Archive

Archive for the ‘Methodology’ Category

It Can Be In How You Look At It….

August 20, 2009 Leave a comment

Krupinski et.al. (2005) found that the use of LCDs to view radiographic images degrades reader performance. This work has led me to wonder if variation in human seated height, combined with the use of LCDs to present stimulus sets, might have a confounding effect on attractiveness research.

Liquid Crystal Displays – LCDs – are becoming increasingly common as computer monitors. I just checked an online electronics retailer and could not find a single CRT monitor for sale. LCDs are smaller, lighter, can last longer, and use less energy than CRTs – so the transition to the newer technology makes sense. Each type of technology has pros and cons, and there are significant performance differences between manufacturers and models. One consistent limitation of LCDs, however, is limited viewing angle: color and brightness changes occur as viewers move off-axis. Technological improvements have resulted in increased viewing angles in the latest, high-end LCDs. However, most of the improvement has been in the lateral rather than in the vertical direction. This image shows typical, vertical angle of view effects on a higher end laptop LCD screen.

LCD-Viewing-Angle-Variation

Left: iphone snap from a viewing angle of approximately 0°; Middle: snap from below; Right: snap from above.

Some observations:

  • Viewing from below primarily darkens the image while viewing from above lightens it
  • Viewing from below darkens the skin and increases contrast
  • Viewing from below can add more dimensionality to a face
  • Viewing from above primarily lightens the skin while decreasing contrast and dimensionality

While just rules of thumb, darkening the skin and increasing contrast are both masculinity enhancing photographic techniques. Lightening the skin and decreasing contrast are femininity enhancing techniques. Adding dimensionality is a technique used to make subjects appear to weigh less. Given that sexual dimorphism and apparent body mass are relevant to evaluations of attractiveness, I suggest that beauty researchers who use LCDs take steps to minimize possible confounding effects of variation in seated height when presenting stimuli on LCDs. As an initial suggestion, perhaps all subjects in beauty studies that utilize LCDs should be positioned to maintain a 0° angle between raters’ eyes and stimulus-image eyes? Possible approaches include:

  • Use of a chin rest (the technique used by Dr. Rhodes and colleagues (2007) in the study from which this image was taken)
  • Use of adjustable height seating
  • Use of adjustable height monitors

Additionally, research to investigate whether there is a measurable effect of seated height on attractiveness ratings (with LCD presentation) seems called for.

Cautions

  • Not all images will show this degree of off-axis variation
  • LCDs will show more or less off-axis variation than is visible here: each make/model is different

In closing, I will emphasize that the variations in the image above involve alterations of viewing angle that are substantially greater than would be expected given the variation in adult human seated height. Therefore, under normal research conditions the effect will not be as dramatic. However, in most/all LCDs, alterations in vertical viewing angle begin to show these changes rather quickly.

Wayne Hooke
ResearchBlogging.orgKrupinski EA, Johnson J, Roehrig H, Nafziger J, & Lubin J (2005). On-axis and off-axis viewing of images on CRT displays and LCDs: observer performance and vision model predictions. Academic radiology, 12 (8), 957-64 PMID: 16023384

Rhodes, G., Peters, M., & Ewing, L. (2007). Specialised higher-level mechanisms for facial-symmetry perception: Evidence from orientation-tuning functions Perception, 36 (12), 1804-1812 DOI: 10.1068/p5688

Categories: Methodology, The Face

Size UK

August 13, 2009 Leave a comment

Wells, et. al. (2007), have published data from 3D body scans taken from a 9617 member, cross-sectional sample of UK adults (stratified by age and SES). The study seems intended to gather commercial (garment sizing information) and health related information, but does have some implications for beauty research.

Implications:

  • Body shape correlates with age in females more obviously than in males.
  • Females more likely to become “apples” – or proportionally more similar to males – with age.
  • BMI is not a good predictor of body shape/proportion or waist circumference: at least at the border of the overweight range. Male subjects with a BMI between 24-25 had waist circumferences ranging from 29.7″ to 43.3″ while female subjects ranged from 28.6″ to 44.8″ (a 3.5 standard deviation range in waist circumference for a narrow range of BMI).
  • BMI is insensitive to age associated body weight redistributions.
  • Height and circumference explains most of the variance in weight in both men (91.7%) and women (94.8%). Thus, visual cues are strong predictors of weight.    

