International Journal of Obesity (2008) 32, 336–342

& 2008 Nature Publishing Group All rights reserved 0307-0565/08 $30.00
www.nature.com/ijo

ORIGINAL ARTICLE
BMI-based body size guides for women and men:
development and validation of a novel pictorial
method to assess weight-related concepts
CV Harris1, AS Bradlyn1, J Coffman1, E Gunel2 and L Cottrell3
1
Health Research Center, School of Medicine, West Virginia University, Morgantown, WV, USA; 2Department of
Statistics, West Virginia University, Morgantown, WV, USA and 3Department of Pediatrics, West Virginia University,
Morgantown, WV, USA

Objectives: To develop and evaluate two BMI-based instruments to determine perceptions of weight status, particularly
perceptions of overweight and obesity, using pictorial images of women and men.
Methods: Pictures of adults with known BMI values were used to construct gender-specific body size guides (BSGs) containing
10 bodies that ranged from underweight to class III obesity. Figures were standardized and a composite face was added to each.
The BSGs were administered to 400 adults to assess the psychometric properties of the instruments and weight perceptions.
Results: High correlations between the BMIs of respondents and the BMIs of the current body selected by respondents provided
strong support for the criterion-related validity of the BSGs, and the logical pattern of responses to items assessing perception of
weight categories supported construct validity for the scales. Test–retest reliability, assessed by correlations for both current and
ideal body, was also high, despite the lengthy 6-month testing interval. Respondents’ perceptions of the bodies within specific
weight categories indicated that a majority failed to recognize the overweight female as overweight and perceived the
overweight male as normal weight. Obese bodies were generally unrecognized as such until the bodies reached the higher levels
of obesity (that is, BMI values 439). Perception of weight was influenced by the respondents’ weight status and gender.
Conclusions: Psychometric analyses indicated the BSGs are valid and reliable instruments. These results, coupled with the face
validity of the scales and the relationship between the bodies and BMI values, indicate the BSGs offer advantages over existing
instruments for researchers of weight perception and body image. Administration of the scales to an adult sample confirmed
that overweight and obesity are under-recognized. Increased efforts to improve public understanding of these terms are needed
and the BSGs may provide useful tools for this purpose.
International Journal of Obesity (2008) 32, 336–342; doi:10.1038/sj.ijo.0803704; published online 14 August 2007

Keywords: overweight; perception; validation; body size

Introduction index (BMI), are poorly understood by the general public.

The prevalence of overweight and obesity is increasing
dramatically in many countries around the globe.1 In the
United States, this increase has been labeled an ‘epidemic’ of
significant public health concern due to the association
between obesity and long-lasting health problems, increased
disability and premature death.2,3 Unfortunately, the con-
cepts of overweight and obesity, as defined by body mass
Correspondence: Dr CV Harris, Health Research Center, West Virginia
University, Suite 1400 HSCFSouth, PO Box 9136, Morgantown, WV
26506-9136, USA.
E-mail: charris@hsc.wvu.edu
Received 30 January 2007; revised 17 May 2007; accepted 3 July 2007;
published online 14 August 2007
Multiple investigations have documented misperceptions of
weight status by adults,4–6 adolescents,7 and parents of
preschool through adolescent children.8–10 Physicians also
have been found to misperceive their own weight status and
that of their patients,11,12 and to infrequently use BMI
measurements in their practices.13,14 Although the US
Surgeon General has issued a ‘call to action,’ and the lay
press and professional literatures are replete with articles
discussing the health risks associated with obesity, these
warnings may miss their intended targets if affected
individuals fail to recognize the personal relevance of the
warnings or health-care providers fail to appropriately
identify and counsel the overweight patients they treat.15,16
Assessing perceptions of the terms ‘overweight’ and ‘obese’
has been approached through a variety of methods. To assess

