pleasant Amplitude (uV^2) neutral unpleasant
0 ms
-10 -5 0 5 10
µV Oz Pz
1000 ms
Minimum norm least squares
Paralleling results regarding mean voltage, no significant main effect or interaction including VALENCE was found for the P1. However, in the N1 (100-130 ms) window, a main effect of emotional valence on MNE dipole strength was found (F(2,20) = 10.11, p<0.01). Post-hoc testing revealed that across recording sites pleasant pictures were related to significantly enhanced dipole strength as compared to both neutral and unpleasant pictures (p<0.01). In posterior-superior sensors, where this activity was most pronounced, we found a linear relationship between dipole strength and valence category:
Figure III.B.5: Grand Mean MNE topography of the N1 (left) and a representative segment of the slow wave (right) for the three affective categories. The lambda for regularization was selected 0.02 (see section II for details on MNE).
Post-hoc testing of the significant interaction VALENCE X SITE (F(10,100) = 6.7, p<0.01) demonstrated differential brain responding at posterior superior sites according to affective picture category, again with pleasant pictures being related to higher activity than
0.2 0.4 0.6 0.8 0.9 1.0 1.2 nA/m^2
MNE for N1 (100-130 ms) MNE for slow wave (800-850 ms)
neutral pictures (p<0.01), which in turn were more activating than unpleasant pictures (p<0.01). Right hemisphere activitiy contributed most to this effect (VALENCE X HEMISPHERE: F(2,20) = 6.3, p<0.05). In the P300 (300-340 ms) window, pleasant pictures were associated with greater dipole strength than neutral and unpleasant pictures across recording sites (VALENCE: F(2,20) = 5.4, p<0.05). However, posterior-inferior, medial-inferior and posterior superior sites exhibited an arousal effect, i.e. emotional pictures were associated with greater dipole strength than neutral ones (VALENCE X SITE, F(10,100) = 6.5, p<0.01).
No valence main effect was seen for the slow wave (750-850 ms). However, a VALENCE X SITE interaction (F(10,100) = 10.6, p<0.01) revealed that in posterior-inferior and medial-inferior recording sites, emotional stimulation was related to higher dipole strength than viewing neutral pictures.
Discussion GBA modulation
Perception, experience and expression of emotions may be represented in neuronal networks connecting subcortical (limbic) structures with neocortical areas (Tucker and Dawson, 1984; Derryberry and Tucker, 1992; Liotti and Tucker, 1995). This model postulates that widespread rather than focal neuronal activity is specifically related to emotional processing, a prediction that is consistent with the present experimental outcome. Processing of emotional pictures induces topographically specific cortical activity in EEG gamma from 30 to 50 Hz. The valence hypothesis of hemispheric asymmetries predicts a significant valence by hemisphere interaction with a shift of relatively higher cortical activity for negative valence over the right hemisphere. Such an interaction was statistically pronounced but restricted to the 40 Hz band (30-50 Hz). However, we found exactly the opposite pattern with higher activity in the 30 to 50 Hz band over right anterior and posterior temporal sites during the processing of pleasant pictures. In accordance with the general valence hypothesis overall power was higher at right temporal and frontal sites than at left hemisphere electrodes.
These findings are in line with observations in stroke patients, which exhibit more pronounced deficits with respect to both emotion perception and expression after right hemisphere damage (see Borod, 1992, for an overview). The more specific assumption that the left hemisphere would produce more gamma activity when pleasant emotional material is to be processed, however, can clearly be rejected on the basis of the present results. As discussed in the above,
few studies exist which have analyzed the higher EEG frequency bands. Most reports limited their analysis to alpha desynchronization (Davidson, 1998). In the present study we did not detect any sensitivity of alpha activity to the valence of emotional processing. Several factors may account for this lack: First, we have subtracted the mean activity phase-locked to picture onset prior to transformation into the frequency domain. The visual evoked response and the P1/N1 complex in particular add to the power in the alpha band (Schürmann et al., 1998) and thus may have contributed to the effects in previous reports. Second, contrary to other studies in the field, we have measured the emotional conditions in relation to a neutral baseline. It might be the case that alpha desynchronization effects appear when positive and negative valence are contrasted without correction for watching neutral pictures. Third, other configuration of electrode sites may produce effects. In order not to violate statistical assumptions, we refrained from comparing every electrode location and any combination with its analogue of the opposite hemisphere. The valence hypothesis in its more specific version would predict higher gamma band activity on right temporal electrodes for unpleasant pictures (Borod, 1992). However, we have found exactly the opposite pattern. The few studies which analyzed higher frequency components showed - in line with the present findings - higher activity for positive valence at right hemisphere temporal electrodes (Tucker, 1984; Tucker and Dawson, 1984; Ray and Cole, 1985; Aftanas et al., 1998). As far as we know, the present study is the first, which tested cortical responses in the respective frequency bands by correcting for (a) general noise, which was operationalized by subtracting the respective power of a pre-stimulus time period, and, (b) calculating the absolute level of hemispheric function for negative and positive valence by subtracting the power while watching neutral slides. This procedure resulted in a significant increase only in 30-50 Hz band when compared to the neutral condition across anterior and posterior temporal sites for the left and right hemisphere and for negative and positive valence. The present experiment could not replicate the finding of pronounced alpha desynchronization over left anterior regions for positive and right hemisphere anterior alpha desynchronization for negative valence. In sum, the present results suggest the activity of a widespread cortical network predominantly in right frontal and temporal and to a lesser degree in left temporal areas. Thus, the present results support the idea of the existence of asymmetric cortico-limbic networks (Liotti and Tucker, 1995) and the gamma band topography of the present study suggests a contribution of the ventral visual pathway in combination with anterior areas. Whether presenting visual emotional material or imagination results in a different topography of gamma band activity is subject to further research. As with pictures, subjects should also be instructed to imagine a neutral scene in
order to control for effects of imagination. The present study, however, can be seen as a first step to uncover a possible relation between gamma band activity and emotion.
