malnutrition, stunting, wasting, cognitive

646 Am J Gun Nutr l989;49:646-53. Printed in USA. © 1989 American Society for Clinical Nutrition

Effects of missing breakfast on the cognitive functions ofschool children of differing nutritional status13

Donald T Simeon and Sally Grant ham-McG regor

ABSTRACt’ We examined the effects of omitting breakfast on the cognitive functions of

three groups ofchildren: stunted, nonstunted controls, and previously severely malnourished.

They were admitted to a metabolic ward twice. After an overnight fast half the children re-

ceived breakfast on their first visit and a cup of tea the second time. The treatment orderwas reversed for the other half. When breakfast was omitted, both the stunted and previously

malnourished groups responded similarly. The malnourished groups had lower scores in flu-ency and coding whereas the control group had higher scores in arithmetic. The children were

divided into wasted and nonwasted groups. Wasted children were adversely affected in the

digit span backwards tests, and wasted members of the malnourished groups were adverselyaffected in efficiency ofproblem solving and those in the control group in digit span forwards.

These results indicate that cognitive functions are more vulnerable to missing breakfast inpoorly nourished children. Am J Clin Nutr 1 989;49:646-53.

KEY WORDS Breakfast, fasting, severefunctions

Introduction

Although school feeding programs have been in exis-tence for many years, evaluations have been few and

have generally lacked scientific rigor (1-3). The mainoutcome variables have been nutritional status (4-9),school attendance (5, 9, 10), school achievement (4, 5,10, 1 1), and classroom behavior (12-14). Benefits havebeen found in all these variables; however, findings havebeen inconsistent except in the case of behavior. Someof the inconsistencies may be explained by differencesin the initial nutritional and socioeconomic status of the

samples.In a study in Jamaica (10), when breakfast was given

to a group of poor and mostly undernourished childrenfor one semester, there was an improvement in schoolachievement but no impact on weight or height. Themechanism ofthe improvement was not clear but in gen-eral where benefits have been found in school achieve-ment or behavior it has been suggested that the allevia-

tion of hunger may play an important part (1, 2, 10, 13).There have been several studies of the effects of short-

term fasting in school children (1 5-17). English studiesof secondary school students (1 5) found that omittingbreakfast had no effect on their performance at arithme-tic, short-term memory, and attention-demanding tasks.A study of North American children aged 9-1 1 y (16)found that omitting breakfast had a beneficial effect onshort-term memory. However it had an adverse effect onthe problem solving ability of children with low IQs but

not in those with high IQs. When the American studywas replicated (1 7), it was found that short-term fastinghad an adverse effect on problem solving in all the sub-jects regardless oftheir IQs. There was also an increase inattention to task-irrelevant information. The beneficialeffect on short-term memory was not replicated. There-

fore, the findings ofthese studies have been inconsistent,with either few or no detrimental effects being demon-

strated.The subjects in each of the studies of the effects of

missing breakfast were adequately nourished. There is adanger that these findings may influence policy decisionson school feeding, not only in developed countries butalso in developing ones. Undernutrition is endemic in

these countries and children who are undernourished aremore likely to be exposed to hunger. Furthermore theymay react differently.

An extreme example of undernutrition is early-child-hood severe malnutrition. Survivors generally havelower IQs (1 8-20) and school achievement levels (2 1, 22)

I From the Tropical Metabolism Research Unit, University of theWest Indies, Kingston, Jamaica.

2 Supported by Nestle Nutrition Research Grant 85/35, the Well-

come Trust, and contributions from Grace Kennedy and CompanyLimited, lCD (Jamaica) Ltd. and Restaurants ofJamaica, Ltd.

3 Address reprint requests to D Simeon, Tropical Metabolism Re-

search Unit, University ofthe West Indies, Mona, Kingston 7, Jamaica.Received December 17, 1987.Accepted for publication April 26, 1988.

at MIT

LIBR

AR

IES

on July 10, 2011w

ww

.ajcn.orgD

ownloaded from

BREAKFAST AND CHILDREN’S MENTAL FUNCTION 647

Efficiency = Z score (errors) + Z score (latency) (1)

and more behavioral problems (2 1 , 23) than matchedcomparison children several years later. It is possible thatthese children may be particularly vulnerable to fasting.

We have, therefore, investigated the effect of missing

breakfast on the cognitive function of the followinggroups: linear-growth-retarded school children (stunted),matched control children who were not growth-retarded,and children who were severely malnourished in earlychildhood.

It is generally considered that low weight-for-height(wasting) indicates recent nutritional experienceswhereas stunting, or low height-for-age, is the best mdi-cator ofthe duration ofundernutrition (24). In addition,low height-for-age is the commonest form of undernutri-tion in developing countries (25). The undernourishedgroup in this study was therefore selected by degree ofstunting. The effect of degree of wasting was also exam-med.

