The authors have declared that no competing interests exist.
Conceived and designed the experiments: MJT. Performed the experiments: KLC MJT. Analyzed the data: KLC PLC MJT. Contributed reagents/materials/analysis tools: MJT PLC. Wrote the paper: MJT PLC KLC.
What is the ideal body size and shape that we want for ourselves and our partners? What are the important physical features in this ideal? And do both genders agree on what is an attractive body? To answer these questions we used a 3D interactive software system which allows our participants to produce a photorealistic, virtual male or female body. Forty female and forty male heterosexual Caucasian observers (females mean age 19.10 years, s.d. 1.01; 40 males mean age 19.84, s.d. 1.66) set their own ideal size and shape, and the size and shape of their ideal partner using the DAZ studio image manipulation programme. In this programme the shape and size of a 3D body can be altered along 94 independent dimensions, allowing each participant to create the exact size and shape of the body they want. The volume (and thus the weight assuming a standard density) and the circumference of the bust, waist and hips of these 3D models can then be measured. The ideal female body set by women (BMI = 18.9, WHR = 0.70, WCR = 0.67) was very similar to the ideal partner set by men, particularly in their BMI (BMI = 18.8, WHR = 0.73, WCR = 0.69). This was a lower BMI than the actual BMI of 39 of the 40 women. The ideal male body set by the men (BMI = 25.9, WHR = 0.87, WCR = 0.74) was very similar to the ideal partner set by the women (BMI = 24.5, WHR = 0.86, WCR = 0.77). This was a lower BMI than the actual BMI of roughly half of the men and a higher BMI than the other half. The results suggest a consistent preference for an ideal male and female body size and shape across both genders. The results also suggest that both BMI and torso shape are important components for the creation of the ideal body.
What makes a human body attractive to the opposite sex? In evolutionary psychology terms it is a judgment of a potential partner’s health and reproductive potential
Previous studies that have attempted to define the importance of these physical cues have had a significant limitation. These studies have used line-drawings, photographs and, more rarely, video clips and 3D laser scans as test stimuli
To overcome these important methodological limitations, we have used an interactive 3D software programme to determine male and female participants’ perceptions of their ideal body and their ideal partners’ body size and shape. The participants could alter the virtual 3D image of the body in more than 90 independent dimensions allowing very subtle changes in body shape. The body could be rotated through 360° to allow our participants to examine the body from different viewpoints. The scaled volume of these 3D models can then be measured and, assuming they have a standard body density, their body weight can then be estimated. Additionally, the scaled circumference of the chest, waist and hips of each body can be measured to allow the waist-to-hip ratio (WHR) and the waist-to-chest ratio (WCR) to be calculated. By taking anthropometric measures from all our participants, we can determine whether the participants’ own physical dimensions influence their choice of their own ideal body. This morphing technique allows us to answer two key questions:
A similar difference of opinion exists for what is the main determinant of male attractiveness. Some studies assert that upper body shape (a broad upper body and a narrow waist, the classic V-shape) is the primary predictor of attractiveness, whereas others point to BMI as the key feature
By asking both men and women to set their ideal bodies we can determine which features they change and how their ideal body differs from their actual bodies. We can see whether they change shape or size or both.
The study was reviewed and approved by the School of Psychology Ethics Committee of Newcastle University.
A total of 80 heterosexual Caucasian undergraduate students aged 18–21 (40 females mean age 19.10 years, s.d. 1.01; 40 males mean age 19.84, s.d. 1.66) were recruited from Newcastle and Northumbria Universities. Participation was voluntary. However some students gained course credit. All participants gave informed consent and the aims and procedure of the study were explained beforehand. None of the students had previous experience with using the software.
The participants used a 3D modelling software package (Daz Studio 3.1 from Daz3d.com) which allows the adjustment of photo-realistic male and female 3D models on a flat panel screen in order to modify different aspects of the body’s features (see
The bodies are displayed in slightly different viewing angles, and each body could be rotated though the whole 360°. Along the right of the picture are some of the 94 sliders which allowed different parts of the body to be independently altered.
Each participant created a total of four 3D bodies; two that represented their ideal body and two that represented their ideal partner’s body. In each of the two conditions, the participants began with a ‘heavy’ body and then a ‘thin’ body, or vice versa. The order was counterbalanced between participants. The two estimates were averaged to render a final model. The use of fat/thin bodies as a starting point was to reduce potential anchor effects which might have occurred if participants had just begun by adjusting a normal weight body. The female “thin” body had a BMI of 14.9 and the “large” body had a BMI of 26.6. The male “thin” body had a BMI of 16.5 and the “large” body had a BMI of 37.7.