Beauty relevant findings:

  • Rank order of strongest predictors of weight in women: height, hip, bust, thigh, and waist.
  • Rank order of strongest predictors of weight in men: height, waist, chest, and thigh.
  • Thigh, arm, and waist girths are strongly(?) related to body fat – implied but not directly addressed in this article.
  • After age 30, mean male waist-hip-chest measurements maintain relatively constant ratios (see Fig. 2A).
  • After age 30, mean female waist circumference increases relative to hip, chest, and bust: leading to a decreased tendency toward hourglass figures with age (see Fig. 2B).

  • Average male waist circumference increases about 0.2″ per decade.
  • Average female waist circumference increases about 1.1″ per decade.

Strengths

  • Total body measurements from 3D scans accurate to 0.2″ were used
  • Large sample
  • Point-cloud data may be accessible to future research on attractiveness

Limitations

  • Sample may not be representative
  • Cross-sectional data do not identify individual developmental trajectories: some of the relationships in the data could be cohort specific

 

Wayne Hooke

ResearchBlogging.orgWells JC, Treleaven P, & Cole TJ (2007). BMI compared with 3-dimensional body shape: the UK National Sizing Survey. The American journal of clinical nutrition, 85 (2), 419-25 PMID: 17284738

When Attractive Isn’t Beautiful

August 6, 2009 12 comments

I was thinking about beauty research recently, and was struck by the thought that there seems to be a reduced range of attractiveness ratings in the stimulus sets used in contemporary studies of female body attractiveness. As a tentative check for what might be selective memory on my part, I did a quick and dirty search of the recent peer reviewed literature.

My peer reviewed search criteria were casual:

  • use a multi-field search of PsycINFO with the search terms “whr” and “attractiveness”;
  • articles had to directly test some component of female body attractiveness;
  • I had to be able to access a full-text version of the article; and
  • in cases of multiple articles using what appears to be the same stimulus set, include only one article.

Attractiveness Ranges in Peer Reviewed Publications

Approximate Range

Average Attractiveness

Source DOI

Comments

1.4-6.8

NR

10.1027/1016-9040.12.3.220  
2.2-6.8

4.9/10

10.1016/j.evolhumbehav.2008.08.007 Range extracted from male data
2.6-7.2

NR

10.1016/j.evolhumbehav.2007.09.002 Range converted from a 1-7 scale
NR

NR

10.1016/j.paid.2007.07.017  
4.9-7.6

NR

10.1016/j.paid.2006.12.003 Range converted from a 0-9 scale
1.4-8

NR

10.1111/j.1467-9450.2006.00526.x Range converted from a 1-9 scale
5.4-7.0

Multiple reported

10.1080/13548500500155982 Range converted from a 1-8 scale

 

Ranges are on a 1-10 scale. The conclusion to be drawn from this limited sampling is concerning: even the most attractive stimulus images in these studies seem to be relatively average. For comparison, I gathered attractiveness ratings from a small, opportunity sample of images from two online photo rating sources.

Attractiveness Ranges in On-Line Photo Rating Sites

Approximate Range

Average Attractiveness

Source URL

Comments

5.7-9.9

9.1

http://www.ratingmylooks.com/new_pics.php

Limited to viewable full bodies on the “Newest Members” page with at least 100 ratings

4.4-9.9

8.2

http://www.hotornot.com/

Limited to first 10 viewable bodies when the “Show Me” pull-down menu is set to “Women Only” aged 18-25 with at least 100 ratings

 

There appears to be a substantial difference, which, at the least demonstrates that average attractiveness ratings do not have to be in the average range of the scale. I have graphed the individual data points (average attractiveness ratings for each image) from the photo rating sites in the scatter plot below. The thick blue line indicates the average rating for the most attractive stimulus image in the peer reviewed studies I list in the table above. Clearly, high average attractiveness ratings are possible but are not typically being given to stimuli in the research literature.

