Cover Image
close this bookEarly Supplementary Feeding and Cognition (Society for Research in Child Development, 1993, 123 pages)
close this folderVI. Results from the cross-sectional follow-up
View the document(introductory text...)
View the documentEffects of experimental treatment on test scores
View the documentEffects of treatment on nonschooled subjects

Effects of experimental treatment on test scores

Statistical analyses were performed on both the entire sample and the "cohort of maximum exposure" (see Fig. 4 above) and then contrasted with results obtained for the cohort of late exposure. In keeping with previous work that has documented sex differences in both the biological and the behavioral response to supplementation (Engle & Levin, 1984), preliminary analyses tested for presence of any treatment x sex interactions. Since none emerged, all final analyses included sex as a main effect.

The data were analyzed in two steps. First, correlations were obtained across all subjects between the psychoeducational test scores and both SES indicators and school variables. Second, regression analyses were calculated on each cohort separately (e.g., entire sample, cohort of maximum exposure, cohort of late exposure) to assess the specific effects of the early supplementary feeding independent of the SES and school variables' effects.

Correlational Analyses

SES Indicators and Test Scores

Table 16 presents the correlations obtained between the SES indicators and the psychoeducational and information-processing test scores. All correlations with the former were significant (p <.001), ranging in size from r =.15 to r =.39; higher SES was associated with higher test performance. Similarly, correlations with the information-processing tests were in the predicted direction; however, they were generally smaller than those involving the psychoeducational tests, and only 13 of 27 coefficients were statistically significant.

TABLE 16: CORRELATIONS BETWEEN ADOLESCENT OUTCOMES AND SOCIOECONOMIC INDICATORS


SOCIOECONOMIC INDICATORS


House Quality

Mother's Education

Father's Occupation

Psychoeducational:


Literacy

.31***

.27***

.16***


Reading

.15***

.19***

.15***


Numeracy

.28***

.29***

.20***


Knowledge

.28***

.30***

.25***


Raven's Matrices

.25***

.21***

.22***


Interamerican:






Reading

.39***

.27***

.32***



Vocabulary

.35***

.25***

.27***

Information processing:


Reaction time:






Simple

-.05

-.06*

-.02



Choice

-.09**

-.04

-.01



Memory

-.07**

-.03

-.04


Trials to criterion

-.15***

-.11***

-.11***


Impulsivity

-.05

-.06*

-.03


Efficiency

-.14***

-.10***

-.07***


% negative correct

.13***

.09**

.06

NOTE - Sample sizes range from 1,162 to 1,281 for all tests, except reading tests (range = 864 - 952) and paired associates (range = 948-1,046).

*p <.05.

**p <.01.

***p <.001.

Schooling and Test Scores

Age of entry into school correlated at a statistically significant level with all the psychoeducational test scores (see Table 17); the earlier the enrollment, the better the performance. As would be expected, in the case of the information-processing tasks, fewer coefficients (four of seven) were statistically significant, and their magnitude was much smaller (range from - 0.10 to 0.16 for significant coefficients).

All correlations between the other school performance variables and psychoeducational test scores were statistically significant; as in the case of the SES indicators, the correlations with the information-processing variables were, generally, smaller, and only 65% of them (13 of 21) reached statistical significance.

TABLE 17: CORRELATIONS BETWEEN SCHOOLING VARIABLES AND ADOLESCENT OUTCOMES


SCHOOLING VARIABLES


Age at Entry

Pass

Fail

Grade Highest

Psychoeducational:


Literacy

-.36***

.69***

-.17***

.58***


Reading

-.24***

.40***

-.19***

.22***


Numeracy

-.33***

.61 ***

-.23***

.50***


Knowledge

-.20***

.39***

-.19***

.29***


Raven's Matrices

-.25***

.29***

-.20***

.21***


Interamerican:







Reading

-.25***

.33***

-.20***

.18***



Vocabulary

-.27***

.36***

-.18***

.22***

Information processing:


Reaction time:







Simple

.02

-.09**

.02

-.03



Choice

.04

-.05

-.02

-.03



Memory

.11***

-.14***

.01

-.15***


Trials to criterion

.11***

-.13***

.05

-.10**


Impulsivity

.01

-.08**

.06*

-.02


Efficiency

.16***

-.27***

.08**

-.22***


% negative correct

-.10***

.23***

-.08**

.17***

*p <.05.

