This article was downloaded by: [Ohio State University Libraries] On: 23 May 2012, At: 12:03 Publisher: Psychology Press Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of Clinical and Experimental Neuropsychology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ncen19
Event-related potentials to emotional and neutral stimuli a
a
Sarah F. Lang , Charles A. Nelson & Paul F. Collins a
a
University of Minnesota,
Available online: 04 Jan 2008
To cite this article: Sarah F. Lang, Charles A. Nelson & Paul F. Collins (1990): Eventrelated potentials to emotional and neutral stimuli, Journal of Clinical and Experimental Neuropsychology, 12:6, 946-958 To link to this article: http://dx.doi.org/10.1080/01688639008401033
PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-andconditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sublicensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.
Journal of Clinical and Experimental Nwropsychology 1990, Vol. 12, NO.6, p ~946-958 .
0168-8634/90/1206-0946$3.00 Q Swets & Zeitlinger
Event-Related Potentials to Emotional and Neutral Stimuli* Sarah F. Lang,Charles A. Nelson, and Paul F. Collins University of Minnesota
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
ABSTRACT The present study examined subjects’ cognitive processing of pictures of emotional and neutral facial expressions, as measured by Event-Related Potentials (ERPs). In Experiment 1, 10 subjects viewed two slides of a woman modelling angry and happy expressions; in Experiment 2, 10 subjects viewed slides of two women modelling neutral expressions. One face appeared on 20% of the trials and the other on 80% of the trials. Subjects counted the low frequency target face. In both experiments, the area of the P300 component was larger at Pz than Cz. In Experiment 1, P300 area was larger when the target was happy; peak amplitude was greater when the target was angry. No differences between neutral target faces were found for P300 amplitude or area in Experiment 2. These results suggest that emotional versus neutral facial expressions elicit different electrophysiological responses; responses are further differentiated to positive versus negative expressions.
The adult’s recognition of facial expressions of emotion has been studied extensively (for review, see Salzen, 1981). In the vast majority of such work, a subject’s recognition has been evaluated by recording a behavioral response (e.g., reaction time) while he/she is presented with photographs of individuals posing discrete expressions. From such work, it has been found that positive emotions
* The authors would like to thank Mary Henschel, Kathy Nugent, and Kim Williamson for their assistance in this research. Research and manuscript preparation were made possible in part by a grant to the first author from an NIMH predoctoral traineeship (MH15755) and to the second author by a grant from the NIH (HD23389) and the McKnight-Land Grant Research Fund of the University of Minnesota, and by a grant from the NIH to the Center for Research in Learning, Perception, and Cognition, University of Minnesota (HD01751). Portions of this manuscript were submitted by the first author in partial fulfillment of the predoctoral requirement for the Institute of Child Development and were presented at the meetings of the Society for Psychophysiological Research, San Francisco, California, in October, 1988. Requests for reprints should be sent to Charles A. Nelson, Center for Research in Learning, Perception, and Cognition, University of Minnesota, 75 East River Road, Minneapolis, MN 55455, USA. Accepted for publication: March 12. 1990.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
EVENT-RELATED POTENTIALS AND EMOTION
941
are recognized more easily than negative ones (Feyereisen, Malet, & Martin, 1986). For example, Kirouac and Dore (1983) found that among Ekman’s (1976) six primary emotions, happiness was recognized the fastest and with the most accuracy. These same authors also measured accuracy of judgment as a function of stimulus exposure time. Happiness was recognized more accurately at shorter exposure times than the other emotions. For example, while the happy expression was correctly identified 80% of the time when it was exposed for 30 ms, the angry expression was recognized with comparable accuracy only when exposed for 50 ms (Kirouac & Dore, 1984). Support for this general pattern of findings can also be found in the developmental literature, where it has been reported that young children identify happy expressions more easily and with greater accuracy than they identify angry expressions (a finding that persists even when the expressions are presented upside down; Kestenbaum & Nelson, 1989). Although the studies discussed thus far have been useful in describing which expressions are more easily discriminated from which others, they have failed to reveal much about the neural transactions that take place when a subject is asked to recognize a discrete expression. This question was addressed in the current study by recording event-related potentiais (Ems),an electrophysiological response that is well-suited to monitor the neurological processing of discrete stimuli, such as pictures of facial expressions. A small literature currently exists on Ems associated with adults’ recognition of emotion. For example, Johnston and colleagues (Johnston, Burleson, & Miller, 1987; Johnston, Miller, & Burleson, 1986) reported a larger P300 response to stimuli rated as pleasant and unpleasant as opposed to stimuli rated as neutral. There were no differences to emotional valence (positivity vs. negativity). Yee and Miller (1987) reported a larger P300 to unpleasant slides than to pleasant ones, and they concluded that the difference may reflect the emotional intensity of the stimuli rather than their valence (the unpleasant slides were of “blood injury and medical procedures”, while the pleasant slides were of “landscapes, young animals, and food”). This interpretation would be in keeping with Johnston et al.’s (1986; 1987) results, which suggest that emotional intensity provokes a larger P300 response than neutrality. In none of these studies did the amplitude of the P300 unequivocally discriminate between positive and negative emotions. Furthermore, these studies used stimuli that were designed to evoke emotional responses in the subject, rather than addressing whether subjects’ recognition of these stimuli would differ. The primary question addressed in the current study was whether ERPs could be used to distinguish between a positive emotion and a negative emotion, and whether responses to these stimuli would differ from responses to neutral stimuli. In a series of two experiments, ERPs were recorded as subjects were presented with faces that varied in either expression (happy vs. angry; Experiment 1) or identity (one neutral face vs. another neutral face; Experiment 2). Although a range of ERP components was recorded, particular attention was paid to subjects’ P300 response. This component is typically studied by means of the “oddball”
948
SARAH F.LANG ET AL.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
paradigm. In this paradigm, a rare (low probability) stimulus and a frequent (high probability) stimulus are presented to the subject in a randomized sequence. The subject is instructed to perform a task (e.g., keep a running count of the target events) each time the rare stimulus appears. Hypotheses The specific hypothesis of the present study was that visual stimuli presented in an oddball paradigm would elicit variations in P300 amplitude corresponding to a model proposed by Johnson (1986). In this model, the amplitude of the P300 is seen to vary on three dimensions: probability (larger to the rare, or low probability, stimulus), (b) stimulus meaning as defined by the task (counting the rare, or target stimulus), and (c) information transmission (larger to the attended vs. the ignored stimulus). Since in this study design the low probability stimulus was at the same time both the counted and the attended stimulus, all three experimental dimensions converged on one stimulus (i.e., the target). Variations in amplitude would thus be multiply determined. It was also hypothesized that a non-task-related facet of stimulus meaning, namely the salience of the particular emotion depicted by the target stimulus, would have some effect on P300 morphology. It was predicted that subjects’ cognitive engagement would vary depending on the emotional content of the attended stimulus, and that this variation would be discernible in some aspect of the P300 component. Specific predictions were not made as to whether latency, amplitude, or scalp distribution would vary or in which direction the effects would be.
