International Journal of Obesity (1999) 23, 1288±1294
ß 1999 Stockton Press All rights reserved 0307±0565/99 $15.00
http://www.stockton-press.co.uk/ijo
Perception of body size in the Tongan
community: differences from and similarities to
an Australian sample
P Craig1*, V Halavatau2, E Comino3 and I Caterson4
1
Metabolism and Obesity Services, Department of Endocrinology, Royal Prince Alfred Hospital, Camperdown, NSW, Australia; 2Central
Planning Department, Government of Tonga, Nuku'alofa, Kingdom of Tonga; 3School of Community Medicine, University of New South
Wales, Sydney, NSW, Australia; 4Human Nutrition Unit, Department of Biochemistry, University of Sydney, Sydney, NSW, Australia
OBJECTIVE: The prevalence of obesity in Tonga is high and increasing. Attempts to address this major health issue
would be assisted by a realistic concept of current and preferred body sizes. In this study, body size perception in
Tongans was compared with that of Australians.
DESIGN: Cross-sectional comparative study.
SUBJECTS: 542 Tongans, 89 of whom were enrolled in a weight-loss programme (WLP); 481 Australians.
Measurements: Subjects were weighed and measured. Two series of photographs (one female, one male), representing speci®c body mass indexes (BMIs) were used to assess body perception.
RESULTS: The BMIs of the Tongans were higher than those of the Australians (Tongans: female 32.6 0.35 (mean
s.e.m); male, 30.4 0.33; Australians: female, 25.6 0.33; male 26.3 0.26 kg=m2). Tongan women underestimated
their body size; Tongan men and Australians overestimated. WLP and younger Tongan women preferred smaller body
sizes for themselves and for men and women in general. There were gender differences in preferred healthy and
attractive female sizes (men: 27 kg=m2; non-WLP women: 25 kg=m2; WLP women: 23 kg=m2) and male sizes (men: 29 ±
30 kg=m2; women: 26 ± 27 kg=m2) chosen by Tongans. Preferred, attractive and healthy body sizes chosen by
Australian men and women were similar; about 24 kg=m2 for males and 21 ± 22 kg=m2 for females.
CONCLUSION: Tongans preferred larger body sizes than did Australians, particularly the men, but WLP women's
preferences were similar to those of Australians. There is evidence that preferences are changing in Tonga with time,
and probably with increasing Western in¯uence.
Keywords: body perception; body weight; cross-cultural; Tongan; Australian
Introduction
The prevalence of non-communicable diseases is
increasing in Tonga with growing Western in¯uence.
Obesity is considered a risk for these conditions, and
has become a major health issue as body size continues to expand in Tonga, along with that of many
other populations in the Paci®c1 and elsewhere.2 ± 7
The 1986 Tongan National Nutrition Survey showed
that 90% of women over 30 years of age had a body
mass index (BMI ˆ weight=height2) greater than
25 kg=m2, and 39% of women and 10% of men
were obese (BMI 30 kg=m2), using standards
derived from Western data.8 However, Polynesians
have a heavy muscular build, and the appropriateness
of similar BMI standards across different populations
has been questioned. It may be that the moderately overweight category for Caucasians (BMI
*Correspondence: P Craig, Metabolism and Obesity Services,
Royal Prince Alfred Hospital, Level 9, Queen Mary Building,
Grose St., Camperdown, NSW 2050, Australia.
E-mail: [email protected]
Received 20 November 1998; revised 10 May 1999; accepted
26 June 1999
25 ± <30 kg=m2) should be considered normal
weight for Tongans;8 while a lower BMI constitutes
the healthiest weight range for Asians.9,10
At the same time, the body size preferences of
different cultural groups may or may not coincide
with the healthy weight ranges. Tongans are very
homogeneous, both culturally and genetically, with a
strong sense of community, and a traditional preference for large body size resulting from a complex
interaction between genetic predetermination and
environment (selection by survival during long sea
voyages across the Paci®c;11 the development of
ceremonial feasts as an important social rite in
response to an uncertain food supply.12,13) It is still
believed that Tongans see obesity as beautiful,14 and a
way of illustrating what their actual body size is and
suggesting a suitable goal would be a useful educational tool.
