Introduction: Pupilometry is the study of how a pupil reacts to different emotions and stimuli. The research on the topic of is scattered and fairly shallow. Related research has been conducted on facial expressions and their reaction and relation to emotion. There are some relationships to the facial expression research and, but these relationships do not tell the whole story. Some interest has been brought up through research in the field of and its predictive powers on emotion. Pupilometry has to do with the dilation and contraction of the pupil relative to the stimulus or emotion being studied.
In the early stages of pupil dilation research, there was intrigue as to the response of the pupil to emotional situations. That focus shifted toward information processing, and soon emotional studies were completely ignored. One emotional study that was done questioned whether pupil size change, as a response to seeing nudity, was a general indicator of arousal. Fifty women and fifty-seven men of the heterosexual college student variety, with a mean age of 21. 4 years were studied. Results of that experiment indicated that pupils did dilate significantly more for nude test group than they did for the clothed group, regardless of the sex of the subject or the sex of the pictured person.
It was the study's conclusion that the pupils dilated because of the increased emotional presence of arousal (Aboyoun and Dabbs 1998). Often, the eyes are poetically quipped as "the window to your soul", "mirror of you heart", or "the gauges showing fleeting feelings and changes" (Whiteside 1974). It has been studied that facial expressions can be inhibited by cognitively trying to hide them. In a study by Ursula Hess and Robert E. Kleck, spontaneous vs. deliberate facial emotions were studied.
It was concluded that people could deliberately control facial expression. Pupil reaction on the other hand is not controllable by cognitive means. The dilation and contraction of the pupils is an involuntary response based on reaction to stimulus and or emotion. Even if a person tried to hide their reaction to some stimulus physically, the size of the pupil could still be measured to observe a person's reaction to a situation (Hess, U. and Kleck 1997). Eckhard Hess states in his book that normal reactions to the emotion "happy" should indicate significantly larger pupil size than the emotion "mad." He goes on to suggest that large doe-eyed pupils are seen as happy, and small beady-eyed pupils are seen as mad (Hess, E.
1974). We will expand this theory even further, looking at the additional emotions of indifferent, sad, and surprised. It is our hypothesis that no difference exists between the means of each emotion. Ho: m Happy = m Mad = m Sad = m Indifferent = m Surprised.
Ha: otherwise. Methods: Initially our group explored potential concepts of how ideas and memories were evoked related to music. We were unable to develop a hypothesis from this area of study that interested us, and that had significant background information. Upon further research, we came upon the study of, and how the size of the pupil may be an indicator of emotion.
We came across a book written by Eckhard Hess. He has tested this theory of correlating the pupil size and emotions. The two emotions that he tested were Happiness and Anger. We have developed a hypothesis similar to Hess's. We have also expanded the testing to include the emotions of surprise, indifference, and sadness. A survey was created that recorded personal information from each subject.
This survey consisted of three double-sided pages. The first page collected personal data, the second through sixth focused on different emotions. For each emotion we have displayed five faces, all of which are identical in appearance (per emotion) changing only the pupil size. The size ranges from 20 pixels to 40 pixels in an increasing increment of 5 pixels, (20... 25... 30...
35... 40, ). This increment is an interval scale. The subject views the 5 faces and then makes a decision based on which face matches the emotion labeled.
Upon completion of the survey, data was tabulated the data and statistical testing was initiated. Analysis of Variance (ANOVA) was initially the statistic of preference, but due to non-normal data and inconsistent variation, two assumptions of the ANOVA test, we were unable to use that test. Instead, pair wise comparisons of each emotion were conducted using paired t-tests. This data was used to fuel our discussion. A sample survey is attached at the end of this paper. Results: Several statistical tests were applied to our data to try and determine whether pupil size is a good predictor of how people perceive emotion.
Initially, using Minitab, we output some basic descriptive statistics for all the different emotions. These outputs included a histogram, box plot, and confidence intervals for both the mean and median of each emotion, among others. Additionally, a box-plot of scores for each emotion was created to get a visual indication of what was going on. The evaluation of the descriptive statistics showed non-normality in both the mad and surprised emotion scores. Normality is intrinsic assumption of both regression and ANOVA. The box-plot also showed non-constant variance, which is another rejection criterion when using regression or ANOVA.