GENERAL METHOD Subjects Twenty-eight subjects participated in two experiments. Eight subjects were excluded, one because of experimenter error and seven because of technical problems resulting in unusable data obtained from one or more leads. Ten subjects per experiment (2 males and 8 females in Experiment 1; 3 males and 7 females in Experiment 2) remained in the study. Subjects were neurologically normal adults between the ages of 20- and 50-years old. Prior to beginning the procedure, subjects were informed that the aim of the study was to learn more about how the human brain responds to pictures of faces. Procedure In each of two experiments, slides from Ekman’s Pictures of FaciaZAfSect (1976) were rear-projected, one at a time, onto a translucent screen situated 125 cm from the subject. Visual angle subtended by the slides was approximately 9 x 14 degrees. An IBM AT computer controlled two Kodak carousel projectors. The slides were presented in random order. The target stimulus appeared on 20% of
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
EVENT-RELATED POTENTIALS AND EMOTION
949
the trials, and the nontarget stimulus appeared on 80% of the trials. Each slide was exposed for 100 ms; the interstimulus interval varied randomly between 500 and 1000 ms. Subjects were tested individually. Once the electrodes had been attached, the subject was seated in a dark, sound-attenuated room. Before the testing began, the subject was shown both stimuli. The experimenter instructed himher to pay close attention to one of the faces (the target), to keep a running count of its occurrence, to focus on the center of the screen, and to blink as little as possible. Two blocks of 200 trials each were presented. A different face served as the target stimulus within each block (see below).
ERP Recordings The electroencephalogram (EEG) was recorded from four channels. Grass AgAgCl disc electrodes were placed over midline occipital (Oz), parietal (Pz), vertex (Cz), and frontal (Fz) scalp sites, according to the International 101’20 system. Scalp leads were referenced t? linked ears. Horizontal and vertical eye movements (EOG) were recorded from miniature eye electrodes placed above and below the right eye in a transverse configuration. Impedances were less than 10 kQ for scalp electrodes. Eye and ear impedances were less than 25 kR. The recording epoch for each trial was 1200 ms. The EEG was sampled every 10 ms beginning 100 ms pre-event (i.e., baseline). Grass Model 12A5 amplifiers were used to amplify the data (100 Hz). The gain for the EEG was 100,000; for EOG, 5,000. Bandpass width was .1 to 30 Hz.
Data Reduction and Inspection Trials were deleted if an eye movement exceeded 25 microvolts within a 100 ms window, or if the EEG (from any scalp lead) exceded A-D values for longer than 50 ms. Remaining trials were averaged by stimulus type (i.e., target vs. nontarget) for each subject. (Stimulus type also corresponded to probability: target stimuli were low probability, and non-target stimuli were high probability.) Within each block, the two stimulus type averages contained equal numbers of trials. Grand means were computed from subject averages for each block in each experiment. Visual inspection of the grand means revealed a positive-going waveform, which was elicited predominantly by the target stimulus and appeared maximal at Pz and Cz. Average peak latency across grand means was 462 ms. A window was set to bracket the waveform (labelled P300). This window (290-800 ms) was defined broadly so as to encompass between-subject variation in peak latency and return to baseline. Area scores, peak amplitude, and latency values were computed from subject averages. All data analyses were restricted to Pz and Cz, where P300 was maximal.
950
SARAH F.LANG ET AL.
EXPERIMENT 1 The goal of the first experiment was to describe what, if any, differences would be evident in the P300 component elicited by pictures of happy and angry facial expressions.
METHOD
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
Stimuli. Ekman slides #66 and #69 were used. These are pictures of a female face modelling
expressions of happiness and anger. Condition was determined by which face subjects counted in a given block of trials. In Condition 1, subjects counted the angry face; in Condition 2. they counted the happy face. Each subject completed both conditions; the order of conditions was counterbalanced, so that half of the subjects began with Condition 1 and half with Condition 2.
Design.
RESULTS AND DISCUSSION Prior to performing statistical analyses, the data were screened for distributional normality and homogeneity of variance. It was determined that the data met the assumptions required for the analysis of variance. For each experiment, three epsilon-corrected analyses of variance (ANOVA) with repeated measures (N= 10) were performed using BMDP-2V. Independent ANOVAs were run for area (integrated above baseline within the P300 window), peak amplitude (measured from baseline), and latency to peak. Post-hoc comparisons were computed using the Newman-Keuls’ Multiple-Range Test; significance was determined at the .01 level. Area above baseline was analyzed using a 3-way ANOVA: Condition (i.e., “count angry” vs. “count happy”) x Stimulus Type (i.e., target vs. non-target) x Lead (i.e., Pz and Cz). There was a significant Condition x Stimulus Type interaction, F(1,9) = 14.92,p < .004. For both conditions, area was larger in response to the target stimulus than the non-target -stimulus. Within stimuli, area was larger in Condition 2 (count happy) than Condition 1 (count angry) for the target stimulus. The difference in area between conditions was not significant for the non-target stimulus. These data are illustrated in Figures 1 and 2. The P300 area ANOVA also showed a Condition x Lead interaction, F(1,9) = 6.59, p < .03. Area was significantly larger at Pz than at Cz for both conditions (across stimulus types). At Pz, area was larger for Condition 2 (count happy) than Condition 1 (count angry); at Cz, the difference in area between conditions was not significant. Since the P300 was prominent in response to the target stimulus but negligible to the nontarget stimulus, P300 amplitudes and latencies were analyzed within the target stimulus. A 2-way ANOVA for Condition (i.e., “count angry” vs. “count happy”) x Lead (i.e., Pz and Cz) was conducted with amplitude as the dependent variable. This revealed main effects of condition and lead. Amplitude
95 1
EVENT-RELATED POTENTIALS AND EMOTION
1
c
I
Pz 25;
204
-5-
L
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
:10;
\,
E-15-
-20-25+
0
I
I
I
I
I
1
200
400
600 ms
800
1000
1200
800
1000
1200
Time __-
cz 254 20;
-20 -25+----7-
€-
0
200
-
-
400
-
-
600
Time ms Fig. 1. Grand averages for all subjects ( N = 10) in Experiment 1, Condition 1 (“count angry”). The dark line represents the response, averaged across trials and subjects, to the target (low-probability) stimulus. The light line represents the response to the non-target (high-probability) stimulus. Pz and Cz represent parietal and vertex scalp, respectively.