For comparison, here’s a typical scatter plot from the research literature:

J RILLING, T KAUFMAN, E SMITH, R PATEL, C WORTHMAN (2009). Abdominal depth and waist circumference as influential determinants of human female attractiveness. Evolution and Human Behavior, 30 (1), 21-31 DOI: 10.1016/j.evolhumbehav.2008.08.007

 

Characteristically, the stimulus sets used in the research on female body attractiveness are nicely distributed around the mean, but attractiveness ratings from either end of the distribution are typically lacking. Off the top of my head, there are several potential problematic consequences:

  1. This sort of restricted range can have a significant effect on correlations. There is a nice java applet at a site housed by Rice University that lets one tinker with restricting the range of several data sets while recalculating regressions/correlations. Using the “Assumptions Met” data set and limiting the range of the data in the applet in order to more-or-less approximate the typically limited range found in beauty research we find a change from r = .60 to r = .38.
  2. Attractiveness may have nonlinear relationships with other variables. Restricted ranges may conceal these relationships and their strength.
  3. If edge or bow effects (a tendency for accuracy/discrimination to be higher at the extremes of the stimulus range) are present in evaluations of attractiveness, limited stimulus ranges will again conceal/alter relationship strengths. As far as I know, this possibility has not been explored in attractiveness research.

     

There are several possible explanations for the apparent restricted attractiveness range that we find in contemporary beauty research. A non-comprehensive list includes:

Medium Characteristics

  • Raters may have expectations of glamorous or artistic presentation for the very beautiful
  • Raters may have expectations of digitally altered figures or of figures in poses that enhance attractiveness

Stimulus Characteristics

  • Most/least beautiful stimuli may be absent from the stimulus sets
  • Certain clothing may conceal beauty relevant characteristics
  • Innocuous postures may communicate nonverbal information that flattens attractiveness ratings

Rater Characteristics

  • Raters in scientific studies may use more rigorous checks on their ratings of stimulus imagery than do raters at online photo rating sites
  • Raters in scientific studies may use different criteria when checking their ratings of stimulus imagery than do raters at online photo rating sites

In summary, there appears to be a restricted range in the attractiveness ratings of stimuli in contemporary research on female body attractiveness. If this is correct, there is a significant chance that some aspects of our understanding of attractiveness might be compromised.

Wayne Hooke
ResearchBlogging.orgSwami, V., Neto, F., Tovée, M., & Furnham, A. (2007). Preferences for Female Body Weight and Shape in Three European Countries European Psychologist, 12 (3), 220-228 DOI: 10.1027/1016-9040.12.3.220

RILLING, J., KAUFMAN, T., SMITH, E., PATEL, R., & WORTHMAN, C. (2009). Abdominal depth and waist circumference as influential determinants of human female attractiveness☆ Evolution and Human Behavior, 30 (1), 21-31 DOI: 10.1016/j.evolhumbehav.2008.08.007

 
Sorokowski, P., & Pawlowski, B. (2008). Adaptive preferences for leg length in a potential partner Evolution and Human Behavior, 29 (2), 86-91 DOI: 10.1016/j.evolhumbehav.2007.09.002

 
SWAMI, V., MILLER, R., FURNHAM, A., PENKE, L., & TOVEE, M. (2008). The influence of men’s sexual strategies on perceptions of women’s bodily attractiveness, health and fertility Personality and Individual Differences, 44 (1), 98-107 DOI: 10.1016/j.paid.2007.07.017

 
Singh, D., & Randall, P. (2007). Beauty is in the eye of the plastic surgeon: Waist–hip ratio (WHR) and women’s attractiveness Personality and Individual Differences, 43 (2), 329-340 DOI: 10.1016/j.paid.2006.12.003

 
SWAMI, V., & TOVÉE, M. (2007). Perceptions of female body weight and shape among indigenous and urban Europeans Scandinavian Journal of Psychology, 48 (1), 43-50 DOI: 10.1111/j.1467-9450.2006.00526.x

