Gaze Data: Viewing Portraits

BC: CS392 / PS392 / FA294, Fall 2006

Christopher Huang

Introduction

I am a TA for Introduction to Feminisms, which is a very interesting class. One lecture on the topic of the beauty industry and its implications on perception inspired me to do this project. I have noticed that we as a culture are very self conscious. We are bombarded by advertisements and media sources depicting unrealistically physically beautiful men and women and the constant barrage of unrealistically physically beautiful people causes us to feel inadequate about our own bodies. See Campaign for real beauty and Social Commentary in a slideshow.

This bombardment of such images causes us to find ways to try to not feel inadequate, whether it makes us buy diet pills, health food, makeup, skin products, other beauty products, obsessively work out, or the extreme, plastic surgery, all which leads to a multi billion dollar beauty industry. I have noticed just from social observation that many times it seems females, at least in the college aged population, are more judgmental about other females compared to males judging females or females judging males or males judging other males. I have not been able to find relevant literature on this topic. I have only found tangential literature that is not really related. My idea for this study was based mostly on my own observations and not on previous literature, which already introduces potential flaws in the study. The study design was completely designed from scratch. If I was able to find literature, I would base my design off of already established research. I was however able to find literature on self esteem which is related to the larger implication questions I am getting at with my study. Thornton and Moore (1993) in Physical Attractiveness Contrast Effect: Implications for Self Esteem and Evaluations of the Social Self showed that self ratings of attractiveness were enhanced by exposure to unattractive people and self ratings decreased when they were exposed to very attractive people. I wanted to explore this topic of superficiality in judging further using some sort of concrete scientific data that certainly is better than anecdotal hearsay and social observation. I wanted to view gaze data on images of this unrealistic beauty that is bombarded on a daily basis to try to explore the overall question of: how much are these images really affecting us? I have no way of answering that question. I can only find out how much we are looking at these images and see what I can infer from that. I am essentially trying to answer 4 questions, each of which should be its own well-designed, more concise study, but I tried to tackle all 4 in this one, which was a mistake on my part as I introduced too many independent variables at once mainly due to my curiosity. I am more interested in the first two questions.

Main Questions

1. Is there a difference in how we look at portraits when we are asked to do an innocuous task (like trying to figure out what nationality someone is) compared to when we are directly asked to judge their attractiveness level? In asking this question, I hope to find clues to help answer the overarching questions: where do we look in everyday interaction, even if we are not being judgmental about individual attractiveness? Do we automatically just look elsewhere besides faces (body, etc.)? If so, does that mean that we are very judgmental people when it comes to appearance, even when our task has nothing to do with judging their attractiveness? I do not and cannot directly establish that we are judgmental, but I can show a sign that we are if I find that the gaze data is similar for the innocuous task compared to when we are directly asked to judge individual physical attractiveness.

2. Is there a gender difference between the way we look at portraits of both men and women? We know from common sense that men tend to be superficial in where we look. We know from common sense that men are more likely to objectify women and look more at areas covered up by bathing suits. Is this true for women as well? Do women look at the same places men do? ? Are males or females more consistent in where they look? The main unanswerable broader question is this: if men and women tend to look at the same places, would this lead to perhaps a greater overall self-consciousness when it comes to appearances if we find that both men and women are looking at the same places?

3. Where do we look when we look at images? If we are looking at an image of a fully clothed person are we less likely to look at areas covered up by bathing suits or other areas on the body? If we are looking at the body just as much when people are clothed, what does that say about everyday interactions? Are we checking out the entire body even when people are clothed?

4. Does looking at double portraits, where there is both a man and a woman tell us anything else that might be relevant? Do men look more at the man or woman? Do women look more at the man or woman?

Predictions

1. People check out the body no matter what the task is. I assume most people will look more at the face, but people will still check out the body.

2. Men and women will show little to no difference in gaze patterns. I would not be surprised if I saw some cases of women looking more at a woman's body.

3. I do not think clothes matters that much. I think we will look at the body no matter what. We might look more at the body when skin is exposed which has scary implications. If we are looking that much, does that mean that the images we are bombarded with of unrealistically attractive men and women are affecting us more? I think we look at bodies no matter what.

4. I did not have a good prediction for this question. I predicted that we would see cases of men looking more at men, men looking more at women, women looking more at men, and men looking more at women. I do not think this can be generalized across genders.

