Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Effects of breakfast on classroom behaviour in rural Jamaican schoolchildren

Susan M. Chang, Susan P. Walker, John Himes, and Sally M. Grantham-McGregor



The effects of giving breakfast on classroom behaviour were examined in 57 undernourished (<= - 1 SD weight-for-age) and 56 adequately nourished (>-I SD weight-forage) children, selected from four rural Jamaican schools. Using a time-sampling method of observation, the children's behaviour was observed twice, once after receiving breakfast and once after receiving a piece of fruit. The impact of breakfast varied among the schools but not between nutritional groups. In the school that was best equipped and organized, the children were more attentive (p <.005) and moved less (p <.05) when they received breakfast than when they had no breakfast. In the other three schools there was no improvement; in two of these schools, the children were less on task when given breakfast (p <.02 and p < .01), and they talked more in one school (p <.05). This suggests that school breakfast may only benefit children's behaviour in the presence of satisfactory classroom infrastructure.


The academic achievement levels of schoolchildren in developing countries are far lower than those of children in industrialized countries [1]. There is increasing concern that children's poor health and nutrition may detrimentally affect their ability to learn in school and subsequently lead to low achievement levels [2, 3]. In particular, undernutrition and poor dietary intakes have been found to be associated with poor attainment levels in several studies [4]. Many countries have implemented school feeding programmes in the expectation that the children's school performance will benefit [5]. However, few programmes have been evaluated. Early studies often lacked scientific rigour, but more recent studies indicate that benefits to school achievement and increased school attendance may occur [68].

There are various mechanisms by which school meals may affect children's school learning and behaviour. It has been theorized that the consumption of meals may meet the specific nutrient needs of the child. For example, if the meals provide iron, the achievement of iron-deficient children may improve [9,10]. Meals may also improve the children's nutritional status, especially in those who were previously malnourished. Numerous studies have shown that school-aged children who had protein-energy malnutrition in early childhood have poor IQ levels and school achievement and greater behavioural problems than children who were not malnourished [11]. There is also increasing evidence that missing breakfast or fasting has negative effects on children's cognition [12,13]. The results of these studies imply that by providing meals to these children the effects of short-term hunger would be reduced. Further, in a small school trial we have shown benefits of breakfast to children's arithmetic scores even though their nutritional status did not improve [6]. This suggests that the benefits may have been due to the relief of hunger.

The outcome variables of most school feeding evaluations have been school achievement, school attendance, dietary intake, and nutritional status measured by anthropometry. Spending sufficient time on task is an essential part of successful learning [14]. It is possible that paying more attention to task and other improvements in classroom behaviour may mediate the benefits of school meals on achievement. Associations have been found between teachers' ratings of children's attention-restlessness and interest-participation in the first year of school and their school grades and test performance in later years [15]. There is, however, little information concerning the extent to which classroom behaviour is affected by providing school meals.

In two early studies, Laird and colleagues [16] and Keister [17] found that mid-morning snacks reduced children's negative behaviour in the classroom. Unfortunately the behaviours were poorly defined. Benton and colleagues [18] found that children who were given a glucose drink were more attentive and showed fewer signs of frustration in an experimental task. However, the results may not be the same in the classroom situation. All three studies were conducted in developed countries, and the findings may be different in children in developing countries, who are more likely to be undernourished.

There is some evidence that the classroom behaviour of undernourished children differs from that of adequately nourished children. In two studies [19, 20] teachers reported that children who had been severely malnourished in early childhood had shorter attention spans and were more distracted and restless than matched controls.

Teachers' assessments may be biased by the children's appearance, socio-economic status, and school grades. A more valid picture of the children's behaviour is obtained by direct observation. Sigman and colleagues [21 ] in Kenya observed that girls with better dietary intake were more attentive in class than girls with poorer intakes. In contrast, in Egypt boys with higher iron intake were observed to be more off task in the classroom [22]. Associations between behaviour and nutritional status may, however, be confounded by poverty. For example, children from impoverished homes may behave differently in school because of their poor environment rather than their poor nutritional status. It is easier to infer causal relationships from nutritional interventions. In a recent report, McDonald and colleagues [23] found that Kenyan children became more off task during a famine. However, other stressors may have played a role. We are unaware of any studies of the effects of school meals on children's classroom behaviour in developing countries.

