Intern. J. Neuroscience, (2871-891, 2002 Taylor &Francis Copyright @ 2002 Taylor & Francis, healthsciences (0020-7454/02 $12.00 + 00 DOI: 10,1080/00207450290025897 NARROW-BAND SPECTRAL MEASUREMENTS OF EEG DURING EMOTIONAL TASKS ERZSEBET MAROSI OSCAR BAZAN GUILLERMINA YANEZ JORGE BERNAL THALIA FERNANDEZ MARIO RODRIGUEZ JUAN SILVA ALFONSO REYES Laboratory of Neurosciences National University of Mexico Fes Iztacala, Mexico The objective of this stud to discover how narrow-band spectral measurements express emotional behavior. Electroencephalographic narrow band absolute and relative spectral powers were calculated for every 1 He of frequency from the recordings of 40 healthy male university stu- dents during emotional tasks. Sentences evoking positive, negative, and neutral emotional states were used as stimuli. Repeated multivariate analyses of variance were computed with IQ as covariate. The results showed only a few significant relations between absolute power and emotion. Relative power reflected better the differences determined by emotional states. The frequencies between 7.6 and 9.5 Hz showed differ- ences between emotional states over the entire scalp with increased power during neutral state. Interaction between emotional stares and cerebral regions revealed that all cerebral areas had an important role, especially frontal, parietal, and temporal regions. Hemispheric wa Received 25 January 2002. This project was supported by grant IN209396 DGAPA, UNAM. The authors express their gratitude to Mrs. Rebecca Stewart for her comments and insightful suggestions, Address correspondence to Dr. Erzséhet Marosi, Apartaclo Postal 82, Atizapin de Zaragoza, Estado de México, C.P. 52971, Mexico. E-mail: marosi@servidor.unam.anx 871872 E. Marosi et al. differences showed less power in the left hemisphere during positive emotional state (joy of love) in the 8.6 10 9.5 Hz band and in the right hemisphere during negative emotional state (frustration) in the 17.6 to 18.5 Hz band. Significant interaction between states, hemispheres, and electrode positions was observed in the frequency range of 17.610 I Hz. As none of our results covered the range of any broad band, we may conclude that narrow-band calculations are more adequate for the study of emotions, because their use reduces the danger that frequency- specific effects go undetected or cancel each other Keywords EEG, emotions of positive and negative valence, narrow band, spectral measurements The study of brain electrical activity during cognitive functioning is one of the most intriguing areas in investigation. Oscillatory brain activity is far from being completely understood. Most cognitive functions are based on highly parallel and distributed information processing. The hypothesis is that synchronization of neuronal dis- charges can serve as the integration of distributed neurons into cell semblies. Event-related oscillation bridges the gap between single neurons and neural assemblies, Selectively distributed oscillatory systems act as resonant communication networks through large popu- lation of neurons with functional relations to memory and integra- tive functions (Basar, Basar-Froglu, Karakas, & Schiirman, 1999). Neurons in cortical networks engage in synchronous activity in dif- ferent frequency bands and the probability of occurrence of syn- chronous activity in a particular frequency range depends on the central state of brain, on the presence of sensory signals, or on the occurrence of any mental activity (Singer, 1993). Apparently, the brain is tuned to a certain frequency depending on cerebral activity and this frequency is identified as a mental state. Low-frequency illations are related to coma or sleep and high-frequency oscilla- tions, especially in the beta and gamma bands, are particularly pro- nounced in awake performing brains. They generally show close relationship with sensory stimulation, attention, or emotional arousal. The study of emotional behavior has its inherent difficulties. One of them is that emotional behavior has multiple components and categorization of these components has not been done. Some inves- tigators focus on behavioral aspects and propose basic and cogni- tive factors. Others use anatomical distinctions and mention corticalEEG During Emotional Tasks 873 or subcortical elements. Furthermore, we assume that emotional be- havior changes from person to person. This idea could be explained by the cognitive aspects of emotion. In other words, the interpreta- tion of our emotions is varied in a broad range, although the bio- logical mechanism associated with emotional behavior must have similar characteristics in every person. A complete theoretical issue is needed to give a good base for further investigations. Electroencephalography