All the participants were tested on the same PC in the Body Image Lab at the Institute of Neuroscience. Participants were asked to adjust the sliders until they were satisfied that the model looked like their ideal body and then they were asked to produce their ideal partner’s body. No time limit was placed upon them. Although there are 94 sliders, many of them are used for comparatively subtle adjustments to features such as the length of the ring finger on the left hand, and were not used. We ourselves had not altered these minor features in the “heavy” and “thin” bodies and we had left them at the default setting. Instead, most participants used a core set of sliders (mean 36.2 sliders, s.d. 7.8) which changed features, such as stomach depth and hip width.
After completion, a set of anthropometric measures were taken from the participants by the lead author (K.L.C.). Height was measured using the Marsden/Invicta Free Standing Height Measure and weight was measured using the Weight Watchers 8944U Heavy Duty Body Fat Analyser Scale. Using a standard tape measure, the waist and hip circumferences were measured, along with bust and under-bust circumferences if female, and chest circumference if male, following the protocols outlined in the Health Survey for England
The final 3D models were exported from Daz Studio, once clothing had been removed, and reopened in 3ds Max (autodesk.com), where they were set either to the height of the participant (for their own ‘ideal’) or to the height of the average British man (1.78 m) or woman (1.64 m) (for ‘ideal partner’). First, the volumes of the 3D models were calculated by the software, scaling the body volume relative to the body height entered by the experimenter. Once the volumes were known, the weights of the models were estimated by multiplying their volumes by the density of either the average young adult female body (1.04 g/cm3) or the average young adult male body (1.06 g/cm3)
Next, 3ds Max was used to slice through each model at predetermined points along its length to measure the circumference of the bodies at the chest, waist and hips in male models, and the bust, under-bust, waist and hips in female models. The software scaled the circumferences (measured in cm) to the dimensions that the bodies would have if they were real. However, the circumference measures generated by 3ds Max for the hips in male bodies and bust and hips in female bodies tend to be larger than the same measurements taken from real bodies. This is because 3ds max calculates the path length around - each slice which includes, for example, the cleft in the bust or buttocks. In comparison, a tape measure looped around the bust or hips will straddle these gaps, and so will produce a shorter distance. To compensate for these effects, we screen grabbed the cross-sectional slices of the bust or hips in 3ds Max and imported them into ImageJ (
A potential weakness of this methodology is the question of whether the participants can reliably manipulate the software controls to produce the body size and shape that they want. To answer this question we ran a test-retest experiment in which participants were asked to repeat the modelling task. We asked 15 Caucasian female participants (average age 22.57, s.d. 2.76) and 15 Caucasian male participants (average age 23.21, s.d. 2.84) to set their ideal body size and shape using the same methodology as described above. They then repeated the same tasks the following day. The Pearson’s correlation between the BMI values of the models that participants set on the two days was highly statistically significant (r = 0.99, p<.001) and a paired-samples t-test of the bodies’ BMIs showed no significant differences between the settings for the bodies on the two days (paired T-test: t(25) = 1.69, p = .103).
In this results section we first show that that there are significant differences in size and shape between the actual bodies of the participants and their ideals. We then show that these ideal bodies differ from the expected shape of real bodies of the same BMI, implying an explicit choice for specific sizes
A summary of the anthropometric data from the participants’ actual and ideal bodies are shown in
Body A and C are the ideal female bodies set by the female and male participants respectively and Body B and D is the ideal male body set by the female and male participants respectively.
BMI | Bust/Chest | Under Bust | Waist | Hips | WCR | WHR | ||
|
||||||||
Average | 21.7 | 87.4 | 75.93 | 72.91 | 99.4 | 0.86 | 0.73 | |
SD | 2.07 | 5.17 | 5.6 | 5.48 | 5.36 | 0.2 | 0.19 | |
|
||||||||
Average | 18.85 | 93.97 | 68.33 | 61.12 | 87.89 | 0.67 | 0.70 | |
SD | 1.75 | 8.24 | 4.06 | 3.38 | 6.52 | 0.09 | 0.04 | |
|
||||||||
Average | 18.82 | 90.02 | 69.2 | 61.95 | 84.82 | 0.69 | 0.73 | |
SD | 1.56 | 4.73 | 5.79 | 5.79 | 4.92 | 0.05 | 0.04 | |
|
||||||||
Average | 24.54 | 97.74 | – | 86.12 | 98.76 | 0.88 | 0.87 | |
SD | 3.38 | 9.21 | – | 9.47 | 7.93 | 0.04 | 0.06 | |
|
||||||||
Average | 25.86 | 111.26 | – | 82.00 | 91.17 | 0.74 | 0.87 | |
SD | 3.95 | 9.44 | – | 9.17 | 9.59 | 0.05 | 0.04 | |
|
||||||||
Average | 24.46 | 104.16 | – | 80.57 | 90.81 | 0.77 | 0.86 | |
SD | 2.9 | 7.43 | – | 7.22 | 7.15 | 0.05 | 0.03 |
Males appear to prefer a more tubular shape in their lower torso (as indexed by a higher WHR) as their ideal. In comparison, females appear to desire a curvier lower torso shape (with a lower WHR values) for their ideal.