ERPs
With respect to ERPs, we were able to replicate the main result reported from similar studies, namely a voltage modulation in time windows later than 300ms, as a function of arousal (Laurian et al., 1991; Mini et al., 1996; Palomba et al., 1997). No effects were found for the earlier time segments, as has been reported previously (Palomba et al., 1997; Diedrich et al., 1997). Using high-density EEG, the maximum of this late arousal effect was found to be centered at electrode site Cz, showing a broad distribution. A further question of this experiment addressed the source configuration of the late arousal modulation. Our results from MNE analysis clearly support the view that it is visual processing that is modulated as a function of picture arousal in the later time windows. However, additional activation over anterior sensors was also observed in this study, allowing for the conclusion that this modulation in visual areas of the cortex is effected by modulatory afferents that may include prefrontal areas, as would be consistent with proposals emphasizing prefrontal activity discussed in section I of this thesis. As an alternative explanation, it has been suggested that the slow modulation is of anterior origin, reflecting participants’ suppression of emotional expression (Diedrich et al., 1997). Although this explanation is not sufficient to account for the present results, a contribution of such frontal inhibitory processes to the ERP is consistent with our data. Of course, it cannot be excluded that the frontal MNE maxima may also be related to small time locked eye movements that went undetected in artifact control. MNE dipole density of P300 showed its maximum at more posterior sites than the late positive wave suggesting that arousal modulation of the late components reflects similar afferent activity acting on increasingly higher levels of visual processing, showing greater effects at higher stages of visual processing. This is consistent with the physiology of amygdala efferents to visual cortex, being more dense to higher-order than to lower-order visual areas (Amaral et al., 1992).
In contrast to results with mean voltage, we found a modulation of the dipole strength of the N1 as a linear function of picture valence when using MNE dipole density as a dependent variable. Pleasant pictures were associated with the greatest degree of activation.
The source of this effect was clearly located in occipital visual cortices, but not in areas V1 or V2 (see Figure III.B.5). This might be regarded as evidence for an early selection mechanism
in terms of Lang’s concept of ‘motivated attention’, paralleling results from studies of spatial selective attention (Hillyard, 1993). More precisely, the N1 modulation may reflect allocation of attention as a function of stimulus properties, i.e. significance for the organism. Instead of being cued by external stimuli such as arrows, the participants would thus rapidly direct their attention to significant stimuli, based on rough visual analysis. Furthermore, this finding might be due to subcortical afferent modulation of visual processing such as for the fast, extra-lemniscal thalamo-amygdoloid-cortical route proposed by LeDoux. This model would of course predict an opposite pattern, showing increased activation in response to aversive pictures. In general, given the poor signal to noise ratio of the present study and the high interindividual variability of the ERP, these interpretations must remain speculative.
Experiment C was conducted in order to overcome some of the problems of this experiment and to shed light on questions of hemispheric asymmetries (see below).
Relation between GBA and ERP
Most interestingly, the topographical distribution of effects regarding slow wave MNE displays remarkable similarity to the findings in the gamma-1 band (see Figures III.B.3 and III.B.5). Although there is no time information available as to the time course of GBA, it might be the case that the GBA topography containing information across a time window of 6000 ms reflects specifically long-lasting higher order processing in temporal and frontal structures. This interpretation seems plausible, given that the integrated activity of prefrontal and temporal areas should occur later in time than mere modulatory activity in earlier stages of processing of the emerging visual percept. Nevertheless, investigation of the time course of GBA in response to emotive stimuli is necessary in order to obtain information as to the validity of these interpretations. Therefore, this issue will be further discussed below, with respect to the findings of experiment C.