**p <.01.

***p <.001.

In sum, bivariate associations between test performance and SES and between test performance and schooling were in the expected directions. In keeping with theoretical assumptions, the psychoeducational test battery was more sensitive to environmental factors (e.g., SES and schooling variables) than was the information-processing test battery.

Hierarchical Regression Analyses

Analytic Strategy

The main question raised in this Monograph is whether exposure to Atole had a differential effect on cognition in adolescence than exposure to Fresco. At issue is not only whether early malnutrition had developmental consequences but also whether early nutrient supplementation had preventative effects. Data analyses could be selectively focused on the presence or absence of main effects to address this issue; however, from a developmental perspective, such an approach is far too restrictive. It fails to consider the possibility of differential effects of treatment that may be related to particular characteristics of the population, and it disregards these data's potential for making a theoretical contribution to our understanding of cognitive development.

Adopting a more sophisticated approach, the data were analyzed using a hierarchical regression model. All independent variables were standardized. As in the correlational analysis, the 1987 data were selected for use as socioeconomic indicators, and a composite SES score was created by summing mother's education, father's occupation, and house quality index. Individual characteristics (sex, age, and attendance with consumption partialed out) were entered first, followed by the SES composite score, then the two school variables (age at entry and maximum grade), and, finally, treatment. In this way, we were able to calculate the amount of variance accounted for by the different types of predictors. In a subsequent step, two interaction terms were entered - treatment x maximum grade and treatment x SES. The results indicate the percentage of variance accounted for (R²) and the F values and regression coefficients for each variable in the step in which it was entered, controlling for all other variables entered prior to this step.

The order in which we entered variables was based on the following basic assumptions: (i) Between-village differences in SES variables overlap with experimental treatment effects and could therefore account for some effects that might otherwise be attributed to treatment. (ii) Level-of-schooling variables have direct effects on the outcome variables and intervene in the relation between experimental treatment and outcome. (iii) Experimental treatment has direct effects on the outcome variables independent of all other effects that may be evident.

A three-way interaction term (treatment x SES x maximum grade) was also entered into the models; the results of these analyses are mentioned only where a significant three-way interaction was obtained. Similarly, since the focus of the analyses was on the effects of the treatment, presentation of results is limited to those SES x maximum grade interactions that were statistically significant. Since none of the treatment x sex or treatment x age interactive terms were shown to be statistically significant by preliminary analyses, they were dropped from the final model.

Entire Sample

Psychoeducational test performance. - Table 18 presents a summary of the results of the regression analyses. In general, males outperformed females on the psychoeducational tests (e.g., reading, numeracy, and RPM). There was also a significant main effect of age on all but the literacy test - older adolescents performed consistently better than younger ones. Likewise, the residual of attendance (after partialing out consumption) was positively associated with test performance, reaching statistical significance on three of the seven outcome variables (reading, numeracy, and RPM). As would be expected, there were main effects of SES, age of entry in school, and maximum grade attained: the higher the SES level, the younger the age at entry, and the higher the grade attained in school, the higher the test scores.

TABLE 18: RESULTS OF HIERARCHICAL REGRESSION ANALYSES FOR THE ENTIRE SAMPLE ON THE PSYCHOEDUCATIONAL TESTS

Step and Variables


Model F

F to enter

Parameter Estimate

Direction of Effect Favors:

Numeracy:


1. Sex

.07

21.29***

13.14***

- 1.901

Males



Age



41.53***

1.227

Older



Attendance a



9.19**

.894

Higher


2. SES

.15

37.07***

78.69***

1.079

Higher


3. Age at entry

.46

120.17***

146.73***

-.735

Younger



Max grade



342.30***

4.362

Higher


4. Treatment (Rx)

.47

108.30***

20.63***

1.924

Atole


5. Rx X grade

.49

82.74***

2.73





Rx X SES



8.12**





Grade X SES



27.38***



Knowledge:


1. Sex

.21

74.93***

3.47

-.419

...