Methods

Sample

The study comprised three groups of children aged 9-10.5y, each with 30 subjects. The children had no obvious mental orphysical handicap. Only one child was selected per household.

Group 1 was the previously severely malnourished group-subjects who were identified from hospital records and hadbeen admitted to the University Hospital of the West Indies

(UHWI) for severe malnutrition during the first 2 y of life;group 2 was the stunted group-children whose heights were

below -2 SDs of the NCHS standards (26); and group 3 wasthe control group-children whose heights were above - 1 SDof the NCHS standards. Groups 2 and 3 were identified fromthree primary schools located in poor areas close to the UHWI.The stunted children were identified first and then the controlchildren were selected from the same class. They were matchedfor gender and area of residence. Where more than one childwas available, the nearest in age was used. There were 19 boysand 11 girls in group 1 and 15 boys and 15 girls in each ofgroups 2 and 3.

Tests ofcognitivefunctions

The battery of tests included both measures of fine-grainedlevels ofcognitive functions and more global or classroom-type

tasks (Table 1). They included three subtests of the WechslerIntelligence Scale for Children (27): arithmetic, digit span, andcoding. The arithmetic test comprised mental arithmetic prob-lems, each having a time limit. The score represents the totalnumber ofcorrect answers.

The digit-span test involved the immediate recall of increas-ingly longer strings ofdigits that were read to the children. One

set had to be recalled forward as they were given and a secondset had to be recalled backwards. The forward and backwarditems were analyzed separately and the children’s scores repre-

sent the total number ofcorrect responses.In the coding test the children had to substitute symbols for

numbers as quickly as possible. The score represents the totalnumber ofcorrect symbols written during a fixed time.

These three tests were chosen because performance on themis affected by attention and distractibility (28), which wethought were likely to be susceptible to the omission of break-

fast. In addition, the arithmetic test measures computationalskills and the coding test measures visual short-term memory

TABLE 1

The tests ofcognitive functions

Test-retestcorrelation

Tests Cognitive functions (r)

Arithmetic Computational skills 0.95

Digit span Auditory short-termmemory

0.87

Coding Visual short-term 0.92

Fluency

memory

Generation ofideas 0.97Listening Comprehension ofspoken 0.94

comprehension paragraphsMFFT*

Errors Problem-solving ability 0.66

Latency Speed ofresponse 0.94

HCItCentral task Visual short-term

memory0.71

Incidental task Attention to task-irrelevant information

0.62j

S Matching Familiar Figures Test.

t Hagen’s Central-Incidental task.� Test reliability for the HCI incidental task was computed using

slit-half reliability.

and clerical speed and accuracy. The forward items ofthe digitspan test measure auditory short-term memory whereas thebackward set includes an immediate processing element.

Two subtests of the Clinical Evaluation of Language Func-tions (29) were also used: fluency and listening comprehension.The fluency test required the children to name as many itemsas possible in two different categories. The score represents thenumber ofitems named in a limited time period. This test wasused to measure generation ofideas and motivation. In the lis-tening comprehension test four short stories were read to thechildren who then had to answer questions about them andthis test measures attention, auditory short-term memory, andcomprehension and resembles work performed in school.

The other two tests used, the Matching Familiar Figures Test

(MFFT) (30) and the Hagen’s Central-Incidental task (HCI)(3 1), were previously shown to be sensitive to the omission of

breakfast (16, 17). For each item of MFFT, the children werepresented with two cards. On one card was the picture of anobject and on the other were a number of variant pictures ofthe same object. The children had to choose which of the pic-

tures on the second card was exactly like that on the first. Thescores included the number of errors made before the correctpicture was identified and the latency time, which was the timetaken before the children’s first response.

The number oferrors made is a measure of problem-solvingability whereas the latency time measures reaction time. TheMFFT comprised an easy and a hard set, which were analyzedseparately.

The MFFT was analyzed in terms of impulsivity and effi-ciency of problem solving. Firstly, Z scores were computedfrom the children’s number of errors and mean latency time.Efficiency scores were computed by adding the Z score of totalerrors to the Z score of mean latency time; impulsivity scoreswere computed by subtracting the Z score ofthe mean latencytime from the Z score ofthe total errors (32), ie,

at MIT

LIBR

AR

IES

on July 10, 2011w

ww

.ajcn.orgD

ownloaded from

648 SIMEON AND GRANTHAM-MCGREGOR

TABLE 2The standard breakfast

Energy Carbo-Weight content Protein Fat hydrate

g kcal g g g

Nutribun* 105 333 14 1 65Milk 242 169 9 3 25Cheese 25 97 6 8 1Total 590 29 12 91

* Used in the Jamaican school feeding program.

Impulsivity = Z score (errors) - Z score (latency) (2)

The efficiency and impulsivity scores compare the children’sperformances relative to each other. A negative efficiency scoreis indicative of greater efficiency (fewer errors and fast time)whereas a positive impulsivity score indicates a more-impul-sive performance (more errors and fast time).