has been used to measure changes in brain activity associated with emotion, as it is a noninvasive technique having high temporal resolution that can be easily synchronized with behavioral events. It gives us an excellent opportunity to relate electric activity of the brain to different emotional states. Most of the research has been focused on evoking emotions and measur- ing associated neural states. Traditionally, EEG studies have used broad-band measurements. These broad bands were determined in an arbitrary way; when electroencephalograms (EEGs) were inter- preted by visual inspection, these broad bands proved to show evi- dent changes. Lately, with the use of EEGs the idea still prevails that broad bands behave in an unitary way (Basar et al., 1999), although more and more studies indicate the necessity to split broad bands into two or more subbands (Klimesh, 1999; Iwaki, Hayashi, & Hori, 1997; Mecklinger, Ktamer, & Strayer, 1992; Aftanas, Lotova, Koshkaroy, & Papoy, 1998). Our previous paper on broad-band mea- surements related to emotions demonstrated mean frequency changes, suggesting that emotional behavior occurs in strictly defined narrow frequency ranges that do not overpass the limits of the broad bands. Because broad-band measurements are average values of changes in different frequencies, calculations may dilute or distort the re- sults. Based on our results, we suggested that narrow bands should be computed to reveal these subtle changes (Marosi et al., 2001). Unfortunately, narrow-band studies on positive and negative emo- tional states are rare. Iliuchenok (1996) presented emotionally pos tive, negative, and neutral words and recorded EEG in 10 deriva- tions (F3, F4, C3, C4, T3, T4, P3, P4, O1, and O2), He computed narrow frequency bands of | Hz. He reported an increase in relative power of 7 to 7.5 Hz for emotionally positive stimuli in frontal central, and temporal areas and 7.5 to 8 Hz changes in frontal, central, and parietal areas. Lately, Krause, Viemeré, Rosenquist,874 E. Marosi et al. Sillanmiiki, and Astrém (2000) studied EEG changes for every 2 Hz of frequencies to three types of emotionally laden film clips (ag- gressive, sad, and neutral). They calculated absolute power data and found significant derivation-emotional state interaction differences between no stimulation and aggressive film content in the 4 to 6 Hz band. Another issue is the valence theory. Behavioral observation of patients with unilateral brain damage (Gainotti, 1972; Sackeim et al., 1982; Lee, Loring, Dahl, Kimford, & Meador, 1993) indicated that left anterior lesions were often associated with depressive re- action and right anterior lesions with unduly cheerful behavior. A variety of recent EEG studies found suppression of alpha power when subjects were induced to experience certain emotions, and this suppression depended on the valence of the evoked emotion. In the case of negative emotions, less power was observed in the right than in the left frontal regions, whereas during the experience of positive emotion, alpha power asymmetry showed the opposite pat- tern (Ahern & Schwartz, 1985; Davidson, 1992; Wheeler, Davidson, & Tomarken, 1991; Aftanas, Koshkarov, Pokrovskaja, Lotova, & Mordvintsev, 1996; Bekkedal, Punskepp, & Ko: 1997). These EEG studies generally were done with broad-band measurements and focused their attention exclusively on the alpha band in order to show frontal desynchronization during emotional states. Other EEG and positron emission tomographic (PET) scan stud- ies revealed that right parietal (Smith, Meyers, Kline, & Bozman, 1987; Tucker & Dawson, 1984; Heller, Nitschke, Etienne, & Miller, 1997; Hagemann, Naumann, Becker, Maier, & Bartussek, 1998) and temporal (Schneider et al., 1995; Lane et al., 1997) areas also played an important role in emotional behavior. The objective of our study was to detect changes of absolute and relative power in 19 narrow frequency ranges (1 Hz) associated with neutral, positive, and negative emotional states in human subjects. If the hypothesis that every mental state, such as attention, relaxation, or emotional arousal, is accompanied by certain frequencies were true, then we should obtain a strictly defined frequency range character- izing emotional states in frontal, temporal, and parietal derivations. We also assume that negative emotions show less power in right hemispheric frontopolar leads, and positive emotions in the left one.EEG During Emotional Tasks 875 METHODS Forty young and healthy male, university students, with ages be- tween 20 and 25 years old, were studied. Considering that emo- tional behavior changes with sex, in this investigation we have chosen only men (Grossman & Wood, 1993; Kring & Gordon, 1998). The study of women is in process and