rose net al 1988
Post on 04-Jun-2018
219 Views
Preview:
TRANSCRIPT
-
8/13/2019 Rose Net Al 1988
1/7
Journal of Consulting and Ginical Psychology1988, Vol. 56, No. 4, 583-589
Copyright 1988 by the American Psychological Association, Inc.0022-Q06X/88/J00.75
Effects of Sugar (Sucrose) on Children's Behavior
Lee A. RosenColorado State University
Mary E. BenderDepartment of Psychiatry
University ofCaliforniaIrvine
Sue SorrellEarly Education Center Outreach Project
Jackson, Mississippi
Sharon R. Booth
Virginia Polytechnic and State University
Melanie L. McGrathDepartment of Psychiatry and Human Behavior
University of Mississippi Medical Center
Ronald S. DrabmanDepartment of Psychiatry and Human Behavior
University of Mississippi Medical Center
We examined the effects of sugar on the behavior of 45 preschool and elementary school children.
Using a double-blind within-subject challenge design, we provided all children with a basic breakfast
that included a challenge drink containing either 50 gof sucrose, a placebo (aspartame) of compara-
ble sweetness, or only a very small amount of sucrose. The results indicated that high amounts of
sugar caused a small increase in the children' s activity level (as rated by their teachers) and a small
decrement in the performance of the female subjects on a simple learning task and that sugar affectedthe cognitive performance of the preschoolers differently than that of elementary school children.
All of these effects, however, were quite small in magnitude and were not considered clinically sig-
nificant. The results did not support the view that sugar causes major changes in children's behavior.
Parents and teachers often report that foods containing large
amounts of refined sugar (sucrose) produce detrimental effectson children's behavior. Anecdotal reports of increased activitylevels, irritability, and impairment in the ability to sustain at-tention and inhibit impulses are common. In addition, several
articles and books written for the general public have popular-
ized the view that sugar exerts negative effects (e.g., Duffy.
1975;Tauraso, 1983). Many of these have stated that sugar is a pri-mary cause of attention-deficit hyperactivity disorder as it is de-
fined in the Diagnostic and Statistical Manual of Mental Disor-ders (DSM-III-R; American Psychiatric Association, 1987) inchildren. In fact, according to a study by Bennett and Sherman
(1983), 45% of pediatricians and family practitioners have rec-ommended a low-sugar diet for at least some of the hyperactivechildren they treat. Scholarly writings have also made un-founded pronouncements regarding sugar's effects. For in-stance, in a comment in the American Psychologist, Buchanan
(1984) labeled sugar the most ubiquitous toxin. Only re-
cently, however, have investigators begun to seriously examine
this important issue.Unfortunately, the few controlled experiments that have been
conducted have reported conflicting results. Some have showndetrimental effects of sugar on behavior (e.g., Conners &Blouin, 1982; Conners etal., 1985; Goldman, Lerman, Contois,
& Udall, 1986), whereas others have shown either salutary
The authors wish to thank the principals and staff of the Christ the
King Elementary School and of the Broadmoor Kindergarten and Pre-
school for their patience and cooperation.
Correspondence concerning this article should be addressed to Lee
A. Rosen, Department of Psychology, Colorado State University, Fort
Collins, Colorado 80523 or to Ronald S. Drabman. Division of Psychol-
ogy, University of Mississippi Medical Center. 2500 North State Street,Jackson. Mississippi 39216-4505.
effects (e.g., Behar, Rapoport, Adams, Berg, & Cornblath, 1984;
Conners et al., 1985) or no effects (e.g., Ferguson, Stoddart, &Simeon, 1986; Milich Pelham, 1986; Wolraich, Milich,Stumbo, & Schultz, 1985). Interestingly, these findings appear
to bear no relation to the types of children studied, to the mea-sures collected, or to the amounts of sugar ingested.
Given the current state of knowledge in this area, firm con-clusions regarding sugar's effects on children's behavior areclearly premature. Thus, the present investigation further eval-
uated the effects of sugar on children's behavior.
