[1] 8w Gender Differences in Electrophysiological Responses to Facial Stimuli

[1] 214w Background: A facial discrimination task was adapted to be used in an event-related potential (ERP) paradigm in order to evaluate whether young women's brain responses to affective stimuli differed from those of young men. Methods: The stimuli used to generate a late positive component of the ERP, designated the "P450," were male and female faces with neutral, sad, or happy facial expressions. Subjects were instructed to respond to the happy and sad faces but not to the neutral faces. The amplitude and latency of the P450 component was evaluated with respect to the gender of the subject, as well as the gender and emotional affect of the facial stimuli themselves.Results: In all subjects, the sad faces elicited longer latency and higher amplitude P450 components as compared to the happy faces. Female subjects were found to generate significantly longer latency and higher amplitude P450 components than male subjects to both happy and sad faces. All subjects were found to respond more quickly to: male happy faces Ͼ female happy faces Ͼ female sad faces Ͼ male sad faces. Conclusions: These data suggest that the morphology of the late positive component of the ERP differs depending on the emotional expression of the stimuli, the gender of the facial stimulus, and the gender of the subject. Biol

[1] 121w O ver the last decade, there has been increasing interest in identifying electrophysiological responses to affective stimuli to provide new tools by which to explore differences between individuals in the processing of affective stimuli. Several studies have utilized facial stimuli to elicit event-related potentials (ERPs) (Vanderploeg et al 1987;Nelson and Nugent 1990;Lang et al 1990;Kesten-baum and Nelson 1992). Most investigators have identified a late positive going component that peaks between 300 and 600 msec, depending on the paradigm, after the presentation of an affective stimulus (Johnston et al 1986;Barret and Rugg 1989;Sommer et al 1991). This component has typically been labeled by some researchers as a P300 late positive component, or a P450 wave (Lang et al 1990;Kestenbaum and Nelson 1992).

[2] 142w Recent studies using facial stimuli have attempted to delineate how this ERP component varies in morphology as a function of stimulus conditions. Almost all studies have reported larger late positive component amplitudes in response to more "emotional" facial expressions compared to neutral expressions (Roschmann and Wittling 1992;Laurian et al 1991;Johnston et al 1986;Vanderploeg et al 1987). A comparison of P450 amplitude responses to different "types" of emotional expressions has yielded less consistent results. For instance, Nelson and Nugent (1990) found that the late positive component amplitudes were larger in target or infrequently presented stimuli, regardless of facial expression. Other investigators have found that P450 component amplitudes are larger when elicited by happy faces than by sad faces (Kesenbaum and Nelson 1992). Yee and Miller (1987) reported a larger P300 amplitude elicited by viewing unpleasant scenes than to pleasant scenes presented on slides.

[3] 134w One of the difficulties in interpreting these investigations is that no two studies utilized the same paradigm. Some studies presented only a few faces repeatedly, leaving open the issue of "adaptation" to the emotional content of the stimuli. Also, differences may exist in the emotional "intensity" of the stimuli utilized. For instance, an angry face may be a more emotionally "intense" stimulus than a happy face, or an unpleasant stimulus (i.e., faces of people with dermatological diseases) may be more intense than a pleasant stimulus (i.e., pictures of babies). Comparisons among the results of these studies may also prove to be difficult, since the stimulus categories utilized may not be equivalent in "meaning" to the subjects. Also, most of these previously conducted ERP studies did not explore gender differences in response to such stimuli.

[4] 153w Recently, Erwin et al (1992) developed a facial recog-nition task that can be administered during physiological neuroimaging and, thus, permits examination of effects of specific dimensions of behavior on regional brain activity. Their research focused on the receptive processing of happy and sad facial affect in normal subjects and several clinical groups (Gur et al 1992;Erwin et al 1992;Heimberg et al 1992). One consistent finding in their studies is that women and men differ in their responses to these facial stimuli. Women are better at correctly detecting happy faces than sad faces and better at detecting expressions on male faces than on female faces. Men, on the other hand, are equally responsive to male happy and sad faces, but are significantly worse at detecting sadness in women's faces than in men's faces. These findings suggest that this behavioral emotion discrimination task may be sensitive in detecting performance measures to both gender and affect.