comparisons between sexes will be published later. All subjects had an IQ higher than 95 as mea- sured by the Weschler WAIS Intelligence Scale (Spanish transla- tion) with a mean of 106. All were right-handed and none used nonprescribed drugs 2 days previous to the recordings. Recordings were carried out during imaginative emotional states using silver-silver chloride (Ag/AgCl) cup electrodes in 20 referen- tial derivations of the International 10-20 System (Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, 02, F7, F8, T3, T4, TS, T6, Fz, Cz, Pz, and Oz) with interlinked auricular reference. The EEGs were recorded using a MEDICID Mind Tracer system. The amplifiers had a bandwidth of 0.5 to 30 Hz and a notch filter of 50 Hz. The stimuli consisted of three types of sentences presented on the screen of a computer. The first type of sentences was designed to evoke negative emotional state (NES), frustration. (For example: You run all over the city to take the bus and you arrive 5 minutes late.) The second type of sentences had positive emotional state (PES). It was joy of love. (For example: You meet your girlfriend whose radiant beauty greatly affects you.) The third type of sentences intended to evoke neutral state (NS). (For example: You hear the barking of a dog from far away.) Stimulus rate and the format of presentation (length, size, type of letter, and luminosity) were the same throughout the experi- ment. Verbal, categorical judgments on the emotional valence of the stimuli (positive, negative, or neutral) was requested from all sub- jects after the recordings. Subjects classified correctly 90.8% of the frustration stimuli. The rest of the frustration stimuli (9.2%) were considered neutral. Subjects classified correctly 99.4% of joy of love stimuli and 0.6% of them were considered as neutral. Eighty-five percent of the neutral sentences were classified correctly. The other 15% were considered slightly positive, although sub- jects admitted that the emotional content was not as intense as PES876 E. Marosi et al. (joy of love). Incorrectly classified sentences were omitted from the analyses. Every stimulus was presented to the experimental subjects twice. The first time it was kept on the monitor until the subject asked for the second exposure. The second time it was exposed during 2000 ms. The segments used for calculation were those of the second exposure. At the first presentation the subjects were asked to read the sentence, imagine the situation, and induce the emo- tional state involved. When they felt intense emotion, they pressed the space bar and maintained the sensation at least 2 s for the re- cording. The intensity of the emotion was monitored by autonomic nervous responses (heart rate and galvanic skin responses). Elec- trodes for skin response were placed in the palm and on the back of the nondominant hand. Seventy different sentences, in each of the three situations, were presented in a balanced order. The task was difficult, so we needed to record 62 subjects to have 40 clean EEGs, without artifacts and with autonomic response marked by tachycardia or skin response. Fifteen percent of increase in the heart rate and in the amplitude of skin response was considered as good autonomic arousal. We included only those segments that showed significant autonomic changes. EEGs were visually inspected (off-line) and segments with arti- facts were excluded from further analysis. The raw EEG data were submitted to the fast Fourier transform (FFT) in order to calculate power spectra for every 1 Hz of frequency bands, beginning with 1.5 and finishing with 20.5 Hz (total = 19 bands of 1 Hz). Twenty- four segments of 2.56 s (61.44 s) of artifact-free EEG were used for the analyses. We computed both absolute (AP) and relative power (RP) values. AP data were log transformed and RP data were trans- formed by log [«/(1 =x)] to ensure a Gaussian distribution (John et al., 1983). In this way, in all bands and all derivations the checking of normality resulted in a W higher than .92 (Shapiro-Wilk test, SAS). EEG parametric scores for every | Hz band were entered into repeated multivariate (MANCOVA) with Greenhouse adjust- ment in the following way: (a) a model of 3 x 2 x 5 (Emotional States x Hemispheres x Derivations), where deviations are fronto- polar (Fpl—Fp2), frontal (F3-F4), central (C3-C4), parietal (P3- P4), and occipital (O1-O2) leads; (b) temporal derivations were entered by a model of 3 x 2 x 3 (Emotional States x HemispheresEEG During Emotional Tasks 877 x Derivations), where deviations are frontotemporal (F7—F8), temporal (T3—-T4), and temporal posterior (T5-T6) leads; and (c) midline leads were entered by a model of 3 x 4 (Emotional States x Derivations), where deviations are Fz, Cz, Pz, and Oz. The