Age



218.31***

1.479

Older



Attendance a



3.00

.212

...


2. SES

.26

77.23***

66.98***

.416

Higher


3. Age at entry

.37

85.92***

69.60***

-.413

Younger



Max grade



83.06***

1.036

Higher


4. Treatment (Rx)

.39

77.21***

16.01***

.819

Atole


5. Rx X grade

.39

60.83***

.38





Rx X SES



4.69*



Vocabulary:


1. Sex

.06

14.36***

1.45

-.705

...



Age



41.37***

1.818

Older



Attendance a



.25

.164

...


2. SES

.14

27.56***

63.19***

1.087

Higher


3. Age at entry

.26

39.65***

50.42***

- 1.153

Younger



Max grade



59.55***

2.742

Higher


4. Treatment (Rx)

.31

42.78***

45.72***

3.717

Atole


5. Rx X grade

.32

35.55***

2.45





Rx X SES



12.34***



Reading achievement:


1. Sex

.05

11.01***

.01

-.038

...



Age



32.59***

1.017

Older



Attendance a



.42

.139

...


2. SES

.13

25.24***

64.80***

.709

Higher


3. Age at entry

.22

31.76***

35.64***

-.634

Younger



Max grade



42.57***

1.525

Higher


4. Treatment (Rx)

.25

32.20***

27.33***

1.915

Atole


5. Rx X grade

.29


9.50**





Rx X SES



29.99***

24.30***


Raven's Progressive Matrices:


1. Sex

.07

23.24***

39.80***

- 1.801

Males



Age



25.88***

.563

Older



Attendance a



4.04*

.326

Higher


2. SES

.11

27.12***

35.96***

.412

Higher


3. Age at entry

.16

27.26***

21.24***

-.308

Younger



Max grade



27.91 ***

.854

Higher


4. Treatment (Rx)

16

23.37***

.17

.121

...

Raven's Progressive Matrices:


5. Rx X grade

.17

18.91***

.22





Rx X SES



5.67*





Grade X SES



.00




6. Rx X grade X SES

.17

15.94***

4.62*



Literacy:


1. Sex

.02

7.30***

.61

-.057

...



Age



.61

-.060

...



Attendance a



20.67***

.178

Higher


2. SES

.10

24.85***

75.64***

.140

Higher


3. Age at entry

.54

166.40***

179.11***

-.044

Younger



Max grade



627.23**

.705

Higher


4. Treatment (Rx)

.54

143.53***

3.45

.095

Atole


5. Rx X grade

.56

109.46***

2.98





Rx X SES



.08





Grade X SES



40.01***



Reading:


1. Sex

.15

40.39***

25.75***

- 1.343

Males



Age



90.09***

1.092

Older



Attendance a



5.32*

.346

Higher


2. SES

.17

33.59**

11.35***

.217

Higher


3' Age at entry

.31

51.36***

77.29***

-.735

Younger



Max grade



67.88***

1.345

Higher


4. Treatment (Rx)

.31

43.96**

.03

.047

...


5. Rx X grade

.32

34.65***

.80





Rx X SES



2.67



NOTE - Sample sizes for the entire sample range from 678 to 868.

a Attendance is the residual value after regressing attendance on consumption.

p <.10.

*p <.05.