was greater (more positive) for Condition 1 (target = angry) than Condition 2 (target = happy), F(1,9) = 9 . 1 5 , < ~ .01. Across conditions, amplitude was greater at Pz than at Cz, F(1,9) = 18.12, p < .002. An identical ANOVA with latency as the dependent variable revealed no significant effects. An additional analysis was conducted to investigate the possibility that the effects of increased area to “happy” and increased amplitude to “angry” might reflect an artifact of averaging, rather than (as we will argue) endogenous differ-
952
SARAH F.LANG ET AL.
!
PZ
0) h
-4J 4
0
>
0
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
c
0 r
I
I
I
I
200
400
600
800
I
I
1000 1200
Time ms
-
CZ
25; 204 154
I
.“,IO-j €- 153
V
-201 -25+------~0 200
------I
400 Time
600 ms
1
1
I
800
1000
1200
Fig. 2. Grand averages for all subjects (N = 10) in Experiment 1, Condition 2 (‘‘count happy”). The dark line represents the response, averaged across trials and subjects, to the target (low-probability)stimulus. The light line represents the response to the non-target (high-probability) stimulus. Pz and Cz represent parietal and vertex scalp, respectively. ences in response to the two emotions. The specific hypothesis entertained was that the observed differences were due to variability in P300 peak latency for the different target conditions. Greater variability in the “count happy” condition, for example, would cause peak smearing, which would result in greater area than in the “count angry” condition. To investigate this possibility, single trial estimates were made for P300 peak latency response under the two conditions, and
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
EVENT-RELATEDPOTENTIALS AND EMOTION
95 3
standard deviations of peak latency were computed within each condition for each subject. The grand mean peak latency standard deviation was 1.5% larger in the “count happy” condition at the Pz lead, and 6.1% larger at the Cz lead. These differences were non-significant in paired t tests conducted separately for the two leads (Pz: t = 0.38, ns; Cz: t = 0.90, ns), which indicates that latency variability was not a significant source of the difference in P300 area and amplitude between conditions. In summary, analysis of the 290-800 ms window revealed a positive-going waveform that was maximal at Pz and that was larger, across leads, to attended stimuli. At Pz, P300 area above baseline was larger to the happy than to the angry face. However, within the attended (target) stimuli, peak amplitude was greater to the angry face. These results clearly point to processing differences in the recognition of a positive vs. a negative emotional expression. Nevertheless, the possibility that these differences reflect the discrimination of some aspect of facial configuration unrelated to emotion could not be ruled out on the basis of these results. To control for possible spurious effects, a second experiment was conducted.
EXPERIMENT 2 In this experiment, subjects attended to one of two neutral faces. It was hypothesized that, in the absence of emotional expression, the P300 would not vary as a function of the face counted. METHOD Stimuli. The stimuli were Ekman slides #6 and #65. These depict two female faces mod-
elling neutral expressions. Design. Subjects were tested under two conditions. In Condition 1, subjects counted face #6; in Condition 2, they counted face #65. Each subject completed both conditions; the order of conditions was counterbalanced,so that half of the subjects began with Condition 1 and half with Condition 2.
RESULTS AND DISCUSSION Prior to conducting statistical analyses, data were screened as in Experiment 1 (see above). Area above baseline within the 290-800 ms window was analyzed using a 3-way ANOVA: Condition (i.e., “count #6” vs. “count #65”) x Stimulus Type (i.e., target vs. non-target) x Lead (i.e., Pz and Cz). The area ANOVA revealed a highly significant Stimulus Type x Lead interaction, F(1,9) = 33.35, p < .0003. Across leads, area was larger to the target stimulus. Within the target stimulus, area was significantly larger at Pz than at Cz. Within the non-target stimulus. there was no difference in area between leads.
954
SARAH F.LANG ET AL.
P300 peak amplitude was again analyzed within the target stimulus in a 2way ANOVA (Condition x Lead). Amplitude did not differ significantly between conditions. There was a main effect of lead, F(1,9) = 21-81, p < .001. As in Experiment 1, amplitude was greater at Pz than at Cz. An ANOVA for peak latencies showed no significant effects. These findings are illustrated in Figures 3 and 4.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
PZ
-20 -25 0
I
I
I
1
200
400
600
800
I
I
1000 1200
Time ms CZ
Fig. 3. Grand averages for all subjects (N = 10) in Experiment 2, Condition 1 ("count #6").The dark lime represents the response, averaged across trials and subjects, to the target (low-probability)stimulus. The light line represents the response to the non-target (high-probability) stimulus. Pz and Cz represent parietal and vertex scalp, respectively.
955
EVENT-RELATED POTENTIALS AND EMOTION
Pz
2
254 201 :5{ 1oi
f p
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
I----
I
--25 20&-r-.-0
200
4hO
660
Time
ms
400 Time
600 ms
800
id00 12100
800
I000
1200
Fig. 4. Grand averages for all subjects (N = 10) in Experiment 2, Condition 2 (“count #65”). The dark line represents the response, averaged across trials and subjects, to the target (low-probability) stimulus. The light line represents the response to the non-target (high-probability) stimulus. Pz and Cz represent parietal and vertex scalp, respectively.
In summary, the P300 waveform in Experiment 2 showed lead and stimulus type main effects that were similar to those seen in Experiment 1. P300 area and amplitude were larger at Pz and larger to the attended stimulus. The most striking difference between the experiments is the absence in Experiment 2 of any condition effects: neither P300 area nor amplitude differed significantly according to which face subjects counted.