[5] 151w The purpose of the present study was to adapt the facial discrimination task developed by Erwin et al (1992) to a visual ERP paradigm. An affective-target paradigm was employed that is similar to those used in previous studies investigating ERP responses to affective stimuli (see Lang et al 1990;Kestenbaum and Nelson 1992;Nelson and Nugent 1990;Laurian et al 1991;Johnston et al 1986); however, the present study differed in that we used a total of 36 faces, ranging in the degree of affect, from very happy, to neutral, to very unhappy. In addition, both male and female facial stimuli were utilized. The study had three specific aims: 1) to test whether sad and happy faces differentially modify the morphology of late positive components using these stimuli; 2) to evaluate whether male or female facial stimuli elicit similar responses; and, lastly 3) to determine if a subject's gender modifies brain responses to facial stimuli.

[1] 200w Only trials in which subjects correctly identified the target stimuli within 1000 msec were used for analysis. Out of a possible 216 trials (neutral trials ϭ 151; sad and happy faces each ϭ 33 trials) presented for each subject, the average number of trials per subject per affective stimuli is presented as: average number of neutral trials used: male subjects n ϭ 135, range ϭ 11; female subjects n ϭ 129, range ϭ 19; average number of happy and sad face trials used per subject: happy face trials n ϭ 28, range ϭ 2; sad face trials n ϭ 24, range ϭ 4. Further breakdown of average number of trials per subconditions was not available. Both Figures 1 and 2 display ERPs elicited from male and female subjects in all leads during the presentation of the facial stimuli (sad, happy, and neutral faces). In our study, the maximum amplitudes for the P450 component were found over the central and parietal areas, with a maximum at Pz for the sad and happy faces and quite symmetrically represented for neutral faces. There were no effects of menstrual cycle phase (luteal vs. follicular) on P450 amplitude or latency; therefore, all data were combined.

[2] 20w Significant main effects were found for facial affect (happy and sad faces) on P450 latency [F(1,1,28) ϭ 10.21, p Ͻ

[3] 159w .003]. As the top row in Figure 3 indicates, univariate F tests revealed that subjects had significantly longer P450 latencies to the sad faces than to the happy faces for all leads except F7 (sad Ͼ happy) [Fz: F(1,130) ϭ 7.050, p Ͻ .009; Cz: F(1,130) ϭ 16.718, p Ͻ .001; Pz: F(1,130) ϭ 13.803, p Ͻ .001; F3: F(1,130) ϭ 4.871, p Ͻ .029; F4: F(1,130) ϭ 14.280, p Ͻ .000; F8: F(1,130) ϭ 12.880, p Ͻ .001]. There was also a significant facial affect (happy vs. sad) ϫ gender of facial affect (male vs. female face) interaction [F(1,1,28) ϭ 8.77, p Ͻ .006]. Univariate F tests revealed these significant differences for all leads except F7 [Fz: F(3,126) ϭ 3.739, p Ͻ .013; Cz: F(3,126) ϭ 7.790, p Ͻ .001; Pz: F(3,126) ϭ 7.334, p Ͻ .001; F3: F(7,124) ϭ 2.981, p Ͻ .034; F4: F(7,124) ϭ 5.603, p Ͻ .001; F8: F(7,124) ϭ 4.474, p Ͻ .005].

[4] 92w In exploratory analyses Tukey's post hoc comparisons revealed that subjects had significantly longer P450 latencies to the female sad faces compared to female happy faces (all leads except F7) and male happy faces (leads Cz, Pz, and F4) (female sad faces Ͼ female and male happy faces). Subjects also had significantly longer P450 latencies to the male sad faces compared to female happy faces (leads Cz and Pz) (male sad faces Ͼ female happy faces). There were no significant differences between male sad or happy faces. No other significant interactions were found.