**p <.01.

***p <.001.

After controlling for potential confounders (in Steps 1-3), there was a main effect of treatment on the numeracy, knowledge, and Interamerican vocabulary and reading achievement tests; in all instances, the difference favored the Atole group. However, in each of these cases, the interactive term treatment x SES (Step 5) was statistically significant; this interactive term was also significant in the case of the Progressive Matrices test. The independent estimates of the slopes for the Atole and the Fresco groups on each of these five tests showed a significant positive slope for the Fresco but not for the Atole group. Figure 5 illustrates the pattern of the relations observed in each of these interactions. Thus, in the Fresco villages, test performance improved as SES improved, whereas no relation between SES and test performance existed in the Atole villages.

In addition, there was a significant maximum grade x treatment interaction on the reading achievement score; this interaction is plotted in Figure 6. In this case, both slopes were statistically significant, but the relation within the Atole group was stronger (b = 2.445, p <.001) than in the Fresco group (b =.827, p <.05). In other words, Atole children in the upper percentiles of grade attainment scored significantly higher than Fresco children, whereas no treatment differences were observed among children at the lowest end of the grade attainment distribution.

Significant SES X maximum grade interactions were obtained on the numeracy and literacy tests (see Table 18 above). In both cases, the significant positive relation between grade attainment and the outcomes fat all levels of SES became less strong as SES levels increased; thus, subjects from low socioeconomic backgrounds showed a stronger relation between level of schooling and outcome.

Finally, there was a significant three-way interaction (F = 4.62, p <.05) on RPM. When the grade x SES interactions on this measure were explored separately for Atole and Fresco villages, neither interaction was statistically significant. However, in Atole villages, there was a positive relation between SES and RPM within the highest levels of grade attainment (b =.274, p =.05) but no relation when the attained grade level was low. Conversely, in Fresco villages, the relation between SES and RPM was significant at low levels of grade attainment (b =.502, p <.0003) but not at higher levels (b =.245, p <.07).

In sum, significant treatment effects were observed on four of the psychoeducational outcomes. Treatment x SES interactions were significant on five outcomes, indicating that the significant differences between Atole and Fresco subjects increased as socioeconomic status declined. Treatment x grade attained interactions revealed that the benefits of Atole were strongest for those children reaching a higher maximum grade.

Information processing. - As was true for the psychoeducational test,, on most information-processing tasks, males outperformed females (i.e., simple and choice reaction time, percentage negative correct, impulsivity, and efficiency on the memory task). Similarly, SES and both school measures were associated with performance on paired associates, memory reaction time, percentage negative correct, and efficiency - in the expected direction.

Of the seven analyses of information-processing outcomes (Table 19), there were two significant main effects of the treatment. Atole subjects performed faster and more efficiently on the memory task than those exposed to Fresco.


FIG. 5. - SES X treatment interaction for vocabulary for the entire sample


FIG. 6. - Maximum grade x treatment interaction for reading achievement in the entire sample.

TABLE 19: RESULTS OF HIERARCHICAL REGRESSION ANALYSES FOR THE ENTIRE SAMPLE ON THE INFORMATION-PROCESSING TESTS

Step and Variables


Model F

F to enter

Parameter Estimate

Direction of Effect Favors:

Simple reaction time:


1. Sex

.02

4.58**

12.20***

.021

Males



Age



.87

-.001

...



Attendance a



.98

-.003

...


2. SES

.02

3.72**

1.14

-.002

...


3. Age at entry

.02

3.15**

.11

-.002

...



Max grade



3.89*

-.007

Higher


4. Treatment (Rx)

.02

2.76**

.37

-.004

...


5. Rx X grade

02

2.33**

1.54





Rx X SES



.14



Choice reaction time:


1. Sex

.01

3.90**

11.29***

.038

Males



Age



.31

.002

...



Attendance a



.08

.002

...


2. SES

.02

3.76**

3.31

-.005

...


3. Age at entry

.02

2.62*

.01

-.002

...



Max grade



.73

-.006

...


4. Treatment (Rx)

.02

2.41*

1.15

-.013

...


5. Rx X grade

.02

1.98*

.92


...



Rx X SES



.04


...

Paired associates trials to criterion:


1. Sex

.006

1.51

3.24

.903

...



Age



.91

-.332

...