There is a vast literature on the measurement of
body image which has been summarised in several
reviews.15 ± 17 In essence, body image is concerned
with two different concepts: the perceived image of
the body in space, and the affective response of
individuals to their bodies. The focus of this study
was the perceptual rather than the affective domain.
Ethnicity and perception of body size
P Craig et al
Methods used to assess body perception have utilised the adjustable body-distorting mirror18,19 and
variable anamorphic lens20 ± 22 or video distorting
techniques,19,23 ± 26 where subjects vary their own
image in response to questions about perceived and
preferred body size. Series of silhouettes, drawings,27,29 or photographs30 ranging from slim to
obese have been used, particularly in cross-cultural
research. However, while these techniques were able
to assess percentage change from the actual image of a
subject, or compare selected silhouettes, none
assigned actual body size to the images. In one
instance a series of silhouettes were produced using
individual subjects with known BMIs,31 but the different body shapes and varying heights and BMI
increments between images made comparison
between them dif®cult.
In a study on the body perception of Australians,
overweight men (BMI 25 kg=m2) overestimated
their body size; while those of acceptable size
(BMI ˆ 20 ± <25 kg=m2), matched for age, perceived
quite accurately.32 Women overestimated and wanted
to be slimmer whether they were overweight or not.32
Polynesian adults from the Cook Islands33 chose
higher ideal BMIs than did Australians (27 and
24 kg=m2 for men and women, respectively, compared
with 24 and 22 kg=m2 for Australians), but while older
women and men still preferred larger body sizes,
younger women appeared to accept the Western
norm of a slim body size.
The aim of this study was to investigate perceived
and preferred body size in Tongan adults, a group who
traditionally prefer large body size, and compare
perceptions with an Australian (largely Caucasian)
sample where Western slim ideals predominate. The
hypotheses under investigation were: (a) the more
similar community preferences and actual body size,
the more accurate is body size perception; (b) body
size preferences decrease with increasing Western
in¯uence; (c) body size preferences re¯ect socially
accepted norms.
Methods
Study populations
Two groups of subjects were included in the study.
The ®rst was a sample from several villages on the
main island groups in Tonga (Tongatapu, Vava'u,
Ha'apai, Eua), with the exception of the northernmost
Nuias (not readily accessible). The subjects came
predominantly from church and community groups,
with some staff from government departments and
commercial organisations.
The Australians were a random sample of the staff
of a large Sydney hospital. Subjects came from a
diversity of cultural backgrounds, but were predominantly of Caucasian origin.
Instrument
Two series of images (one female, one male) were
prepared using a photograph of known BMI, the width
of the image being altered while maintaining its height.
The BMI of each image was calculated using a polynomial regression estimation. The method has been
described in detail elsewhere.32 The female series was
validated by measuring the bi-trochanteric widths of
the images (the widest point of the gluteal region or
buttocks) and comparing these to bi-trochanteric
widths measured on photographs of 25 females of
known BMI corrected for height (range: 16.3 ±
53.6 kg=m2). This was repeated using the bi-iliac
widths (just below the waist, commonly referred to
as the breadth at the hip bones) for male images and
photographs of 26 men (range: 15.5 ± 54.9 kg=m2). The
bi-iliac breadth has been shown to be the best predictor
of BMI in males, while the bi-trochanteric breadth best
predicts BMI in women (T Olds, personal communication, 1996; based on the South Australian anthropometric database). The high predictability of BMI from
bi-iliac (males: R2 ˆ 0.92) and bi-trochanteric breadths
(females: R2 ˆ 0.83) in this sample justi®ed use of this
method for validating the series of photographs. The
lower values for females resulted from the greater
variation in female body shape.
This photographic method was shown to be transferable to other models, allowing use of culturally
appropriate series with the different cultural groups.