[5] 288w A between-subjects ANOVA revealed no significant differences in reaction time (RT) between males and females (group). It did reveal a significant main effect for facial Figure 2. Grand averages of ERPs elicited from male and female subjects in response to different facial stimuli (sad, happy, and neutral faces) (n ϭ 35). Significant differences were observed between male and female subjects. Female subjects had significantly higher amplitudes and longer P450 latencies to facial expressions than did male subjects. affect. All subjects responded more quickly to happy faces than to sad faces (happy faces Ͻ sad faces) (629 versus 649 msec) [F(1,33) ϭ 25.150, p Ͻ .000]. Reaction times were also broken down into gender/affective type (male sad/happy and female sad/happy faces). In this case only 16 trials of RT data were available for each of the four categories; therefore, the data should be considered exploratory. A reaction time main effect for stimuli gender type and stimuli gender type by group interaction was found [F(3,99) ϭ 15.571, p Ͻ .000; F(3,99) ϭ 4.706, p Ͻ .004]. Subjects were found to respond more quickly to female happy faces than to female or male sad faces (female happy Ͻ female sad ϭ male sad) [639 vs. 650 msec; 639 vs. 673 msec; F(2,33) ϭ 4.606, p Ͻ .017; F(2,33) ϭ 8.941, p Ͻ .001] and to male happy faces than to female or male sad faces (male happy Ͻ female sad ϭ male sad) [F(2,33) ϭ 7.461, p Ͻ .002; F(2,33) ϭ 19.115, p Ͻ .000]. Also, subjects responded more quickly to female sad faces than to male sad faces [F(2,33) ϭ 8.311, p Ͻ .001] and to male happy faces than to female happy faces [F(2,33) ϭ 11.029, p Ͻ .000].

[6] 114w Figure 4 shows reaction times between male and female subjects as a function of the gender of the facial stimuli. The results show that male subjects responded more quickly to male happy faces than to female happy faces (606 vs. 646 msec), whereas female subjects responded more quickly to female happy faces than to male happy faces (646 vs. 670 msec) [happy male versus happy female: F(1,17) ϭ 21.738, p Ͻ .000 and happy female versus happy male: F(1,16) ϭ 15.831, p Ͻ .001, respectively]. Therefore, each gender identified more quickly with the happy affect of their own gender, whereas both genders identified a sad female expression more quickly than a sad male expression.

[7] 105w Male and female subjects also differed in the way errors were made. Male subjects made more errors in identifying a neutral expression as a sad expression than did female subjects (42 vs. 16 error trials), whereas female subjects made more errors in identifying a neutral expression as a happy expression than did male subjects (31 vs. 15 error trials). Male subjects were also more likely to identify a happy expression as a neutral expression than the female subjects (16 vs. 4 error trials). Although none of these results was statistically significant, male subjects tended to rate faces as less happy than did the female subjects.

[8] 155w All subjects filled out a posttest questionnaire immediately following the completion of the facial discrimination task. Both male and female subjects reported that the sad faces were more difficult to identify (n ϭ 14), with neutral faces a very close second (n ϭ 12). Only 1 subject rated the happy faces as the most difficult to identify. Both genders overwhelmingly agreed that the happy expressions were the easiest to identify (n ϭ 21). Male subjects described more difficulty in identifying female facial expressions (n ϭ 11) than male expressions (n ϭ 3), whereas female subjects described much more difficulty in identifying male facial expressions (n ϭ 9) than female expressions (n ϭ 5). Almost all male subjects rated the mouth (n ϭ 12) as crucial in identifying the facial expression, half the female subjects thought the mouth (n ϭ 7) to be crucial, the other half rated the eyes (n ϭ 7) as most crucial.

[9] 102w In summary, sad faces produced significantly larger amplitude and longer P450 latencies when compared to the happy faces. Female sad faces elicited the longest latency P450 components, whereas the happy female faces elicited the shortest P450 latencies. The reaction time analysis suggested that male subjects responded more quickly to male happy faces than to female happy faces, whereas the female subjects responded more quickly to female happy faces than to male happy faces. Thus, each gender identified more quickly with the happy expression of their own gender, whereas both genders identified a sad female expression more quickly than a sad male expression.

