Recently, greater emphasis has been placed on smile esthetics in dentistry. Eye tracking has been used to objectively evaluate attention to the dentition (mouth) in female models with different levels of dental esthetics quantified by the aesthetic component of the Index of Orthodontic Treatment Need (IOTN). This has not been accomplished in men. Our objective was to determine the visual attention to the mouth in men with different levels of dental esthetics (IOTN levels) and background facial attractiveness, for both male and female raters, using eye tracking.
Facial images of men rated as unattractive, average, and attractive were digitally manipulated and paired with validated oral images, IOTN levels 1 (no treatment need), 7 (borderline treatment need), and 10 (definite treatment need). Sixty-four raters meeting the inclusion criteria were included in the data analysis. Each rater was calibrated in the eye tracker and randomly viewed the composite images for 3 seconds, twice for reliability.
Reliability was good or excellent (intraclass correlation coefficients, 0.6-0.9). Significant interactions were observed with factorial repeated-measures analysis of variance and the Tukey-Kramer method for density and duration of fixations in the interactions of model facial attractiveness by area of the face ( P <0.0001, P <0.0001, respectively), dental esthetics (IOTN) by area of the face ( P <0.0001, P <0.0001, respectively), and rater sex by area of the face ( P = 0.0166, P = 0.0290, respectively). For area by facial attractiveness, the hierarchy of visual attention in unattractive and attractive models was eye, mouth, and nose, but for men of average attractiveness, it was mouth, eye, and nose. For dental esthetics by area, at IOTN 7, the mouth had significantly more visual attention than it did at IOTN 1 and significantly more than the nose. At IOTN 10, the mouth received significantly more attention than at IOTN 7 and surpassed the nose and eye. These findings were irrespective of facial attractiveness levels. For rater sex by area in visual density, women showed significantly more attention to the eyes than did men, and only men showed significantly more attention to the mouth over the nose.
Visual attention to the mouth was the greatest in men of average facial attractiveness, irrespective of dental esthetics. In borderline dental esthetics (IOTN 7), the eye and mouth were statistically indistinguishable, but in the most unesthetic dental attractiveness level (IOTN 10), the mouth exceeded the eye. The most unesthetic malocclusion significantly attracted visual attention in men. Male and female raters showed differences in their visual attention to male faces. Laypersons gave significant visual attention to poor dental esthetics in men, irrespective of background attractiveness; this was counter to what was seen in women.
Dental attractiveness for men is not tied to facial attractiveness levels.
At IOTN 10, more attention was paid to the mouth than to the eyes in men.
Women view the eyes more than men do when viewing men.
In recent years, increased emphasis has been placed on esthetics in dentistry. It has been suggested that dentists should plan treatment that considers not only functional, but also esthetic, objectives because most patients said that they were interested in improving the appearance of their teeth.
Shaw and Shaw et al were some of the first to begin to understand the effect of dental esthetic alterations and the way that they affect how people are judged. They showed that 11-year-old children with normal incisors were judged to be more desirable as friends, better looking, more intelligent, and less aggressive. With young adults, they showed that although background facial attractiveness was more important than dental condition, a normal dental appearance (normal incisor relationship) was judged to be more socially attractive over a range of personal characteristics.
In 2015, using an online survey conducted by the Harris Poll, 14,962 responses from a randomly selected study group of people, ages 18 and older, were analyzed. Twenty-nine percent of low-income adults and 28% of young adults (18-34 years) believed that the appearance of their mouth and teeth affects their ability to interview for a job. Twenty-five percent of all adults avoid smiling, 23% feel embarrassed, and 20% experience anxiety due to the condition of their mouth and teeth. Finally, 82% of all responders agreed that “it is easier to get ahead in life if I have straight, bright teeth.” Overall, it is clear that the general population feels that dental esthetics are important, and they also have an influence on psychosocial judgments.
Other studies have been conducted to explain what is an esthetic and visually pleasing smile. The aims in these studies have been to determine what the raters found most esthetic and what deviations from ideal were acceptable. For example, ideal upper to lower midlines have no deviation, but it is acceptable to vary 2 to 3 mm from the ideal and still be considered esthetic. These authors used a specific perspective (circumoral only, lower face only, or full-face views) with different rating groups (laypersons, dental professionals, different sexes). These studies may have been biased because they directed the raters’ attention to the dentition.