Fifty-®ve females (30 Polynesians, 25 Australians)
were asked to select the photograph closest to their
own body size from both Paci®c Island and Australian
female series. The procedure was repeated using the
two male series with 50 males (25 Cook Islanders, 25
Australians).33 Weighted kappa values of 0.66 for
females and 0.75 for males33 (kappa values of
0.41 ± 0.60 represent moderate agreement, 0.61 ±
0.80 substantial agreement, and those greater than
0.81 excellent agreement beyond chance34) and intraclass correlations of 0.88 for females and 0.91 for
males between the Caucasian and Paci®c Island series
of photographs showed the degree of concordance
between the different cultural series to be highly
acceptable.
Procedure
Subjects were weighed to the nearest 0.1 kg on a
portable solar-powered electronic scale in Tonga and
a stationary electronic scale in Australia. All subjects
were measured to the nearest 0.5 cm using the same
portable height measuring instrument. BMI was calculated from measured weights and heights.
Graded sets of 10 photographs, representing calculated BMIs (female range: 18.8 ± 52.5; male range:
17.5 ± 51.3 kg=m2), were used to measure body perception. Two series of photographs appropriate for the
ethnic group (one female, one male) were laid out in
increasing order of size. Subjects were asked to
indicate which size represented their own body size
1289
Ethnicity and perception of body size
P Craig et al
1290
(perceived size), their preferred own body size (preferred size), and which sizes they considered the most
healthy and most attractive for their own and the
opposite sex.
Tongans were slightly older than Australians (Tongans Ð women, 40.8 0.81 y; men, 40.0 0.98 y;
Australians Ð women, 38.7 0.68 y; men, 37.1
0.71 y; P < 0.05).
Statistics
Body size
Results were expressed as mean standard error. The
differences were tested for signi®cance using Student's t-test.35 Pearson correlation, or partial correlation, was used in describing the relationship between
variables.
The study was approved by the Central Sydney
Area Health Service Ethics Committee and all subjects in Australia completed and signed an informed
consent. The Tongan segment of the study was conducted with the approval and assistance of the Tongan
Food and Nutrition Committee.
Results
Subjects
The proportion of males from each age group in the
Tongan sample and the most recent census data36 was
similar; the study women were slightly older. A
slightly higher proportion came from the smaller
islands. A total of 542 subjects (299 females and
243 males) were measured. A small subsample
(n ˆ 89) were enrolled in a nationwide weight loss
programme (WLP).
Although the Australian sample of 481 subjects
(261 females, 220 males) cannot be claimed to
represent the Australian population, the BMI distribution was similar to that of the recent National Nutrition Survey (NNS),37 the males being somewhat
lighter (females Ð BMI < 20: NNS ˆ 6%, sample ˆ
7%; BMI ˆ 20 ± <25: NNS ˆ 41%, sample ˆ 50%;
BMI ˆ 25 ± <30: NNS ˆ 29%, sample ˆ 27%; BMI 30: NNS ˆ 18%, sample ˆ 17%; w2 ˆ 2.64, P > 0.05.
Males Ð BMI < 20: NNS ˆ 2%, sample ˆ 2%; BMI ˆ
20 ± <25: NNS ˆ 32%, sample ˆ 44%; BMI ˆ 25 ±
<30: NNS ˆ 45%, sample ˆ 36%; BMI 30: NNS ˆ
19%, sample ˆ 19%; w2 ˆ 6.06, P < 0.05).
Table 1
Tongan women had a greater BMI as a group than
Tongan men; among Australians, men had a greater
BMI than women (Table 1). Tongans had a greater
BMI than Australians (Table 1).
Perceived and preferred body size
Mean measured, perceived and preferred body sizes
for women and men appear in Table 1. Tongan
women underestimated their own body size (measured
vs perceived) by about 1.5 kg=m2 (t ˆ 4.87, P < 0.001,
95% con®dence interval (CI) ˆ 0.9, 2.2). Older (those
over 50) and larger women (BMI 35) were more
accurate. All women underestimated their body size to
the same extent, whether they were in the WLP or not.
Tongan men overestimated their weight to similar
degree (t ˆ 5.16, P < 0.001, 95% CI ˆ 1.2, 2.7),
although younger men (under 30) and those with a
BMI under 30 were accurate. Australians overestimated their own body size, women by 2.3 kg=m2
(t ˆ 8.47, P < 0.001, 95% CI ˆ 1.7, 2.8), and men to
the same extent as the Tongan men (t ˆ 3.67,
P < 0.001, 95% CI ˆ 0.6, 2.0). Younger and smaller
(BMI < 25) Australian men selected accurately,
whereas all Australian women overestimated regardless of their body size. All groups preferred to have
smaller body sizes than their current perceived sizes.