[1] 87w The results of this study suggest that facial affect presented briefly can alter the morphology of the P450 component of the ERP in a fashion specific to the affect (happy/sad faces) and gender (male/female) of the stimuli. In addition, men and women subjects differed in their P450 responses to facial expression. The results of this study support the use of this facial discrimination task (Erwin et al 1992) as a sensitive neurobehavioral probe that can be administered during physiological testing, particularly in ERP studies of emotion discrimination.

[2] 127w Several studies have reported that P450 amplitude can vary with the presentation of stimuli with different emotional content (Lang et al 1990;Kestenbaum and Nelson 1992;Nelson and Nugent 1990;Laurian et al 1991;Johnston et al 1986). In the present study, the amplitude of the P450 component was larger in response to sad faces as compared to happy faces. There are several potential explanations for these findings. It has been suggested that stimulus "intensity" may be an important variable in determining P300 amplitude (Papanicolaou et al 1985;Roth et al 1980;Polich 1989). Thus, the differences in P450 amplitude elicited by the sad faces in the present study and other studies may be due to the fact that sad faces may be more "intense" or provocative to the brain than happy faces.

[3] 93w Other investigators have suggested that P450 amplitude for facial stimuli varies as a result of task demands (Laurian et al 1991;Nelson and Nugent 1990;Johnston et al 1986). In the present study subjects reported that sad faces were more difficult to identify. Although clever attempts for delineating ERP components related to the valence of the stimuli have been conducted (Laurian et al 1991;Johnston et al 1986), future studies in which the task can be isolated and independent of stimulus intensity and meaning would help to clarify how affective stimuli elicit differences in P450 amplitude.

[4] 121w P450 latency component differences were not found over the menstrual cycle in the current study. Johnston and Wang (1991) have found the P3 amplitude component to be sensitive to the menstrual cycle. In their study, both the pleasant and unpleasant categories elicited larger P3 amplitudes than the neutral category. During the "high progesterone" phase of the menstrual cycle, the investigators found marked increases in the P3 amplitude, but only to stimuli that were determined to be emotionally "pleasant" (i.e., pictures of babies). Two previous studies using affectively neutral stimuli (auditory discrimination tasks) have failed to detect fluctuations in either the amplitude or latency of the P3 component of the ERP over the menstrual cycle (Fleck and Polich 1988;Ehlers et al 1996).

[5] 159w Previous studies investigating ERP responses did not report significant P450 latency differences between affective stimuli. Significantly shorter latencies for the happy faces when compared to the sad faces (happy Ͻ sad) were found using the present paradigm. This finding was also confirmed by the reaction time data, as subjects had significantly shorter reaction times to happy as compared to sad faces. In addition, subjects reported finding happy faces easier to detect than sad faces. One reason for the significant differences in the P450 latencies found in the present study may be related to the type of paradigm utilized. Several previous studies presented only a few faces as stimuli, which may have, after several presentations, become relatively easy to detect, independent of their emotional affect. In the present study, 36 faces were utilized that had varying degrees of affective content. Using a more difficult paradigm may have fostered more heterogeneity in the latency data, thus allowing for detection of differences.

[6] 94w The literature indicates that gender differences in ERP response to neutral stimuli are either nonsignificant or only modest (see Polich 1986Polich , 1992Polich , 1993;;Polich and Martin 1992;O'Connor et al 1994); however, significant gender differences in the P450 ERP component to affectively laden facial stimuli were found in the present study. Female subjects were found to have longer latency and higher amplitude P450 ERP components. Assuming that P450 amplitude is a measure of the "intensity" of response to the stimuli, these data suggest that women may be more "sensitive" to emotional stimuli than men.

[7] 152w Subjects also differed in the amount of time it took them to behaviorally indicate they had identified correctly the emotional affect of the face. Each gender responded more quickly to the happy expression of their own gender, whereas both genders responded to a sad female expression more quickly than to a sad male expression. While these data should be considered preliminary due to the small number of trials tested, they support previous behavioral data collected using this paradigm. Erwin et al (1992) found in their experiment that when measurements were calculated on "sensitivity" (a true positive response being when the subject responds within the happy range for a happy face, or sad range when a sad face was presented), male subjects were more "sensitive" at detecting sad emotions in men. Men did not show this sensitivity in detecting female emotional expressions. Furthermore, women were more sensitive at detecting emotional expressions in men.