Experiment C: Differential effects of affective valence and arousal in the cerebral hemispheres: A hemifield study of ocillatory brain activity and ERPs
Introduction
The current study employed a hemifield paradigm in order to further explore the hemispheric differences in emotional processing that have been reported in experiment B.
Presentation of stimuli to the visual hemifields is a technique that has extensively been used in studies of visual spatial attention e.g. (Luck et al., 1990; Mangun and Buck, 1998). The advantages of this procedure are manifold: First, it is possible to experimentally manipulate the cerebral hemisphere to be first involved in processing of a stimulus presented in the contralateral hemifield (Posner et al., 1987). A second benefit especially for studies of high-frequency brain activity lies in the fact that lateralized processing as indicated by GBA can more easily be distinguished from EMG artifacts that are not expected to occur selectively in the hemisphere contralateral to the stimulus. Further, differential response timing in the hemispheres, operationalized as the latencies of ERP components or spectral events, may provide important information regarding the processing steps within and between cerebral hemispheres (Hillyard and Anllo-Vento, 1998). Recently, this approach has been shown useful for the study of GBA modulation by visual selective spatial attention (Gruber et al., 1999). Although hemifield paradigms have substantially increased our understanding as to how lateralized networks mediate attentional processes, this technique has only seldom been used in affective research (Kayser et al., 1997; Pizzagalli et al., 1999). Research into attentional modulation of visual processing has provided strong support for the view that attentional processes are moderated to a large extend by networks being located mainly in the right hemisphere (Corbetta, 1998). Given theoretical and empirical parallels between attentional and affective processes as described in section I.A.1, it seems reasonable to expect a right-hemisphere (left hemifield) preponderance also of the effects of motivated attention.
That is, differences between affectively arousing (pleasant and unpleasant) and neutral pictures should be more pronounced when stimuli are presented in the left hemifield, or alternatively, these effects should be enhanced at right hemispheric sites, across stimulus locations. This hypothesis can be derived from results from experiment B and is consistent with the literature showing evidence for a right hemisphere advantage in ‘affective’ as
compared to ‘neutral’ processing (see section I.B). Another finding from Experiment B showed selective co-activation of the left hemisphere in response to aversive stimuli. Thus, a further aim was to examine whether this effect could be replicated.
The experimental design of the present study also allows an investigation into effects of hemifield / hemisphere interactions with affective categories on ERP measures. More precisely, ERP asymmetries that are related to changes on affective dimensions can be examined. Accordingly, a right hemisphere / left hemifield ERP amplitude increase for affective pictures was expected. Furthermore, we aimed at replicating the finding of a late positive wave modulation. As with experiment B, the question was addressed whether this latter effect is an index of (1) increased intensity of visual processing or reflects (2) higher processes such as emotional expression inhibition (Diedrich et al., 1997).
Methods Participants
10 right-handed volunteers (7 women, 3 men; age range from 22-40, mean age 26.1) with normal or corrected-to-normal vision consented to participate. They received class credits or a small financial bonus for participating.
Stimuli and Procedure
The identical set of stimuli as in experiment B was used, with three categories of pictures differing in affective valence (pleasant, neutral, unpleasant). All pictures were presented on a 19 inch computer monitor with a refresh rate of 70 Hz. Pictures subtended a visual angle of 8 degrees horizontally, the eccentricity of the center of the pictures to either side being 2.8 degrees. The distance between the screen and the subjects’ eyes was 1.7 m. A chin rest was used in order to keep these settings constant within and between participants.
Pictures were presented for 1000 ms, with an inter-stimulus interval that was randomly varying between 3000 and 6000 ms. A fixation point was marked in the center of the screen.
The order of presentation was pseudo-randomized both regarding hemifield and affective category. A straight sequence of more than two pictures of the same subcategory (e.g.
mutilation, families etc.) was prevented. Two blocks of 300 pictures were presented, resulting in a total of 100 trials per affective category and hemifield. Subjects were first presented with 2 examples of the affective stimuli for each affective category, respectively, that were not part
of the experimental series. They were also instructed to maintain gaze on the fixation point and to avoid exploratory eye-movements and eye blinks during picture presentation. These requirements were practiced until the participants were familiar with the procedure and did not report problems. Subsequently, the electrode net was applied and participants entered an electrically shielded chamber, where the EEG recordings were conducted. After the EEG recordings, subjects viewed the 60 different pictures again in a pseudo-randomized order and were asked to rate the respective picture on two categories, affective valence and arousal, using a paper and pencil version of the Self-Assessment Manikin (SAM; Bradley and Lang, 1994). In this last block, no hemifield presentation was done, but subjects looked at each picture without being constrained in any way.