deri- vations for the 3 x 2 x 5 and 3 x 2 x 3 models were selected in this way to ensure the symmetry of the model. Post hoc comparisons of means were carried out by the Tukey honest significant differences. test. IQ values of each subject were entered as covariate. (IQ was used as covariate, because of its influence on cognitively induced emotions.) RESULTS The MANCOVA showed significant derivation main effect in all bands and significant hemisphere main effect in some of them, al- though herewith we present only those data that were related to emotional states (that is, state main effect and state-hemisphere and state-hemisphere-derivation interactions). Figure | shows all the significant findings when absolute power values were computed. Absolute power is a measurement of area under the curve that reflects amplitudes of the EEG waves. Figure 1A shows significant state main effect in the band of 3.6 to 4.5 Hz at temporal regions with higher amplitudes during NES and NS than during PES (p < .0371, F(2, 38) = 3.50, E = .9188). Figure 1B reveals significant state-derivations interaction (p < 0001, F(4, 36) = 9.53) in the same band, with PES and NES at lower amplitudes at frontotemporal than at temporal leads and with higher power values during NES. NS amplitudes were higher at frontotemporal regions. The post hoc test showed the following significant differences between means: F7-F8 in PES < T3-T4 in PES; T3-T4, T5-T6 in NES; and T5-T6, F7-F8 in NS. Figure 1C shows significant state-derivation interaction in the 6.6 to 7.5 Hz band at midline electrode positions (p < .0472. F(6, 34) = 2.18). The post hoe test did not reveal significant differences in means. Figure 1D shows significant state-derivation interaction in the 14.6 to 15.5 Hz band at midline electrode positions (p < .0046, F(6,ainjosqe ais + on avis 2 ow ais = ais ais = rand ret ea pine assy ag ean me (com cps) em fa oa asi pen oan 8 cvs mets rSe a wos ne (som 2 8 lave) uso 05 ro ipa a srk au moa MS (Gomme ten 2 a 878EEG During Emotional Tasks 879 34) = 3.27). We observed similar power values during PES and NES, with lower amplitudes during NES. The post hoc test revealed significant differences between Fz during PES and NS. These results show that absolute power values did not reflect hemispheric effect during the different emotional states. The 3.6 to 4.5 Hz band showed differences related to emotional states and state-derivation interaction at temporal regions. Midline electrodes showed state-derivation interactions in the 6.6 to 7.5 Hz band with higher differences at Pz, and in the 14.6 to 15.5 Hz band with higher differences at Fz. Figures 2 to 6 show all the significant findings when relative power values were computed. Relative power is the percentage of the total power in every band. Figure 2A shows significant state-hemisphere interaction in the frequency range of 4.6 to 5.5 Hz when the five superior electrodes were calculated (p < .0191, F(2, 38) = 4.63). We can observe in- creased relative power in the left hemisphere during PES and NS and in the right one during NES. In this band, PES had more power than NS. The Tukey post hoc test revealed significant differences between PES in the left hemisphere and NS in the right one. Figure 2B reveals significant state-hemisphere interaction in the same band for temporal leads (p < .0320, F(2, 38) = 3.94). Tempo- ral leads behaved in the same way as the five superior leads (Figure 2A). The post hoc test showed that PES in the left hemisphere had more power than PES and NS in the right one. Figure 2C shows state main effect in the frequency range of 6.6 to 7.5 Hz at temporal regions (p < .0047, F(2, 38) = 6.63). The post hoc test revealed higher power values during NS than PES and NES. Figure 3 shows the significant findings in the 7.6 to 8.5 Hz frequency band. We obtained significant state main effect for the five superior leads (p < .0230, F(2, 38) = 4.38, E =.9624 (Figure 3A). The Tukey test revealed significant differences between NES and NS. Figure 3B shows state main effect for temporal regions (p < .0310, F(2, 38) = 3.99, E =.9594), with NS having more relative power than the other two states Figure 3C displays significant state main effect at midline elec- trode positions (p < .0408, F(2, 38) = 3.63, E = .9662), with NS having more relative power than the two emotional states.ai = avis -0- se avis > ‘covets coe a 38 nN 880 pm 599 ea vem Ae pina S97 ma voce fame ‘oom mcr eg) oP idOf plied ZH $"8 01 9) (Gos oc) oot ene Ba Wve i se9 baw 153853 Nim AVIS ‘oom era eno Po ais (eee oscar ore Bs ‘vom 9509192 Mavs (tom nodont non melt ‘car -2) 0c20>4:0¢¥ 62 4 ‘hora 54922 We 103543 BENS "tee nied a) een 2 881882 E, Marosi et al. Figure 4 shows the significant findings in the frequency band of 8.6 to 9.5 Hz. We can