The HCI task comprised a number of cards each of whichhad a picture ofan animal and an object. For the central task aline of cards was placed in front of the child and his attentionwas drawn to the animals. The cards were then covered and thechild had to recall the position ofspecific animals. The score onthe central task was the number ofcorrect locations identified.Then the children had to recall which object was on the cardwith each animal. The incidental score represents the numberof items correctly matched to the animals. The central taskmeasures visual short-term memory whereas the incidentaltask is a measure of attention to task-irrelevant information.

By its nature the incidental task was only valid for the first testsession and therefore it was not repeated on the second visit.

Before the study all tests except the HCI incidental task wererepeated in 10 children within 1 wk. The test-retest correlationswere significant and ranged from 0.66 to 0.97. The Spearman-Brown split-halfreliability (33) for the HCI incidental task was0.62(Table 1).

Procedure

After the subjects were identified, their homes were visitedand consent to participate in the study was obtained from theirparents. Details of their social background were obtained andtheir housing was scored on a scale that gave equal weightingto quality ofwater supply, sanitation, and crowding (34). Therating ranged from 3.0 (worst housing) to 9.0 (best housing).The presence of certain household possessions, including tele-vision, radio, stove, and refrigerator, was also counted.

The subjects were then admitted overnight, two at a time, toa special research ward on two occasions 1 wk apart. They en-tered in the afternoon and at 1700 they received a standarddinner of chicken, bread, and french fried potatoes that pro-vided “-940 kcal. On the afternoon oftheir first visit they were

given the Peabody Picture Vocabulary Test (PPVT) to measure

their IQs (35).At 0800 the next day they were given either a standard break-

fast (Table 2) providing 590 kcal (“.�25% oftheir daily caloricrequirement) or a cup (“.‘ 1 85 mL) oftea sweetened with aspar-tame. The breakfast comprised items from the Jamaican gov-ernment school feeding program. It is common for many poorJamaican children to have only a cup of tea before going to

school. Therefore it was hoped that by providing tea the chil-dren would not have been alerted to the fact that something

unusual was happening. However the design was not truly dou-ble-blind. The level ofaspartame used (< 4 mg/kg) was not cx-pected to affect their mental functions (36).

By systematic random assignment, halfthe children receivedbreakfast on the morning of their first visit and the tea on themorning oftheir second visit. The treatment order was reversedfor the other halfofthe children. At 1 100 the tests of cognitivefunctions were given by a tester who was unaware of the chil-dren’s group or fasting state. Each test session lasted for“-#45 mm.

After the tests the children were given lunch and then takenhome. The subjects weights and heights were measured follow-ing established procedures (37) on the morning of their firstvisit.

Permission to conduct this study was obtained from the Eth-ics Committee ofthe UHWI.

St atistical analysis

The weights and heights were expressed as a percentage ofthe expected value (NCHS standards) for the children’s age andsex. Group differences in mean age, anthropometric variables,IQ scores, and social-background variables were examined byanalyses of variance (ANOVAs) or chi-square tests where ap-propriate (38).

The experimental findings for the tests of cognitive func-tions, with the exception of the HCI incidental task, were cx-amined by factorial repeated-measures analyses of varianceand covariance. The previously severely malnourished and thestunted groups behaved similarly when they missed breakfast,

and there was no significant difference between these twogroups for any ofthe cognitive tests. They were therefore com-

bined and compared with the control group to test for treat-ment-by-group effects.

The scores on the children’s first and second visits were thedependent variables in the repeated-measures analyses. There-fore the test-retest effect was the within-subject factor. Treat-ment and nutritional group were used as between-subjects fac-tors. To examine the effects ofwasting, the sample was dividedinto a wasted and nonwasted set, defined for this study as chil-

dren with weight-for-height below and above 90% of the cx-pected value, respectively (24). There were 30 wasted childrenwho were evenly distributed among the original three groups.

The ANOVAs with repeated measures were then rerun withwasting as an added between-subjects factor.

IQ was used as a covariate because there was a group differ-ence but it did not significantly change the main findings.Weight-for-height was also used as a covariate when treatment

and group were used as factors and it too did not change themain findings. Consequently, only the results ofthe ANOVAswith repeated measures will be reported.

The HCI incidental task was not repeated on the children’s

second visit. Therefore the scores of the children who hadbreakfast on the first visit were compared with those who didnot by ANOVA. Factors included whether they had fasted ornot, nutritional group, and wasting. When IQ was used as acovariate it was not significant.