956
SARAH F.LANG ET AL.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
GENERAL DISCUSSION The results from Experiment 1 and Experiment 2 confirm the initial hypotheses. Both experiments produced a P300 component that varied in peak amplitude as a function of stimulus probability and task demand. Also as predicted, P300 varied as a function of the emotional content of the stimulus. In Experiment 1, P300 area was larger when subjects counted the happy face, and amplitude was greater when they counted the angry face. In Experiment 2, neither P300 area nor amplitude differed between the two neutral faces. In both experiments, P300 showed stimulus type effects that were predicted by the P300 literature. Regardless of the stimulus conrenr. P300 area above baseline and peak amplitude were larger to the rare stimulus. In these experiments, probability and stimulus value (task-relevance) variables converged on the target stimulus, which served both as the low probability as well as the attended, or counted stimulus. The literature has demonstrated consistently that low probability and task-relevant stimuli elicit a larger P300 response. The average latency of P300.462 ms, falls within the 300-600 ms range cited in the literature. The absence of latency effects in both experiments suggests that stimulus evaluation time was constant across each pair of faces; that is, subjects found the faces equally discriminable. Since results were not compared between experiments, it is not known whether the emotional vs. neutral faces required significantly different rates of processing; inspection of the latencies would suggest that they did not. In Experiment 1, P300 area and amplitude were sensitive to condition, or stimulus content. In contrast, Experiment 2 showed no condition effects. Presumably, the difference between the experiments had to do with the additional variable of emotional expression that was present in the faces used as stimuli in Experiment 1 and absent in the faces used in Experiment 2. However, the possibility cannot be excluded that the differences between experiments have to do instead with variation in processing depending on whether target and non-target stimuli represent expressive variants of a single face (as in Experiment 1) or a single expression modelled by two different faces (as in Experiment 2). Although such an explanation could account for the global discrepancy between the two experiments in terms of condition effects, it does not address the significant differences between conditions within Experiment 1, namely that within the target stimulus and at Pz,P300 area was larger to the happy face, and P300 amplitude was significantly greater when the target face was angry.
P300 Area Although P300 amplitude was reduced to the happy target face relative to the angry target face, P300 area was larger to the happy face. Visual inspection of the grand mean waveforms (see Figures 1 and 2) suggests that the area difference is due to increased positivity later in the epoch. This positivity appears in the count happy condition but not in the count angry condition. P300 appears to
EVENT-RELATED POTENTIALS AND EMOTION
957
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
resolve, or return to baseline, faster in the count angry condition, a pattern that is also apparent in the grand mean waveforms for both conditions in Experiment 2 (see Figures 3 and 4). The resolution is slower in the count bappy condition: the response appears to be sustained longer.
P300 Amplitude Johnson’s (1986) “triarchic model” suggests several explanations for the amplitude difference found in the present study. First, P300 amplitude may reflect differences in task and stimulus complexity. If amplitude increases as processing becomes more complex, then greater amplitude to the angry face may indicate that subjects were more engaged by an angry expression than by a happy one, or that the visual configuration itself is more complex in an angry face than in a happy one. If this were the case, however, one would expect to find corresponding differences in latency, namely, longer latency to peak in the “count angry” than in “count happy” condition. The absence of latency effects makes the “complexity explanation” somewhat unlikely. Second, P300 amplitude differences may be a function of stimulus significance. Johnson (1986) cites studies in which P300 amplitude varied according to the monetary value attached to a stimulus and to stimulus (sensory) intensity. Amplitude was positively correlated with monetary value and with loudness. If these variables can be translated into terms of emotional expression, then one might attribute the greater amplitude elicited by the angry face to its greater intensity, relative to the happy face. This explanation would be in keeping with Yee and Miller’s (1987) interpretation of their results, in which P300 was larger and reaction time was faster to unpleasant stimuli. The authors suggest that subjects found the unpleasant slides more intense than the pleasant ones. Accordingly, P300 amplitude would be sensitive to emotional intensity. Third, P300 amplitude to the angry and happy faces may depend on deployment of attention, which in turn may vary with the emotion depicted by the stimulus. The angry expression may elicit more focused attention, and possibly more arousal, from subjects than the happy expression. This could be because an angry expression is more compelling than a happy expression, or because an occasional angry face seen in the context of frequent happy ones is more arousing than a rare happy face seen in the context of angry ones. In conclusion, although the findings reported herein should be considered preliminary, the results of this study suggest that there are measurable differences in the neurophysiological processing of positive vs. negative emotional expressions. REFERENCES Ekman, P. (1 976). Pictures offacial affect. Palo Alto, CA: Consulting PsychologistsPress. Feyereisen, P., Malet, C., & Martin, Y. (1986). Is the faster processing of expressions of happiness modality-specific? In H.D. Ellis, M.A. Jeeves, F. Newcombe, & A. Young (Eds.),Aspects offuce processing (pp. 349-355). Dordrecht, The Netherlands: Martinus Nijhoff Publishers.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
958
SARAH F.LANG ET AL.
Johnson, R. (1986). A triarchic model of P300 amplitude. Psychophysiology, 23,367-384. Johnston, V.S., Burleson, M.H., &Miller, D.R. (1987). Emotional value and late positive components of ERPs. In R. Johnson, Jr.. J.W.Rohrbaugh, & R. Parasuraman (Eds.), Current trends in Event-Related Potential research (EEG Supplement 40)(pp. 198203). Amsterdam: Elsevier. Johnston, V.S.,Miller, D.R., & Burleson. M.H. (1986). Multiple P3s to emotional stimuli and their theoretical significance. Psychophysiology, 23,684-694. Kestenbaum. R., & Nelson, C. (1989). The process of recognizing emotions from childhood to adulthood: A reaction time study with upright and inverted faces. Manuscript submitted for publication. Kirouac, G., & Dore, F.Y. (1983). Accuracy and latency of judgment of facial expressions of emotions. Perceptual and motor skills, 57.683-686. Kirouac, G., & Dore, F.Y.(1984). Judgment of facial expressions of emotion as afunction of exposure time. Perceptual and Motor Skills, 59, 147-150. Salzen, E.A. (1981). Perception of emotion in faces. In G. Davies, H. Ellis, & J. Shepherd (Eds.), Perceiving and rememberingfaces (pp. 133-169). London: Academic Press. Yee, C. M. & Miller, G. A. (1987). Affective valence and information processing. In R. Johnson, Jr.. J.W.Rohrbaugh, & R. Parasuraman (Eds.), Current trends in Event-Reluted Potential research (EEG Supplement 40) (pp. 300-307). Amsterdam: Elsevier.