observe significant state main effect for the five superior leads (Figure 4A: p < .0142, F(2, 38) = 5.03, E = .9727), for temporal leads (Figure 4C: p < .0304, F(2, 38) = 4.01, E = .9853), and for midline electrode positions (Figure 4D: p < 0161, F(2, 38) = 4.86, E = .9342), with all cerebral regions having more power during NS than during the two emotional states. We can also see significant hemisphere-state interaction (Figure 4B) for the five superior leads (p < .0318, F(2, 38) = 3.95). The post hoc test revealed the following differences in means: by hemi- spheres: PES in the left hemisphere had less power than PES in the right one; and by states: NS had more power than NES and PES in the left hemisphere. Figure 5A shows significant state-derivation interaction in the 8.6 to 9.5 Hz frequency band at midline electrode positions (p < .0392, F(6, 34) = 2.27). The Tukey post hoc test revealed the fol- lowing signi’ t differences: by emotional state: Pz in NS > Pz in NES and Oz in NS > Oz in PES; by derivation: posterior regions (Pz and Oz) in NS had higher power values than anterior regions (Fz and Cz) in NES and NS. Figure 5B reveals significant state-derivation interaction in the 12.6 to 13.5 Hz band for the five superior electrode positions (p < .0352, F(6, 34) = 2.12. The post hoc test revealed the following significant differences in means: by emotional state: Fp!—Fp2 in the three states had less power than C3-C4 in PES and P3-P4, O1-O2 in the three states; by derivation: posterior regions had more power than anterior ones in the three states, Figure 5C shows significant state-derivation interaction in the 13.6 to 14.5 Hz band for midline electrode positions (p < .0435, F(6, 34) = 2.22). The post hoc test revealed the following signifi- cant differences between means: by emotional state: Fz in PES and Oz in the three states > Cz in NS and Pz in the three states; by derivation: Fz and Cz in NES and NS had less power than Oz in the three states. Figure 5D shows significant state-hemisphere interaction in the 17.6 to 18.5 Hz band for the five superior electrode positions (p < .0285, F(6, 34) = 4.09). The Tukey post hoc test revealed that left NES > left and right PES, right NES, and left NS.pataioysumen ayy swoys six “Sanya JONOd 2 Sead oe es pani vers eau vem fan (coe ie ay mo ‘eunepieod easier1 avis = Joy purg sao vores pine 8-814 0a nem ‘woman on dO eee eh 208 Bea imag csc eu ewan (com msc a en DE fF SI Sad “SanpeA som Jo} YAOONWIN J0 siinsos mrautUatS *§ AMADA oueaaaa anit a « avis = Se maple ae Sean voce eg afl. . =. ae a er mem gera na arms me ‘com vo eae 884EEG During Emotional Tasks 885 Figure 6A shows significant state-hemisphere-derivation inter- action in the 17.6 to 18.5 Hz frequency band for the five superior electrode positions (p < .0353, F(8, 32) = 2.68). The post hoc test revealed the following significant differences between means: by emotional states: the relative power at F3 in NES was major than in NS; Fpl! in the three states was major than in PES at central, pari- etal, and occipital regions; by hemispheres: F3 in NS had less power than F4; C3 in NES had more power than C4; and P3 in NES had more power than P4; by derivation: anterior regions had more power than posterior ones. Figure 6B shows significant state-hemisphere-derivation inter- action in the 18.6 to 19.5 Hz band for temporal leads (p < .0401, F(6, 34) = 2.97). The post hoc test revealed the following signifi- cant differences between means: by states: T3 in PES and NES and F7-F8 in the three states had higher RP than T5—T6 in the three states, by hemisphere: T3 had higher RP than T4 in PES and NES; by derivation: T3 in PES and NES and F7-F8 in the three states had higher relative power than T5—T6 in the three states. These results show that the 7.6 to 9.5 Hz band in all leads and the 6.6 to 7.5 Hz band in temporal regions reflected state main effect, showing that these frequencies are closely related to emo- tional states. In all cases neutral state had more relative power than the two emotional states (PES and NES). We obtained hemispheric differences by emotional states in the frequency ranges of 4.6 to 5.5, 8.6 to 9.5, and 17.6 to 18.5 Hz. In the 4.6 to 5.5 Hz band, PES had more power in the left hemisphere and frustration in the right one in all homologous leads. In the band of 8.6 to 9.5 Hz, the five superior leads showed more power in the right hemisphere than in the left during PES. The five superior elec- trodes had more power in the left hemisphere than in the right one during NES, in the band of 17.6 to 18.5 Hz. State-derivation interactions were observed in the frequency ranges of 8.6 to 9.5 and 12.6 to 14.5 Hz. In the 8.6 to 9.5 Hz band, midline leads showed differences by state especially at Pz and Oz, where the two