Results

The characteristics of the children and their socialbackgrounds are shown in Table 3. There were nodifference between groups in the mean age and weight-for-height ofthe children. However, the control childrenhad higher IQ scores than did the other two groups (AN-

at MIT

LIBR

AR

IES

on July 10, 2011w

ww

.ajcn.orgD

ownloaded from

BREAKFAST AND CHILDREN’S MENTAL FUNCTION 649

TABLE 3

Age, IQ scores, anthropometric indices, and demographic

characteristics ofthe sample by group

Previouslymalnourished

(n = 30)

Stunted

(n = 30)

Control(n = 30)

Age(mo) 117.2± 6.4* 116.8± 6.7 115.1 ± 5.5

IQ scoret 62.3 ± 16.8 62.8 ± I 1.3 72.4 ± 1 1.8

Height-for-age(%4 95.6 ± 4.5 89.3 ± 1.4 99.7 ± 2.4

Weight-for-height (%) 92.3 ± 6.6 93.2 ± 5.4 93.6 ± 7.6

Weight-for-age

(%)f 81.6 ± 1 1.5 68.6 ± 5.6 92.2 ± 10.1

Socioeconomic

characteristicsHousingrating 5.8± 1.6 5.6± 2.1 6.1± 2.2

Householdpossessions 1.8± 1.3 2.0± 1.6 2.6± 1.4

Female caretakereducation

(%)�None 12 0 3

Gradesl-4 32 23 17

GradesS-9 44 63 70

Secondaryschool 12 13 10

*1± SD.

t ANOVA, p < 0.01.j: ANOVA,p < 0.001.§ One child in the previously malnourished group had no female

caretaker and n = 29.

OVA, F = 5.3, p < 0.01 ; post-ANOVA comparisons:control vs stunted, p < 0.01 , and control vs previouslymalnourished, p < 0.0 1). There were group differencesas expected in height-for-age (ANOVA, F = 89.6, p< 0.00 1) and weight-for-age (ANOVA, F = 46.7, p< 0.001). The mean height-for-age and weight-for-age ofthe previously malnourished group were greater thanthose of the stunted group (post-ANOVA comparisons,p < 0.001 for both). However they were less than those ofthe control group(post-ANOVA comparisons, p < 0.001for both).

All groups had poor housing and few household pos-sessions. There was no difference between the groups.The female caretakers in the different groups had similareducational levels and few had attended secondaryschools.

The scores of the cognitive tests with and withoutbreakfast are shown in Table 4 and the statistical findingsof the ANOVAs with repeated measures with group,treatment, and wasting as factors are given in Table 5.

Fluency (Fig 1)

In the fluency test the groups behaved differently onmissing breakfast. There was no change in score in the

control group whereas the score of the combined mal-nourished group deteriorated in the fasted state, produc-

ing a significant treatment-by-group effect (post-AN-OVA comparisons offed vs fasted states, combined mal-nourished groups, p < 0.001). When wasting was used asa factor, there was no new significant finding.

Arithmetic (Fig 1)

The control children had a higher score in the arithme-tic test when they had no breakfast whereas the com-bined malnourished group was not affected. This pro-duced a significant treatment-by-group effect (post-AN-OVA comparisons of fed vs fasted states, control group,p < 0.001). When wasting was used as a factor, there wasno new significant finding.

Coding (Fig 1)

There was no significant group, wasting, treatment-by-

group, treatment-by-wasting, or treatment-by-group-by-wasting effect for the coding test. However, the treatmenteffect approached significance (p < 0.08). When thescores ofthe groups were analyzed separately, the controlgroup was not affected whereas the combined malnour-ished group had lower scores in the fasted state (ANOVAwith repeated measures for combined malnourishedgroup, treatment, p < 0.05).

Digit-span-backward items (Fig 2)

When treatment, group, and wasting were used as fac-tors, the wasted children of the total sample had lowerscores in the fasted state whereas the nonwasted childrenwere not affected. This produced a significant treatment-by-wasting effect (post-ANOVA comparison of fed vsfasted states, wasted children of the total sample, p< 0.05).

Digit-span-forward items (Fig 3)

When treatment, group, and wasting were used as fac-tors, there was a significant treatment-by-group-by-wast-ing effect for the digit-span-forward test. The wasted chil-dren of the control group had lower scores in the fastedstate whereas the nonwasted members ofthis group hadhigher scores.

Neither the wasted nor the nonwasted members of thecombined malnourished group were affected when theywent without breakfast (post-ANOVA comparisons offed vs fasted states: wasted children of control group,p < 0.02, and nonwasted members ofthe control group,p < 0.05).

MFFT-easy items (Fig 3)

When the efficiency scores of the easy items of theMFFT were examined with treatment, group, and wast-ing as factors, there was a significant treatment-by-groupeffect. The controls improved on missing breakfast (ie,they responded faster and made fewer errors) whereas

the combined malnourished group was not affected.There was also a treatment-by-wasting effect with thenonwasted children not being affected whereas thewasted children deteriorated on missing breakfast. How-ever, the latter effect was due solely to the wasted mem-

at MIT

LIBR

AR

IES

on July 10, 2011w

ww

.ajcn.orgD

ownloaded from

* Results ofANOVA oftest scores on children’s first admission.