Those Tongan subjects enrolled in the WLP preferred smaller body sizes than those who were not in
the WLP Ð 23.5 kg=m2 compared with 24.9 kg=m2 for
women (t ˆ 2.35, P ˆ 0.02); 27.9 kg=m2 compared
with 29.5 kg=m2 for men (not statistically signi®cant).
Older (partial correlation ˆ 0.25, P < 0.001) and heavier Tongan women (partial correlation ˆ 0.13,
P ˆ 0.03) selected larger preferred body sizes,
whereas Tongan men's preferred sizes were not
related to either age or body size.
Heavier Australian women (partial correlation ˆ
0.39, P < 0.001; controlling for age) and men (partial
Women's and men's measured, perceived and preferred body sizes
Group
Women
All Tongans
WLP Tongans
Other Tongans
Australians
Men
All Tongans
WLP Tongans
Other Tongans
Australians
Number
Measured BMI (mean s.e.)
Perceived BMI (mean s.e.)
Preferred BMI (mean s.e.)
299
64
235
261
32.6 0.36
35.0 0.75
31.9 0.38
25.6 0.33d
31.1 0.51a
33.4 1.20a
30.4 0.56a
27.9 0.49a
24.6 0.25b
23.5 0.54b
24.9 0.28b
21.3 0.21b,d,e,
243
25
218
220
30.4 0.33
34.0 0.69
29.9 0.30
26.3 0.26d
32.1 0.55a
39.6 1.77a
31.3 0.55a
27.5 0.51a
29.4 0.33b
27.9 1.04b
29.5 0.34b
24.4 0.24b
d,e,
All Tongans ˆ all Tongans in sample; Other Tongans ˆ subjects not in WLP.
a
Signi®cantly different from measured BMI; bsigni®cantly different from perceived BMI; using paired t-tests;
c
Signi®cantly different from WLP; dsigni®cantly different from Tongans; esigni®cantly different from WLP Tongans;
using unpaired t-tests.
Ethnicity and perception of body size
P Craig et al
correlation ˆ 0.29, P < 0.001) chose larger preferred
body sizes. Age did not affect choice.
Health and attractive body preferences
Female sizes. The attractive and healthy female sizes
chosen by Tongan women did not differ signi®cantly
from each other and were similar to their own
preferred size (Table 1 and Table 2). Women participating in the WLP chose smaller attractive (t ˆ 3.80,
P < 0.001) and healthy females sizes (t ˆ 3.05,
P < 0.001). Tongan men chose healthy and attractive
sizes for females which were signi®cantly larger than
those chosen by the women (BMI ˆ approximately
26.5 vs 25.0 kg=m2; P ˆ 0.001). Younger Tongans
chose the same female sizes, but these increased
with age at a greater rate for men than for women,
from about 24 to 29 kg=m2, and 24 to 26 kg=m2,
respectively (data not shown).
In contrast, Australian women and men chose
smaller attractive, rather than healthy, sizes for
females (t ˆ 7.87, P < 0.001; and t ˆ 3.85, P < 0.001,
respectively) (Table 2); women's attractive sizes were
signi®cantly smaller t ˆ 2.12, P ˆ 0.03) than those
chosen by men. Larger women (partial correlation ˆ
0.23, P < 0.001) chose slightly larger attractive female
sizes, but age did not affect choice. Neither age nor
measured size altered the female sizes chosen by men.
Male sizes. Attractive Tongan male sizes chosen by
both women (t ˆ 4.51, P < 0.001) and men (t ˆ 3.95,
P < 0.001) were larger than healthy male sizes. Men's
preferred sizes were similar to their chosen attractive
rather than healthy male sizes. Again, Tongan men
preferred larger healthy and attractive male body sizes
than did women (approximately 29 kg=m2 compared
with 26 kg=m2 for healthy; 30 kg=m2 compared with
27 kg=m2 for attractive; P < 0.001) (Table 2). Older
subjects, both women and men, chose larger attractive
male sizes (t ˆ 0.24, P < 0. 001; t ˆ 0.19, P ˆ 0.003,
respectively). Healthy and attractive male sizes
chosen by female WLP participants were smaller,
Table 2
and were similar to those chosen by Australian
women.