Attendance a



.39

.187

...


2. SES

.02

4.48***

13.30***

-.455

Higher


3. Age at entry

.04

5.55***

6.20**

.335

Younger



Max grade



8.83**

-.874

Higher


4. Treatment (Rx)

.05

4.92***

1.18

-.588

...


5. Rx X grade

.05

4.42***

.11





Rx X SES



5.03*



Memory reaction time:


1. Sex

.005

1.48

.25

-.036

...



Age



3.07

.090

...



Attendance a



1.11

-.046

...


2. SES

.01

2.37*

5.02*

-.042

Higher


3. Age at entry

.04

5.41***

6.61**

.031

Younger



Max grade



16.12***

-.180

Higher


4. Treatment (Rx)

.05

5.92***

8.72**

-.241

Atole


5. Rx X grade

.05

4.61***

.06





Rx X SES



.05



% negative correct on memory task:


1. Sex

.04

11.93***

25.26***

-.083

Males



Age



.55

-.009

...



Attendance a



9.98**

.030

Higher


2. SES

.05

10.54***

6.15**

.010

Higher


3. Age at entry

.08

12.28***

7.71**

-.006

Younger



Max grade



22.39***

.046

Higher


4. Treatment (Rx)

.08

10.66***

.93

.017

...


5. Rx X grade

.08

8.34***

.53





Rx X SES



.03



Impulsivity on memory task:


1. Sex

.02

6.78***

12.65***

.341

Males



Age



4.90*

-.062

Older



Attendance a



2.79

-.091

...


2. SES

.02

5.12***

.18

-.010

...


3. Age at entry

.02

3.44**

.07

.006

...



Max grade



.12

-.020

...


4. Treatment (Rx)

.03

3.33**

2.61

.168

...


5. Rx X grade

.03

2.62**

.00





Rx X SES



.32



Efficiency on memory task:


1. Sex

.02

5.26***

6.33**

.270

Males



Age



3.69*

.117

Older



Attendance a



9.19**

-.182

Higher


2. SES

.03

7.27***

13.09***

-.093

Higher


3. Age at entry

.09

14.33***

16.25***

.067

Younger



Max grade



38.79***

-.377

Higher


4. Treatment (Rx)

.10

13.50***

7.87**

-.308

Atole


5. Rx X grade

.10

10.51***

.19





Rx X SES



.04



NOTE. - Sample sizes for the entire sample range from 678 to 878. a Attendance is the residual value after regressing attendance on consumption.

*p <.05.

**p <.01.

***p <.001.

In addition, there was one treatment x SES interaction on the paired associates task. Whereas SES was not associated with performance in the Atole subjects, in Fresco villages there was a significant inverse relation (b = -.569, p <.05). As SES improved, fewer trials were necessary to reach criterion for the Fresco subjects.

In sum, when significant, treatment effects favored Atole subjects. Compared to the observed effects on psychoeducational test performance, treatment effects were less consistently observed across outcomes and tended to be main rather than interactive effects.

Cohort of Maximum Exposure

Psychoeducational tests. - As shown in Table 20, the results of analyses restricted to this cohort closely resembled those obtained for the entire sample. Test performance tended to be better for older subjects, males, and those with higher levels of SES and grade attainment. There were main effects of treatment on numeracy, knowledge, and the Interamerican reading and vocabulary tests favoring Atole subjects. The interactions between treatment and SES were again statistically significant in the cases of numeracy, knowledge, RPM, and the Interamerican reading and vocabulary tests. The interactions for four of these outcomes (with the exception of reading, where there was a significant three-way interaction) are depicted in Figure 7. In all but the case of RPM, the slopes were statistically significant for the Fresco but not for the Atole group. At the lower ends of the SES distribution, Atole subjects performed significantly better than Fresco subjects; at higher levels of SES, there were no differences between Atole and Fresco subjects. On RPM, a slightly different pattern was observed: Atole children performed significantly better than Fresco children at the lowest end of the SES distribution (at or below the tenth percentile) but below Fresco children at its highest end (at or above the ninetieth percentile).