The Experiment

Subject selection. Subjects were selected on a convenience volunteer basis. I asked my friends to come in for the study, which might be a problem if I unconsciously asked the people who I knew would support my predictions to be in it. I do not think I did though. To the best of my knowledge, assuming no one has been lying to me, all of the participants in the study were heterosexual. It would obviously change things if a few were homosexual although that would also be a potential additional variable to introduce in further study. Wood (2004) in The Gay Male Gaze, Body Image Disturbance and Gender Oppression Among Gay Men found that gay males experience greater levels of body dissatisfaction than other groups because of higher levels of gender nonconformity. There is a threat to internal validity in that the research was done by me and not a blind research assistant, clueless to what my hypothesis was. I may have unconsciously ran the experiment in ways that would support my predictions. Random selection of study subjects is ideal, although it is very unrealistic. I am trying to generalize to all men and women, but at best I can generalize to BC men and women, and I cannot even say that I represented the entire range in population of BC men and women. External validity, in particular, my ability to generalize to my target population of all men and women is threatened by poor subject selection. Due to my limited number of potential subjects, the experiment was treated as a within-subjects experiment (instead of a between subjects design), meaning there was no random assignment of tasks. This way I could see differences between people, however I introduce the problem of multiple treatment interference and potential attrition and fatigue during the 10 minute long study. Gaze data may have been affected by the fact that the task was repeated, or the fact that people get bored during that 10 minutes.

Procedure. Subjects were asked to come in the Eye Tracking Lab in Higgins 026 at Boston College which uses the Tobii Eye Tracker X50 with Clearview 2.6.3 as companion software. They were told that the study was on racial prejudice and stereotypes in an effort to give them a deceptive, distracting task attempting to remove the possibility that subjects might be aware that their gaze patterns were being watched. If I was a participant walking into an eye tracking lab, as opposed to a diversity and racial stereotypes lab, and was given the task of free viewing, I would be suspicious because I would know what the dependent variable is right away. I would probably change my gaze patterns knowing I was being watched. I asked all the subjects after they did the tasks whether they had inclination as to what the study was actually about and whether they thought their gaze patterns were accurate to how they would normally look at such images. Only one female mentioned that she shied away from looking at the body because she knew she was being watched. Two males noted that they were suspicious that it was an eye tracking lab. Others responded that they really believed that the study was on racial stereotypes and thought they looked at the photos naturally. Assuming no one lied to me, my deception worked well.

Subjects were calibrated to the Eye Tracker using only 5 calibration points. The decision to only use 5 points was due to a lack of understanding of how to use ClearView. It was set at 5 as a default when I first began a new project and I had already run roughly 10 subjects before I realized that I should have used more. 16 is ideal as it would provide more accurate gaze data. Some of the individual hot spot plots looked slightly off. There would be a hotspot on the upper neck instead of where I would expect it: the face, but I kept it at 5 for the rest of the subjects to maintain consistency.

Subjects were read instructions.

Condition 1: what do you think the ethnicity of this person is? Name a country where he or she might be from, not including the US unless you think the person is Native American

15 Images were presented for 15 seconds each in the sequence (the last fully clothed portrait was presented as the 15th portrait, but other than that, the sequence follows what is presented on the stimuli page) I chose to keep the same sequence instead of randomizing the image sequence because I did not see disadvantages to that; however it is potentially a confounding variable. 15 seconds might be too long a time though, as I may have unconsciously chose a time that would support my predictions. It seems that people might get bored after completing the task and look all around the photo. However, I am trying to generalize from these images to making statements about how we look at images in real life and how we look at real people in real life; in both of those cases, 15 seconds seems like a short time especially in the case of human to human interaction. Further studies into these questions might choose a time closer to 5 seconds to see if there are differences, or limit the data to 0-5s on what I have already collected using Clear View. As of now, my gaze data might be more reflective of the independent variable of time photo was presented, as opposed to just the photo itself. A blank page appeared giving them time to answer the question. The subject was asked to press the space bar to go to the next image. When they were finished with the first task, they were asked to repeat the task, this time under condition 2.

Condition 2: rate the physical attractiveness of this person based on a 1-10 scale, 10 being the highest

For the first few subjects, I wrote down exactly what they said verbally in response to the 2 different condition questions. I stopped doing this because it was not the dependent variable I was measuring. Also, I did not receive permission to use these photographs. I shot all of the Fully Clothed images (they asked me to shoot their portrait) and the third half clothed photo, but for a different reason. I never asked permission to use their photographs for this study and thought that reporting objectifying attractiveness data would be unethical, so I chose to not write anything down. Further studies on this matter might want to focus on attractiveness ratings as a variable to see if that affects where people look. Subjects were then debriefed and told what the actual questions in my study were. Subjects were asked if they thought had any inclination as to what the study was about and whether they looked naturally at the photos.