Australian men also chose larger attractive than
healthy male sizes (t ˆ 3.67, P < 0.001), but these
were the same as those chosen by Australian
women. Age and own measured BMI did not in¯uence choice of healthy or attractive male sizes among
Australians.
Discussion
The Tongans were fairly representative of the population, in terms of age, sex and distribution across the
islands. Sampling predominantly from church groups
from a variety of denominations ensured a range of
occupational backgrounds among the subjects; 95% of
the Tongan population attend church regularly. The
Australian sample had a similar body size distribution
to the Australian population, and included a wide
range of occupational backgrounds from professional
to manual labourers. Perceived mean ideal female and
male sizes selected by this sample were very similar
to those chosen by other Australian groups32,33,38
(ranging from 21.4 to 22.7 kg=m2 for women, and
23.8 to 24.5 kg=m2 for men), demonstrating a striking
consistency in Australian perceptions.
One advantage of the method used for assessing
body perception was the assignment of a BMI to each
photograph, allowing comparison of chosen images
with actual and recommended BMIs. Increasing the
BMI by a constant increment between photographs
strengthened statistical analysis. It was not feasible to
use photographs of individuals of known BMI to form
the test instrument; the height variation between
images and different body types made comparison
between them dif®cult. The method used had three
disadvantages. Firstly, it was easier to relate to one's
own image when responding to questions about perceived and preferred own body size, but a prepared
series was adequate for questions concerning ideal
attractive and healthy sizes. Using the subjects' own
image would have required transport and use of a
Ideal healthy and attractive body size for females and males chosen by Tongan and Australian women and men
Group
Women
All Tongans
WLP Tongans
Other Tongans
Australians
Men
All Tongans
WLP Tongans
Other Tongans
Australians
Number Healthy female (mean s.e.) Attractive female (mean s.e.) Healthy male (mean s.e.) Attractive male (mean s.e.)
299
64
235
261
24.9 0.27
23.3 0.54
25.4 0.31c
21.9 0.20d,e
25.4 0.27
23.2 0.60
25.8 0.35c
20.5 0.18b,d,e
25.7 0.31
23.9 0.71
26.3 0.34c
23.8 0.23d
26.8 0.31b
24.9 0.68b
27.4 0.36b,c
23.6 0.20d
243
25
218
220
26.5 0.36a
26.1 1.20
26.6 0.38
21.6 0.20d,e
26.7 0.36a
27.2 1.23
26.6 0.38
21.1 0.20a,b,d,e
28.7 0.34a
28.0 1.07
28.8 0.36
23.5 0.23d,e
29.8 0.38a,b
29.4 1.37
29.9 0.40b
24.1 0.24b,d,e
All Tongans ˆ all Tongans in sample; Other Tongans ˆ subject not in WLP.
a
Signi®cantly different from women of same ethnic group; bsigni®cantly different from healthy ideal for same sex (paired t -tests);
c
Signi®cantly different from WLP participants; dsigni®cantly different from Tongans; esigni®cantly different from WLP Tongans using
unpaired t -tests.
1291
Ethnicity and perception of body size
P Craig et al
1292
computer, unrealistic in the context of a ®eld study.
Secondly, at the extremes actual body shape varies
more than it is possible to depict by simply altering
the width of an image. This criticism applies even
when using the subject's own photograph, and is a
shortcoming of the video distorting method, not of the
current modi®cation. This problem was minimised
because the majority of women focus on the buttocks
as the width of identi®cation and subjects were asked
to consider width compared with height. Thirdly, the
technique is unable to account for variation in body
shapes, which is more of a problem among women,
who have greater body variability than men. Culturally appropriate series, as used in this study, take
some account of this.
The hypotheses investigated in this study concerned
perceptual accuracy, the effects of Westernisation and
social pressure on body size preferences.