Treatment x maximum grade interactions were also observed on the reading and the two achievement tests. In all cases, slopes were positive and significant for Atole but not for Fresco subjects. On both achievement tests (for an illustration, see Fig. 8), differences between treatment groups increased with grade attainment, with children from Atole villages scoring significantly higher at the upper ends of the grade distribution than subjects exposed to Fresco. On the reading test (Fig. 9), Fresco subjects outperformed Atole subjects at the lower end of the distribution (at or below the twenty-fifth percentile), but there were no differences between groups in performance at the upper end of the grade distribution.

A grade x SES interaction was observed on the literacy test. As was true for the entire sample, the relation between grade attainment and literacy became weaker as SES improved, although, at all levels of SES, grade attainment was significantly and positively related to the outcome.

Finally, a significant three-way interaction was obtained on the Interamerican reading test. The SES x grade interaction was significant in Fresco but not Atole villages; within the Fresco villages, associations between SES and achievement were positive and significant for both low (b = 1.885, p <.0001) and high (b =.834, p <.0001) levels of grade attainment.


FIG. 7a. - SES X treatment interactions for numeracy, RPM, knowledge, and vocabulary in the cohort of maximum exposure. Numeracy


FIG. 7b. - SES X treatment interactions for numeracy, RPM, knowledge, and vocabulary in the cohort of maximum exposure. Knowledge


FIG. 7c. - SES X treatment interactions for numeracy, RPM, knowledge, and vocabulary in the cohort of maximum exposure. Raven's Progressive Matrices


FIG. 7d. - SES X treatment interactions for numeracy, RPM, knowledge, and vocabulary in the cohort of maximum exposure. Vocabulary

Information-processing tests. - Results obtained for the maximum exposure cohort were again similar to those for the entire sample (see Table 21). Gender (i.e., males), higher socioeconomic status, earlier school entry, and higher grade attainment were associated with enhanced performance on the information-processing test battery. Three significant main effects of treatment were observed. Compared to Fresco subjects, Atole subjects reached criterion faster on the paired associates task and responded faster and more efficiently on the memory task.

TABLE 20: RESULTS OF HIERARCHICAL REGRESSION ANALYSES FOR THE COHORT OF MAXIMUM EXPOSURE ON THE PSYCHOEDUCATIONAL TESTS

Step and Variables


Model F

F to enter

Parameter Estimate

Direction of Effect Favors:

Numeracy:


1. Sex

.02

3.46*

2.12

- 1.066

...



Age



6.69**

1.556

Older



Attendance a



1.55

.529

Higher


2. SES

.10

11.15***

33.42***

1.042

Higher


3. Age at entry

.46

57.06***

84.09***

-.721

Younger



Max grade



184.59**

4.456

Higher


4. Treatment (Rx)

.47

50.83***

7.75**

1.708

Atole


5. Rx X grade

.48

41.44***

.01





Rx X SES



10.06***



Knowledge:


1. Sex

.06

8.71***

.17

.140

...



Age



25.42***

1.627

Older



Attendance a



.56

.135

...


2. SES

.12

13.95***

27.93***

.408

Higher


3. Age at entry

.24

21.55***

34.14***

-.481

Younger



Max grade



30.78***

.932

Higher


4. Treatment (Rx)

.26

20.04***

8.57**

.919

Atole


5. Rx X grade

.27

16.63**

.79





Rx X SES



6.74**



Raven's Progressive Matrices:


1. Sex

.06

8.07***

19.40***

- 1.844

Males



Age



4.32*

.793

Older



Attendance a



.50

.171

...


2. SES

.09

9.80***

14.19***

.396

Higher


3. Age at entry

.13

10.09***

6.59***

-.170

Younger



Max grade



13.06***

.871

Higher


4. Treatment (Rx)

.13

8.66***

.22

.212

...


