Experiments & Data

Quotation from Three Dimensional Photography
The Principles Of Stereography 1953: 

 “I have had several women say about a stereogram,
“Oh! I like that. It doesn’t make me look as fat as I usually do in a photograph!”

Herbert C. McKay  (page 63)

This fattening effect is highly significant, and easily seen when comparing life-size projections of people in 2D & 3D:

3D photographs of  volunteers wearing swimsuits were projected life-size in 2D or 3D to different groups. The lens separations were 63mm (orthostereoscopic) and the 2D images were projected synoptically (with identical 2D images in both projectors).   The model on the left was consistently judged to be slightly overweight in 2D and correct in 3D. The model on the right was judged as correct in 2D but underweight in 3D. In all cases, the 3D bodyweight judgements most closely matched their actual Body Mass Index figures of 21 for the left model and 18 for  the right. Whenever a size judgement differed between 2D and 3D images of the same person, the 2D projections were always judged to be larger and heavier than both their 3D projection and their actual Body Mass Index.

Side-By-Side Viewing Vs. Random Presentations

In the early feasibility studies, the 3D projections were all shown as a set of 20 images before or after the otherwise identical 2D projections of the same set of people. Viewers with normal stereoscopic vision always saw bodies and faces projected life-size,  and in 3D they were always seen as significantly slimmer. In subsequent studies, all of the stereoscopic images of people in these experiments were randomised and presented as a mixture of 2D and 3D life-sized projections. A significant slimming effect in 3D was still reported, even when viewers were unaware the images were randomised and alternaternately projected in 2D or 3D. This experiment is the most important because (unlike the feasibility study) subjects never saw the same person in 2D and 3D. Yet the same slimming effect was still observed and remained highly significant.

Experimental 3D & 2D CGI Shapes
Used To Detect Shape Change

It was observed that the 3D projections of people were not only slimmer than when seen in 2D, but also appeared to have much slimmer necks, waists and legs.
So to remove obvious human associations, CGI “peanut” shapes were made in 2D with various sized waists and in 3D with a fixed WHR (Waist-Hip Ratio) of 0.7, following healthy human female proportions.

These images can be seen in 3D below by cross-eyed “free-viewing”.

3D Peanut Stimuli With 0.7 Waist-Hip Ratio

Same Magnification, But Variable Waist Size 2D Versions Of The Peanut Shape (Always Projected As 0.7 WHR In 3D)

Results

The images above are from the range of 13 waist-hip ratio cards used. When trying to find a matching 2D waist hip ratio shape to the orthostereoscopically projected 3D image, almost all viewers selected a slimmer 2D card than the shape projected. However, when synoptic* 2D versions of the peanut were projected, the mean result was slightly larger than life sized.

* The synoptic 2D projections were viewed using polarised 3D glasses on an adjacent screen to the 3D projections. They were pixel-matched to avoid keystoning, yet delivered larger than life size estimations. Many experiments were conducted, with virtual Inter-ocular distances for the stereo slides ranging from 30 to 300mm. Reported here are four disparities ranging from 0 to 120mm, with 60mm representing the average separation for children or adults and the condition most closely approximating human stereo vision

These three seemingly identical images above are different. The left and right are averages of the actual sizes seen in 2D and stereo 3D.  The animation below shows the range of perceived size changes the subjects reported seeing. The larger waist represents the average of all of their 2D size matches.  The middle waist has the 0.7 Waist-Hip ratio projected in 3D. The slimmest waist (right) is the average seen by all subjects in 3D when viewing an image with the most natural condition: 60 mm of virtual camera separation (IOD) .

Remarkably, the “breathing” effect of the waist seen in the animation above was very clearly seen in the original 3D experiments. Simply by gazing quickly from the 3D to the 2D projections (or by closing one eye when looking at any of the 3D targets) would show the waist fattening as obviously as is seen in the morph above.

3D Target Slimmed In A Dose-Response Curve

These graphs show that as increasingly wider stereoscopic disparities (left to right) were projected in 3D, the 0.7 WHR target area at the waist appears significantly smaller in size comparisons with an otherwise identical 2D target.  The widening stereo base images were always projected with the same shape and 0.7 proportions because the object had rotational symmetry. Yet the subjects consistently perceived these changes in 3D inter-ocular distances as increasingly slimmer waist size estimations. These size estimations followed a strong dose-response curve, with the slimmest waist size estimations always associated with the widest stereo-base projections.

Why 3D is Perceived As Slimmer Than 2D

Evidence for the ecological validity of the slimming effect of 3D on waists and necks was provided in a mathematical proof by Dr Richard Latto of Liverpool University. He calculated that the occluded area viewed stereoscopically (shown as the central shadow area above from the 60mm lens separations at 1.6 m distance from the target) has a significantly slimmer waist-hip ratio than the 2D occluded area (right image). This slimming of the occluded background area correlates to the perceived slimming seen in the 3D experiments and also follows a dose-response curve: The wider the separation between the optical channels, the slimmer the waist area appears to be. 

Research Question Two:

Are 2D size estimation errors caused only by the loss of stereoscopic size, shape and distance information?   Are there other photographic distortions of body shape that are as significant, or even larger than the 3D to 2D effects previously reported?

Leonardo’s Constraint 2

“The further that a spherical body is from the eye, the more of it you will see.”