emotional states had significantly less power than the neutral state. The five superior electrodes (12.6 to 13.5 Hz) had anterior leads with less power than posterior ones and PES, with more power at central regions than the other two states. Midline886 E. Marosi et al. A) Plot of Means (five superior leads) ‘S-way interaction in the 17.6-18.5 Hz band F=(8,32) 2.68; p<.0953 22 228 20 23s oe Bow j a0 ass —o- STATE a PES “0+ STATE a NES *@ STATE an he TET RGM LEFT RG? URFT AGT LEFT RIGHT LEFT RIGHT Fot-tee ree 3c Pope (01-02 1B.) Plot of Means (temporal leads) ‘Sway interaction in the 18.6-19.5 Hz band Fe(4, 96) 2.97; p<.0401 2% 235 i 2 2m _ i ass Ey ~o- STATE i aad PES -& STATE a NES ve STATE an LEFT RIGHT LEFT RIGHT, eT RIGHT NS 1314 1818 Free FIGURE 6. Significant results of MANCOVA for relative power values, Y axis shows the transformed relative power values. PES is joy of love, NES is frustration, and NS is neutral state, (A) State-hemisphere-derivation interaction in the 17.6 to 18.5 Hz band for frontopolar, frontal, central, parietal, and occipital leads. (B) State: sphere-derivation interaction in the 18.6 to 19.5 Hz band for temporal leadsEEG During Emotional Tasks 887 electrodes showed state-derivation differences in the frequency range of 13.6 to 14.5 Hz, with Fz and Cz having more power during NS than during PES and NES. State-hemisphere-derivations interactions were observed in the 17.6 to 18.5 Hz band for the five superior electrode positions and in the 18.6 to 19.5 Hz band for the temporal leads, with complex interaction between them. DISCUSSION The observations in the present study are of interest from several perspectives. First, they suggest that traditional broad bands do not reflect reliably EEG changes during emotional states, as none of our results covered completely the range of any broad band. Specific changes dependent on emotional states occurred in the 7.6 to 9.5 Hz band over the entire scalp. These frequency ranges correspond to high theta and low alpha power. These findings are partly sup- ported by the findings of Iiuchenok (1996) who found that 7.6 to 8.5 Hz frequency range reflected best emotions on EEG. Theta rhy- thm of monkeys, and maybe of humans too, is generated in septal- hippocampal-cortical connections, and is considered to contribute to the EEG as well. Long-term potentiation of these neurons is currently a popular model of learning and memory and has an im- portant role in cognitive performance (Stewart & Fox, 1990). The measurement of absolute power (amplitudes) is commonly used in the studies of electric activity during emotional behavior, but relative power values (percentages) are rarely considered. Ac- cording to our results, relative power revealed more emotion depen- dent changes than absolute power. Consequently, we can recom- mend using both, if it is possible, and, if not, relative power is more convenient. It is a common practice to study frontal desynchronization, as the frontal lobes are considered to have a preferential role in emotional behavior. Most previous papers on this topic only mention frontal activation in the alpha band during emotional tasks (Ahern & Schwartz, 1985, Schellberg, Besthorn, Klos, & Gasser, 1990; Sobotka, David- son, & Senulis, 1992; Tucker & Dawson, 1984; Davidson, 1992).888 E. Marosi et al. Considering the complexity of brain, it is difficult to imagine that a higher nervous function, like emotion, should cause activation of the frontal lobes only. Our results showed important frontal involvement (state-derivation interaction) mainly at Fz, although temporal leads also had important participation. Emotional behavior is considered to be processed by the limbic system, and previous PET scan (Schneider et al., 1995; Lane et al., 1997) and EEC studies (Davidson, 1992; Sobotka et a., 1992; Wheeler et al., 1993; Tomarken, Davidson, & Henriques, 1990) suggest frontal and temporal cortex involvement in phasic emotional states with right hemispheric superiority (Spence, Shapiro, & Zaidel, 1996). The role of temporal areas in emotional behavior is also shown in patients with epilepsy of temporal lobe. Our data also revealed that parietal leads had an important partici- pation in emotional behavior. These data are in accordance with the findings of Smith et al. (1987). These authors focused their study on the differential parietal processing of emotional stimuli under affective and cognitive conditions. They found bilateral parietal in- volvement with higher differentiation in the right hemisphere. We also observed the participation of occipital areas (Oz) in the 8.6 to 9.5 Hz frequency range. A “cross-modality” experiment in humans and cats of Basar and Schuerman (1994) found marked alpha components only for visual stimulation. The authors propose