Journal of Clinical and Experimental Neuropsychology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ncen19
Event-related potentials to emotional and neutral stimuli a
a
Sarah F. Lang , Charles A. Nelson & Paul F. Collins a
a
University of Minnesota,
Available online: 04 Jan 2008
To cite this article: Sarah F. Lang, Charles A. Nelson & Paul F. Collins (1990): Eventrelated potentials to emotional and neutral stimuli, Journal of Clinical and Experimental Neuropsychology, 12:6, 946-958 To link to this article: http://dx.doi.org/10.1080/01688639008401033
PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-andconditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sublicensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.
Journal of Clinical and Experimental Nwropsychology 1990, Vol. 12, NO.6, p ~946-958 .
0168-8634/90/1206-0946$3.00 Q Swets & Zeitlinger
Event-Related Potentials to Emotional and Neutral Stimuli* Sarah F. Lang,Charles A. Nelson, and Paul F. Collins University of Minnesota
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
ABSTRACT The present study examined subjects’ cognitive processing of pictures of emotional and neutral facial expressions, as measured by Event-Related Potentials (ERPs). In Experiment 1, 10 subjects viewed two slides of a woman modelling angry and happy expressions; in Experiment 2, 10 subjects viewed slides of two women modelling neutral expressions. One face appeared on 20% of the trials and the other on 80% of the trials. Subjects counted the low frequency target face. In both experiments, the area of the P300 component was larger at Pz than Cz. In Experiment 1, P300 area was larger when the target was happy; peak amplitude was greater when the target was angry. No differences between neutral target faces were found for P300 amplitude or area in Experiment 2. These results suggest that emotional versus neutral facial expressions elicit different electrophysiological responses; responses are further differentiated to positive versus negative expressions.
The adult’s recognition of facial expressions of emotion has been studied extensively (for review, see Salzen, 1981). In the vast majority of such work, a subject’s recognition has been evaluated by recording a behavioral response (e.g., reaction time) while he/she is presented with photographs of individuals posing discrete expressions. From such work, it has been found that positive emotions
* The authors would like to thank Mary Henschel, Kathy Nugent, and Kim Williamson for their assistance in this research. Research and manuscript preparation were made possible in part by a grant to the first author from an NIMH predoctoral traineeship (MH15755) and to the second author by a grant from the NIH (HD23389) and the McKnight-Land Grant Research Fund of the University of Minnesota, and by a grant from the NIH to the Center for Research in Learning, Perception, and Cognition, University of Minnesota (HD01751). Portions of this manuscript were submitted by the first author in partial fulfillment of the predoctoral requirement for the Institute of Child Development and were presented at the meetings of the Society for Psychophysiological Research, San Francisco, California, in October, 1988. Requests for reprints should be sent to Charles A. Nelson, Center for Research in Learning, Perception, and Cognition, University of Minnesota, 75 East River Road, Minneapolis, MN 55455, USA. Accepted for publication: March 12. 1990.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
EVENT-RELATED POTENTIALS AND EMOTION
941
are recognized more easily than negative ones (Feyereisen, Malet, & Martin, 1986). For example, Kirouac and Dore (1983) found that among Ekman’s (1976) six primary emotions, happiness was recognized the fastest and with the most accuracy. These same authors also measured accuracy of judgment as a function of stimulus exposure time. Happiness was recognized more accurately at shorter exposure times than the other emotions. For example, while the happy expression was correctly identified 80% of the time when it was exposed for 30 ms, the angry expression was recognized with comparable accuracy only when exposed for 50 ms (Kirouac & Dore, 1984). Support for this general pattern of findings can also be found in the developmental literature, where it has been reported that young children identify happy expressions more easily and with greater accuracy than they identify angry expressions (a finding that persists even when the expressions are presented upside down; Kestenbaum & Nelson, 1989). Although the studies discussed thus far have been useful in describing which expressions are more easily discriminated from which others, they have failed to reveal much about the neural transactions that take place when a subject is asked to recognize a discrete expression. This question was addressed in the current study by recording event-related potentiais (Ems),an electrophysiological response that is well-suited to monitor the neurological processing of discrete stimuli, such as pictures of facial expressions. A small literature currently exists on Ems associated with adults’ recognition of emotion. For example, Johnston and colleagues (Johnston, Burleson, & Miller, 1987; Johnston, Miller, & Burleson, 1986) reported a larger P300 response to stimuli rated as pleasant and unpleasant as opposed to stimuli rated as neutral. There were no differences to emotional valence (positivity vs. negativity). Yee and Miller (1987) reported a larger P300 to unpleasant slides than to pleasant ones, and they concluded that the difference may reflect the emotional intensity of the stimuli rather than their valence (the unpleasant slides were of “blood injury and medical procedures”, while the pleasant slides were of “landscapes, young animals, and food”). This interpretation would be in keeping with Johnston et al.’s (1986; 1987) results, which suggest that emotional intensity provokes a larger P300 response than neutrality. In none of these studies did the amplitude of the P300 unequivocally discriminate between positive and negative emotions. Furthermore, these studies used stimuli that were designed to evoke emotional responses in the subject, rather than addressing whether subjects’ recognition of these stimuli would differ. The primary question addressed in the current study was whether ERPs could be used to distinguish between a positive emotion and a negative emotion, and whether responses to these stimuli would differ from responses to neutral stimuli. In a series of two experiments, ERPs were recorded as subjects were presented with faces that varied in either expression (happy vs. angry; Experiment 1) or identity (one neutral face vs. another neutral face; Experiment 2). Although a range of ERP components was recorded, particular attention was paid to subjects’ P300 response. This component is typically studied by means of the “oddball”
948
SARAH F.LANG ET AL.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
paradigm. In this paradigm, a rare (low probability) stimulus and a frequent (high probability) stimulus are presented to the subject in a randomized sequence. The subject is instructed to perform a task (e.g., keep a running count of the target events) each time the rare stimulus appears. Hypotheses The specific hypothesis of the present study was that visual stimuli presented in an oddball paradigm would elicit variations in P300 amplitude corresponding to a model proposed by Johnson (1986). In this model, the amplitude of the P300 is seen to vary on three dimensions: probability (larger to the rare, or low probability, stimulus), (b) stimulus meaning as defined by the task (counting the rare, or target stimulus), and (c) information transmission (larger to the attended vs. the ignored stimulus). Since in this study design the low probability stimulus was at the same time both the counted and the attended stimulus, all three experimental dimensions converged on one stimulus (i.e., the target). Variations in amplitude would thus be multiply determined. It was also hypothesized that a non-task-related facet of stimulus meaning, namely the salience of the particular emotion depicted by the target stimulus, would have some effect on P300 morphology. It was predicted that subjects’ cognitive engagement would vary depending on the emotional content of the attended stimulus, and that this variation would be discernible in some aspect of the P300 component. Specific predictions were not made as to whether latency, amplitude, or scalp distribution would vary or in which direction the effects would be.