Results/Discussion

Data presented uses the relative fixation length option in the analysis section of ClearView showing the percentage. I could have used the absolute time too which might have been the better option given that each photograph was presented for the same amount of time (15 seconds) but in my decision process I chose the relative fixation length, which is certainly better for cases when the length of photo presentation is not uniform. Both options are presenting the same data though and it should not matter too much which one I chose. Further studies might analyze absolute time data to find further observations about the data.

One female subject was completely eliminated and not included in any of the data because of a problem with her pupils being extremely dilated. More than a third of her images had NO DATA ON SCREEN when viewing the hotspots. I had a total of 11 male subjects and 13 female, but cut one of them.

Main Questions

1. Is there a difference in how we look at portraits when we are asked to do an innocuous task (like trying to figure out what nationality someone is) compared to when we are directly asked to judge their attractiveness level?

In asking this question, I hope to find clues to help answer the overarching questions: where do we look in everyday interaction, even if we are not being judgmental about individual attractiveness? Do we automatically just look elsewhere besides faces (body, etc.)? If so, does that mean that we are very judgmental people when it comes to appearance, even when our task has nothing to do with judging their attractiveness? I do not and cannot directly establish that we are judgmental, but I can show a sign that we are if I find that the gaze data is similar for the innocuous task compared to when we are directly asked to judge individual physical attractiveness.

No major differences found. People look the same places when asked to judge ethnicity compared to attractiveness but there are a few subtle differences: we see slightly more gaze data looking at the body with the attractiveness condition.

Now what if we break it down further by gender in comparing the ethnicity condition to the attractiveness condition? Will we see anything new by breaking it down further?

Men: No major differences found. People look the same places when asked to judge ethnicity compared to attractiveness but there are a few subtle differences: we see slightly more gaze data looking at the body with the attractiveness condition.

Women: No major differences found. People look the same places when asked to judge ethnicity compared to attractiveness but there are a few subtle differences: we see slightly more gaze data looking at the body with the attractiveness condition.

Potential problems with these conclusions. There is a huge flaw in this study in that much of it is based on qualitative research. Further study into the question should focus on density plots in a way of quantifying the data. I personally made the judgment that these differences found, and there clearly are differences found, are subtle and not significant. I have no scientific backing for that: only my own analysis which might be influenced by my desire to demonstrate findings that fit with my prediction. Again, I am trying to be objective about this. Also, as previously discussed, the 15 seconds the photo was presented might have been too long. The gaze plots in the more surprising ethnicity condition might be better explained by boredom. However, I still show that in 15 seconds, eyes wander to other places besides the face in the innocuous ethnicity condition even if there is slightly more wandering in the attractiveness condition.

Discussion. I have demonstrated that in the 15 seconds eyes wander to other places besides the face in the innocuous ethnicity condition even though there is slightly more wandering in the attractiveness condition. What are the implications of this? We look the same places when we are judging attractiveness as we do in an innocuous situation. It shows that we are potentially just as judgmental (at least we look just as much) in innocuous situations as we are when we are directly judging attractiveness. The innocuous ethnicity condition is supposed to generalize to every day interaction with humans, although as previously discussed there are flaws in that external validity. Assuming that we can generalize this finding, that we look at the body even in innocuous situations as much (or almost just as much), it might help explain the overall self-consciousness that happens within society.

2. Is there a gender difference between the way we look at portraits of both men and women?

We know from common sense that men tend to be superficial in where we look. We know from common sense that men are more likely to objectify women and look more at areas covered up by bathing suits. Is this true for women as well? Do women look at the same places men do? ? The main unanswerable broader question is this: if men and women tend to look at the same places, would this lead to perhaps a greater overall self-consciousness when it comes to appearances if we find that both men and women are looking at the same places?

No major differences found. Men and women look the same places although there are a few subtle, not significant differences found. Men look at male bodies too. Women look at female bodies too.

The last conclusion is based on the assumption that I have already proved that there is no difference between the innocuous ethnicity and attractiveness conditions. But what if I did not prove that? Further Breakdown:

Innocuous Ethnicity Condition

No major differences found. Men and women look the same places although there are a few subtle, not significant differences found. Men look at male bodies too. Women look at female bodies too.

Attractiveness Condition

No major differences found. Men and women look the same places although there are a few subtle, not significant differences found. Men look at male bodies too. Women look at female bodies too.

Potential problems with these conclusions. There is a huge flaw in this study in that much of it is based on qualitative research. Further study into the question should focus on density plots in a way of quantifying the data. I personally made the judgment that these differences found, and there clearly are differences found, are subtle and not significant. I have no scientific backing for that: only my own analysis which might be influenced by my desire to demonstrate findings that fit with my prediction. Again, I am trying to be objective about this. Also, as previously discussed, the 15 seconds the photo was presented might have been too long. Further study into these questions should include a way of quantifying the data into numerical values (density plots, etc)

Also, since these are composites, it does not factor in individual differences. For example, if you combine two gaze plots of one person who looked at only the body and one other person who looked only at the face, the composite of those two would lead us to believe that everyone is looking at both the face and body when in fact one is looking at the face and one is looking at the body. There is danger in relying on these composites, which leads us to the next question.