Without normality and constant variance, these tests cannot be trusted as to their outputs. Instead, we used paired t-tests between all of the different combinations of emotions. This allowed us to analyze each emotion and how it scored relative to each other emotion. The following are the results of the t-tests: Paired T for Happy - Sad Mean St Dev SE Mean Happy 135 3. 259 1.
321 0. 114 Sad 135 3. 022 1. 395 0. 120 Difference 135 0. 237 1.
870 0. 16195% CI for mean difference: (-0. 081, 0. 555) T-Test of mean difference = 0 (vs not = 0): T-Value = 1. 47 P-Value = 0. 143 Paired T for Happy - MadN Mean St Dev SE Mean Happy 135 3.
259 1. 321 0. 114 Mad 135 3. 230 1. 666 0. 143 Difference 135 0.
030 2. 262 0. 19595% CI for mean difference: (-0. 355, 0. 415) T-Test of mean difference = 0 (vs not = 0): T-Value = 0. 15 P-Value = 0.
879 Paired T for Happy - IndifferentN Mean St Dev SE Mean Happy 135 3. 259 1. 321 0. 114 Indifferent 135 2. 793 1. 134 0.
098 Difference 135 0. 467 1. 597 0. 13795% CI for mean difference: (0. 195, 0. 738) T-Test of mean difference = 0 (vs not = 0): T-Value = 3.
40 P-Value = 0. 001 Paired T for Happy - SurprisedN Mean St Dev SE Mean Happy 135 3. 259 1. 321 0. 114 Surprise 135 3.
007 1. 595 0. 137 Difference 135 0. 252 2.
051 0. 17695% CI for mean difference: (-0. 097, 0. 601) T-Test of mean difference = 0 (vs not = 0): T-Value = 1. 43 P-Value = 0. 156 Paired T for Sad - MadN Mean St Dev SE Mean Sad 135 3.
022 1. 395 0. 120 Mad 135 3. 230 1. 666 0. 143 Difference 135 -0.
207 2. 338 0. 20195% CI for mean difference: (-0. 605, 0. 190) T-Test of mean difference = 0 (vs not = 0): T-Value = -1. 03 P-Value = 0.
304 Paired T for Sad - IndifferentN Mean St Dev SE Mean Sad 135 3. 022 1. 395 0. 120 Indifferent 135 2.
793 1. 134 0. 098 Difference 135 0. 230 1.
723 0. 14895% CI for mean difference: (-0. 064, 0. 523) T-Test of mean difference = 0 (vs not = 0): T-Value = 1.
55 P-Value = 0. 124 Paired T for Sad - SurprisedN Mean St Dev SE Mean Sad 135 3. 022 1. 395 0. 120 Surprise 135 3. 007 1.
595 0. 137 Difference 135 0. 015 2. 168 0. 18795% CI for mean difference: (-0. 354, 0.
384) T-Test of mean difference = 0 (vs not = 0): T-Value = 0. 08 P-Value = 0. 937 Paired T for Mad - SurprisedN Mean St Dev SE Mean Mad 135 3. 230 1. 666 0. 143 Surprise 135 3.
007 1. 595 0. 137 Difference 135 0. 222 2. 288 0. 19795% CI for mean difference: (-0.
167, 0. 612) T-Test of mean difference = 0 (vs not = 0): T-Value = 1. 13 P-Value = 0. 261 Paired T for Mad - IndifferentN Mean St Dev SE Mean Mad 135 3.
230 1. 666 0. 143 Indifferent 135 2. 793 1.
134 0. 098 Difference 135 0. 437 1. 976 0. 17095% CI for mean difference: (0. 101, 0.
773) T-Test of mean difference = 0 (vs not = 0): T-Value = 2. 57 P-Value = 0. 011 Paired T for Surprised - IndifferentN Mean St Dev SE Mean Surprise 135 3. 007 1. 595 0. 137 Indifferent 135 2.
793 1. 134 0. 098 Difference 135 0. 215 2. 060 0. 17795% CI for mean difference: (-0.