GENERAL METHOD Subjects Twenty-eight subjects participated in two experiments. Eight subjects were excluded, one because of experimenter error and seven because of technical problems resulting in unusable data obtained from one or more leads. Ten subjects per experiment (2 males and 8 females in Experiment 1; 3 males and 7 females in Experiment 2) remained in the study. Subjects were neurologically normal adults between the ages of 20- and 50-years old. Prior to beginning the procedure, subjects were informed that the aim of the study was to learn more about how the human brain responds to pictures of faces. Procedure In each of two experiments, slides from Ekman’s Pictures of FaciaZAfSect (1976) were rear-projected, one at a time, onto a translucent screen situated 125 cm from the subject. Visual angle subtended by the slides was approximately 9 x 14 degrees. An IBM AT computer controlled two Kodak carousel projectors. The slides were presented in random order. The target stimulus appeared on 20% of
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
EVENT-RELATED POTENTIALS AND EMOTION
949
the trials, and the nontarget stimulus appeared on 80% of the trials. Each slide was exposed for 100 ms; the interstimulus interval varied randomly between 500 and 1000 ms. Subjects were tested individually. Once the electrodes had been attached, the subject was seated in a dark, sound-attenuated room. Before the testing began, the subject was shown both stimuli. The experimenter instructed himher to pay close attention to one of the faces (the target), to keep a running count of its occurrence, to focus on the center of the screen, and to blink as little as possible. Two blocks of 200 trials each were presented. A different face served as the target stimulus within each block (see below).
ERP Recordings The electroencephalogram (EEG) was recorded from four channels. Grass AgAgCl disc electrodes were placed over midline occipital (Oz), parietal (Pz), vertex (Cz), and frontal (Fz) scalp sites, according to the International 101’20 system. Scalp leads were referenced t? linked ears. Horizontal and vertical eye movements (EOG) were recorded from miniature eye electrodes placed above and below the right eye in a transverse configuration. Impedances were less than 10 kQ for scalp electrodes. Eye and ear impedances were less than 25 kR. The recording epoch for each trial was 1200 ms. The EEG was sampled every 10 ms beginning 100 ms pre-event (i.e., baseline). Grass Model 12A5 amplifiers were used to amplify the data (100 Hz). The gain for the EEG was 100,000; for EOG, 5,000. Bandpass width was .1 to 30 Hz.
Data Reduction and Inspection Trials were deleted if an eye movement exceeded 25 microvolts within a 100 ms window, or if the EEG (from any scalp lead) exceded A-D values for longer than 50 ms. Remaining trials were averaged by stimulus type (i.e., target vs. nontarget) for each subject. (Stimulus type also corresponded to probability: target stimuli were low probability, and non-target stimuli were high probability.) Within each block, the two stimulus type averages contained equal numbers of trials. Grand means were computed from subject averages for each block in each experiment. Visual inspection of the grand means revealed a positive-going waveform, which was elicited predominantly by the target stimulus and appeared maximal at Pz and Cz. Average peak latency across grand means was 462 ms. A window was set to bracket the waveform (labelled P300). This window (290-800 ms) was defined broadly so as to encompass between-subject variation in peak latency and return to baseline. Area scores, peak amplitude, and latency values were computed from subject averages. All data analyses were restricted to Pz and Cz, where P300 was maximal.
950
SARAH F.LANG ET AL.
EXPERIMENT 1 The goal of the first experiment was to describe what, if any, differences would be evident in the P300 component elicited by pictures of happy and angry facial expressions.
METHOD
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
Stimuli. Ekman slides #66 and #69 were used. These are pictures of a female face modelling
expressions of happiness and anger. Condition was determined by which face subjects counted in a given block of trials. In Condition 1, subjects counted the angry face; in Condition 2. they counted the happy face. Each subject completed both conditions; the order of conditions was counterbalanced, so that half of the subjects began with Condition 1 and half with Condition 2.
Design.
RESULTS AND DISCUSSION Prior to performing statistical analyses, the data were screened for distributional normality and homogeneity of variance. It was determined that the data met the assumptions required for the analysis of variance. For each experiment, three epsilon-corrected analyses of variance (ANOVA) with repeated measures (N= 10) were performed using BMDP-2V. Independent ANOVAs were run for area (integrated above baseline within the P300 window), peak amplitude (measured from baseline), and latency to peak. Post-hoc comparisons were computed using the Newman-Keuls’ Multiple-Range Test; significance was determined at the .01 level. Area above baseline was analyzed using a 3-way ANOVA: Condition (i.e., “count angry” vs. “count happy”) x Stimulus Type (i.e., target vs. non-target) x Lead (i.e., Pz and Cz). There was a significant Condition x Stimulus Type interaction, F(1,9) = 14.92,p < .004. For both conditions, area was larger in response to the target stimulus than the non-target -stimulus. Within stimuli, area was larger in Condition 2 (count happy) than Condition 1 (count angry) for the target stimulus. The difference in area between conditions was not significant for the non-target stimulus. These data are illustrated in Figures 1 and 2. The P300 area ANOVA also showed a Condition x Lead interaction, F(1,9) = 6.59, p < .03. Area was significantly larger at Pz than at Cz for both conditions (across stimulus types). At Pz, area was larger for Condition 2 (count happy) than Condition 1 (count angry); at Cz, the difference in area between conditions was not significant. Since the P300 was prominent in response to the target stimulus but negligible to the nontarget stimulus, P300 amplitudes and latencies were analyzed within the target stimulus. A 2-way ANOVA for Condition (i.e., “count angry” vs. “count happy”) x Lead (i.e., Pz and Cz) was conducted with amplitude as the dependent variable. This revealed main effects of condition and lead. Amplitude
95 1
EVENT-RELATED POTENTIALS AND EMOTION
1
c
I
Pz 25;
204
-5-
L
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
:10;
\,
E-15-
-20-25+
0
I
I
I
I
I
1
200
400
600 ms
800
1000
1200
800
1000
1200
Time __-
cz 254 20;
-20 -25+----7-
€-
0
200
-
-
400
-
-
600
Time ms Fig. 1. Grand averages for all subjects ( N = 10) in Experiment 1, Condition 1 (“count angry”). The dark line represents the response, averaged across trials and subjects, to the target (low-probability) stimulus. The light line represents the response to the non-target (high-probability) stimulus. Pz and Cz represent parietal and vertex scalp, respectively.