650 SIMEON AND GRANTHAM-MCGREGOR

TABLE 4

Scores ofthe tests ofcognitive functions with and without breakfast by group*

Previously malnourished Stunted Control

Breakfast No breakfast Breakfast No breakfast Breakfast No breakfast

fluency 23.9 ± 6.8 21.8 ± 7.1 22.2 ± 6.1 20.5 ± 4.9 22.4 ± 5.4 22.7 ± 4.1Arithmetic 7.5 ± 2.9 7.5 ± 3.1 7.5 ± 1.9 7.6 ± 1.9 7.9 ± 1.8 8.5 ± 2.2

Coding 21.7 ± 13.3 20.6 ± 12.9 22.9 ± 12.3 21.5 ± 10.8 25.7 ± 9.4 25.6 ± 10.3Digit span

Forwarditems 5.3 ± 1.9 5.2 ± 2.0 5.3 ± 1.7 5.3 ± 1.5 5.2 ± 1.8 5.1 ± 1.7Backwarditems 2.8 ± 2.4 2.7 ± 2.2 2.6 ± 1.6 2.7 ± 1.6 2.8 ± 1.3 2.8 ± 1.4

HCI

Central 6.0 ± 1.9 5.7 ± 1.4 5.8 ± 1.8 5.3 ± 1.5 5.6 ± 1.5 5.7 ± 1.9Incidental 2.5 ± 1.9 2.3 ± 1.3 1.5 ± 1.5 2.2 ± 2.2 2.0 ± 1.8 1.4 ± 1.8

Listeningcomprehension 10.5 ± 5.1 10.2 ± 5.2 10.0 ± 4.0 9.9 ± 2.7 10.9 ± 4.2 1 1.5 ± 4.1

MFFT easy itemsErrors 7.9 ± 4.3 8.0 ± 4.9 8.5 ± 4.5 8.6 ± 5.2 8.8 ± 5.4 7.6 ± 4.8Latency 8.4 ± 5.6 8.2 ± 5.4 8.1 ± 5.9 8.1 ± 4.9 9.5 ± 8.3 8.1 ± 6.9Efficiency -0.14 ± 0.85 -0.01 ± 0.84 -0.08 ± 0.97 0.1 1 ± 0.78 0.22 ± 1.20 -0.09 ± 1.23

Impulsivity -0.07± 1.55 -0.03± 1.75 0.11 ± 1.54 0.12± 1.74 -0.04± 2.07 -0.09± 1.82MFFT hard items

Errors 16.4 ± 5.9 15.8 ± 6.1 17.9 ± 5.8 17.9 ± 5.6 16.7 ± 6.1 15.0 ± 6.3Latency 9.7 ± 8.0 1 1.4 ± 13.0 9.6 ± 10.0 9.0 ± 7.5 10.0 ± 10.4 12.3 ± 17.6Efficiency -0.11± 1.22 -0.03± 1.08 0.14± 1.42 0.13± 1.06 -0.03± 1.48 -0.10± 1.43Impulsivity -0.10± 1.40 -0.11± 1.67 0.17± 1.47 0.42± 1.10 -0.07± 1.64 -0.31± 1.90

#{149}1± SD.

bers ofthe combined malnourished groups and there was bers of the combined malnourished group were lessa significant treatment-by-group-by-wasting effect. Nei- efficient whereas the nonwasted members of this groupther the wasted nor the nonwasted members ofthe con- were not affected when they had no breakfast (post-AN-trol group were significantly affected in the fasted state OVA comparisons of fed vs fasted states: controls, pwhen examined separately. However, the wasted mem- < 0.05; wasted children ofthe total sample, p < 0.01; and

TABLES

Results ofthe multifactor ANOVA with repeated measures for the tests ofcognitive functions

Treatment byTreatment by Treatment by group by

Treatment Group Wasting group wasting wasting

F p F p F p F p F p F p

Fluency 6.8 0.01 0.2 NS 0.0 NS 4.8 0.04 0.2 NS 0.1 NS

Arithmetic 2.7 0.10 1.8 NS 2.2 NS 4.6 0.04 0.1 NS 2.0 NSCoding 3.2 0.08 2.5 NS 0.2 NS 0.6 NS 0.0 NS 0.7 NSDigit span

Forward 0.0 NS 0.1 NS 0.2 NS 0.1 NS 0.6 NS 7.8 0.01Backward 0.5 NS 0. 1 NS 0.2 NS 0. 1 NS 4.6 0.04 0. 1 NS

HCI

Central 1.5 NS 0.0 NS 0.0 NS 0.8 NS 0.4 NS 0.9 NSIncidental* 0.0 NS 1.2 NS 0.0 NS 1.0 NS 0.2 NS 0.5 NS

Listeningcomprehension 0. 1 NS 1.5 NS 0. 1 NS 0.6 NS 0.3 NS 0.3 NS

MFFT easy itemsEfficiency 2. 1 NS 0.04 NS 2. 1 NS 4.4 0.04 5.8 0.02 5.7 0.02Impulsivity 0.0 NS 0. 1 NS 1.3 NS 0.2 NS 0.0 NS 0.0 NS