Method
Subjects
The experiment was conducted at two private, church-run schools:
a preschool and an elementary school. At the preschool, 30 children
participated (20 boys and 10 girls, mean age = 5 years, 4 months) who
were white and from middle- to upper-middle-class families. Fifteen
elementary school children also part icipated (6 boys and 9 girls, mean
age = 7 years, 2 months) who were first- and second-grade (regular edu-
cation) Black children from predominately working class families.None of the children had been identified for mental deficiency; however.
2 of the preschool boys were reported (by the preschool director) to be
hyperactive, and I of these was being treated with Ritalin.
Procedure
Pretesting. Prior to the onset of the experiment, each child's mother
and teacher were asked to indicate whether they thought the child was
behaviorally sensitive to sugar. Nineteen of the 30 preschool mothers
indicated that they thought their child was behaviorally responsive to
sugar, and 1 1 of the 30 preschoolers were identified by their teachers as
possible responders. Eight of the 15 elementary school children were
identified by their mothers as behaviorally responsive 10 sugar, and 7 of
the 15 were identified as such by their teachers. In addition, each child's
teacher rated the child using the Conners Teacher Rating Scale (CTRS;
583
-
8/13/2019 Rose Net Al 1988
2/7
584 ROS ET AL
Conners, 1969). Seven preschool children and 1 elementary school
child scored above 15 on the Hyperactivity Index of this scale, which is
a cutoff traditionally used as an indicator of attention deficit disorder.
Experimental conditions. Three conditions were manipulated in this
investigation: a high-sugar condition, a low-sugar condition, and a con-
trol aspartame condition (low in sugar but with a sweet taste). The ex-
periment was conducted on 15 consecutive school days at each site, with5 days spent on each condition. The specific condition for any particular
day was randomly selected for each child (i.e., the 5 days of each condi-
tion that each child received were not necessarily 5 consecutive days)
and was independent from the conditions selected for the other subjects.
Double-blind control was observed throughout the experiment by keep-
ing experimenters, teachers, and parents unaware of the children's con-
ditions. In addition, chi ldren were kept unaware of the conditions
through the use of the sweet-tasting (but low in sugar) control condition.
These conditions were delivered by manipulati ng the amount of sugar
the children received in a breakfast they were fed at their schools (the
children fasted from bedtime the night before). 1 The breakfast each
child received is outlined in Table 1. The amount of carbohydrates (su-
crose, other simple sugars, starches, etc.) in the basic breakfast totaled
40.75 g. In addition, each child received 4 oz of an orange-flavored
drink. The amount of sugar each child received vvas manipulated by
varying the amount of sucrose in this breakfast drink from 0 to 50 g of
sucrose (approximately equal to the sucrose in 2 candy bars). The high-
sugar drink contained 50 g of sucrose; the Ion-sugar drink contained
6.25 g of sucrose (just enough to make it palatable): and the sweet-but-
low control drink contained no sucrose (see Table I). The total amount
of carbohydrates and other nutrients contained in the entire breakfast
(basic breakfast plus drink) for each condition is also displayed in Table
1. The high-sugar breakfast contained 90.75 g of carbohydrates (mostly
sucrose and other simple sugars); the low-sugar breakfast contained
47.0 g, and the sweet-but-low control breakfast contained 40.75 g. This
control condition was made possible through the use of an aspartame-
sweetened (Nutrasweet) drink. The intensity of sweetness for the two
sweet-tasting dri nks was equated so that the 4-oz sweet-but-low break-
fast drink contained 122 mg of aspartame. Because children sometimesfailed to eat their entire breakfasts (although virtually all children did
consume their drinks), the amount of each food stuff that was actually
ingested was measured for each child following each breakfast by sub-
tracting the amount of food left from the original amount served. This
value was used to determine the ingested dose of sucrose (and the total
amount of carbohydrate ingested) for each child for each day. These
data are presented in the Results section.