Furnham, A., & Reeves, E. (2006). The relative influence of facial neoteny and waist-to-hip ratio on judgements of female attractiveness and fecundity Psychology, Health & Medicine, 11 (2), 129-141 DOI: 10.1080/13548500500155982

Categories: Methodology, The Body

Routine Vision Screening for Eye Tracking Studies

August 1, 2009 Leave a comment

Eye tracking studies are likely to increase in frequency and influence in the study of human beauty. Many eye tracking studies do not report screening for refractive error. According to estimates derived from data gathered in the National Health and Nutrition Examination Survey (1999-2004), “clinically important refractive error affects half of the US population 20 years or older.” (Vitale, et.al., 2008) As a result, I would suggest routine vision screening of all subjects in eye tracking studies.

Wayne Hooke 

Susan Vitale; Leon Ellwein; Mary Frances Cotch; Frederick L. Ferris III; Robert Sperduto. “Prevalence of refractive error in the United States, 1999-2004″
Arch Ophthalmol 2008; 126: 1111-1119.

 

Blurry, resized Snellen Chart courtesy of Wikipedia Commons.

Categories: Methodology

Beauty is in the Eye Movements of the Beholder

July 21, 2009 1 comment

An ongoing debate today is the relative contribution of total body fat and waist-hip ratio to evaluations of female body attractiveness. Converging research has tended to support a more significant role for total body fat (as measured by BMI or VHI) than for WHR. Using a novel eye tracking methodology, Cornelissen, et.al. (2009) have found that eye movements made during evaluations of body attractiveness are more similar to eye movements made when assessing total body fat rather than to those made when assessing waist-hip ratio.

Figure 1 The left and central columns of bodies show contour plots of the fixation distributions for the attractiveness (ATT), body fat (FAT) and WHR estimation conditions for both genders overlaid onto the image of the reference body. In order to facilitate inspection of the data across all conditions, fixation density in the left and central columns has been converted to a percentage score, indicated by colour bars (0-100), with red indicating the highest density. The right column of bodies shows the differences in the fixation density (i.e. differences in raw scores) between the genders for the three conditions. Positive differences are shown as red/yellow colours; negative differences are shown as blue/cyan colours. The black contours demarcate regions within which these differences are statistically significant as determined by the GLMMs. (Figure and description from: Cornelissen, p. 9)

 

Strengths

  • Carefully mapped, high resolution eye tracking data was used.

Cautions

  • Male evaluators minimally fixated on the edges of the torso when evaluating WHR. This is hard to make sense of.
  • Methodology assumes that foveal detail is central in processing body attractiveness. While this seems likely, it remains possible that significant data is accessed in more peripheral visual areas. Foveal detail takes in approximately 2° of the visual field while the functional visual field varies from 4-90° (Murata: PMID: 15151160; or Questia Excerpt ) – in this study, hip width was reliably within the functional visual field when fixation was at the midpoint of the body (4-4.5°). Cornelissen et.al. provided a check for this potential confound by presenting the same stimuli to different evaluators, this time establishing that the focal point was in the center of the torso while presenting the image for only 100ms. Evaluators were not able to accurately estimate WHR or fWHR (the apparent WHR that is determined simply by using the 2 dimensions available to the evaluator in the photographic display). This check does provide support for the methodology used in the primary study, but some may not view this check as compelling.
  • A Tobii 1750 Eye Tracker was used to track eye movements. The Tobii 1750 is accurate to 0.5°; drifts less than 1.0°; and has head movement compensation error of less than 1.0°. Assuming these errors are potentially additive, individual measurements could be off by no more than 2.5°. I have never used eye tracking technology and so have no expertise with it: but, the published accuracy parameters seem uncomfortably large when attempting to track eye movements across a 4.0-4.5° target field.
  • Analysis was conducted on fixation density (a count of fixations on regions of the body) – no sequential analyses are reported in this publication (e.g.: waist → hip → breast → upper arm → waist).
  • Optimal BMI/VHI ratios appear under represented and may not be present in the stimulus set. Average BMI was 22.3 (sd = 2.3).
  • The mean WHR in this study is .74 (sd = .04) with a range of .64 – .84: we can estimate that 7-8 of the 46 stimulus images contained optimal or curvier waist lines. This preponderance of straighter waist lines could decrease visual interest in this area and might have influenced the outcome of the study.