To quantify these effects we computed a between-subjects (i.e. gender: male versus female) ANOVA and a within-subjects (i.e. condition: actual versus ideal) mixed ANOVA separately for WHR and WCR.
For WHR, we found a statistically significant main effect of gender (F1,117 = 506.48, p<.0001), but not of condition (F1,117 = 0.26, p = .613). The interaction between gender and condition was statistically significant (F1,117 = 17.34, p<.0001). In order to identify which differences contributed to the interaction term, we computed post-hoc differences of least square means using the Tukey-Kramer correction to compensate for multiple statistical comparisons. We found that the female participants’ ideal WHRs were significantly less than their actual WHRs (t117 = 3.30, p = .0069; r = 0.29) and that the males’ ideal WHRs were significantly higher than their actual WHRs (t117 = −2.59, p = .05; r = 0.23).
For WCR, we found a statistically significant main effect of gender (F1, 156 = 37.15, p<.0001) and condition (F1, 156 = 249.13, p<.0001). However, the interaction between gender and condition was not statistically significant (F1,156 = 1.31, p = .254). We computed post-hoc differences of least square means using the Tukey-Kramer correction to compensate for multiple statistical comparisons and found that females’ and males’ ideal WCRs were significantly lower than their actual WCRs (t156 = 11.97, p<.0001; r = 0.69 and t156 = 10.35, p<.0001; r = 0.64 respectively).
Our analysis suggests that the ideal body size and shape of both the male’s and female’s ideals differs from the corresponding actual bodies. However, a possible confound is that in real life, body shape and body size tend to co-vary in a non-linear way (i.e. a body with a particular BMI will have a particular shape), with different parts of the body changing size at different rates with changing BMI. We have already illustrated this relationship in women’s bodies in several previous studies
The plots illustrate that increasing BMI is associated not only with a generalized increase in torso width, reflected by the systematic separation of one profile from the next, but also with a non-linear component to the change in body shape. This non-linear component is illustrated by the male torso outline in sub-regions A (near the waist) and B (the lower hip). In region A, as BMI increases from 15 to 35, the contour of the waist changes from convex to concave and in region B, the slope of the line from lower to higher hip slices becomes less and less steep. There are similar non-linear shape changes in the female torso in sub-regions C (the upper chest) and D (upper hip).
A simple regression can then be used to estimate the relationship between each slice width and BMI. The key feature to appreciate about
In the current study we seek to answer the question of how different are people’s own ideal body shapes compared to the shape they currently have, as well as the ideal body they would seek in a partner. The complex shape changes illustrated in
Since the BMI of both genders’ ideals is different from their actual BMI, we can calculate what proportion of the change in torso shape of their ideal body is attributable just to the change in BMI alone. In other words we can predict the component of shape change in the ideal which is predicted by the BMI of the ideal body shape selected. We can then compute the difference between the bust/chest, under-bust, waist and hip circumferences of the ideal image and the equivalent circumferences computed on the basis of the BMI of the ideal and then ask whether, on average, these are significantly different from zero. If this population of differences is not significantly different from zero, this suggests that the shape of the body that participants choose as their ideal is no different from merely choosing a higher or lower BMI. However, if the population of differences in circumferences is significantly different from zero, this means that the shape of the bodies that participants choose as their ideal is different from what they would achieve by merely selecting a higher or lower BMI.
The regression analyses to estimate the BMI shape change effect are based on circumference measures taken from 120 male and 120 female volunteers. The females were measured at bust, under-bust, waist and hips and the males at chest, waist and hips. The average age of the female volunteers was 20.3 years s.d. 3.5 and the average age of the male volunteers was 20.7 years s.d. 2.1. For each gender, we computed separately the regression between BMI and chest/bust, under bust waist and hip respectively, and then used these regression equations to estimate the expected circumferences in the ideal bodies chosen, based purely on their BMI.