Cognitive measures. On 2 days (determined randomly )of each condi-
tion, approximately 20-30 min following the completion of breakfast,
children were tested on several measures sensitive to cognitive function-
ing. 2 First, children completed a paired-associate learning task in which
they were asked to learn and then to recall a number of animal names
paired with certain zoo numbers. This procedure has been advocated
by Swanson and Kinsbourne (1976,1979), and our use of the procedure
in this context was consistent with their instructions (i.e., for pretesting,
for the use of alternate forms, etc.). The number of errors in recall was
scored for each of the two administrations conducted dur ing each con-
dition and was averaged to derive a mean score on this task. Second,
children worked on a matching task in which each child was asked to
indicate the one picture of a set of six that exactly matched a model
picture. This task was similar to the traditional Matching Familiar Fig-
ures Test (Kagan, Rosman, Day. Albert. & Phillips. 1964) and used the
same sti mulus items and procedures (except that only 3 items were used
during each administration instead of the traditional 12). This allowed
us to use alternate forms so that no child viewed any particular item
more than once. Both latency to first' response and number of errors
were scored for each administration and were averaged to derive a single
mean score on this measure for each condition. Third, dur ing each day
of each condition, children completed various product-measure testsdesigned to measure thei r performance on simple academic tasks. The
Table 1Nutrient Content of Breakfast
Food Protein Fat Carbohydrate Calories
Basic BreakfastToasted oats
(7oz)
Whole milk(6oz)
Toasted bread( slice)
Margarine(Itsp)
Grape jelly
( tsp)Total
Drink (4 oz)Low sugarControl
High sugarTotal nutrient
contentLow sugarControl
High sugar
2.8 g
6.4 g
2.0 g
0.0 g
I I . 2g
0.0 g
0.0 g
0.0 g
l l . 2 g11.2g
l l . 2 g
l.4g
6.5 g
1-Og
2.3 g
11.2g
0.0 g
0.0 g
0.0 g
l l . 2 g11.2g
l l . 2 g
14.0 g
8.25 g
14.5 g
0.0 g
4.0 g40.75 g
6.25 g0.0 g
50.0 g
47.0 g40.75 g
90.75 g
77.0
119.0
75.0
0.2
15.6286.8
27.04.0
202.0
313.8290.8
488.8
Note. Dashes indicate that only trace amounts were present.
preschool children completed a tracing task that was scored for the
mean number of correctly traced items for each child for each condi-
tion. This task consisted of several work sheets, more than the child
could possibly complete, on which the child was told to trace as many
letters (As and Os) as possible. The elementary school children worked
simple arithmet ic worksheets that were scored for the mean number
of problems completed and the mean accuracy for each child for each
condition. Each child was told to work as quickly and accurately as pos-sible. These worksheets were geared to the individual child's academic
level.
Teacher ratings. Each day teachers completed the Abbreviated Con-
ners Teacher Rating Scale (ACTRS) and a 10-point global rating scale.
These scales were completed immediately preceding the child's lunch
time (or morning snack) and were to reflect the child's behavior for the
entire morning. The ACTRS is a 10-item checklist that is sensitive to
hyperactive behavior (Conners, 1973). This measure was used by Behar
et al. (1984) in their study of sugar's effects, and a variant of it was used
by Ferguson et al. (1986). Scores on this measure range from 0 to 30,
with a score of 15 or above generally accepted as one sign of attention
deficit disorder (Ross & Ross, 1982). The ACTRS has been used widely
in studies examining the effects of food additives and stimulant medica-
tion on attention deficit disorder children (Rumsey & Rapoport, 1983).
The global rating scale is a 1-10-point Likert-type scale on which the
respondent rates any global changes in the child's behavior. For exam-
ple, the anchor for a score of I is skw, lethargic, very inactive and quiet.
' Parents were carefully instructed not to feed their children before
school during the period that the experiment was in progress, and each
child was regularly quizzed regarding prior ingestion of food that morn-
ing. If the child stated that she or he had eaten that morning, the parents
were contacted to verify this information and the child was rescheduled
for testing. This, however, occurred rarely.