Through these studies, many important confounding factors have been uncovered. A model sex effect has been shown that affects different preferences in dental esthetic characteristics. Also, in general, female models have been shown to be judged more critically. Furthermore, the sex of the rater may also be influential on the dental esthetic ratings made.
A number of studies have demonstrated that dental professionals tend to be more critical of dental esthetic aberrations than the lay population, although this is not always the case. Additionally, the perspective that the raters are viewing the model in may be important. When a full-face perspective is used, the background facial attractiveness of the model and its possible effect on dental esthetic perceptions must also be considered. Chang et al demonstrated different preferences for models of different facial attractivenesses for dental variables that had a facial context.
To achieve an objective measure of a person’s visual attention to the face, eye tracking can be used to record where persons are looking when they view a face. Eye trackers can provide a quantitative measure of real-time visual attention. The majority of viewing time is spent in fixations (90%), whereas the remaining portion involves saccades, which are the fast eye movements to reposition the fovea, and occur when visual attention is directed to a new area.
Eye-tracking cameras use video-processing software to track the pupil with infrared or near infrared light, and corneal reflection is used to record visual attention. Viewers are calibrated to several predetermined positions before viewing the images. Typically, it is the fixation that is used to determine visual attention. This method maintains the most essential information for understanding cognitive and visual processing behavior.
Eye tracking has been used in dentistry. Hickman et al showed that in well-balanced faces of orthodontically treated patients, no single area had a significantly greater amount of visual attention, and that the mouth was only a small part of the visual attention at 10%. Subsequently, the eye-tracking characteristics of female models with different levels of dental esthetics and background facial attractiveness were reported by Richards et al. As a follow-up to this study, Johnson et al looked at some esthetic borderline IOTN treatment need levels in women.
Wang et al showed a significant deviation in the scan path of pretreatment patients compared with normal and posttreatment patients, and that orthodontic treatment normalizes the scan path.
In this study, we examined male models in the same manner as did Richards et al and Johnson et al. Our specific aim was to ascertain whether there is a point on the aesthetic component of the Index of Orthodontic Treatment Need (IOTN-AC) when the severity of dental esthetics would be enough to attract the most visual attention in male faces. A secondary aim was to determine whether background facial attractiveness level (attractive, average, unattractive), or the sex of the rater, had any effect on the visual attention to dental esthetics.
Material and methods
The first preliminary step for the development of this project was to obtain models for background facial images to be used in the study. Potential models were recruited at Ohio State University, Columbus. Two frontal full-face digital images were obtained for each consenting subject (EOS Digital Rebel XT camera, Canon, Melville, NY)- 1 social smile where the participants showed the teeth and 1 photo with no teeth visible. Facial attractiveness was rated from the image with no teeth visible to avoid the potential that the facial attractiveness of the models would be affected by their natural dentition.
Model images were excluded that were judged by the researchers to have a significant distraction, such as a facial tattoo, extreme facial hair or hairstyle, asymmetry, abnormal piercing, and so on. Models were 18 to 30-year-old white men in an attempt to eliminate the variable of ethnicity and to control for sex. In total, 109 facial images were collected and included for the background facial attractiveness ratings.
The images showing no teeth were rated by peer young adults who had no background or formal education in dentistry (laypersons). They rated the images as attractive, average, or unattractive, or 3, 2, and 1, respectively. Reliability was determined by repeating all images in random order during the same rating session. Twenty raters evaluated the images, and reliabilities were good for intrarater (weighted kappa, 0.67; 95% confidence interval [CI], 0.64-0.70) and poor for interrater (weighted kappa, 0.35; 95% CI, 0.34-0.37).
The images were selected based on the mean value of the ratings. The 3 images with the highest mean closest to 3 were selected as attractive, the 3 closest to 2 were selected as average, and the lowest 3 image means were selected as unattractive. This provided the greatest spread among the images for the level of facial attractiveness.