Eletrophysiological recordings
The standard EGI montage was used as described in section II. Data were sampled at a rate of 500 Hz, constrained by online band-pass filtering between 0.1 to 200 Hz. Artifact-free epochs of 300 ms pre-onset and 724 ms post-onset of the stimulus were obtained using the procedure by Junghöfer et al, as described in section II of this thesis. The mean voltage of a 200 ms pre-stimulus baseline was subtracted to correct for offsets. After artifact correction, a mean of 67 trials for each affective category/hemifield combination was obtained, with a minimum value of 45 and a maximum value of 89 artifact-free trials.
Data reduction and analysis Data analysis GBA
Spectral responses to affective pictures were examined using wavelet transform of the artifact-free epochs. A constant m = f0/σ f = 7 was selected in order to achieve good time and frequency resolution in the examined frequency range from 9.8 to 68.8 Hz. This resulted in a frequency step ∆f= 0.97. Wavelets of this family were normalized in order to have equal amounts of energy. A 250 ms prestimulus period was used as baseline for the time-frequency information, and the mean of this time window was subtracted from the time-frequency matrix for each frequency and time point. After wavelet transform, spectral power in six frequency bands was obtained by computing the average power for alpha (9–13 Hz), beta (14-20 Hz), Low Gamma ((14-20-30 Hz), Mid Gamma (30-45 Hz), High Gamma (45-65 Hz). Four time windows were examined that were selected on the basis of Grand Mean time-frequency distributions to contain the most pronounced event-related spectral changes: (1) an early
window, ranging from 70 to 90 ms after stimulus onset, (2) a 170 – 220 ms window (3) a 280-340 ms window and (4) a late window ranging from 480 to 550 ms. To allow for comparisons between experiments, we used lateral electrode clusters for statistical analysis. However, compared to experiment B, both posterior and anterior clusters were enlarged to account for the hemifield procedure and to be sensitive for signals both related to posterior and to anterior sources (see Figure III.C.1).
Figure III.C.1: Electrode groups formed for statistical testing. Encircled sites: regional means for spectral measures. Posterior encircled groups also served for statistical testing of P1, N1 and P300.
Groups with arrows were selected to test significance of anterior N2 and late positive wave.
Figure III.C.1 displays the relative position of the selected clusters with respect to sites of the international 10-20 system. The mean spectral power within each frequency band and time window was subject to four factor repeated-measure ANOVA with the factors HEMIFIELD (left, right), VALENCE (pleasant, neutral, unpleasant), HEMISPHERE (left, right), and RECORDING SITE (temporal anterior, temporal posterior) for each frequency
2 1
statistical testing of late positive wave
band, respectively. Effects of picture categories on SAM scores were evaluated using repeated measurement ANOVA with the factor VALENCE (pleasant, neutral, unpleasant).
Data analysis ERP
Two methods were employed for analysis of ERP data. First, after averaging separately according to the factorial structure of the experiment, three pronounced components were identified on the basis of grand means: A P1 component, peaking around 120 ms, a N1 being maximal around 160ms and a P300 like wave that showed latencies around 280 ms (see table III.C.1). Peak latencies and amplitudes of three components were extracted using the time windows identified on the basis of grand average data. For this purpose, voltages and latencies at electrodes corresponding to the sites T5 and T6 of the international 10-20 system were examined. Thus, a 2 X 3 X 2- ANOVA was computed for each parameter of each component, having the within factors HEMIFIELD (left, right) AFFECTIVE CATEGORY (pleasant, neutral, unpleasant) and SITE (T5, T6). This procedure allowed for investigating the latencies, which may reveal differences in response timing depending on affective valence or arousal.
A second approach used ANOVAs on regional voltage means for data reduction.
Based on grand mean traces and interindividual peak variability, time windows for the five major components were identified: P1 (100-130ms), N1(155-185ms), P300(280-340ms), anterior N2 (220-260ms), and a late positive wave (400-600ms). In order to achieve comparability with respect to GBA analyses, ERP ANOVAs were computed for the identical set of electrode clusters as described above for GBA analysis. Appropriateness of this
Based on grand mean traces and interindividual peak variability, time windows for the five major components were identified: P1 (100-130ms), N1(155-185ms), P300(280-340ms), anterior N2 (220-260ms), and a late positive wave (400-600ms). In order to achieve comparability with respect to GBA analyses, ERP ANOVAs were computed for the identical set of electrode clusters as described above for GBA analysis. Appropriateness of this