Brain and Cognition 62 (2006) 43–57 www.elsevier.com/locate/b&c Dissociation of event-related potentials indexing arousal and semantic cohesion during emotional word encoding Daniel G. Dillon, Julie J. Cooper, Tineke Grent-‘t-Jong, Marty G. WoldorV, Kevin S. LaBar ¤ Center for Cognitive Neuroscience, Duke University, Durham, NC 27708-0999, USA Accepted 23 March 2006 Available online 6 May 2006 Abstract Event-related potential (ERP) studies have shown that emotional stimuli elicit greater amplitude late positive-polarity potentials (LPPs) than neutral stimuli. This eVect has been attributed to arousal, but emotional stimuli are also more semantically coherent than uncategorized neutral stimuli. ERPs were recorded during encoding of positive, negative, uncategorized neutral, and categorized neutral words. DiVerences in LPP amplitude elicited by emotional versus uncategorized neutral stimuli were evident from 450 to 1000 ms. From 450 to 700 ms, LPP eVects at midline and right hemisphere frontal electrodes indexed arousal, whereas LPP eVects at left hemisphere cen- tro-parietal electrodes indexed semantic cohesion. This dissociation helps specify the processes underlying emotional stimulus encoding, and suggests the need to control for semantic cohesion in emotional information processing studies. 2006 Elsevier Inc. All rights reserved. Keywords: AVect; Amygdala; Categorization; Emotion; Evoked potentials; P300 1. Introduction Schupp, Cuthbert et al., 2004; Schupp, Ohman et al., 2004). This argument is based on cognitive studies that show an Emotional stimulus encoding has been well-studied association between increased perceptual processing of using event-related potential (ERP) methodology. These stimuli (after initial identiWcation) and increased LPPs (Rit- investigations have consistently shown that, relative to neu- ter & Ruchkin, 1992). It is proposed that emotional stimuli tral stimuli, emotional stimuli elicit late positive-polarity (e.g., pictures of opposite sex nudes or dangerous animals) potentials (LPPs) of increased amplitude (Cuthbert, are particularly salient to humans because they are closely Schupp, Bradley, Birbaumer, & Lang, 2000; Diedrich, Nau- tied to reproduction and survival (i.e., they are evolution- mann, Maier, & Becker, 1997; Dolcos & Cabeza, 2002; arily signiWcant stimuli). Because of this salience, emotional Johnston, Miller, & Burleson, 1986; Naumann, Bartussek, stimuli elicit activity in the brain’s appetitive and defensive Diedrich, & Laufer, 1992; Palomba, Angrilli, & Mini, 1997; motivational systems (Lang, Bradley, & Cuthbert, 1998) Schupp et al., 2000; Schupp, Cuthbert et al., 2004; Schupp, and tend to recruit attentional resources and sustained Ohman et al., 2004). Emotional ERP eVects generally attentional processing, which, in turn, leads to increased encompass the P3 and/or a subsequent slow-wave (SW) and amplitude LPPs. This hypothesis receives support from the are broadly distributed, having been observed over frontal, fact that highly arousing emotional stimuli, which are pre- central, and parietal scalp regions. sumably especially motivationally signiWcant, elicit LPPs of It has been suggested that increased LPPs reXect greater amplitude than less arousing emotional stimuli increased attentional resources devoted to emotionally (Cuthbert et al., 2000; Schupp et al., 2000; Schupp, Jungho- arousing stimuli (Cuthbert et al., 2000; Schupp et al., 2000; fer, Weike, & Hamm, 2003). Another factor that may contribute to the amplitude V * Corresponding author. Fax: +1 919 681 0815. di erence in LPPs elicited by emotional and neutral stimuli E-mail address: [email protected] (K.S. LaBar). is semantic cohesion. A group of neutral items (e.g., the 0278-2626/$ - see front matter 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.bandc.2006.03.008 44 D.G. Dillon et al. / Brain and Cognition 62 (2006) 43–57 words rug, theory, and folder) is likely to possess lower inter- There are two main assumptions inherent in the experi- item associativity than a group of emotional items (e.g., the mental design. First, ERPs that distinguish both emotional negative words violence, anger, and threat) and are thus less stimulus classes from the categorized neutral stimulus class semantically cohesive. Numerous studies have shown that (school-related words) should reXect arousal-related pro- ERPs are sensitive to semantic factors during stimulus cesses (because the emotional and categorized neutral stim- encoding. For example, studies of semantic priming have uli have been equated for semantic cohesion). Based on the demonstrated that primed words (compared with unprimed emotional ERP literature, we expected arousal-related ERP words) elicit increased positivities during relatively late eVects to be