Accuracy in body size perception in relation to
community preferences
Overall, no cultural or gender group accurately perceived their own body size, although older, larger
Tongan women and younger, smaller men were accurate. These subgroups could be considered to be those
in which community expectations and actual body
sizes coincided. Although it was possible that the
greater accuracy of older, heavier Tongan women
was due to the more limited choice at the extremity
of the photographic series, it seems more likely that
these older women have retained a more realistic or
cultural preference for large body sizes. The younger,
and generally lighter, men were those whose average
sizes could be considered within the range considered
preferable in their cultural group (28 kg=m2 for Tongans, 25 kg=m2 for Australians). The greater pressure
on Western women to conform to the slim ideal could
explain the fact that all Australian women overestimated whether they were overweight or not.
Effect of Westernisation on body size preferences
Tongan women's preferred sizes for self, healthy and
attractive females (about 25 kg=m2) were signi®cantly
different from those chosen by Australians (21 ±
22 kg=m2). Preferred female sizes increased with age
for Tongans, but not for Australians, suggesting a
changed preference over time and a growing acceptance of smaller body sizes. Sizes chosen by WLP
women (23 kg=m2) were similar to those chosen by
Australians and Cook Island Maori aged under 50;33
sizes chosen by Tongans over 40 were similar to those
of the older Maori group.33 The gradual decrease in
preferred sizes among the younger Tongan women
suggests growing in¯uence of Western preferences for
slimmer women; the Cook Islands are more Westernised than Tonga.
The preference for smaller body size by the WLP
participants may be an effect of the programme itself,
or may result from those with these preferences being
more likely to participate in the WLP. If the preference for diminishing body size has resulted from
Westernisation, there is stronger evidence for the
second option; participants in the WLP were younger
(but not lighter) than those from the wider community
(P ˆ 0.02).
Tongan men's preferred own and attractive sizes
were also signi®cantly different from Australians.
Smaller preferred attractive sizes among the young
Tongans could also be in¯uenced, at least in part,
from growing Western in¯uence in Tonga.
Social norms and body size preferences
The consistency of healthy, attractive and preferred
body sizes chosen by Australians in this and other
studies suggested that acceptable body sizes of around
24 kg=m2 for males and 21 ± 22 kg=m2 for females
were fairly established in the Australian community.
The larger body sizes chosen by heavier Australian
men may have been an expedient response to reality
or preference.
Gender differences aside, the `healthy male size' as
chosen by Tongans was comparatively consistent. It
was men's attractive sizes which changed with age.
The lack of a widely accepted cultural norm of
`attractive male size' within the culture may be
pivotal in terms of acceptance of a need to change,
rather than what may be regarded as a generally
Western concept of a `healthy' body size. Among
Tongan women, preferred sizes resembled healthy
sizes. It may be that health messages promoting a
smaller body size have reached women to a greater
extent than men.
Participation in the WLP was not associated with
altered self-perception among Tongan women.
Assuming that the age effect on preferences was
evidence of growing Western in¯uence, it seems
that the process of change in female preferences and
perceptions occurs in the order of community preferences diminishing, and later perceptions changing. In
contrast all Australian women wanted to be thinner
and all overestimated, in a cultural climate where the
thinner body is better. Cook Island women also
perceived themselves with remarkable accuracy, but
the younger ones preferred smaller body sizes,30
adding credence to this interpretation.
It was interesting to note that expected sizes were
larger for males than for females at all ages, despite
Tongan women being, on the average, larger than
their men. Western pressure for ever smaller body size
standards may have resulted in the preferred BMIs of
around 23 kg=m2 for women and 24 ± 25 kg=m2 for
men selected by Tongan WLP women. These appear
to be verging on the unrealistic for the Tongan
physique, and there may be a need for caution with
messages which seem to promote or condone smaller
female body sizes as the ideal.