Group | Body Slice | Average Difference in Circumference (cm) | t-test value | p value | r value | Power |
|
Chest | 11.04 (0.86) | 12.76 | <.0001 | 0.90 | >.99 |
Waist | −12.92 (0.69) | −18.67 | <.0001 | 0.95 | >.99 | |
Hips | −9.64 (0.59) | −16.46 | <.0001 | 0.93 | >.99 | |
|
Chest | 6.93 (0.92) | 7.52 | <.0001 | 0.77 | >.99 |
Waist | −10.99 (0.52) | −21.08 | <.0001 | 0.96 | >.99 | |
Hips | −7.60 (0.46) | −16.55 | <.0001 | 0.94 | >.99 |
The difference in the slice circumferences from the two bodies are shown along with the standard error in brackets. The DF for the t-test was 39.
To further explore this result we carried out t-tests for each set of circumferences (i.e. chest, waist, hips) for the populations of differences (see
The results from the T-tests of location above show that, the average shape of the ideal female bodies set by male and female participants differs significantly from the shape that would be predicted based solely on the BMI of the ideals. Next, we test whether the shapes of these ideals differ when comparing the settings made by male versus female participants. To address this question, we used a 2-factor, repeated-measures GLMM, where factor 1 was the gender of the participant (male, female) and factor 2 was the circumference (chest, waist and hip). There was no main effect of gender (F1,234 = 0.01, P = .938). The main effect of circumference was significant (F2,234 = 523.42, p<.0001) as was the interaction between gender and circumference (F2,234 = 12.78, p<.0001). To determine which individual ideal shape measures differed between male and female participants, we calculated post-hoc differences of least square means using the Tukey-Kramer correction to compensate for multiple statistical comparisons. The difference between male and female settings of chest circumference was statistically significant (p<.0001), whereas the differences for waist and hip were not.
An independent t-test shows that the ideal male BMI set by the female participants is not significantly different from that set by the male participants (t(78) = 1.81, p = 0.074; effect size r = 0.20; power to detect at two-sided alpha of 0.05 = 0.44). The WHR of the two bodies were also not significantly different: t(78) = 1.43, p = .229; effect size r = 0.16; power to detect at two-sided alpha of 0.05 = 0.20), but WCR was significantly different (t(78) = −3.09, p = .003; effect size r = 0.33; power to detect at two-sided alpha of 0.05 = 0.67).
Group | Body Slice | Average Difference in Circumference (cm) | t-test value | p value | r value | Power |
|
Bust | 10.78 (2.97) | 3.62 | .0008 | 0.50 | >.94 |
Under Bust | −3.73 (0.60) | −6.19 | <0.0001 | 0.70 | >.99 | |
Waist | −6.43 (0.47) | −13.68 | <.0001 | 0.91 | >.99 | |
Hips | −0.63 (0.77) | −0.81 | .42 | 0.13 | .12 | |
|
Bust | 6.88 (0.81) | 8.46 | <.0001 | 0.80 | >.99 |
Under Bust | −2.81 (0.81) | −3.48 | .001 | 0.50 | .92 | |
Waist | −5.53 (0.48) | −11.56 | <.0001 | 0.88 | >.99 | |
Hips | −3.64 (0.57) | −6.40 | <.0001 | 0.72 | >.92 |
The difference in the slice circumferences from the two bodies are shown along with the standard error in brackets. The DF for the t-test was 39.
T-tests of location for the populations of differences, where the null hypothesis was a mean of zero, are all statistically significant at p<.05, even after applying a Bonferroni correction for multiple comparisons, with the exception of female settings for the hip circumference.
The results from the T-tests of location above show that the average shape of the ideal female bodies set by male and female participants differs significantly from the shape that would be predicted based solely on the BMI of the ideals. Next, we test whether these ideal body shapes differ when comparing the settings made by male versus female participants. As before, to address this question, we used a 2-factor repeated measures GLMM, where factor 1 was the gender of the participant (male, female) and factor 2 was the circumference (bust, under-bust, waist and hip). There was no main effect of gender (F1,78 = 1.67, P = .201). The main effect of circumference was significant (F3,234 = 63.68, p<.0001), but there was no significant interaction between gender and circumference (F3,234 = 2.43, p = .066).