This study provides further, substantive, converging evidence to support the relatively greater importance of total body fat over waist hip ratio in the evaluation of female body attractiveness. Suggestions for further research using eye movement tracking methodologies might include:

  • Sequential analyses of eye movement data
  • Utilization of stimulus images with dimensions that take-up significantly more than 4-4.5° of the visual field
  • Utilization of more optimal/attractive stimulus images
  • Utilize images that hold BMI/body fat constant, but still vary substantially in appearance

Wayne Hooke
ResearchBlogging.org
Cornelissen, P., Hancock, P., Kiviniemi, V., George, H., & Tovée, M. (2009). Patterns of eye movements when male and female observers judge female attractiveness, body fat and waist-to-hip ratio Evolution and Human Behavior DOI: 10.1016/j.evolhumbehav.2009.04.003

Categories: Methodology, The Body

3D Modeling: Promising Technology for Exploring Bodily Attractiveness

April 18, 2009 1 comment

3D modeling technology, like that used below, shows great promise in researching bodily attractiveness. Current technology allows actual bodies to be scanned and represented in 3D format. Here are two examples from published studies.

Alternatively, purely digital creations can be used.

 

The shapes of the bodies can be retained and variously presented: as either static images or rotating 3D movies. Features unrelated to body shape/size/proportion can be easily removed (like clothing, skin tone, complexion, facial features, etc.). Further, subtle, undetectable digital alterations in relevant features (such as shape/size/proportion) are easily implemented on the 3D images – which should enable researchers to explore more fine-grained aspects of bodily attractiveness. Depending on the research question, this technology offers several advantages over line drawings and photographs and I encourage beauty researchers to consider the use of this technology.

Wayne Hooke

 

Image Sources:

Mesh Skinned 3D model:

Fan, J., Dai, W., Qian, X., Chau, K.P., Liu, Q. (2007). Effects of shape parameters on the attractiveness of a female body. Perceptual and Motor Skills, 105(5), pp. 117-132.

Gray Skinned 3D model:

Brown, W.M., Price, M.E., Kang, J., Pound, N., Zhao, Y., Yu, H. (2008). Fluctuating asymmetry and preferences for sex-typical bodily characteristics. Proceedings of the National Academy of Sciences, 105(35), pp. 12938-12943. DOI: 10.1073/pnas.0710420105

Sydney 3D Model:

Available at: http://my.smithmicro.com/dr/poser.html

Categories: Methodology, The Body

No Evidence of the Good Genes Hypothesis Found

February 20, 2009 13 comments

Using photographs of real men, Peters, et.al (2009) found no evidence of a preference for either masculinized or symmetric male faces or bodies in ovulating women.

Previous studies that have found a relationship between ovulation and attraction to masculine features have used computer-morphed images that are weak in ecological validity. This study used photographs of actual men, like the ones below.

Masculinity, attractiveness, and symmetry ratings of the stimuli appear to approximate a normal distribution, strengthening the ecological validity of this study. The only noteworthy limitation in this design is that there were no objective measurements of masculinity or symmetry – only subjective ratings were used.

The authors were also careful to use precise measurements of ovulation to ensure that the ratings of women in the ovulatory phase were well-within the previously identified six-day long sexually active phase of the menstrual cycle.

Wayne Hooke

Marianne Peters, Leigh W. Simmons, Gillian Rhodes (2009). Preferences across the Menstrual Cycle for Masculinity and Symmetry in Photographs of Male Faces and Bodies PLoS ONE, 4 (1) DOI: 10.1371/journal.pone.0004138
ResearchBlogging.org

Follow

Get every new post delivered to your Inbox.

Join 25 other followers