Ethnicity and perception of body size
P Craig et al
Analyses of the body composition of these subjects
demonstrated consistently larger fat-free mass and
lower percentage body fat in Tongans at the same
BMI compared with Australians (our results, unpublished data). Adjusting Tongan healthy weight ranges
to an equivalent percentage of body fat to that found
in the Australians would permit higher BMI levels for
Tongans. What is more, the preferred sizes selected by
the majority of Tongan subjects Ð of the order of
26 kg=m2 for females and 28 kg=m2 for males Ð
were quite compatible with measured normal percentage body fat values. Achieving these levels in the
Tongan population, where the average BMIs were
substantially higher, remains a challenge.
Conclusion
This study demonstrated differences between Tongan
and Australian subjects in body size perception and
preferences, with the Tongans preferring larger sizes,
particularly the males. Among older Tongan women
and younger Tongan and Australian men, where
actual body size resembled cultural expectations of
body size, perceptions were more accurate. Increasing
Western in¯uence in Tonga appeared to result in
expectations of decreasing body size, particularly
among women. The greater gap between expectations
and reality among those women who participated in
the weight loss programme may be verging on unrealistic. If healthy weight ranges for Tongans were
increased to account for differing body composition,
preferred body sizes chosen by the majority of
Tongans were within the acceptable range.
Acknowledgements
We are grateful to the Food and Nutrition Committee
for granting approval for this study to be conducted in
the Kingdom of Tonga. We would like to thank many
community leaders in Tonga for their assistance with
organising subjects for study, and all the subjects,
both in Tonga and Australia, for their time and
cooperation.
References
1 Proceedings compiled by Malolo M. Information Document,
South Paci®c Commission, No. 58. South Paci®c Commission:
Noumea. 1stRegional workshop for nutritionists, dietitians and
nutrition educators: planning, implementing and evaluating
nutrition education programmes (Noumea). South Paci®c
Commission 1991; pp 24 ± 25.
2 Kuskowska-Wolk A, Bergstrom R. Trends in body mass index
and prevalence of obesity in Swedish men 1980 ± 89.
J Epidemiol Commun Health 1993; 47: 103 ± 108.
3 Kuskowska-Wolk A, Bergstrom R. Trends in body mass index
and prevalence of obesity in Swedish women 1980 ± 89.
J Epidemiol Commun Health 1993; 47: 195 ± 199.
4 Australian Society for the Study of Obesity. Healthy Weight: a
national obesity strategy. Australian Society for the Study of
Obesity: Sydney 1995.
5 Prentice AM, Jebb SA. Obesity in Britain: gluttony or sloth?.
B Med J 1995; 311: 437 ± 439.
6 Seidell JC, Verschruren WMM, Kronhout D Levels and trends
of obesity in The Netherlands, 1987 ± 1991. Int J Obes 1995;
19: 924 ± 927.
7 Flegal KM, Carroll MD, Kuczmarski RJ, Johnson CL Overweight and obesity in the United States; prevalence and trends,
1960 ± 1994. Int J Obes 1998; 22: 39 ± 47.
8 Maclean E, Bach F, Badcock J. The 1986 national nutrition
survey of the Kingdom of Tonga: Summary report. South
Paci®c Commission: Noumea. Technical Paper No. 200.
1992.
9 Wang J, Thornton JC, Russell M, Burastero S, Heyms®eld S,
Pierson RN. Asians have lower body mass index (BMI) but
higher percent body fat than do whites: comparisons of
anthropometric measurements. Am J Clin Nutr 1994; 60:
23 ± 28.
10 Singh RB, Mori H, Chen J, Mendis S, Moshiri M, Zhu S, Kim
SH, Sy RG, Faruqui AM. Recommendaitons for the prevention
of coronary artery disease in Asians: a scienti®c statement of
the International College of Nutrition. J Cardiovasc Risk 1996;
3: 489 ± 494.
11 Houghton P. People of the great ocean. Cambridge University
Press: Cambridge 1996.
12 Bell FLS. The place of food in the social life of central
Polynesia. Oceania 1931; II: 117 ± 135.
13 Barrow T. Women of Polynesia. Seven Seas: Wellington,
1967.
14 Anon. Plumpness may be beautiful Ð but it's deadly. Extract
from. The Tonga Chronicle, 11 March 1993. Paci®c Island
Nutr 1993; 16(June): 5.
15 Cash TF, Brown TA. Body image in anorexia nervosa and
bulimia nervosa: a review of the literature. Behav Modif 1987;
11: 487 ± 521.