- The 20-30-min delay between the completion of breakfast and the
onset of cognitive testing was similar to the delay used in other studies
that have examined the effects of sugar on children's behavior (e.g., Wol-
raich et al., 1985) and is the length of time needed for most children's
blood sugar levels to peak (Pickens, Burkeholder, & Womack, 1967).
-
8/13/2019 Rose Net Al 1988
3/7
EFFECTS OF SUGAR 8
Table 2
Mean Dose of Carbohydrate by Site
Carbohydrate dose
Drink Breakfast'
Site SD SD
PreschoolLow sugarControlHigh sugar
Elementary schoolLow sugarControlHigh sugar
OverallLow sugarControlHigh sugar
0.3102.49
0.2301.81
0.2802.26
0.0500.43
0.0900.67
0.0600.51
2.051.724.11
1.541.193.63
1.881.543.95
0.390.360.67
0.480.371.00
0.420.360.78
a Sucrose plus other carbohydrates. b n = 20. Mean weight = 20.0 kg
(SD = 3.5). n = 15 . Mean weight = 27.6 kg (SD = 9.9). n = 45.Mean weight = 22.5 kg (SD = 5.6).
and the anchor for a score of 10 is extremely active, constantly moving,loud or disruptive Scales of this type (e.g., theGlobal Impressions Scale)have been advocated for use in studies of the behavioral effects of dieton children (Rumsey Rapoport, 1983).
Observational measures. Observations were scheduled for each childfor each day of the experiment. Children were observed after their cog-nitive testing was completed but not before '/ hr following breakfast.
Children were observed in a variety of sites, and for this reason d ifferentbehaviors were targeted for observation depending on the site, with theexception of fidgeting, a behavior thai was common to all sites. At the
preschool, children were observed during a freeplay period and, thus,the following behaviors were scored: a] fidget, defined as any repetitivemotor act not part of an approved task (e.g., foot tapping, scratchingface, tapping fingers); (b) change in activity, defined as movement from
one a c t i v i t y area to another (e.g., from block area to art area): (c) activemovement, defined as gross body movements propelling Ihe entire body
(e.g., walking, running, jumping): (d) vocalization, defined as any audi-
ble sound from the child's mouth (e.g., talking, whistling, singing); and(e) aggression, defined as any intense movement toward another person,either directly or by using a material object as an extension of the hand(e.g., hitting, throwing rock) that results in contact. At the elementaryschool, the children were observed during in-seat independent work
time. The behaviors observed at this site were (a) fidget and (b) on-task
behaviors, defined as those appropriate to the classroom task assignedby the teacher (e.g.. sitting and working quietly, watching the teacher
give group instructions). Observational data were collected by severalobservers watching the children each morning of the experiment. Ob-servations were broken into continuous 10-s intervals, with each childobserved 2-4 min each day (for a total of 10-20 min for each condition).If any of the target behaviors occurred during the 10-s interval, the en-tire interval was scored as an instance of that behavior (with the excep-tion of on-task behavior, for which an interval was scored as on-task onlyif no instance of off-task behavior was observed during that interval).
Reliability. Reliability checks were conducted on each observer foreach type of behavior observed. An occurrence reliability score for eachbehavior was derived by dividing the number of agreements on the oc-currence of a particular behavior by the number of these agreementsplus the number of disagreements. In addition, when it was possible, a
Cohen's kappa coefficient was calculated to correct for chance agree-ments (see Kent & Foster, 1977). At the preschool site, reliability on thefidget code was collected on 50i of the fidget data, yielding an occur-rence reliability of 89% (range = 63-100; = .93). Reliability was as-sessed on 46% of the data for change in activity, active movement, vocal-ization, and aggression. The occurrence reliabilities for these behaviors
were 99% (range = 92-100; = not computable), 88% (range = 83-100; = .88), 82% (range = 66-100: = .82), and 94% (range = 83-100;K =not computable), respectively. At the elementary school site, 40 o of thefidget data were checked for reliability, yielding an 84% (range = 70-95; = .79) occurrence reliability; 80% of the on-task data were checked,yielding an 8 5 < S > (range = 67-99; = .82) occurrence reliability.