was greater (more positive) for Condition 1 (target = angry) than Condition 2 (target = happy), F(1,9) = 9 . 1 5 , < ~ .01. Across conditions, amplitude was greater at Pz than at Cz, F(1,9) = 18.12, p < .002. An identical ANOVA with latency as the dependent variable revealed no significant effects. An additional analysis was conducted to investigate the possibility that the effects of increased area to “happy” and increased amplitude to “angry” might reflect an artifact of averaging, rather than (as we will argue) endogenous differ-
952
SARAH F.LANG ET AL.
!
PZ
0) h
-4J 4
0
>
0
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
c
0 r
I
I
I
I
200
400
600
800
I
I
1000 1200
Time ms
-
CZ
25; 204 154
I
.“,IO-j €- 153
V
-201 -25+------~0 200
------I
400 Time
600 ms
1
1
I
800
1000
1200
Fig. 2. Grand averages for all subjects (N = 10) in Experiment 1, Condition 2 (‘‘count happy”). The dark line represents the response, averaged across trials and subjects, to the target (low-probability)stimulus. The light line represents the response to the non-target (high-probability) stimulus. Pz and Cz represent parietal and vertex scalp, respectively. ences in response to the two emotions. The specific hypothesis entertained was that the observed differences were due to variability in P300 peak latency for the different target conditions. Greater variability in the “count happy” condition, for example, would cause peak smearing, which would result in greater area than in the “count angry” condition. To investigate this possibility, single trial estimates were made for P300 peak latency response under the two conditions, and
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
EVENT-RELATEDPOTENTIALS AND EMOTION
95 3
standard deviations of peak latency were computed within each condition for each subject. The grand mean peak latency standard deviation was 1.5% larger in the “count happy” condition at the Pz lead, and 6.1% larger at the Cz lead. These differences were non-significant in paired t tests conducted separately for the two leads (Pz: t = 0.38, ns; Cz: t = 0.90, ns), which indicates that latency variability was not a significant source of the difference in P300 area and amplitude between conditions. In summary, analysis of the 290-800 ms window revealed a positive-going waveform that was maximal at Pz and that was larger, across leads, to attended stimuli. At Pz, P300 area above baseline was larger to the happy than to the angry face. However, within the attended (target) stimuli, peak amplitude was greater to the angry face. These results clearly point to processing differences in the recognition of a positive vs. a negative emotional expression. Nevertheless, the possibility that these differences reflect the discrimination of some aspect of facial configuration unrelated to emotion could not be ruled out on the basis of these results. To control for possible spurious effects, a second experiment was conducted.
EXPERIMENT 2 In this experiment, subjects attended to one of two neutral faces. It was hypothesized that, in the absence of emotional expression, the P300 would not vary as a function of the face counted. METHOD Stimuli. The stimuli were Ekman slides #6 and #65. These depict two female faces mod-
elling neutral expressions. Design. Subjects were tested under two conditions. In Condition 1, subjects counted face #6; in Condition 2, they counted face #65. Each subject completed both conditions; the order of conditions was counterbalanced,so that half of the subjects began with Condition 1 and half with Condition 2.
RESULTS AND DISCUSSION Prior to conducting statistical analyses, data were screened as in Experiment 1 (see above). Area above baseline within the 290-800 ms window was analyzed using a 3-way ANOVA: Condition (i.e., “count #6” vs. “count #65”) x Stimulus Type (i.e., target vs. non-target) x Lead (i.e., Pz and Cz). The area ANOVA revealed a highly significant Stimulus Type x Lead interaction, F(1,9) = 33.35, p < .0003. Across leads, area was larger to the target stimulus. Within the target stimulus, area was significantly larger at Pz than at Cz. Within the non-target stimulus. there was no difference in area between leads.
954
SARAH F.LANG ET AL.
P300 peak amplitude was again analyzed within the target stimulus in a 2way ANOVA (Condition x Lead). Amplitude did not differ significantly between conditions. There was a main effect of lead, F(1,9) = 21-81, p < .001. As in Experiment 1, amplitude was greater at Pz than at Cz. An ANOVA for peak latencies showed no significant effects. These findings are illustrated in Figures 3 and 4.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
PZ
-20 -25 0
I
I
I
1
200
400
600
800
I
I
1000 1200
Time ms CZ
Fig. 3. Grand averages for all subjects (N = 10) in Experiment 2, Condition 1 ("count #6").The dark lime represents the response, averaged across trials and subjects, to the target (low-probability)stimulus. The light line represents the response to the non-target (high-probability) stimulus. Pz and Cz represent parietal and vertex scalp, respectively.
955
EVENT-RELATED POTENTIALS AND EMOTION
Pz
2
254 201 :5{ 1oi
f p
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
I----
I
--25 20&-r-.-0
200
4hO
660
Time
ms
400 Time
600 ms
800
id00 12100
800
I000
1200
Fig. 4. Grand averages for all subjects (N = 10) in Experiment 2, Condition 2 (“count #65”). The dark line represents the response, averaged across trials and subjects, to the target (low-probability) stimulus. The light line represents the response to the non-target (high-probability) stimulus. Pz and Cz represent parietal and vertex scalp, respectively.
In summary, the P300 waveform in Experiment 2 showed lead and stimulus type main effects that were similar to those seen in Experiment 1. P300 area and amplitude were larger at Pz and larger to the attended stimulus. The most striking difference between the experiments is the absence in Experiment 2 of any condition effects: neither P300 area nor amplitude differed significantly according to which face subjects counted.