that this alpha re- sponse might reflect primary sensory processing, whereas the theta responses might be more involved in associative-cognitive process- ing, as they are less dependent on sensory input. Our data showed that the neutral state generally had higher rela- tive power values than emotional ones. This finding is in accor- dance with the observation of Bekkedal et al. (1997), who described lower power values during the emotional states as compared with the neutral one. This result can be related to the difficulty of the task in the bands of 7.6 to 9.5 Hz. Gevins et al. {1997) described that a difficult task elicited smaller slow parietocentral alpha, than did the easy task. Supposedly, in our study imagining a neutral situation was the easier task, as subjects need not cope with induc- ing strong emotions, Decreased theta was observed by Klimesch et al. (2001) during high alertness, which supports our finding of hav- ing less theta power during emotional situations than in neutral ones. Another point to consider is the hemispheric activation duringEEG During Emotional Tasks 889 different emotional states. Previous behavioral and sodium amytal studies on the hemispheric specialization of the emotional expres- sion (Gainotti, 1972; Sackeim et al., 1982; Lee at al., 1993) re- ported that emotions of negative valence are processed by the right frontal cortex, and those of positive valence by the left frontal cor- tex. A variety of research suggests that, when subjects are induced to experience certain negative emotions, there is a suppression (i.e., more activation) of alpha power in the right than in the left frontal region, whereas during the experience of positive emotion alpha power asymmetry described in this region shows an opposite pat- tern (Tucker & Dawson, 1984; Wheeler et al., 1993; Sobotka et al., 1992). Clinical observations of patients with unilateral brain dam- age also support this finding. Our study revealed less relative power in the left hemisphere during positive emotion in the 8.6 to 9.5 Hz band and in the right hemisphere during negative emotion in 17.6 to 18.5 Hz band. However, absolute power values in the 4.6 to 5.5 Hz band showed an opposite pattern, where positive and neutral states evoked higher amplitudes in the left hemisphere and negative emotional state did not show significant hemispheric differences. Emotion is a superior cognitive function, and the brain electric processes engaged in emotions can be assumed to show substantial interindividual differences. Our results are based on group measure- ments, but general tendencies must be manifested. However, in the future individual peak frequency adjustment, proposed in the thought- ful paper by Klimesh (1999), should be considered. A better understanding of the neural events, which underlie emo- tional behavior, will depend upon the ability to resolve the dynamics of cortical activation in much finer detail. For the moment we can conclude that the use of narrow frequency bands reduces the danger that frequency-specific effects go undetected or cancel each other. Additional studies will be required to confirm definitely the utility of narrow-band measurements over broad-band studies. REFERENCES Aftanas, L. 1, Koshkarov, V. 1, Pokrovskaja, V, L., Lotova, N. V., & Mordvintsev, Y, N. (1996). Event-related desynchronization patterns to emotion-related feedback stimuli. International Journal of Neuroscience, 87, 151-173,890. E. Marosi et al. namical coupling between different brain areas during evoked emotions: An EEG in- ion. Biological Psychology, 48, 121-138. Ahem, G. L.. & Schwartz, G. E. (1985). Differential lateralization for positive and nega- tive emotion in the human brain: EEG spectral analysis. Neuropsychologia, 23, 745: 756. Basar, Basar-Eroglu, C., Karakas, & Schiirman, M. (1999), Oscillatory brain theory: ‘A new trend in neuroscience. /EEE Engineering in Medicine and Biology, 18, 36-66. Basar, E., Basar-Eroglu, C., Karakas, S., & Shilrman, M. (1999). Are cognitive processes manifested in event related gamma, alpha, theta and delta oscillations in the EEG? Neuroscience Letters, 259, 165-168. ar, E., & Schuermann, M. (1994). Funetior sponses in humans and cats recording in Cybernetics, 7: Bekkedal, M. Y. V., Punskepp, J.. & Rossi, J. (1997), Changes in frontal EEG activity in response to auditory stimuli. Sociery for Neuroscience, 33, 519.17. Davidson, R. J, (1992). Anterior cerebral asymmetry and the nature of emotions. Brain and Cognition, 20, 125 Gainotti, G. (1972). Emotional behavior and hemispheric side of lesion. Cortex, 8, 41-55. Gevins, A. (1998). The future of electroencephalography in assessing neurocognitive fune~ tioning. EEG and Clinical Neurophysiology, 106, 165-172. Grossman, M., & Wood, W. (1993). Sex