Results
Dose Levels
To determine the differences in sucrose intake between the
high, low. and control conditions, the amount of sucrose in the
breakfast drink that each child ingested each morning was di-
Table 3
Means and Standard Deviations by School, Sex, and Condition
Boys
Low sugar
Measure
PreschoolPAL errorsMT latencyMT errorsTracing completed
Elementary school 6
PAL errorsMT latencyMT errorsMath completedMath accuracy
M
14.05.86.2
96.6
10.85.86.3
31.694.8
SD
6.92.92.7
57.1
5.12.12.35.0
10.3
Control
A/
14.65.36.5
89.3
11.4
6.95.8
35.795.0
SD
6.41.91.9
56.3
4.14.22.07.87.0
High sugar
A/
13.35.46.9
108.5
8.66.94.9
32.895.8
SD
5.02.62.6
68.3
2.92.62.29.95.7
Low sugar
M
11.96.27.2
82.6
9.77.85.5
51.992.3
SD
6.64.62.2
45.6
4.43.02.0
25.815.0
Girls
Control
.(/
13.85.57.2
88.5
9.68.75.4
52.691.3
SD
4.95.12.0
64.0
3.93.41.7
25.915.2
High sugar
M
14.44.45.9
98.2
11.68.56.4
53.890.3
SD
4.73.42.1
66.0
4.83.72.5
24.516.8
Noie. MT = matching task; PAL = paired-associate learning. For boys, n = 20; for girls, n = 10. b For boys, n = 6: for girls, n - 9.
-
8/13/2019 Rose Net Al 1988
4/7
586 ROSEN ET AL
ow
4
3
2
Control High
B O Y S
GIRLS
P IR E D S S O C I T E L E R N I N G T S K
Figure I. Mean number of errors on the paired-associate
learning task for boys and girls during each condition.
vided by that child's weight (mean weight = 22.5 kg). This
yielded the dose per kg of sucrose for each day for each child.
These amounts were averaged, providing the mean amount of
sucrose ingested during each condition (across children and
across days). The means for the high, low, and control condi-
tions were 2.26, 0.28, and 0.00 g/kg sucrose, respectively. In
addition, total carbohydrate intake (sucrose from breakfast
drink plus other carbohydrates included in the basic breakfast)
for the high, low, and control conditions averaged 3.95, 1.88,
and 1.54 g/kg, respectively. Finally, in no case was an individual
child's sucrose intake or total carbohydrate intake on a low-
sugar condition day or on a control condition day higher than
the intake during a high-sugar condition day; in fact, they were
always quite discrepant. (See Table 2 for the mean carbohydrate
dose of preschool and elementary school children).
Overall Effects
Due to the large number of dependent measures, multivariate
analyses of variance (MANOVAS) were conducted. We con-
ducted 2 x 2 x 3 MANOVAS for both cognitive measures and
teacher ratings to compare the effects of site (preschool vs. ele-
mentary school) by sex (boy vs. girl) by sugar level (high vs. low
vs. control), with repeated measures across the last variable.
When these analyses indicated that significant differences ex-
isted, the univariate differences contributing to this effect were
examined for significance using a repeated-measures analysis
of variance (ANOVA) and a Tukey post hoc test.
Cognitive measures. Results of the cognitive testing are dis-
played in Table 3. The multivariate analysis of the cognitive
measures indicated a significant Sex x Sugar Level interaction
for the cognitive measures as a whole, F(6, 36) = 2.85, p < .05.
An examination of the univariate differences contributing to
this multivariate finding indicated a significant interaction be-
tween sex and sugar level on the paired-associate learning task,
F\2, 82) = 3.69, p < .05 (with the more conservative Green-
house-Geisser adjusted df , p < .05). An analysis of the mean
paired-associate learning task scores for boys and girls for each
sugar level indicated that girls made significantly (Tukey p
top related