Frontal facial images of various levels of dental esthetics were then selected from the database of the Ohio State University orthodontic clinical archive. These images were classified in accordance with the IOTN-AC by researchers. The IOTN-AC is based on 10 levels of malocclusion defined by esthetic impairment, a method that was tested for reliability and validity by Richmond and described by Borzabadi-Farahani. Level 1 represents the least treatment need, level 7 is the upper limit of borderline need, and level 10 is a clear need for treatment based on esthetic grounds. To verify the researchers’ classifications of the IOTN-AC level, experienced orthodontists (15 full-time and part-time university faculty) were surveyed to verify agreement with the level of malocclusion as defined by the IOTN-AC. All images were rated twice to determine reliability. The details of this method have been previously described. The intrarater kappa statistic was 0.72 with 91.7% agreement, and the interrater kappas statistic was 0.56 with 83.1% agreement for all images.
Next, the selected background facial attractiveness model smiling images were combined with the verified IOTN dental esthetic-level images. The models for each background attractiveness level were drawn randomly for combination with 1 of the 3 IOTN malocclusion levels: 1, 7, or 10. Composite images were created using Photoshop Elements 11 software (Adobe Systems, San Jose, Calif) by removing the models’ existing dentition and creating a new layer for the verified IOTN esthetic level, with adjustments for scale and color as needed. The backgrounds of all images were standardized with a basic gray hue, and the scale of each model’s head in the background was visually standardized between all images. In total, 9 composite images were generated, 1 for each malocclusion and attractiveness level: attractive IOTN 1, attractive IOTN 7, attractive IOTN 10, average IOTN 1, average IOTN 7, average IOTN 10, unattractive IOTN 1, unattractive IOTN 7, and unattractive IOTN 10 ( Fig 1 ). In a previous study with facial images of women, reliability in each dental and facial attractiveness level was established; thus, for this study with men, only 1 image per level was used.
Interest areas for the 9 composite images were defined (Experiment Builder; SR Research, Mississauga, Ontario, Canada). The areas mapped were eyes, nose, mouth, ears, eyebrows, chin, hair, and cheeks ( Fig 2 ). To simplify the analysis, the areas were classified as eye, nose, mouth, chin, ear, and other (including hair, eyebrows, cheeks, and so on).
The eye-tracking camera used for the experiment, Eyelink 1000 (SR Research) has a reported average accuracy of 0.25° to 0.5°. For this reason, some space was left between the mapped areas to allow for this degree of error.
Institutional review board approval was obtained for the study (Behavioral Sciences IRB #2012B0414). Raters were recruited on a university campus. Inclusion criteria for raters were set as follows: 18 to 30 years old, able to understand English, no known neurologic conditions, normal or corrected to normal color vision (no hard contact lenses), willing to attend 20 to 25-minute eye-tracking sessions, no recent consumption of drugs or alcohol, no current medications that could affect cognitive abilities, and no mascara or willing to remove it as needed for the session. Dental professionals and dental students were excluded, with the intent of a layperson perspective for the study.
Written informed consent was obtained for all raters. Deception was used so that the raters were not more prone to look at the mouth. Raters were told that this was a study to look at how people view faces, and a debriefing form was given after the study to inform the participants of the actual nature of the study. Additionally, a short questionnaire was given that inquired about the rater’s age, ethnicity, and sex.
Raters were positioned in a table-mounted head-stabilizing device. Each rater was calibrated on a 9 or 5 point calibration with the eye tracker and instructed how the program would run. The eye-tracking data were captured using Data Viewer software (SR Research) for duration of fixations (total time in milliseconds spent viewing each interest area) and density of fixations (total number of fixations in each interest area).
The 9 composite male images were presented in a different order to each subject, and the randomization was determined by the computer program. Five sample images were viewed before the experimental images to allow the participants to familiarize themselves with the program and how it would run before any data were collected. All images were repeated a second time, again in a computer-determined random order, for reliability. To summarize, each subject saw 5 practice images, rated 9 images, and rerated 9 images for reliability for a total of 23 images. Each was viewed for 3 seconds. A flowchart of the procedures is included ( Fig 3 ). Between each image, a cross was displayed in a random changing location on a blank screen to eliminate the possibility that the location of the gaze was fixed from 1 image to the next. A total of 87 subjects consented for the experiment, and 23 were excluded. Eight were excluded for the inability to calibrate them in the eye tracker, 1 was excluded for not completing the demographic information, and 14 were excluded for being of a different race than white. This gave a total of 64 participants meeting the inclusion criteria; 41 were male, and 23 were female.