Are males or females more consistent in where they look?

Questionable conclusion: it seems men might be more consistent in where they look where there seems to be a wider range for women.

Individual examples of large range. Face vs looking everywhere. I have included their data on double images if you are curious what their individual data looked like, although this is not relevant to the question at hand

Example 1: Female who looked mostly at the face.

Example 2: Female who looked mostly at the face.

Example 1: Female who looked everywhere.

Example 2: Female who looked everywhere.

Example 3: Extreme example of female who looked everywhere.

Potential problems with these conclusions. There are huge flaws in my attempt to answer this question. I failed to look at individual examples of males due to time constraints. My conclusions are questionable and again I have nothing but observational qualitative data to support my claims. I am limited by the number of subjects I was able to get as well. If I were to truly do a range study, I would need at least 50 if not more participants, 25 male, 25 female. Ideally, the more the better. Range questions are very difficult to answer with a limited number of participants.

Discussion. It would make sense that there is a greater range for females. From my own observations, women are taught different things about what should be viewed in attractiveness. Women are taught to be more proper and less superficial. In recent years, it seems that women have been taught to be more superficial by society. Men, unfortunately, have always been taught to be superficial and to look everywhere when looking at people. It seems a lot more socially acceptable for men to objectify than for women to objectify, assuming a link between looking at the body and objectification. No good conclusion can be drawn from looking at range data. However, this part of the study does point out a major flaw in grouping multiple hotspot gaze plots together. As I previously mentioned, it is potentially misleading because the composite is just that: a composite and potentially inaccurately generalizes when there are individual differences that cannot be ignored. This is a problem with any research in gaze data from eye tracking. Further study would explore this issue of range more, perhaps seeing if attractiveness as an independent variable will affect range. Range obviously needs to be quantified. I would certainly need to look at range data for more than just singular images of Brad Pitt and Adriana Lima, however I chose only those two because they are both considered by many to be equally highly attractive.

3.Where do we look when we look at images?

If we are looking at an image of a fully clothed person are we less likely to look at areas covered up by bathing suits or other areas on the body? If we are looking at the body just as much when people are clothed, what does that say about everyday interactions? Are we checking out the entire body even when people are clothed?

This was not one of my major questions I sought out to answer; however it is an important one in terms of its implications. Assuming that I have already demonstrated that there is no significant difference between the innocuous ethnicity condition and the attractiveness condition, and assuming that I have demonstrated that there is no significant difference between genders (both are questionable), the next logical step is to see if there is a difference in gaze patterns between the fully clothed stimuli and the half clothed stimuli.

We look everywhere no matter how much people are clothed.

Discussion. These are frightening results. Given what we have already concluded (questionable conclusions, but nonetheless, conclusions), these results suggest that even in everyday interactions, we are checking out bodies, perhaps helping to explain how superficial we are as a society. There is a problem, as previously discussed, in external validity. How well do these portraits of fully clothed or half clothed people generalize to real life? I was originally going to include videos of people talking to see if I could find data showing that we are potentially superficial in real world situations, but that would have added another variable to an already complex study. It would be interesting to study everyday interactions more in depth instead of making the potentially fallible leap from portrait viewing to everyday interactions.

4. Does looking at double portraits, where there is both a man and a woman tell us anything else that might be relevant? Do men look more at the man or woman? Do women look more at the man or woman?

No conclusions can be made, just observations..

This topic of double portraits should be studied further. It should be its own separate study. As of now, it only adds a little bit to my already complicated study.

Final Thoughts

As discussed throughout, there are flaws in this study. Despite them, I was able to demonstrate that 1) there is little difference in gaze data when we do an innocuous task compared to being directly asked to be superficial and judge physical attractiveness. 2)There is very little gender difference if any at all. 3)We look at the body just as much when people in portraits are fully clothed and when they are not. These have scary implications in dealing with how self conscious we are as a society. There are questions as to how good the external validity of this experiment is. The internal validity is already in question, especially the choice of the stimuli, and generalizing across independent variables (male vs female for instance) by using composites of individual data, but externally, we do not know how well the innocuous condition of guessing ethnicity accurately portrays everyday interaction. We also do not know how well portrait viewing translates to everyday interaction. Still, the results I have demonstrated have haunting implications. If all of my conclusions are accurate, it suggests that we are a very superficial and judgmental society, not immune to any sort of bombardment of images that portray images of unobtainable physical beauty.