In the present study we compared the short-term effects of providing school breakfast on the cognitive function and classroom behaviour of undernourished and adequately nourished children. We previously reported that the verbal fluency performance of the undernourished children improved when they received breakfast, but that the performance of the adequately nourished children was not affected [24]. In this paper we report observations of classroom behaviour in a subsample of the children. We hypothesized that the children's behaviour would improve when they had breakfast: specifically, that they would spend more time on task, talk less to their classmates, participate more in teaching activities, and move less around the classroom.


Study schools

The study was conducted in four government primary schools (SV, AR, GH, and Ml) situated in a remote mountainous area of Jamaica. They were within 15 kilometers of each other and served similar poor farming communities. These schools were chosen because reliable transport was generally unavailable and many of the children had to walk some distance to school. Therefore it is likely that even those children who had breakfast at home would be hungry on arrival at school. In addition, at least 25% of the children had weights below -1 SD of the National Center for Health Statistics (NCHS) references [25]. There is limited evidence suggesting that children in rural schools are more likely to benefit from school feeding [5]. In Jamaica most children are enrolled in school, but attendance is often irregular.



For the study of the children's cognitive functions [24], two groups of children aged 8 to 11 years were identified from grades 3 and 4 from the four schools. One hundred children, consisting of all those whose weights-for-age were equal to or less than -1 SD of the NCHS references [25], formed the undernourished group. This definition of undernutrition was the lowest at which we could identify sufficient children. However, most of the children were only mildly undernourished. The children were generally not wasted. Twenty-one percent of the children in the undernourished group had body mass index (BMI) values below the 5th percentile of reference values [26], and their low weight-for-age was primarily due to low height-for-age. For each undernourished child, an adequately nourished child (weight-forage greater than -1 SD of the NCHS reference) of the same sex was selected from the class. To avoid missing data, only children who had attended school for at least 50% of the days in the previous term were enrolled in the study. One child with an obvious physical impairment and those children whose academic abilities were reported by their teachers to be substantially delayed were excluded. For the observation of classroom behaviour, a subsample of 60 children from each group was selected from the original sample.


Observation instrument

Our main interest was in the time the children spent paying attention to school work (on task). However, during the pilot phase we observed that the tasks given to the children were of short duration and sometimes more than one task was set in the same class. We therefore structured the situations in each class to facilitate collecting sufficient data to determine differences between children. The teachers were asked to actively teach the children for two 30-minute sessions in the morning and to set them two tasks, either from the blackboard or their textbooks, that required them to work independently for 30 minutes each. No other instructions were given, and the teachers chose the topic and method of teaching. Extending the duration of the tasks may have somewhat modified the children's behaviour, but this is unlikely because neither the type of task set nor the teaching style was changed. The crossover design of the study meant that the children were exposed to the same classroom situations whether or not they had breakfast.

The schedule was 30 minutes of teaching, which usually began around 9 a.m., followed by 30 minutes of a set task. The schedule was then repeated. Each morning, two subjects were observed at the same time to prevent the observer from constantly looking at one child. It also allowed for the children's behaviour to be observed over a longer period of time. A time-sampling method of observation was used. One subject was observed during a 10-second period, and the child's behaviour was recorded in the following 10 seconds on pre-coded sheets. The behaviour of the second subject was then observed and recorded. Each child was therefore observed for 180 10-second periods each morning, half the periods during teaching and half during set task situations. They were observed on two days when they had breakfast and on two days when they did not.

The behaviours recorded were attention to task (ON TASK), talking to another child (TALKS), gross motor movements (GROSS MOTOR), and participation in the class (PARTICIPATE/RESPONSE) (see the Appendix for definitions). A child was recorded as being on task for a 10-second observation period if he or she was observed to be on task for at least 5 seconds during that period. For the remaining behaviours only the initiation of the behaviour was recorded, yielding an estimate of the frequency of these behaviours.

The observers were four university graduates. Before the study, reliability between pairs of observers was assessed by the Cohen kappa statistic [27]. The kappa values for all behaviours were greater than .90 in each of 10 consecutive children for each pair. To record reliability values throughout the study, a further 20 children were observed by pairs of observers, and the kappa values were greater than .89 for all the behaviours.