broadly distributed, to onset at about latency intervals (e.g., peaking between 550 and 650 ms in 300–400 ms post-stimulus, and to last for several hundred Bentin, McCarthy, & Wood, 1985). In addition, the ampli- milliseconds (Cuthbert et al., 2000; Dolcos & Cabeza, 2002; tude of the N400 component elicited by individual words is Schupp et al., 2000). inversely related to the ease with which those words are inte- Second, ERPs that distinguish categorized neutral stim- grated into an existing semantic context (Kutas & Besson, uli from uncategorized neutral stimuli should reXect pro- 1999; Kutas & Hillyard, 1980; Kutas & Schmitt, 2003; cesses related to semantic cohesion. Because semantic Kutas & Van Petten, 1994). Typically, these studies have cohesion has not been directly investigated in previous neu- used non-emotional stimuli and have highlighted the seman- roscientiWc studies of emotional stimulus encoding, the pre- tic relationships between stimuli to a greater degree than cise timing and spatial location of ERP eVects related to have studies of emotional item encoding. Nonetheless, the semantic cohesion was diYcult to predict. However, given results suggest that the increased semantic cohesion pos- that hemispheric lateralization of activity in semantic-per- sessed by groups of emotional stimuli (relative to neutral ceptual networks varies by stimulus type, we predicted that stimuli) may contribute to increased amplitude LPPs, per- semantic cohesion eVects would be lateralized to the left haps in concert with reductions in N400 amplitude. hemisphere since words were used as stimuli. With respect Recent ERP studies (Maratos, Allan, & Rugg, 2000; to timing, previous ERP research on word encoding indi- McNeely, Dywan, & Segalowitz, 2004; Windmann & cates that ERP components are not modulated by contex- Kutas, 2001) have investigated the role played by semantic tual inXuences until approximately 260 ms post-stimulus cohesion during recognition memory testing for emotional (Halgren, 1990). Furthermore, the duration of other ERP and neutral stimuli, with somewhat mixed results (see Sec- components known to be sensitive to the context in which tion 4). The primary goal of the present study was to deter- lexical stimuli are presented (e.g., the N400) is inversely mine whether semantic cohesion contributes to the related to the ease with which the stimuli can be integrated amplitude diVerence in LPPs elicited by emotional and neu- into the context. Because we used relatively broad contexts, tral items during initial encoding. To achieve this aim, the we predicted that LPPs sensitive to semantic cohesion scalp electroencephalogram was recorded while partici- would be evident over left hemisphere sites in approxi- pants encoded four classes of words. Three of the word mately the 260–700 ms time window. This prediction is in classes are commonly used in emotion research: positive, line with ERP studies of semantic priming, which have negative, and uncategorized neutral. These stimuli were identiWed increased amplitude LPPs for target words (in speciWcally chosen so that the emotional words would be semantically related prime-target pairs) in similar time both more arousing and more semantically coherent than ranges (Bentin et al., 1985). Dissociating the inXuences of the uncategorized neutral words. The novel aspect of the arousal and semantic cohesion during encoding will help study design involves the inclusion of a fourth stimulus characterize the mechanisms underlying emotional advan- class—categorized neutral words—which were organized tages in stimulus processing, with potential implications for around a “school” theme. These categorized neutral stimuli understanding how emotional content beneWts other cogni- were selected to be equivalent to the uncategorized neutral tive functions, such as memory. stimuli with respect to arousal and valence, but equivalent to the positive and negative stimuli with respect to semantic 2. Method cohesion. “School-related” stimuli were selected for the cat- egorized neutral class because “school” constitutes a rela- 2.1. Participants tively broad semantic category whose exemplars could be matched to emotional stimuli with respect to word fre- Twenty-one healthy right-handed participants (12 quency, imageability, and semantic cohesion. We expected females, 9 males) with a mean age of 23 (SD D 4.29) partici- that it would be diYcult to achieve such matching with pated in the experiment. Data from six participants were other typical semantic categories—e.g., fruits, animals, or excluded from the analysis. One of these did not follow furniture—because those categories would tend to be too instructions, one ended the experiment prematurely narrow or too imageable in comparison to the