This Leonardo da Vinci quotation could be equally be interpreted as “the nearer you are to a convex object, the less of its surface you will see.” It seems that Leonardo was well aware of the variability of an image caused by changes in observer-to-subject distance. He may even have had the ability to magnify a distant image with lenses* and see in detail the distance effects he described.

* A Brief History of the Astronomical Telescope IV: Did Leonardo Invent the Telescope 100 Years Before Galileo?
National Geographic July 5, 2011

The Effects Of Proximity On Shape Perception:
Waists And Necks
Become Relatively Fatter With Increasing Camera-To Subject Distances

These illustrations show the effect of camera proximity on shape perception. The peanut shape (with a waist-hip ratio of 0.7) is illuminated by a virtual lamp at three proximities (of 0.5 metres wide angle lens simulation, a typical standard lens distance of 1.6 metres and a telephoto distance of 5 metres).  Imaged from infinity, the occluded area has the same proportions as the target. But with proximity, the shadow area at the waist becomes progressively slimmer relative to the hips and shoulders. This is a similar effect to the previous demonstration and shows that waists (or human necks) can be slimmed by proximity or by widening of the optical channels. 

Waist-Hip And Jaw-Neck Ratios

In the previous illustrations, the shadowed area simulates the actual 2D imaged proportions of a 3D virtual target as it would be recorded at the focal plane inside a real camera. The waist-hip ratio varies because increasing proximity also increases the distance ratio between the nearest and furthest areas of the target.

This effect is even larger in photographs of faces and necks.

The telephoto occluded area confirms the observation by James Cutting that telephoto lenses reproduce the proportions of the face and neck much closer to their actual physical measurements than standard lenses.

However, humans do not have telephoto eyes. So normal facial proportions are always perceived in reality “wide-angle” in close-up with our stereoscopic vision . So real faces are never perceived with the flattening and fattening effects of telephoto lenses.

Size Estimations In Portraiture With Varying Lens Types: Faces Are Flattened And Fattened With Increased Distance.


Jaw To Neck Size Changes With Increasing Distances When Photographed By Wide-Angle, Standard And Tele Lenses


Experiment Viewing Male And Female Portraits

2 male and 3 females were photographed at 5 distances, of which 3 are shown here.

Viewers were asked to estimate the five models bodyweight in a randomised presentation from the full range of camera to subject distances. Unlike here, they never saw the same person photographed at different distances using two wide-angle, one standard and two telephoto leses. The final data was collated from 25 respondents:

Method
2 male and 3 female psychology students were photographed at 5 different distances. Viewers estimated their bodyweight in a randomised mixed presentation. So they never saw the same person photographed at different distances.

Respondents were asked apply one of seven descriptors to each of five portraits: Very Overweight, Overweight, Slightly Overweight, Correct, Slightly Underweight, Underweight,. Very Underweight. These descriptors were given rising numerical (Likert) values from 1-7, with 1 equalling the lowest weight, 4 Correct weight and 7 Very Overweight. These values were averaged and assigned to the camera-to-subject distance at which each photo was taken.

Results
The data shows significantly rising weight estimations with increased camera-to-subject distances.

Gender Effects And Camera-To-Subject Distance: Wider Necks Can Increase Masculinity In Men While Making Slim Women Look Overweight Under The Same Photographic Conditions

Six Females And Six Males (BMI Matched As Underweight, Correct Or Overweight) Photographed At Six Rising Camera To Subject Distances

Method

The original seven Likert scale bodyweight descriptors were used again, as was randomised presentation.  However, each portrait subject’s weight and height (BMI) were also converted into a weight estimation chart individually calibrated in + or – kilograms. Each participant was asked to estimate the models weights, first using the Likert descriptors and then actual weight in kilograms using individual BMI weight estimation charts.

Rising Weight Estimations Trend Repeated, But The Average Weight Of All Models Was Judged To Be Overweight. Fattening Effect Levelled Off After 2.7 Metres (graphs below)

 Combined Male And Female Apparent Weight on Camera:

Both experiments above show the male/female fattening effect begins at
the 1 Meter Camera-to-Subject Distance. However,

The Same Data Separated By Gender shows only females fattened, with all seen as significantly over their actual body weight at all focal lengths

CONCLUSIONS:

All Females Were Judged To Be Significantly Overweight In All Conditions.
Almost All Males However Were Judged To Be At Or Below Their Actual Bodyweight.

The data for both genders does not correlate to their real-world average bodyweight: The female group actually averaged as normal weight in reality, but were all portrayed as significantly overweight in these experiments. The male group however was very slightly overweight on average, but was judged to be underweight compared to their actual BMI.

Jaw-Neck Ratio in
Males and Females

The overwhelming conclusion from these studies is that 2D photography is innately distorting to body image and is dramatically fattening to females. Currently, the only way to correct for these distortions is by using stereoscopic 3D photography with near life-size reproduction. Participant judgement of whole bodies based only on face and neck images showed a clear effect: Widening of the neck caused by telephoto lenses masculinized males but fattened females. The ecological validity of this observation matches what we know in reality. Testosterone and exercise can widen a male neck. But females typically can only develop a widened neck by significant increases in body fat. It is unsurprising therefore that artificially widenend necks in females drive perception to ascribe it to increased body fat, and not increased masculinity.

For more information, please contact bernieharper@gmail.com

< Research: BackgroundResearch: Conclusions: Next >

Hits: 197