personal weight perception, researchers characteristically ask
individuals to identify themselves as ‘underweight, over-
weight or about right’;4–6 it is notable that the term obese has
rarely been included among the descriptors available for self-
assessment. This method of assigning weight-based descrip-
tors to individuals has also been used to assess physician
perception of patient weight. From this literature it appears,
in general, that men show a greater likelihood of misperceiv-
ing their overweight status as normal than women, and that
misperceiving overweight as normal occurs with greater
frequency in the elderly, the less educated and those with
limited incomes.12,17 Women physicians recognized the
overweight status of their patients more readily than men,
but physicians of both genders were less likely to recognize
overweight status among patients who were male, consid-
ered themselves healthy, or reported being physically
active.12 These findings are helpful in the identification of
groups that are more likely to underassess weight status and
highlight some of the biases inherent in the assignment of
emotionally valenced terms such as overweight and obesity,
but they do not provide clear indications about the thresh-
old for perception of different weight classifications.
Alternatively, body image has been assessed via figure-
rating scales (FRS), which contain line drawings or silhou-
ettes. The most widely used FRS was developed by Stunkard
et al.18 more than 20 years ago. FRS instruments typically
obtain perceptions of body image through selection of
figures that are identified as healthy, anorexic or obese, or
obtain ratings of body satisfaction/dissatisfaction via selec-
tion of figures representing the raters’ current (‘real’) versus
preferred (‘ideal’) body. Following the FRS model provided by
Stunkard, researchers have developed additional instruments
for specific groups including children,19 adolescents,20
African Americans,21 and obese individuals.22 Although
several of these instruments have been validated through
correlations between scale values and measures of weight,
none of these instruments contain figures with a precise
relationship to BMI. Because overweight and obesity are
defined by BMI, it is important to be able to link pictorial
representations of these concepts to this index.
In an attempt to link BMI to figural stimuli, Stunkard and
co-workers23 collected self-reported height and weight, as
well as current and preferred figure ratings from a large
sample of adults. These data were used to generate a BMI
value for each figure and to determine which figures
corresponded to BMI-based definitions of thin and obese.
This research represents a substantial advance for the widely
used Stunkard FRS; however, as acknowledged by the
authors, the study was limited by the use of self-reported
height and weight to generate the BMI values for the figures
and did not address participant perceptions of overweight and
obesity. These perceptions were addressed using an innova-
tive scale with six photographic silhouettes of individuals
with known BMI, ranging from 20 to 38.24 The authors
reported that 99% of their sample of women correctly
identified the obese silhouettes, but the majority (97%) also
Body size guides
CV Harris et al
337
equated overweight with obesity. Unfortunately, this scale
does not contain any silhouettes of underweight figures and
its psychometric properties are unknown.
The current study describes the development of pictorial,
BMI-based body size guides (BSGs) for women and men. The
development of simple, easy to use instruments that
incorporate pictures of actual human bodies with known
BMI values from underweight to obese could address some of
the limitations associated with current measures, provide a
clear, visual tool for health education and further our
understanding of the perceptions of overweight and obesity.
Methods
BSG development
Adult women and men were recruited through fliers and
advertisements in local media. Interested individuals com-
pleted a brief telephone screening to obtain information
about height, weight and exercise patterns, and were
scheduled to be photographed if their reported height/
weight and calculated BMI corresponded to preset criteria.
Specifically, we selectively recruited individuals who did not
engage in regular exercise and whose presumed BMI values
were approximately 3 points apart; informal pilot testing had
indicated that adults were able to discriminate between
bodies with this degree of difference.
Women were photographed in black leotards and men
were photographed in tank tops and gym shorts. Height and
weight were obtained in this attire, without shoes and socks,
using the Easy Glide Bearing Stadiometer (Perspective
Enterprises, Portage, MI, USA) and the Seca model 882
digital scale. Using Photoshop software, images were stan-
dardized by adding a gender-matched composite face to each
body and removing any unusual or distinguishing features
(that is, tattoos, scars, or excessive body hair). The composite
faces were created using facial elements (eyes, nose, hair and
so on) from photographs of three or more volunteers.
Separate collections of images were created of the women
and the men. Each resulting BSG (BSG-W for women’s
images and BSG-M for men’s images) contained 10 images of
individual bodies that ranged from underweight (BMIo18.5)
to class III obesity (BMIX40). A greater number of images
depicting higher BMI values were included to provide
maximum utility for obesity research. To assess the dis-
criminability between bodies, the 10 individual photos for
each scale were placed on separate sheets of paper. Eight
graduate students and faculty members from a variety of
disciplines sorted the pictures from smallest to largest with
100% accuracy, indicating there was a discernable difference
between the bodies.
Questionnaire items were created to determine which
bodies were perceived to be healthy, underweight, normal
weight, overweight and obese. Additional items prompted
respondents to identify the body most like theirs (current
body) and the body they would most like to have (ideal body);
International Journal of Obesity
 Body size guides
CV Harris et al
338
these two items were completed only by respondents whose
gender matched the bodies portrayed. Finally, items designed
to assess social context prompted respondents to identify the
bodies which looked most like the adults in their families,
their adult friends and the adults in their communities.
Respondent sample and methods
Questionnaire respondents were 335 women and 65 men
who completed the instruments as part of their participation
in an obesity prevention trial (N ¼ 112), a structured exercise
program (N ¼ 9), or a general survey of health beliefs
(N ¼ 279). Respondents ranged in age from 19 to 77 years
with a mean of 36.8 years (s.d. ¼ 10.9); 98% were white. The
educational attainment of the respondents ranged from less
than high school (29%) to college or graduate degree
completion (42%). The BMI values of questionnaire respon-
dents were calculated from the direct measurement of height
and weight for individuals in the obesity trial and exercise
program (N ¼ 121) and from self-reported height and weight
for individuals in the survey study (N ¼ 279). These values
are referred to as measured BMI and self-reported BMI,
respectively, throughout the article. Respondents were
classified as underweight (1%), normal weight (28%), over-
weight (26%), or obese (45%) based on their measured or
self-reported BMI values using the standard cutoffs.
We certify that all applicable institutional and govern-
mental regulations concerning the ethical use of human
Figure 1 Women’s BSG images, weight classifications and perceptions by respondents. Weight classifications are based on measured BMI; weight perceptions,
held by a majority of respondents (450%), are indicated by the lines below each body. r Copyright 2005 Harris and Bradlyn.
International Journal of Obesity
volunteers were followed during this research. This research
was approved by the West Virginia University Institutional
Review Board.
Statistical analyses
The analyses involving weight classification included only
respondents who were normal weight, overweight, or obese
by BMI due to the small number (N ¼ 4) of underweight
respondents in our sample. Analyses were conducted with
male and female respondents combined, unless noted.
Differences in perceptions between genders and individuals
in the various weight classifications were examined using
2  3 analyses of variance (ANOVAs) with gender (female and
male) and weight class (normal, overweight and obese) as the
factors. Tukey’s HSD tests were used for post hoc analyses.
Analyses of items that allowed respondents to identify more
than one body were conducted using both the smallest and
largest body identified by each respondent.
Results
Psychometric properties
The construct validity for the instruments was determined
by examining the patterns of responses to items that assessed
perceptions of which bodies were healthy, underweight,

