associated significantly with performance, with Atole subjects having significantly faster reaction times (b = - 0.321, P<0.01) than Fresco subjects. Neither of the interactive terms was significant.

TABLE 3 Results of hierarchical regression analyses for memory reaction time¹

Step	Variables	R2	F(eq.)	F-to-enter	b	Direction of effect favors
1	Sex	0.002	0.24	0.03	-0.018
	Age			0.00	-0.051
	Attendance a	0.004		0.69	0.046
	Socioeconomic status		0.43	1.01	-0.024
3	Age at entry	0.02	1.51	3.03	0.037
	Maximum grade			4.25*	-0.117	Higher grade
4	Treatment	0.04	2.64**	9.25**	- 0.321	Atole

¹ Adapted from Pollitt et al. (1993).
* P<0.05.
** P<0.01.
*** P<0.001.

Results of hierarchical analyses for other information processing outcome variables were similar(Table 4). In general, the percent of variance accounted for by the models was small (between 3 and 10%), with males, higher SES, earlier school entry and higher grade attainment associated with enhanced performance. In seven analyses, there were three significant main effects of treatment. Atole subjects responded faster and more efficiently than Fresco subjects on the memory task and reached criterion faster on the paired associates task. None of the interactive terms was significant.

TABLE 4 Summary of results of hierarchical regression analyses for information processing tests 1

	R2	F value
Dependent variable	Full model	Treatment†
Simple RT	03	0.03
Choice RT	02	1.01
Trials to criterion	06	3.65*
Memory RT	04	9.25**
Memory efficiency	10	8.40**
Memory impulsivity	04	2.06
Memory percent correct	07	1.02

¹ Adapted from Pollitt et al. (1993).

† After controlling for age at testing, gender, attendance, socioeconomic status, age at school entry and maximum grade attained.

*P<0.05.
** P<0.01.