956
SARAH F.LANG ET AL.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
GENERAL DISCUSSION The results from Experiment 1 and Experiment 2 confirm the initial hypotheses. Both experiments produced a P300 component that varied in peak amplitude as a function of stimulus probability and task demand. Also as predicted, P300 varied as a function of the emotional content of the stimulus. In Experiment 1, P300 area was larger when subjects counted the happy face, and amplitude was greater when they counted the angry face. In Experiment 2, neither P300 area nor amplitude differed between the two neutral faces. In both experiments, P300 showed stimulus type effects that were predicted by the P300 literature. Regardless of the stimulus conrenr. P300 area above baseline and peak amplitude were larger to the rare stimulus. In these experiments, probability and stimulus value (task-relevance) variables converged on the target stimulus, which served both as the low probability as well as the attended, or counted stimulus. The literature has demonstrated consistently that low probability and task-relevant stimuli elicit a larger P300 response. The average latency of P300.462 ms, falls within the 300-600 ms range cited in the literature. The absence of latency effects in both experiments suggests that stimulus evaluation time was constant across each pair of faces; that is, subjects found the faces equally discriminable. Since results were not compared between experiments, it is not known whether the emotional vs. neutral faces required significantly different rates of processing; inspection of the latencies would suggest that they did not. In Experiment 1, P300 area and amplitude were sensitive to condition, or stimulus content. In contrast, Experiment 2 showed no condition effects. Presumably, the difference between the experiments had to do with the additional variable of emotional expression that was present in the faces used as stimuli in Experiment 1 and absent in the faces used in Experiment 2. However, the possibility cannot be excluded that the differences between experiments have to do instead with variation in processing depending on whether target and non-target stimuli represent expressive variants of a single face (as in Experiment 1) or a single expression modelled by two different faces (as in Experiment 2). Although such an explanation could account for the global discrepancy between the two experiments in terms of condition effects, it does not address the significant differences between conditions within Experiment 1, namely that within the target stimulus and at Pz,P300 area was larger to the happy face, and P300 amplitude was significantly greater when the target face was angry.
P300 Area Although P300 amplitude was reduced to the happy target face relative to the angry target face, P300 area was larger to the happy face. Visual inspection of the grand mean waveforms (see Figures 1 and 2) suggests that the area difference is due to increased positivity later in the epoch. This positivity appears in the count happy condition but not in the count angry condition. P300 appears to
EVENT-RELATED POTENTIALS AND EMOTION
957
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
resolve, or return to baseline, faster in the count angry condition, a pattern that is also apparent in the grand mean waveforms for both conditions in Experiment 2 (see Figures 3 and 4). The resolution is slower in the count bappy condition: the response appears to be sustained longer.
P300 Amplitude Johnson’s (1986) “triarchic model” suggests several explanations for the amplitude difference found in the present study. First, P300 amplitude may reflect differences in task and stimulus complexity. If amplitude increases as processing becomes more complex, then greater amplitude to the angry face may indicate that subjects were more engaged by an angry expression than by a happy one, or that the visual configuration itself is more complex in an angry face than in a happy one. If this were the case, however, one would expect to find corresponding differences in latency, namely, longer latency to peak in the “count angry” than in “count happy” condition. The absence of latency effects makes the “complexity explanation” somewhat unlikely. Second, P300 amplitude differences may be a function of stimulus significance. Johnson (1986) cites studies in which P300 amplitude varied according to the monetary value attached to a stimulus and to stimulus (sensory) intensity. Amplitude was positively correlated with monetary value and with loudness. If these variables can be translated into terms of emotional expression, then one might attribute the greater amplitude elicited by the angry face to its greater intensity, relative to the happy face. This explanation would be in keeping with Yee and Miller’s (1987) interpretation of their results, in which P300 was larger and reaction time was faster to unpleasant stimuli. The authors suggest that subjects found the unpleasant slides more intense than the pleasant ones. Accordingly, P300 amplitude would be sensitive to emotional intensity. Third, P300 amplitude to the angry and happy faces may depend on deployment of attention, which in turn may vary with the emotion depicted by the stimulus. The angry expression may elicit more focused attention, and possibly more arousal, from subjects than the happy expression. This could be because an angry expression is more compelling than a happy expression, or because an occasional angry face seen in the context of frequent happy ones is more arousing than a rare happy face seen in the context of angry ones. In conclusion, although the findings reported herein should be considered preliminary, the results of this study suggest that there are measurable differences in the neurophysiological processing of positive vs. negative emotional expressions. REFERENCES Ekman, P. (1 976). Pictures offacial affect. Palo Alto, CA: Consulting PsychologistsPress. Feyereisen, P., Malet, C., & Martin, Y. (1986). Is the faster processing of expressions of happiness modality-specific? In H.D. Ellis, M.A. Jeeves, F. Newcombe, & A. Young (Eds.),Aspects offuce processing (pp. 349-355). Dordrecht, The Netherlands: Martinus Nijhoff Publishers.
Downloaded by [Ohio State University Libraries] at 12:03 23 May 2012
958
SARAH F.LANG ET AL.
Johnson, R. (1986). A triarchic model of P300 amplitude. Psychophysiology, 23,367-384. Johnston, V.S., Burleson, M.H., &Miller, D.R. (1987). Emotional value and late positive components of ERPs. In R. Johnson, Jr.. J.W.Rohrbaugh, & R. Parasuraman (Eds.), Current trends in Event-Related Potential research (EEG Supplement 40)(pp. 198203). Amsterdam: Elsevier. Johnston, V.S.,Miller, D.R., & Burleson. M.H. (1986). Multiple P3s to emotional stimuli and their theoretical significance. Psychophysiology, 23,684-694. Kestenbaum. R., & Nelson, C. (1989). The process of recognizing emotions from childhood to adulthood: A reaction time study with upright and inverted faces. Manuscript submitted for publication. Kirouac, G., & Dore, F.Y. (1983). Accuracy and latency of judgment of facial expressions of emotions. Perceptual and motor skills, 57.683-686. Kirouac, G., & Dore, F.Y.(1984). Judgment of facial expressions of emotion as afunction of exposure time. Perceptual and Motor Skills, 59, 147-150. Salzen, E.A. (1981). Perception of emotion in faces. In G. Davies, H. Ellis, & J. Shepherd (Eds.), Perceiving and rememberingfaces (pp. 133-169). London: Academic Press. Yee, C. M. & Miller, G. A. (1987). Affective valence and information processing. In R. Johnson, Jr.. J.W.Rohrbaugh, & R. Parasuraman (Eds.), Current trends in Event-Reluted Potential research (EEG Supplement 40) (pp. 300-307). Amsterdam: Elsevier.