MFFT hard itemsEfficiency 0.2 NS 0. 1 NS 0.3 NS 0.0 NS 0. 1 NS 0.2 NSImpulsivity I .8 NS 0.0 NS 0.4 NS 2.2 NS 0.5 NS 0.5 NS

at MIT

LIBR

AR

IES

on July 10, 2011w

ww

.ajcn.orgD

ownloaded from

MEAN SCORE MEAN SCORE

23

22

8.5

8.0

,,

,

25 1-+ 2j

2111�

yes no yes no yes

BREAKFAST BREAKFAST

and coding. #{149} #{149},combined malnourished group,

no

MEAN SCORE

3.0

2.8

2.6

�0�

�0�

yes no

BREAKFAST

BREAKFAST AND CHILDREN’S MENTAL FUNCTION 651

FIG 2. Mean scores by wasting set in digit-span-backwards test.. ., wasted set, n = 30; x - - - x, nonwasted set, n = 60.

FLUENCY

21 �1’

BREAKFAST

AR! THME TIC

FIG 1. Mean scores by group in fluency, arithmetic,n = 60; X - - - X, control group, n = 30.

COOING

MEAN SCORE

wasted members of the combined malnourished group,

p < 0.001).

Other tests

There was no significant effect when breakfast wasomitted in the MFFT hard-items test, listening compre-

hension, and the HCI task.

OIGIT SPAN � BACKWARDS

Discussion

The control group was not adversely affected in any ofthe cognitive tests when breakfast was omitted. Theyeven performed better in arithmetic and were moreefficient in problem solving. In contrast, the previouslymalnourished and the stunted groups were adverselyaffected in fluency (a measure ofgeneration ofideas andmotivation) and coding (visual short-term memory).Relative to the control children they were also adverselyaffected in arithmetic. The adverse effects of missingbreakfast on the previously malnourished and stuntedgroups remained after controlling for both their level ofIQ and degree of wasting. This suggests that the effectswere independent of IQ and recent nutritional experi-ence.

Wasted children, regardless ofwhich nutritional groupthey belonged to, were also adversely affected when theyhad no breakfast, although this was shown by cognitivetests different from those which showed the stunted andthe previously malnourished groups to be affected.Wasted children were adversely affected in the digit-span-backwards test. This is a measure ofauditory short-term memory with an immediate processing compo-nent.

Consequently, the children affected detrimentally bymissing breakfast as shown by most cognitive tests werethose who were wasted as well as being either stunted orpreviously malnourished. These children were also theonly ones detrimentally affected in the efficiency of prob-1cm solving. The wasted children of the control groupwere adversely affected in �the digit-span-forward test. It

at MIT

LIBR

AR

IES

on July 10, 2011w

ww

.ajcn.orgD

ownloaded from

56

54

52

50

‘4

/

4,

z/

//

BREAKFASTno yes

BREAKFAST

HFFT - (FFICIEP�Y

S.,

DIGIT S��N - F�ls(4ROS and hormone production are adversely affected duringsevere malnutrition (39) and it is uncertain whether theseabnormalities are fully reversed after recovery (40-42).During short-term fasting the hormonal and metabolicchanges serve to supply the brain with adequate substrate

�* for its metabolism. Therefore, under short-term fastingconditions malnourished children may be under agreater metabolic stress, which could lead to a deteriora-tion in cognitive performance.

.,�.. The arousal state of the children may be altered by

/ short-term fasting, which may, in turn, affect their per-formance. There is evidence that there is a curvilinear

.-�--------� relationship between the level ofarousal and the quality

of performance (43) with optimal performance at mod-‘4 crate levels of arousal and poorer performances at both

low and high levels. Clearly, the association is complexand performance depends not only on the level ofarousal but also on the type and difficulty ofthe task and

_)?es no nature of the subjects (44). It is possible that malnour-ished children may have levels of arousal different fromthose ofthe control children.

Because the previous reports of missing breakfast in-

0 �S volved only adequately nourished children, comparison- of their results with those ofthis study are limited to the

O�3O results of our control group. Even this group may nothave been adequately nourished because their mean per-

0#{149}15 cent expected weight-for-height was only 93.6 (Table 3).

� One report from England (1 5) indicated that missing

� 0 00 breakfast had no effect on the performance of arithmetic,

-0�15 short-term memory, or attention. In contrast, there wasa deterioration in problem-solving ability and improve-

z -O#{149}30 ments in short-term memory and attention to task-irrele-� vant information in the North American studies(16, 17).x -0 45 We were unable to confirm any of the findings of these

latter two studies although they had designs similar toours. The reasons for the different findings are not clear.