16 Ben-Tovim BI, Walker MK. Women's body attitudes: a
review of measurement techniques. Int J Eating Disorders
1991; 10: 155 ± 167.
17 Hsu LKG, Sobkiewicz TA. Body image disturbance: time to
abandon the concept for eating disorders?. Int J Eating Disorders 1991; 10: 15 ± 30.
18 Traub AC, Orbach J. Psychophysical studies of body image.
Arch Gen Psychiat 1964; 11: 53 ± 66.
19 Freeman RJ, Thomas CD, Solyom L, Hunter MA. A modi®ed
video camera for measuring body image distortion: technical
description and reliability. Psychol Med 1984; 14: 411 ± 416.
20 Glucksman ML, Hirsh J. The response of obese patients to
weight reduction. Psychosom Med 1969; 31: 1 ± 7.
21 Garner DM, Gar®nkel PE, Stacey HC, Moldofsky H. Body
image disturbance in anorexia nervosa and obesity. Psychosom
Med 1976; 38: 327 ± 335.
22 Gleghorn A, Penner LA, Powers PS, Schulman R. The
psychometric properties of several methods of body image.
J Psychopathol Behav Assess 1987; 2: 203 ± 218.
23 Allebeck P, Hallberg D, Espmark S. Body image Ð an apparatus for measuring disturbances in estimation of size and
shape. J Psychosom Res 1976; 20: 583 ± 589.
24 Collins JK, McCabe MP, Jupp JJ, Sutton JE. Body percept
change in obese females following weight reduction therapy.
J Clin Psychol 1983; 37: 507 ± 511.
25 Whitehouse AM, Freeman CPL, Annandale A. Body size
estimation in anorexia nervosa. Br J Psychiat 1988; 153:
23 ± 26.
26 Lautenbacher S, Roscher S, Strian F, Pirke K-M, Krieg J-C.
Theoretical and empirical considerations on the relation
between body image, body scheme and somatosensation.
J Psychosom Res 1993; 37: 447 ± 454.
27 Furnham A, Alibhai N. Cross-cultural differences in the
perception of female body shapes. Psychol Med 1983; 13:
829 ± 837.
28 Fallon AE, Rozin P. Sex differences in perceptions of
desirable body shape. J Abnormal Psychol 1985; 94:
102 ± 105.
1293
Ethnicity and perception of body size
P Craig et al
1294
29 Bell C, Kirkpatrick SW, Rinn RC. Body image of anorexic,
obese and normal females. J Clin Psychol 1986; 42: 431 ± 439.
30 Massara EB, Stunkard AJ. A method of quantifying beauty
and the obese. Int J Obes 1979; 3: 149 ± 152.
31 Whitney E, Hamilton E, Rolfes S. Understanding nutrition
1990; 5th edn., West: St Paul: p 255.
32 Craig PL, Caterson ID. Weight and perceptions of body image
of women and men in a Sydney sample. Community Health
Stud 1990; 14: 373 ± 383.
33 Craig PL, Swinburn BA, Matenga-Smith T, Matangi
H, Vaughan G. Do Polynesians still believe that big is
beautiful? Comparison of body size perceptions and preferences of Cook Island Maori and Australians. NZ Med J 1996;
109: 200 ± 203.
34 Landis JR, Koch GG. The measurement of observer agreement
for categorical data. Biometrics 1977; 33: 159 ± 174.
35 Armitage P, Berry G. Statistical method in medical research.
3rd edn. Blackwell: Oxford 1990.
36 Statistics Department, Kingdom of Tonga. Statistical Abstract
1997. Statistics Department, Government of Tonga: Nuku'alofa, Tonga 1998.
37 Australian Bureau of Statistics National nutrition survey.
Australian Bureau of Statistics and Department of Health
and Family Services: Canberra 1995.
38 Craig PL, Caterson ID. Body size, age, ethnicity, attitudes and
weight loss. In:. Obesity in Europe. Ailhaud G, Guy-Grand B,
Lafontan M, Ricquier D (eds.) 1992; Obesity in Europe 91:
421 ± 426.