[8] 114w In our study, for both male and female subjects, significantly increased amplitudes were found in response to the sad faces in the right frontal regions (F4 and F8) but not in the left (F5 and F7). The results from several studies have indicated hemispheric lateralization in processing negative and positive affect (Laurian et al 1991;Roschmann and Wittling 1992;Wheeler et al 1993). It has been suggested that the right hemisphere may preferentially process sad or negative affect (Mandal et al 1991;Wheeler et al 1993), whereas the left hemisphere may be more associated with positive affective stimuli (Jones and Fox 1992;Wheeler et al 1993). The results of the present study support such findings with electrophysiological data.

[9] 106w Other findings indicating frontal asymmetry have been found using cortical cerebral blood flow (CBF) measurements in subjects participating in a mood induction task (Schneider et al 1994). These investigators found that the only region showing specific lateralized changes in CBF that differentiated sad from happy states was the frontal pole. Similarly, a study investigating emotion-related hemisphere asymmetries, using slides depicting normal human faces (neutral stimuli) in one condition and faces with dermatological diseases (emotional stimuli) in the second condition, found differences between the emotional and the neutral stimuli in the 510 -560-msec range, and it was also restricted to the frontal region (Roschmann and Wittling 1992).

[10] 145w In summary, adapting the facial discrimination task of Erwin et al (1992) to an ERP paradigm yielded physiological measures that were sensitive to the gender of the observer, the valence of the emotion portrayed, and the gender of the facial stimuli. Subjects were found to more easily detect happy faces than sad faces. P450 amplitudes were found to be increased to sad faces as compared to happy faces, and P450 latencies were significantly shorter for happy faces than sad faces. Female subjects showed a slower detection of the facial affect and higher amplitude P450 components than did the male subjects. Also, our results indicated that the P450 ERP component may have hemispheric specialization in frontal areas to affect. These studies further suggest that the use of physiological neuroimaging, such as in EEG and ERP techniques, may further enhance the understanding of the neuropsychology of emotion.

[1] 280w A total of 35 subjects (18 male and 17 female) between the ages of 18 and 25 (years) were recruited from the University of California, San Diego. The female subjects were recruited during different times in their menstrual cycle. The time of their last menses prior to testing was recorded, and their phase was determined by counting from the first day of their last menses to the actual test date. Those women participating 3-15 days after the first day of their menses were designated as being in the follicular phase (n ϭ 9), and those 20 -28 days after their last menses were designated as being in the luteal phase (n ϭ 8). Subjects were interviewed and selected using a questionnaire that assessed demographic, medical, and psychiatric profiles, drinking and drug use patterns, and family history of alcohol and drug abuse and psychiatric disorders (modified from Schuckit et al 1987). No subjects, or their first degree relatives, met the DSM-IV criteria for any psychiatric disorders. Subjects were excluded if they were on any medications, including birth control pills, or were nursing or pregnant. Subjects were run between the hours of 7:30 AM and 11:30 AM, and were asked to refrain from drinking caffeine the morning of the study and from consuming alcohol the night prior to the morning of the study. Prior to the start of the experiment, subjects were informed that the purpose of the study was to learn more about the effects of various computerized cognitive tests on human brainwaves. Each subject filled out a quick questionnaire, concerning aspects of the protocol, immediately following the completion of the facial discrimination task. All subjects signed an approved consent form.

[2] 120w DATA COLLECTION. A facial discrimination task (Erwin et al 1992) was adapted to a computerized visual ERP paradigm. The stimuli were happy, neutral, and sad faces presented on a computer screen for 1000 msec with an intertrial interval of 1000 -1500 msec. The prestimulus interval was 150 msec. Subjects were instructed to press a button whenever a happy or sad face was displayed (each ϭ 15% of the trials) and not to respond to the neutral faces (70% of the trials). There were 36 total faces (12 each of happy, neutral, and sad) presented in random order for a total of 216 trials. The only constraint on randomness was that the same face could not appear in two consecutive trials.