-075 #{149} � no yes no Differences in subjects however are associated with

BREAKFAST BREAKFAST differences in arousal levels (44), and the social and eco-

FIG 3. Mean scores by group by wasting in digit-span-forwards and nomic backgrounds ofthe control children in the presentMFFT efficiency. #{149}- #{149},combined malnourished and wasted, n study would certainly have been more depnved than= 20; X - X, combined malnourished nonwasted, n = 40; those of the children in the other reports. This factor. - - - ., control group and wasted, n = 10; X - - - X, control group could lead to differences in arousal and hence perfor-nonwasted, n = 20. mance.

Because school meals are usually served at lunch time,there is a need to investigate whether the effects of lunch

is difficult to understand why the wasted children of the are the same as those of breakfast in undernourishedmalnourished groups were not similarly affected. children. In well-nourished adults they are not (45). The

The children’s performances on the HCI task and lis- timing ofschool meals may therefore be important. Car-

tening comprehension tests were not affected when they tying out the experiment in a hospital ward raises ques-missed breakfast. This may be because ofthe lack ofsen- tions ofecological validity (46) and the study needs to besitivity of these tests in the sample of children. Alterna- replicated in the school situation.tively, it is possible that the omission ofbreakfast in this In conclusion, under controlled conditions previouslysample was not sufficient to affect the functions mea- severely malnourished and stunted children and thosesured by these tests. who were wasted were adversely affected in a number

This study did not investigate the mechanisms by of cognitive functions when they missed breakfast. Thiswhich missing breakfast may have affected cognitive contrasted with children who were neither stunted norfunctions. However variations in metabolic stress and wasted and were not adversely affected. If the effectsarousal state are two mechanisms that may be responsi- found here are replicated and found to continue over able either singularly or in combination and should be cx- period oftime, then missing breakfast could be a seriousplored in follow-up studies. Carbohydrate metabolism contributor to poor school achievement in undernour-

652 SIMEON AND GRANTHAM-MCGREGOR

0LI

at MIT

LIBR

AR

IES

on July 10, 2011w

ww

.ajcn.orgD

ownloaded from

BREAKFAST AND CHILDREN’S MENTAL FUNCTION 653

ished children. Presently the indication is that where re-

sources are limited, school meals should be targeted to

undernourished children who are more likely than ade-quately nourished children to benefit in school-achieve-

ment levels.

We sincerely thank Mrs Patricia Fletcher for the testing of the chil-

dren, Mr William Schofield for statistical advice, Prof Ernesto Pollittfor his advice, and the Nutrition Products Limited ofJamaica for thebreakfasts used in the study.

References

1. Read M. Malnutrition, hunger and behavior. II Hunger, schoolfeedingprograms and behavior. J Am Diet Assoc 1973;63:386-91.

2. Pollitt E, Gersovitz M, Gargiulo M. Educational benefits of theUnited States schoolfeeding program: a critical review ofthe litera-ture. Am J Public Health 1978;68:477-8l.

3. Dickie N, Bender A. Breakfast and performance. Hum Nutr ApplNutr 1982; 36A:46-56.

4. Tisdall F, Robertson E, Drake T, et al. The Canadian Red Crossschool meal study. Can Med Assoc J 195 1;64:477-89.

S. Liberman H, Hunt I, Coulson A, Clark V, Swendseid M, Ho L.Evaluation ofa gheuo school breakfast program. J Am Diet Assoc1976;68: 132-8.

6. Baker I, Elwood P, Hughes J, Jones M, Sweetnam P. School milkand growth in primary school children. Lancet 1978;2:575.

7. Coyne T, Dowling M, Condon-Paoloni D. Evaluation of preschoolmeals programs on the nutritional health of aboriginal children.Med J Aust 1980;2:369-75.

8. Rewal S. Results of a school lunch program in India. Food NutrBull 1981;3:42-7.

9. Paige D, Cordano A, Huang S. Nutritional supplementation of dis-advantaged elementary school children. Pediatrics 1976;58:697-703.

10. Powell C, Grantham-McGregor 5, Elston M. An evaluation of giv-ingthe Jamaican school meal to a class ofchildren. Hum NutrClin

Nutr 1983; 37C:381-8.

1 1. Gietzen D, Vermeersch J. Health status and school achievementofchildren from head start and free school lunch programs. Public

Health Rep 1980;95:362-8.12. Keister M. Relation ofmid-morning feeding to behavior of nursery

school children. J Am Diet Assoc l950;26:25-9.13. Laird D, Levitan M, Wilson V. Nervousness in school children as

related to hunger and diet. Med J Rec 193 1 ; 134:494-9.14. Dwyer T, Elias M, Warram J, Stare F. Effects of a school snack

program on certain aspects of school performance. Fed Proc

1972;3 1:7 18(abstr).15. Dickie N, BenderA. Breakfast and performance in schoolchildren.

BrJ Nutr 1982;48:483-96.16. Pollitt E, Leibel R, Greenfield D. Brieffasting, stress and cognition

in children. Am J Clin Nutr l981;34:1526-33.