Sample size and power were determined using data from the study of Richards et al as follows: 2 outcome variables were of interest: area of maximum number of fixations (density) and area of maximum duration (duration). Assuming a correlation of 0.4, for area of maximum number of fixations (density) and assuming a standard deviation of 2.178 observations, a sample of 64 subjects would enable detection of a mean difference of 1.0 with a power of 0.91. With a sample size of 64 subjects and assuming a standard deviation of 687.3 ms and a correlation of 0.4,28, a difference of 300 ms could be detected for area of maximum duration with a power of 0.88. All power calculations were based on a nondirectional alpha risk of 0.05.
The results were analyzed with factorial repeated measures analysis of variance. The independent variables were level of dental attractiveness (3 levels: IOTN 1, 7, 10), background attractiveness (3 levels: unattractive, average, attractive), rater sex (2 categories: male, female), and area of the face (3 categories: eye, mouth, nose). Tukey-Kramer post hoc testing was used for multiple comparisons. A log scale transformation was used for density and square-root transformation to achieve a better fit of the respective models.
Reliability was evaluated by repeating all images during the viewing session. The first viewing of the image was used for the data analysis, and the second was to evaluate reliability. Intrarater and interrater reliabilities were evaluated with the intraclass correlation coefficient.
Density was the measure of the total number of fixations in an interest area, and duration was the total time in milliseconds that the observer spent viewing a given interest area. Reliability was assessed; an intraclass correlation coefficient less than 0.40 was poor, 0.40 to 0.59 was fair, 0.60 and 0.74 was good, and greater than 0.75 was excellent. Intrarater and interrater reliabilities were good and excellent in the areas of eye, nose, and mouth for both density and duration of fixations (0.6-0.91). For this reason, these were the areas included in the data analysis. The reliability for the areas of chin, ear, and other had poor reliability in either density or duration, and were eliminated from the data analysis.
For density, significant interactions were seen for the variables of model facial attractiveness and area of the face ( P <0.0004), dental esthetic level and area of the face ( P <0.0001), and area of the face and rater sex ( P = 0.0098). Additionally, the main effects of area of the face were also significant ( P <0.0001). For facial attractiveness by area of the face ( Fig 4 ) in men of average attractiveness, the mouth received significantly more visual attention than it did in the unattractive ( P = 0.0023) and attractive ( P <0.0478) men. In average men, the mouth exceeded both the eye and the nose ( P <0.0001). For unattractive and attractive men, the hierarchy for visual attention was the same: the greatest density at the eyes followed by the mouth and the nose. The eyes were significantly different from the nose in both cases ( P <0.0001).
For dental esthetic level and area of the face ( Fig 4 ), with no treatment need (IOTN 1), the eye received significantly more visual attention than did the mouth ( P <0.0001). For subjects with borderline dental esthetics (IOTN 7), the mouth had significantly more visual attention than it did for subjects in IOTN 1 ( P = 0.0014) and significantly more than the nose ( P = 0.0001), but it was not statistically distinguishable from the eye. At IOTN 10, the mouth received significantly more visual attention than it did in the borderline models (IOTN 7) ( P <0.0001), and it significantly exceeded the eye ( P <0.0001) and the nose ( P <0.0001).
For area of the face and rater sex ( Fig 4 ), both male and female raters gave significantly more visual attention to the eyes than the nose ( P <0.0007 and P = 0.0001, respectively). Women gave significantly more visual attention to the eyes than did the men ( P = 0.0226). Male raters showed significantly more visual attention to the mouth than the nose ( P = 0.0001).
For duration, the pattern was the same as for density. Significant interactions were found in the same areas: model facial attractiveness and area of the face ( P <0.0013), dental esthetic level and area of the face ( P <0.0001), and area of the face and rater sex ( P = 0.0087). Additionally, the main effects of dental esthetics (occlusion) and area of the face were significant ( P <0.0018 and P <0.0001, respectively).
For facial attractiveness by area of the face ( Fig 5 ), again significant differences were seen in duration, with men of average attractiveness receiving more visual attention to the mouth than the unattractive ( P = 0.0447) and attractive ( P <0.0030) models. Again, the hierarchy for unattractive and attractive men was eye, mouth, and nose, with the eye receiving significantly more visual attention than the nose ( P = 0.0001, unattractive; P <0.0001, attractive), but only more than the mouth for attractive men. For average men, the hierarchy was mouth, eye, and nose (but the eye and mouth were not significantly distinguishable here).