Figure 2 Men’s BSG images, weight classifications and perceptions by respondents. Weight classifications are based on measured BMI; weight perceptions, held
by a majority of respondents (450%), are indicated by the lines below each body. r Copyright 2005 Harris and Bradlyn. BMI, body mass index; BSG, BMI-based
body size guides for women.
normal weight, overweight and obese. Figures 1 and 2
provide the images used, the weight classifications and the
weight perceptions of the bodies by a majority (450%) of
respondents. As illustrated, respondents chose successively
larger female and male bodies to correspond to the increas-
ing weight classifications. Only the normal weight female
bodies (B and C) were identified as healthy by a majority of
respondents, while the larger of the two normal weight male
bodies (C) and the overweight male (D) were identified as
healthy for men. Only body A was identified as underweight
for both genders. Although the bodies selected for each
classification were generally similar across the instruments,
as indicated in Figures 1 and 2, larger male than female
bodies were included within the normal and overweight
classifications by a majority of respondents.
Criterion-related validity was assessed through Pearson’s
product moment correlations between the BMI values of the
respondents and the BMI values of the current bodies selected
by respondents. Correlations between measured BMI and
current body BMI were 0.94 for men and 0.86 for women
(Po0.001); correlations between self-reported BMI and current
body BMI were 0.86 for men and 0.88 for women (Po0.001).
Respondents’ selections for current body are provided in
Figure 3. As illustrated, all bodies were selected by at least
one respondent with the exception of one class II obesity
body for men (G) and one class III obesity body for women (J).
Body size guides
CV Harris et al
339
Pearson’s product moment correlations also were used to
determine test–retest reliability in a subset of 84 respondents
who completed the BSGs twice over a 6-month interval.
Despite the lengthy interval, the BMI values of respondents’
current body correlated 0.91 and ideal body correlated 0.77
(Po0.001).
Effects of weight group and gender on perceptions
Perceptions of ideal body. Respondents identified a single,
same gendered body they considered ideal. Although none of
the bodies received a X50% endorsement, the smaller of the
two normal weight bodies (body B) received the greatest
endorsement by both genders (36.4% of men and 42.8% of
women). Both gender and weight classification impacted the
selection: females identified smaller ideal bodies than males
(F(1,388) ¼ 12.22, Po0.0005), normal weight respondents
chose smaller bodies than overweight and obese respon-
dents, and overweight respondents chose smaller bodies
than obese respondents (F(2,388) ¼ 32.74, Po0001). There
was no significant gender–by-weight class interaction.
Family, friends and community. Examination of the effects of
gender and weight class on the perceptions of weight in
female family members and friends (BSG-W) revealed
significant main effects for gender (family members
International Journal of Obesity
 Body size guides
CV Harris et al
340
F(1,385) ¼ 3.92, Po0.05; friends F(1,384) ¼ 4.41, Po0.05)
and weight class (family members F(2,385) ¼ 42.18,
Po0.0001). Female respondents selected larger bodies to
represent the adult women in their families and their women
friends than did the male respondents, and obese respon-
dents selected larger bodies than normal and overweight
respondents. There was no significant interaction between
gender and weight class, and these factors did not have an
effect on the female community bodies selected. Responses
regarding male family members, friends and community
bodies revealed no main effects for gender or weight class,
but a gender by weight class interaction was found for both
family members (F(2,341) ¼ 9.10, P ¼ 0.0001) and friends
(F(2,337) ¼ 4.51, P ¼ 0.01). In general, the differences be-
tween the obese and normal weight men were greater than
the differences between obese and normal weight women in
perceptions of their male family members and friends. Obese
men, as a group, selected the largest bodies for these items
and normal weight men selected the smallest bodies.