5. Rx X grade

.15

8.05**

2.18





Rx X SES



8.39**



Reading achievement:


1. Sex

.01

1.68*

.02

.038

...



Age



3.29**

1.624

Older



Attendance a



1.74

-.420

...


2. SES

.12

10.76***

37.43***

.845

Higher


3. Age at entry

.18

11.70***

12.67***

-.621

Younger



Max grade



11.58***

1.239

Higher


4. Treatment (Rx)

.22

13.47***

20.05***

2.558

Atole


5. Rx X grade

.29

14.43***

13.14***





Rx X SES



14.91***





Grade X SES



.10




6. Rx X grade X SES

.30

12.75***

7.88**



Vocabulary:


1. Sex

.03

3.16*

.01

-.122

...



Age



8.01**

3.290

Older



Attendance a



1.48

-.571

...


2. SES

.11

10.66***

32.26***

1.168

Higher


3. Age at entry

.21

14.25***

18.98***

-1.064

Younger



Max grade



19.17***

2.327

Higher


4. Treatment (Rx)

.26

16.20***

22.35***

3.930

Atole


5. Rx X grade

.30

15.13***

6.13**





Rx X SES



11.28***



Reading:


1. Sex

.09

10.37***

8.83**

-1.152

Males



Age



21.95***

1.849

Older



Attendance a



.32

.121

...


2. SES

.10

8.94***

4.36*

.204

Higher


3. Age at entry

.28

21.20***

47.41**

-.848

Younger



Max grade



35.24***

1.412

Higher


4. Treatment (Rx)

.28

18.12***

.03

.062

Atole


5. Rx X grade

.30

15.22***

5.36*





Rx X SES



2.49



Literacy:


1. Sex

.01

1.49

1.10

.096

...



Age



1.03

.016

...



Attendance a



2.32

.081

...


2. SES

.08

8.85***

30.61***

.125

Higher


3. Age at entry

.55

82.34***

102.52***

-.040

Younger



Max grade



319.76***

.666

Higher


4. Treatment (Rx)

.55

70.54***

.44

.047

...


5. Rx X grade

.56

52.25***

.54





Rx X SES



.14





Grade X SES



13.92***



NOTE. - Sample sizes for the cohort of maximum exposure range from 335 to 416.

a Attendance is the residual value after regressing attendance on consumption.

*p <.05.

**p <.01.

***p <.001.

Although one significant three-way interaction was obtained on the percentage negative correct on the memory task, when the relation between SES and grade was analyzed by treatment, the interactions were nonsignificant in both Atole and Fresco villages. There were no significant two-way interactions.

Overall Summary of Results

Focusing on treatment effects obtained in analyses of the psychoeducational tests, we see that, in both the entire sample and the cohort of maxi


FIG. 8. - Maximum grade x treatment interaction for reading achievement in the cohort of maximum exposure.


FIG. 9. - Maximum grade x treatment interaction for reading in the cohort of maximum exposure.

Fewer effects of treatment were observed on the information-processing tasks. Atole treatment resulted in faster and more efficient processing on the memory task in both the entire sample and the cohort of maximum exposure; the one significant interaction (treatment x SES on paired associates) was observed only in the entire sample. On the outcomes that were affected, treatment differences accounted for between 1% and 2% of the variance m performance.

Results of the hierarchical analyses also provide information regarding the other variables entered into the model and their associations with the outcome variables. Overall, the full models accounted for between 15% (RPM) and 55% (literacy) of the variance in psychoeducational test performance and between 2% and 10% of the variance in information-processing outcomes. In general, males outperformed females on both the psychoeducational tests (i.e., reading, numeracy, and RPM) and the information-processing tasks (i.e., simple and choice reaction time, memory impulsivity, and percentage negative correct). Age was positively associated with performance on almost all the psychoeducational tests but not with performance on the information-processing tasks. Similarly, socioeconomic status and schooling were related to performance on most tests, but these associations were stronger for the psychoeducational than for the information-processing tasks. As indicated also by the correlational analyses, earlier school entry and attainment of a higher maximum grade were associated with improved test performance.