136, 0. 566) T-Test of mean difference = 0 (vs not = 0): T-Value = 1. 21 P-Value = 0. 228 This data tells us that there are only two detectable differences in the scores of the emotions with an alpha rejection level of 5%. To put it another way, we are 95% confident that on future testing, the scores for mad and happy will be roughly 0. 1 to 0.
7 higher than the scores for indifference. Discussion: Initially, we intended to use within a subjects ANOVA to test the five different emotions as levels of the independent variable. Unfortunately we could not do so for two reasons. First, the software package we initially attempted to use was not robust enough to handle all of the subjects we had. The next package used was not readily accessible to testing within subjects as the independent variable (At least not to the experimenters).
The second problem with using the ANOVA analysis was more convincing than the first. This problem was with the data collected. ANOVA has four assumptions, 1) residuals of the data are normally and independently distributed with a mean of 0 and a standard deviation of 1, 2) The variance of the data is constant, 3) the means are constant, and 4) residuals are independent of one another. Our data turned out to be contrary to assumptions 1) and 2), hence damaging the validity of any ANOVA that would be done on our data.
In the place of ANOVA, we conducted pair wise comparisons of each emotion with each other emotion. The test used was a paired t-test, because each subject had a score for each emotion. Our results are neither dazzling nor surprising. The statistics show that our hypothesis is held true, that no difference actually exists by and large among all of the emotions studied. There were significant differences between the emotions of happy and indifferent, and mad and indifferent. Cleaver Post-how analysis might suggest that this is due to the fact that happy and mad are higher arousal emotions than indifferent.
This assumption would seem to draw a parallel to Aboyoun and Dabbs nudity experiment described in the results. If this were the case, additional studies may lead to investigating the level of arousal and its correlation to pupil size. For example, perhaps heart rate could be used as a measure of arousal. Different levels of "mad" and "happy" stimulus could be created and tested on the effects of heart rate in a reliability study.
Then the study could be re-run, assuming a strong correlation existed between the heart rate and levels of emotion, to see if pupil size actually is a good predictor of arousal. I do not think there were any confounding errors due to order. Each person administering the tests (5 people in the group administered them) had a different survey. The surveys varied in the order that the emotions were presented to the subjects.
Perhaps a confounding error might have occurred due to the fact that some people asked whether the test was looking for pupil response. If asked if the pupil was the only thing different the experimenter told the subject yes. If not asked, the experimenter did not give any indication to what the survey was testing. Another threat to validity of this test was that in some cases subjects "learned from other subjects." In other words, sometimes collaboration took place. The tests were generally given in a classroom setting, but strict supervision was not in place. If the experimenter was at one end of the classroom handing out surveys, he or she was not watching the rest of the subjects taking the test.
One possible explanation for the results we obtained is that people do not use pupil size as an indicator of emotion. More likely is that people use pupil size along with other physiological facial expressions, verbal and nonverbal cues, and posture to extrapolate an emotion of another person. Works Cited: Aboyoun, D. C, and Dabbs, J. M.
"The Hess pupil dilation findings: Sex or Novelty?" Social Behavior & Personality. Vol 26 (4). p. 415-419. 1998. Cornelius, Randolph R.
The Science of Emotion: Research and Tradition in the Psychology of Emotion. Prentice Hall, Upper Saddle River, NJ. 1996. Ekman, Paul E.
, Rosenberg, Erika L. What the Face Reveals; Basic and Applied Studies of Spontaneous Expression Using the Facial Action Coding System. Oxford University Press, New York, 1997. Goode, Erica E. , Schr of, Joanne M. , and Burke, Sarah.
"Where Emotions Come From. Psychology 97/98 p. 54-60, 62. 1998. Hess, Eckhard Heinrich. The tell-tale eye: how your eyes reveal hidden thoughts and emotions.
Van Nostrand Reinhold Co. New York. 1975. Hess, Ursula and Kleck, Robert. "Differentiating Emotion Elicited and Deliberate Emotional Facial Expressions." Series in Affective Science. P.
271-288. 1997. Russell, James A. , and Fernandez-Dols, Jose Miguel. The Psychology of Facial Expression.
Cambridge University Press, 1997. Whiteside, Robert L. Face Language: How to Read Anyone's Face Like a Book. Frederick Fell Publishers, Inc.
, NY, 1974.