Perception of health and weight classification. BSG-W and

BSG-M items that assessed perceptions of which bodies were
healthy, underweight, normal weight, overweight and obese
allowed multiple bodies to be selected. As shown in Table 1,
the ANOVAs conducted for BSG-W responses revealed
significant main effects of weight class on perceptions of
healthy, normal weight and overweight bodies and of gender
on perception of obese bodies. There was no significant
gender by weight interactions. Tukey’s HSD analyses revealed
that, in general, individuals in the higher weight groups
selected larger bodies as healthy, normal and overweight
than their lower weight peers. In addition, women selected
larger bodies as obese than did men.
Analyses of BSG-M responses revealed a single main effect
of weight class on the perception of healthy male bodies:
obese respondents identified larger bodies as healthy than
did normal weight respondents (F(2,345) ¼ 3.17; P ¼ 0.04);
there was no effect of gender on perception of health.

Conclusions

This study provides strong support for the psychometric

properties of these BMI-based pictorial instruments, indicat-
ing they provide reliable and valid information about
Figure 3 Percent of respondents, by gender, who selected each body to perceptions of body weight. The high correlations between
represent their current size. respondents’ BMI values whether measured or self-reported,

Table 1 Mean BMIs of BSG-W bodies selected by respondents

Perception of bodies

a
Healthy Underweight Normal weightb Overweightc Obesed

Smallest Largest Smallest Largest Smallest Largest Smallest Largest Smallest Largest

Normal 20.49 24.34 17.49 17.49 21.23 23.89 33.14 54.50 37.93 54.94
Overweight 20.60 24.38 17.74 17.74 21.62 24.05 31.31 51.64 37.86 55.25
Obese 21.31 25.79 18.21 18.21 22.80 25.35 33.96 53.25 37.96 55.28
Female 20.49 24.78 17.88 17.88 21.79 24.39 32.53 53.24 39.41 55.93
Male 21.11 24.89 17.74 17.74 21.98 24.47 33.07 53.02 36.43 54.39