Cohort of Late Exposure

As noted at the beginning of this chapter, a cohort of subjects identified as having been exposed to the treatment only after 24 months of age was analyzed separately to test for the effects of early versus late exposure to treatment. After controlling for all other variables, treatment yielded two main effects: Atole subjects in this cohort performed significantly better on tests of numeracy and knowledge than Fresco subjects. As observed in the other analyses, there was also a significant treatment x grade interaction on the reading achievement test.

TABLE 21: RESULTS OF HIERARCHICAL REGRESSION ANALYSES FOR THE COHORT OF MAXIMUM EXPOSURE ON THE INFORMATION-PROCESSING TESTS

Step and Variables


Model F

F to enter

Parameter Estimate

Direction of Effect Favors:

Simple reaction time:


1. Sex

.01

1.67

4.54*

.018

Males



Age



.09

.001

...



Attendance a



.37

-.003

...


2. SES

02

1.68

1.72

.003

...


3. Age at entry

.03

2.02

3.58

-.012

...



Max grade



1.73

-.007

...


4. Treatment (Rx)

.03

1.73

.03

-.002


Choice reaction time:


1. Sex

.01

1.30

3.65*

.027

Males



Age



.08

.001

...



Attendance a



.17

.003

...


2. SES

.01

1.31

1.34

-.004

...


3 Age at entry

.02

1.04

.12

-.001

...



Max grade



.91

-.088

...


4. Treatment (Rx)

.02

1.04

1.01

-.016

...

Paired associates trials to criterion:


1. Sex

.01

.72

.40

-.450

...



Age



.02

-.713

...



Attendance a



1.73

.538

...


2. SES

.01

.86

1.29

-.208

...


3. Age at entry

.05

3.15**

7.37**

.485

Younger



Max grade



7.98**

-1.210

Higher


4. Treatment (Rx)

.06

3.24**

3.65*

-1.476

Atole

Memory reaction time:


1. Sex

.002

.24

.03

-.018

...



Age



.00

-.051

...



Attendance a



.69

.046

...


2. SES

.004

.43

1.01

-.024

...


3. Age at entry

.02

1.51

3.03

.037

...



Max grade



4.25*

-.117

Higher


4. Treatment (Rx)

.04

2.64**

9.25**

-.321

Higher

Impulsivity on memory task:


1. Sex

.03

3.47*

4.88*

.285

Males



Age



.01

.176

...



Attendance a



5.52*

-.174

Higher


2. SES

.03

2.97*

1.44

-.039

Higher


3. Age at entry

.03

2.15*

1.03

.065

Younger



Max grade



.05

-.017

Higher


4. Treatment (Rx)

.04

2.14*

2.06

.208

Higher

% negative correct on memory task:


1. Sex

.03

4.82**

10.26***

-.074

Males



Age



.02

-.038

...



Attendance a



4.17*

.027

Higher


2. SES

.05

4.84***

4.78*

.013

Higher


3. Age at entry

.07

5.19***

7.67**

-.023

Younger



Max grade



3.66

.026

...


4. Treatment (Rx)

.07

4.59***

1.02

.026

...

Efficiency on memory task:


1. Sex

.01

1.50

3.35

.251

...



Age



.01

.076

...



Attendance a



1.13

-.085

...


2. SES

.03

2.74*

6.43** -

.088

Higher


3. Age at entry

.08

5.54***

11.93

.139

...



Max grade



9.80

-.248

...


4. Treatment (Rx)

.10

6.04***

8.40** -

.428

Atole

NOTE. - Sample sizes for the cohort of maximum exposure range from 335 to 416.

a Attendance is the residual value after regressing attendance on consumption.

*p <.05.

**p <.01.

***p <.001.

Although the pattern of findings was similar to that of the cohort of maximum exposure, the number and magnitude of significant findings were greatly reduced. The results from these analyses do not provide sufficient evidence for specifying the role of timing of the intervention.