17. Pollitt E, Lewis N, Garcia C, Shulman R. Fasting and cognitivefunctioning. J Psychiat Res 1983; 17:169-74.

18. Birch H, Pineiro L, Alcalde E, Toca T, Cravioto J. Relation ofkwashiorkor in early childhood and intelligence at schoolage. Pedi-atr Res 197 1;S:579-85.

19. Hertzig M, Birch H, Richardson 5, Tizard J. Intellectual levels ofschoolchildren severely malnourished during the first two years oflife. Pediatrics 1972;49:8l4-24.

20. Hoorweg J, Stanfield J. The effects ofprotein energy malnutritionin early childhood on intellectual and motor abilities in later child-hood and adolescence. Dcv Med Child Neurol 1976; 18:330-50.

2 1 . Richardson S. The long range consequences of malnutrition in in-

fancy: a study ofchildren in Jamaica WI. In: Wharton B, ed. Topics

in pediatrics 2: nutrition in childhood. Tunbridge Wells: PitmanMedical Pub, 1980:163-76.

22. Galler J, Ramsey F, Solimano G. The influence ofearly malnutri-

tion on subsequent behavioral development. III Learning disabili-ties as a sequel to malnutrition. Pediatr Res 1984; 18:309-13.

23. Galler J, Ramsey F, Solimano G, Lowell W. The influence of early

malnutrition on subsequent behavioral development. II Classroom

. behaviour. J Am Acad Child Psychiatry 1983;22: 16-22.

24. Waterlow J. Classification and definition ofprotein-energy malnu-trition. In: Beaton G, Bengoa J, eds. Nutrition in preventive mcdi-cine. Geneva: World Health Organization, 1976:530-55. (WHO

monograph series 62.)

25. Keller W, Fillmore C. Prevalence ofprotein-energy malnutrition.World Health Stat Q 1983;36:129-67.

26. United States Department ofHealth, Education and Welfare, Pub-lic Health Service, Health Resources Administration. NCHSgrowth charts. Rockville, MD: Health Resources Administration,

1976.

27. Wechsler D. Wechsler intelligence scale for children-revised.New York: The Psychological Corporation, 1974.

28. Kaufman A. Intelligent testing with the WISC-R. New York: JohnWiley and Sons, 1979.

29. Semel E, Wiig E. Clinical evaluation of language functions. Co-

lumbus, OH: Charles Merrill, 1980.

30. Kagan J, Rosman B, Day D, Albert J, Phillips W. Information pro-

cessing in the child: significance ofanalytic and reflective attitudes.Psychol Monogr 1964;78:578-615.

31. Hagen J. The effect ofdistraction on selective attention. Child Dcv1967;38:685-94.

32. Salkind N, Wright J. The development of reflection-impulsivityand cognitive efficiency. Hum Dcv 1977;20:377-87.

33. Cronbach L. Coefficient alpha and the internal structure of tests.

Psychometrika 1951; 16:297-334.

34. Powell C, Grantham-McGregor S. The ecology of nutritional sta-

tus and development in young children in Kingston, Jamaica. Am

J Clin Nutr l985;4l:1322-31.35. Dunn L, Dunn L. Peabody picture vocabulary test-revised. Cir-

dc Pines, MN: American Guidance Service, 1981.

36. Ferguson H, Stodthrt C, Simeon J. Double-blind challenge studiesof behavioral and cognitive effects of sucrose-aspartame ingestionin normal children. Nutr Rev 1986;44(suppl):144-S0.

37. Jelliffe D. The assessment ofthe nutritional status ofthe commu-nity. Geneva: World Health Organization, 1966.

38. Armitage P, Berry G. Statistical methods in medical research. Ox-

ford: Blackwell Scientific, 1987.

39. Alleyne G, Trust P, Bores H, Robinson H. Glucose tolerance andinsulin sensitivity in malnourished children. Br J Nutr 1972;27:585-92.

40. James W, Coore H. Persistent impairment ofinsulin secretion andglucose tolerance after malnutrition. Am J Clin Nutr 1970;23:

386-9.41. Cook G. Glucose tolerance after kwashiorkor. Nature l970;2l5:

1295-6.

42. Becker D, Pimstone B, Hansen J, Hendricks S. Insulin secretionin protein-calorie malnutrition. I Quantitative abnormalities and

response to treatment. Diabetes 1971;20:542-S1.43. Kahneman D. Attention and effort. New Jersey: Prentice-Hall Inc,

1973:28-49.

44. Eysenck M. Arousal learning and memory. Psychol Bull 1976;83:389-404.

45. Craig A. Acute effects of meals on perceptual and cognitive effi-

ciency. Nutr Rev 1986;44(suppl): 163-71.

46. Bronfenbrenner U. The ecology of human development. Cam-bridge, MA: Harvard University Press, 1979.

at MIT

LIBR

AR

IES

on July 10, 2011w

ww

.ajcn.orgD

ownloaded from