Abbreviations: BMI, body mass index; BSG-W, BMI-based body size guides for women. aNormal oobese for smallest healthy body selected (F(2,391) ¼ 3.85,
P ¼ 0.02); normal, overweight o obese for largest healthy body selected (F(2,391) ¼ 8.30, P ¼ 0.0003). bNormal, overweight oobese for smallest (F(2,382) ¼ 16.19,
Po0001) and largest normal weight (F(2,382) ¼ 11.85, Po0.0001) bodies selected. cOverweight oobese for smallest overweight body selected (F(2,382) ¼ 3.16,
P ¼ 0.04); normal 4overweight for largest overweight body selected (F(2,382) ¼ 3.44, P ¼ 0.03). dMen owomen for smallest (F(1,277) ¼ 5.99, P ¼ 0.01) and largest
obese (F(1,277) ¼ 4.86, P ¼ 0.02) bodies selected.

International Journal of Obesity


and the BMIs of the current bodies selected indicate
respondents are able to identify bodies that are similar to
theirs with a reasonable degree of accuracy. This was true for
both genders. Moreover, the test–retest values obtained
across a several months interval were comparable to, or
greater than, those provided for both the Stunkard and more
recently developed scales.22,25 In addition to strong psycho-
metric properties, the current instruments have bodies with
direct rather than inferred relationships to BMI and bodies
representing all three classes of obesity. The BSGs satisfy the
criteria offered by Stunkard26 that additional scales are
justified only to the extent that they demonstrate greater
validity (that is, higher correlations between the scales and
BMI values) than the original measure.
This study also serves to highlight public misperception of
the terms overweight and obesity and to extend the growing
body of research on this important health issue. Although
underweight and normal weight bodies were accurately
identified by a majority of respondents, the overweight
female body was not identified as such and the overweight
male body was misidentified as normal. Moreover, obesity
was only recognized at the higher levels in both female and
male bodies.
Consistent with earlier investigations, this research de-
monstrated the effects of respondents’ weight classification
and gender on perceptions of weight and health and
reaffirmed that individuals who are overweight or obese are
less likely to recognize overweight/obesity than their normal
weight peers and more likely to label overweight bodies as
healthy. The replication of weight and gender effects on
perceptions provides further support for the validity of the
new measures.
There are several limitations to the current study, includ-
ing the small sample size of men and the under-representa-
tion of racial and ethnic groups. A number of previous
investigations have demonstrated racial and ethnic differ-
ences in body image and weight perception, indicating
additional research with the BSG instruments will be needed.
Modifications of the facial features and skin tone of the
scales may be necessary to address issues of acceptability or
validity with more diverse populations; these issues are
presently being examined by our research team.
More difficult to address are potential concerns related to
the relatively small number of bodies in the non-obese
classifications in these BSGs. The number of non-obese
bodies was a function of our use of a 3-point BMI differential
to ensure discriminability between bodies, the limited BMI
range defined for normal and overweight, and the naturally
occurring absence of BMI values below 15 in the general
population. The resulting restriction on the number of
underweight, normal weight and overweight bodies may
limit the applicability of the BSGs to eating disordered and
underweight populations.
In summary, the BSGs are psychometrically sound, easy to
use instruments to assess perceptions of weight and body
image. Further research is needed to explore their utility in
Body size guides
CV Harris et al
341
diverse populations and to assess their potential to improve
recognition of overweight and obesity by the general public
and medical providers.
Acknowledgements
We acknowledge Rachel Yeater and Julie Olexa for their
assistance in conducting this research, and Carol Grimes for
her work on the development of the BSG images. This
research was supported in part by the Centers for Disease
Control and Prevention Grant no. H75/CCH322130-01.