To determine the reliability of the observation instrument over time, the two 30-minute periods of observation from each type of situation were combined for each day, and Pearson product-moment correlations were calculated between values on adjacent days. The Spearman-Brown formula was then used to estimate the reliability of the mean of two days combined [28]. These are shown in table 1. In the teaching situation all R values were at least .60. except for gross motor behaviour, which had an R of .55, and in the set task situation all R values were at least .60.

TABLE 1. Reliability of behaviours observed on day 1 versus day 2, adjusted to estimate reliability of the mean of the two days, for teaching and set task situations (n = 113)

Behaviour Situation
Teaching Set task
ON TASK .78 .74
TALKS .60 .69

All p < .001.
a. K=2r/1+r[28]. b. Not measured in set task situation.


Socio-economic status

A questionnaire was administered to the children to evaluate their socio-economic status. A housing score was calculated from the type of sanitation facility and water supply, the number of people who shared the index child's bed, and whether the family owned any of four household items (radio, television, refrigerator, and gas stove). The book score was the total number of exercise books and textbooks that the interviewer counted in the child's bag on the day of the interview. The quality of the children's uniforms and shoes was rated on a three-point scale (poor, fair. good) by the interviewers; interobserver agreement was greater than 90%. The details of the scores are given in table 2. The three indices have been used previously in other studies to assess the socioeconomic status of Jamaican schoolchildren and have been found to have good reliability and to be significantly associated with school achievement [29, 30]. Ten percent of the children's homes were visited and the validity of their responses was good [24].

TABLE 2. Socio-economic scores calculated from the social questionnaire

Rating Components Score
Book score textbooks and exercise books total number
Housing score (0- 8) type of toilet 0 = pit/none
1 = outside
2 = inside
type of water 0 = river
1 = outside
2 = inside
no. of persons per child's bed 0 = 3-6 persons
1 = 2 persons
2 = 1 person
no. of household possessions (radio, TV, refrigerator, stove) 0 = 0-1 item
1 = 2 items
2 = 3-4 items
Uniform score (0-4) clothing condition 0 = poor
1 = fair
2 = good
shoe condition 0 = poor
1 = fair
2 = good



The study was an experimental one with a crossover design, in which each child was compared with himself or herself with and without breakfast. Since it was not possible to feed some children in a class and not others, children were assigned to treatment by class. At the beginning of the study each class within each school was randomly assigned to receive breakfast or a quarter of an orange for each child, given to the children in their classrooms before the teaching began. The nutritional composition of the breakfast and fruit is given in table 3. After receiving treatment for at least one week, all subjects were observed for two consecutive mornings. After an interval of at least two weeks, the type of treatment was reversed and the subjects were observed for another two days after at least one week of treatment.

Table 3. Nutritional composition of breakfast and fruit

Meal Composition
Weight (g) Energy (kcal) Protein (g)
2 slices bread 68 183 5.9
1 slice processed cheese 28 108 7.0
1 cup chocolate milk 227 229 8.4
Total   520 21.3
Fruit (placebo) 1/4 orange 68 18 0.4



The data for each behaviour for the two days in each treatment condition (breakfast or fruit) were summed for the set task and teaching situations. If the resulting totals were not normally distributed, appropriate transformations were used [31].

Since teacher-student interactions occurred almost exclusively in the teaching situation, "participation" was analysed only in this situation. Because the students usually stood up to speak to the teacher, gross motor behaviour was adjusted for participation in the teaching situation. A linear regression of gross motor behaviours on response was calculated, and the residuals were used as a measure of gross motor behaviour independent of response. To examine the associations between the anthropometric measures and the behaviours, the behaviours were summed over the four days and Pearson product correlation coefficients were calculated. Treatment effects were determined using repeated measures analyses of variance for the teaching and set task situations separately. The within-subject factor was behaviour with and without breakfast. The between-subject factors were school, sex, and nutritional group. The analyses were also conducted with age as a covariate. Since age was not significant in any of the analyses and the findings were not altered, the analyses are reported without age.

Where there were significant interactions with treatment, post-analysis of variance comparisons were calculated to determine the effect of treatment on the behaviours. All statistical analyses were performed with the Statistical Package for the Social Sciences (SPSS Inc., Chicago, Illinois, USA).

Contents - Previous - Next