[3] 89w ERP RECORDING. Subjects were fitted with an electrode cap containing tin electrodes. The impedance of the electrodes was below 5 k⍀. Electroencephalographic (EEG) electrodes were placed across the midline (leads Fz, Cz, and Pz) and in frontal areas (leads F3, F4, F7, and F8) according to the International 10/20 system (bandwidth 0.5-35 Hz). Active leads were references to linked earlobes (A1 and A2), and eye movement was monitored by recording electro-oculogram (EOG) in the lateral infraorbital region. Trials that contained excessive eye movement artifact were eliminated prior to averaging.

[4] 126w The ERP data were digitized at a rate of 256 Hz. Trials in which subjects responded below the 300-msec and above the 1000-msec latency window were excluded. ERP components were quantified using a computerized peak detection routine that identifies baseline-to-peak amplitudes (V) within specified latency windows. The routine is user driven, and each peak detection is verified by the user. The baseline was determined by averaging the 150 msec of prestimulus activity obtained for each trial. The latency (in msec) was defined as the time from stimulus onset to the peak amplitude within a latency window. Visual inspection of the data revealed a late positive going potential at approximately 400 -550 msec (labeled P450). The latency window used to detect the P450 was 400 -550 msec.

[5] 32w DATA ANALYSIS. Grand means were computed from subject averages for each trial, and data analysis was computed across all leads. Peak amplitudes, latency values, and area scores were computed from subject averages.

[6] 246w The effects of facial affect (happy and sad faces), gender of the facial stimuli (female and male faces), gender of subject (female and male subjects), and electrode site on P450 amplitude and latencies were assessed with a 2 ϫ 2 ϫ 2 ϫ 7 repeated-measures multivariate analysis of variance (repeated-measures MANOVA; SPSS, Version 6.1). When the repeated-measures MANOVA procedure indicated significant effects, univariate F tests were conducted. Independent post hoc comparisons were conducted using the Tukey honest significant difference test. In an exploratory analysis, when a two-way interaction was noted for gender of face and facial affect, independent analysis of female sad and happy faces, and male sad and happy faces was conducted using one-way ANOVAs. In this case only 16 ERP trials were available for analyses for each of the four categories. Because this number of trials may be more unstable than those averaged over 32 trials, these data should be interpreted as preliminary. On all comparisons, significance was determined at the .05 level. A repeated two-way ANOVA was conducted for the reaction times, post hoc comparisons were conducted using contrast matrixes, and Bonferroni corrections were applied. Furthermore, for exploratory purposes, the female subjects were divided into two groups, those who participated in the study during their "follicular" phase and those who participated during their "luteal" phase. A two-way repeated-measures ANOVA was conducted with the group (follicular vs. luteal) as the between-subjects factor and facial affect (sad or happy facial stimuli) as the within-subjects factor.

[1] 93w ON P450 LATENCY. The repeated-measures MANOVA revealed significant main effects for gender on P450 latency [F(1,1,28) ϭ 8.94, p Ͻ .006]. Univariate F tests revealed that female subjects had significantly longer P450 latencies compared to male subjects for all leads (female Ͼ male subjects) [Fz: F(1,130) ϭ 11.674, p Ͻ .001; Cz: F(1,130) ϭ 4.368, p Ͻ .039; Pz: F(1,130) ϭ 7.778, p Ͻ .006; F3: F(1,130) ϭ 16.914, p Ͻ .001; F4: F(1,130) ϭ 10.096, p Ͻ .002; F7: F(1,130) ϭ 4.133, p Ͻ .044; F8: F(1,130) ϭ 12.450, p Ͻ .001].

[2] 53w The repeated-measures MANOVA on the P450 amplitude data revealed there was a significant main effect for facial affect (happy and sad faces) on P450 amplitude [F(1,1,28) ϭ 4.32, p Ͻ .05]. As the bottom row in Figure 3 .025; F7: F(1,130) ϭ 4.348, p Ͻ .039; F8: F(1,130) ϭ 4.561, p Ͻ .035].

[3] 18w The repeated-measures MANOVA revealed no significant main effect for gender of facial affect. No other interactions were found.