differences in intensity of emotional experience; A social role interpretation. Journal af Personality and Social Psychology, 65, 1010-1022. Hagemann, D., Naumann, E., Becker, G., Maier, S., & Bartussek, D. (1998). Frontal brain ‘asymmetry and affective style: A conceptual replication. Psychophysiology, 35, 372- 388. ial aspects of evoked alpha and theta re- cross modality” experiments, Biological Heller, W., Nitschke, J. B., Etienne, M. A., & Miller, G. A. (1997). Patterns of regional brain activity differentiate types of anxiety. Jounal of Abnormal Psychology, 106, 376-385. Hiuchenok, 1. R. (1996). Differences in the EEG frequency characteristics during the per- ception of positive-emotional, negative-emotional and neutral words. Zinrnal Nervoi Deiat Im 1. Pavlova, 46, 457-468. Iwaki, T., Hayashi, M., & Hori, T. (1997). Changes in alpha band EEG activity in the frontal area after stimulation with music of different affective content. Perceptual and Motor Skills, 84. 6. John, E. R., Prichep. L., Abn, H., Easton, P., Fridman, J., & Kaye, H. (1983). Neurometric evaluation of cognitive d riders in children, Progress in Neurobiology, 21, 239-290. Klimesch, W. (1999), EEG alpha and theta oscillations reflect cognitive and memory performance: A review and analysis. Brain Research Reviews, 29, 169-195. Klimesch, W., Doppelmayr, M., Wimmer, H., Schwaiger, J., Rdhm, D., Gruber, W.. & Hutzler, F. (2001). Theta band power changes in normal and dyslexic children. Clini- cal Neurophysiology, 112, 1174-1185. Krause, C. M., Viemerd, V., Rosenguist, A., Sillanmaki, L., & Astrom, T. (2000). Relative EEG desynchronyzation and synchronization in human to emotional film content: An analysis of the 4-6, 6-8, 8-10 and 10-12 Hz frequency bands, Neuroscience Letters, 26, 9-12. Kring. A. M., & Gordon, A. H. (1998). Sex differences in emotion: expression, experience and physiology. Journal of Personality and Social Psychology, 74. 686-703. Lane, R. D., Reiman, E, M., Bradley, M. M., Lang, P. J.. Ahtern, G. L. Davidson, R. J., & Schwartz, G. E, (1997). Neuroanatomical correlates of pleasant and unpleasant emo- tion. Neuropsychotogia, 35, 1437-1444. . P., Loring D, W., Dahl, J. L., Kimford, J., & Meador. M. D. (1993). Hemispheric ization for emotional expression. Neuropsychiatry. Neurophysiology and Be- havioral Neurology. 6, 143-148EEG During Emotional Tasks 891 Marosi, E., Rodriguez, H., Rodriguez, M., Silva, J., Bernal, J., Femandez, T.. Yanez, G. Guerrero, V., & Reyes, A. (2001). Broad band spectral measurements of EEG during emotional tasks. Infernational Journal of Neuroscience, 108, 251-279. Mecklinger, A., Ktamer, A. F., & Strayer, D. L. (1992). Event related potentials and EEG components in a semantic memory search task. Psychophysiology, 29, 104-119, Sackeim, H. A., Greenberg, M. A., Weiman, A. L., Gur, R. C., Hungerbuhler, J. P.. & Geschwind, N. (1982). Hemispheric asymmetry in the expression of positive and nega- tivity emotions. Archives of Neurology, 39, 210-218. Schellberg, D., Besthorn, C., Klos, T., & Gasser, T. (1990). EEG power and coherence while male adults watch emotional video films. International Journal of Psychophysi- ology, 9, 279-291 Schneider, F., Gur, R. E., Mozley, L. H., Smith, R. J., Mozley, P. D., Censits, D. M., Alavi, A., & Gur, R. C. (1995). Mood effects on limbic blood flow correlate with emotional self rating: A PET study with oxygen-15 labeled water. Psychiatry Re- search, 61, 265-283. Singer, W. (1993). Synchronization of cortical activity and its putative role in information processing and learning. Annual Review of Physiology 349-374. Smith, B. D., Meyers, M., Kline, R., & Bozman, A. (1987). Hemispheric asymmetry and emotion: lateralized parietal processing of affect and cognition. Biological Psychol- ogy, 25, 247-260. Sobotka, S. Davidson, R. J., & Senulis, J. A. (1992). Anterior brain electrical asymme- tries in response to reward and punishment. EEG and Clinical Neurophysiology, 83, 236-247. Spence, $., Shapiro, D., & Zaidel, E. (1996). The role of the right hemisphere in the physiological and cognitive components of emotional processing. Psychophysiology, 33, 112-22. Stewart M., & Fox, S .E. (1990). Do septal neurons pace the hippocampal theta rhythm? Trends in Neuroscience, 13, 163-168. Tomarken, A. J.. Davidson, R. J., & Henriques, J. B. (1990). Frontal brain asymmetry predicts affective responses to films. Journal of Personality and Social Psycholo: 59, 791-801. Tucker, D. M., & Dawson, S. L. (1984). Asymmetric EEG changes as method actors generated emotions, Biological Psychology, 19, 63-75. Wheeler, E., Davidson, R. J., & Tomarken, A. J. (1993). Frontal brain asymmetry and emotional reactivity: A biological substrate of affective style. Psychophysiology, 30, 82-89,Copyright © 2002 EBSCO Publishing