References
1 Obesity: preventing and managing the global epidemic. Report of
a WHO consultation. World Health Organ Tech Rep Ser 2000; 894:
i–xii, 1–253.
2 Mokdad AH, Serdula MK, Dietz WH, Bowman BA, Marks JS,
Koplan JP. The spread of the obesity epidemic in the United
States, 1991–1998. J Am Med Assoc 1999; 282: 1519–1522.
3 The Surgeon General’s Call to Action to Prevent and Decrease
Overweight Obesity. US Dept HHS/Public Health Service/Office of
Surgeon General: Washington, DC, 2001, pp v–60.
4 Kuchler F, Variyam JN. Mistakes were made: misperception as a
barrier to reducing overweight. Int J Obes Relat Metab Disord 2003;
27: 856–861.
5 Paeratakul S, White MA, Williamson DA, Ryan DH, Bray GA. Sex,
race/ethnicity, socioeconomic status, and BMI in relation to self-
perception of overweight. Obes Res 2002; 10: 345–350.
6 Chang VW, Christakis NA. Extent and determinants of discre-
pancy between self-evaluations of weight status and clinical
standards. J Gen Intern Med 2001; 16: 538–543.
7 Strauss RS. Self-reported weight status and dieting in a cross-
sectional sample of young adolescents: National Health and
Nutrition Examination Survey III. Arch Pediatr Adolesc Med 1999;
153: 741–747.
8 Jackson D, McDonald G, Mannix J, Faga P, Firtko A. Mothers’
perceptions of overweight and obesity in their children. Aust J
Adv Nurs 2005; 23: 8–13.
9 Maynard LM, Galuska DA, Blanck HM, Serdula MK. Maternal
perceptions of weight status of children. Pediatrics 2003; 111:
1226–1231.
10 Boutelle K, Fulkerson JA, Neumark-Sztainer D, Story M. Mothers’
perceptions of their adolescents’ weight status: are they accurate?
Obes Res 2004; 12: 1754–1757.
11 Perrin EM, Flower KB, Ammerman AS. Pediatricians’ own weight:
self-perception, misclassification, and ease of counseling. Obes
Res 2005; 13: 326–332.
12 Caccamese SM, Kolodner K, Wright SM. Comparing patient and
physician perception of weight status with body mass index. Am J
Med 2002; 112: 662–666.
13 Cleator J, Richman E, Leong KS, Mawdsley L, White S, Wilding J.
Obesity: under-diagnosed and under-treated in hospital out-
patient departments. Int J Obes Relat Metab Disord 2002; 26:
581–584.
14 Lemay CA, Cashman S, Savageau J, Fletcher K, Kinney R, Long-
Middleton E. Underdiagnosis of obesity at a community health
center. J Am Board Fam Pract 2003; 16: 14–21.
15 Sciamanna CN, Tate DF, Lang W, Wing RR. Who reports receiving
advice to lose weight? Results from a multistate survey. Arch
Intern Med 2000; 160: 2334–2339.
16 Hamilton JL, James FW, Bazargan M. Provider practice, over-
weight and associated risk variables among children from a
International Journal of Obesity
 Body size guides
CV Harris et al
342
multi-ethnic underserved community. J Natl Med Assoc 2003; 95:
441–448.
17 Kuchler F, Lin BH. The influence of individual choices and
attitudes on adiposity. Int J Obes Relat Metab Disord 2002; 26:
1017–1022.
18 Stunkard AJ, Sorensen T, Schulsinger F. Use of the Danish
adoption register for the study of obesity and thinness. Res Publ
Assoc Res Nerv Ment Dis 1983; 60: 115–120.
19 Collins EM. Body figure perceptions and preferences among
preadolescent children. Int J Eat Disord 1991; 10: 199–208.
20 Sherman DK, Iacono WG, Donnelly JM. Development and
validation of body rating scales for adolescent females. Int J Eat
Disord 1995; 18: 327–333.
21 Patt MR, Lane AE, Finney CP, Yanek LR, Becker DM. Body image
assessment: comparison of figure rating scales among urban black
women. Ethn Dis 2002; 12: 54–62.
22 Williamson DA, Davis CJ, Bennett SM, Goreczny AJ, Gleaves DH.
Development of a simple procedure for assessing body image
disturbances. Behav Assess 1989; 11: 433–446.
23 Bulik CM, Wade TD, Heath AC, Martin NG, Stunkard AJ, Eaves LJ.
Relating body mass index to figural stimuli: population-based
normative data for Caucasians. Int J Obes Relat Metab Disord 2001;
25: 1517–1524.
24 Holdsworth M, Gartner A, Landais E, Maire B, Delpeuch F.
Perceptions of healthy and desirable body size in urban
Senegalese women. Int J Obes Relat Metab Disord 2004; 28:
1561–1568.
25 Thompson MA, Gray JJ. Development and validation of
a new body-image assessment scale. J Pers Assess 1995; 64:
258–269.
26 Stunkard A. Old and new scales for the assessment of body image.
Percept Mot Skills 2000; 90: 930.

International Journal of Obesity