An independent t-test shows that the ideal female BMI set by the female participants is not significantly different from the ideal female BMI set by the male participants (t(78) = 0.09, p = 0.93; effect size r = 0.01; power to detect at two-sided alpha of 0.05 = 0.05). The WBR of the two bodies were also not significantly different (t(78) = −3.64, p<.001; effect size r = 0.38; power to detect at two-sided alpha of 0.05 = 0.91), but WHR was significantly different (t(78) = −3.64, p<.001; effect size r = 0.38; power to detect at two-sided alpha of 0.05 = 0.91).
Both male and female participants created an ideal body that was significantly different in body size relative to their own. The female participants significantly reduced the body size and the male participants increased it. Although some studies have suggested BMI is the primary predictor of female attractiveness and that shape is of marginal importance (e.g.
The female participants’ ideal female body has a BMI which is significantly lower than their actual BMI. Consistent with this lowered BMI, there is a general narrowing of the torso, with the hips, waist and chest (excluding the bust) reducing in circumference (i.e. the volume of the body is reduced). The actual BMI values of the female participants all fall within the normal BMI range (18.5-24.9), with the majority around the middle part of this scale
In contrast to the narrowing of the rest of the female body, the “ideal” bust increases in size (as indexed by bust circumference). Previous studies have linked relative bust size to circulating estrogen levels, with the suggestion that a large bust and a narrow waist should indicate high levels of estrogen and therefore be regarded as attractive
The increase in bust size and narrowing of the torso between the female participants’ actual body and their ideal changes the upper body shape (as indexed by WCR and illustrated in
Unlike the female thin ideal body, the ideal male body is comparatively heavy, falling at the boundary of the normal to overweight categories of the BMI scale. However, these are not bodies that look over-weight, but instead are big and muscular. In fact, our calculation of their BMI is probably an under estimation, because we are assuming that the bodies have the average density for young men (i.e. the average balance of fat to muscle). As muscle is approximately 20% denser than fat, this would under-estimate the mass of a more muscular body such as the male ideals set in this experiment. This result is consistent with previous studies, which have suggested that muscularity (and the associated perception of dominance and strength) is the primary determinant of male attractiveness
The male participant’s ideal body shows an increase in chest circumference (relative to their actual body) and a reduction in the waist and hips to produce a V-shaped upper body. Previous studies have also suggested that men prefer a body that is more muscular than the one they actually possess
The preferences for the ideal female body are broadly similar between the two genders. They both prefer the same low BMI and a relatively curvaceous body with WCR and WHR with values around 0.7. There is also general agreement between the genders on the ideal male body; this male ideal has a relatively large body with a V-shape upper torso and a narrow waist and hips. This is consistent with attractiveness rating studies which to show a strong correlation between male and female attractiveness ratings of male and female bodies (i.e. both genders seem to rate bodies of both genders the same way)
An alternative explanation would be that the ideals are influenced by a common media environment which pushes them towards the same concept of the ideal body. However, there are subtle gender-specific differences in the media images seen in the magazines targeted at men and women. For the male body, magazines aimed at a male audience contain male models which are more muscular than those aimed at a female audience
This is partially what we find here. The male body selected by the male participants is indeed more muscular than the ideal male body chosen by the female participants. However, in the case of the ideal female body both men and women prefer a female body with the same low BMI, but the female participants prefer a larger bust size than the male participants. This directly contradicts what would be expected from the size and shape of the female models in their respective gender-specific media; the men should prefer a heavier female body than the women and a larger bust.
Previous studies have focussed on body size in women’s bodies. These suggest that although women overestimate the level of female thinness desired by men (e.g.
That still leaves the question of why the difference exists in male and female preferences for upper body shape; female participants prefer a larger bust in their ideal female body than men, and male participants prefer a larger chest in their ideal male body than women. This may be linked to within gender competition for status and prestige
An alternative socio-cultural explanation would emphasise how a culture-specific female ideal body size and shape potentially exerts a particularly strong influence on women’s concept of what they should aspire to
The combination of the 3D morphing software and the regression analysis shows that the ideals for both genders have a specific body size (as indexed by BMI) and shape. For both sexes, the primary predictor of female beauty is a relatively low BMI combined with a relatively curvaceous body, whereas the features important for the male ideal are a slightly heavier, muscled body with a specific V-shaped upper body. Although, the results suggest a largely consistent preference for an ideal male and female body size and shape across both genders, but with subtle differences based on an own gender exaggeration of upper body shape.
Thanks to Ellen Roberts and Laura Moxon for help in collecting some of the data.