Emotion and Attention Interaction Studied through Event-Related Potentials Luis Carretie1, �� Manuel Mart��n-Loeches2, �� Jose �� Antonio Hinojosa2, and Francisco Mercado1 Abstract & Several studies on hemodynamic brain activity indicate that emotional visual stimuli elicit greater activation than neutral stimuli in attention-related areas such as the anterior cingulate cortex (ACC) and the visual association cortex (VAC). In order to explore the temporo-spatial characteristics of the interaction between attention and emotion, two processes characterized by involving short and rapid phases, event-related potentials (ERPs) were measured in 29 subjects using a 60-electrode array and the LORETA source localization software. A cue/target paradigm was employed in order to investigate both expectancy-related and input processing- related attention. Four categories of stimuli were presented to subjects: positive arousing, negative arousing, relaxing, and neutral. Three attention-related components were finally analyzed: N280pre (from pretarget ERPs), P200post and P340post (both from posttarget ERPs). N280pre had a prefrontal focus (ACC and/or medial prefrontal cortex) and presented significantly lower amplitudes in response to cues announcing negative targets. This result suggests a greater capacity of nonaversive stimuli to generate expectancy-related attention. P200post and P340post were both elicited in the VAC, and showed their highest amplitudes in response to negative- and to positive-arousing stimuli, respectively. The origin of P200post appears to be located dorsally with respect to the clear ventral-stream origin of P340post. The conjunc- tion of temporal and spatial characteristics of P200post and P340post leads to the deduction that input processing-related attention associated with emotional visual stimulation in- volves an initial, rapid, and brief ���early��� attentional response oriented to rapid motor action, being more prominent towards negative stimulation. This is followed by a slower but longer ���late��� attentional response oriented to deeper processing, elicited to a greater extent by appetitive stimulation. & INTRODUCTION Current Data and Open Issues Though the neural activity associated with emotion and related to attention have mainly been studied separately, they constitute closely interrelated and mutually de- pendent processes. In natural environments, attention is preferentially sustained by stimuli that have affective significance, in contrast to routine, emotionally neutral events (e.g., Lang, Bradley, & Cuthbert, 1997). Some recent experiments using hemodynamic neuroimaging methods provide interesting data indicating that certain attention-related brain areas increase their activity in response to emotional visual stimuli. Firstly, there is a group of studies showing that several areas of the visual association cortex (VAC) are more highly activated in response to both positive (or pleasant) and negative (or unpleasant) images than in response to neutral pictures (Lane, Chua, & Dolan, 1999 Lang et al., 1998 Reiman et al., 1997). Another group of data, also obtained through functional hemodynamic methods, has indi- cated activation in the anterior cingulate cortex (ACC) in response to emotional visual stimulation, positive and negative (Northoff et al., 2000 Teasdale et al., 1999 Canli, Desmond, Zhao, Glover, & Gabrieli, 1998). Both groups of results have been interpreted as indicating increased attention towards affective stimuli. Results obtained in relation to hemodynamic changes associated with attention towards visual emotional stim- ulation should be complemented by data recorded using temporally agile physiological signals. Both attention and emotion are characterized by involving short, rapid subprocesses, some of the most important occurring within the first 500 msec after stimulus onset (e.g., Mangun & Hillyard, 1995 Halgren & Marinkovic, 1994). High temporal-resolution techniques are neces- sary to study several unexplored important issues. Firstly, it is necessary to study which aspects or types of attention are associated with the areas found to be activated in response to emotional visual stimuli. Usual tasks employed in neuroimaging studies are character- ized by the presentation of sets of pictures (with rela- tively long interstimulus intervals) during several seconds or even minutes, and probably involve different types of attention. In this regard, though the involve- 1Universidad Autonoma �� de Madrid, 2Universidad Complutense de Madrid �� 2001 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 13:8, pp. 1109- 1128

ment of both the VAC and ACC in attention to visual stimulation has received solid theoretical and experi- mental support (Sturm et al., 1999 Luck, Chelazzi, Hill- yard, & Desimone, 1997 Shulman et al., 1997 Mangun & Hillyard, 1995 Posner & Raichle, 1995), they have been proposed to reflect different aspects of attention and to be preferentially activated with different task requirements or periods. According to Posner and Pe- tersen (1990), we can distinguish between ������a general alert state and one in which attention is clearly oriented and engaged in processing information������ (p. 33). Tasks or phases within a task in which a significant level of expectancy or alertness is necessary, such as those related to target detection, have been proposed to elicit the former type of attention (which may appear even in the absence of stimulation), and involve the activation of the anterior cingulum (Posner & Petersen, 1990). The involvement of the ACC in alertness has been supported experimentally (e.g., Sturm et al., 1999). On the other hand, tasks requiring recognition of already presented visual stimuli are associated with the second type of attention, involving an increase in the activation of the VAC (Posner & Petersen, 1990 Posner & Raichle, 1995). Several experiments clearly support this involvement of the VAC in input processing-related attention towards visual stimuli (see review by Mangun & Hillyard, 1995). It is important to distinguish both types of attention in response to emotional images because they have sepa- rate functions. In order to achieve this distinction, it is necessary to analyze separately the activity in expect- ancy-related and in input processing-related periods of the experimental task through high temporal resolution signals. Secondly, the temporal characteristics of subprocesses within each type of attention to emotional stimulation should be described. For example, it is necessary to determine whether or not the attentional reactions to positive and to negative stimuli are produced at the same moment. The fact that hemodynamic neuroimaging methodologies have not found temporal differences in responses to the two types of emotional stimulation may be due to the fact that they record an accumulation of activity over relatively long periods. In this regard, several behavioral studies suggest that negative events elicit more rapid or more prominent emotional responses (involving cognitive and physiological changes) than neutral or positive events (see reviews by Cacioppo & Bernston, 1994 Cacioppo & Gardner, 1999 Taylor, 1991). Recent brain activation data obtained with high temporal resolution methodologies (magnetoencepha- lography, in this particular case) support this ���negativity bias��� (Northoff et al., 2000). Attention-related brain ac- tivity in relation to this bias has not yet been studied directly, though the privileged access of negative stim- ulation to attentional resources has been indirectly sug- gested by studies on autonomic physiological responses (Ohman, �� Esteves, Flykt, & Soares, 1993) and behavioral studies using reaction time (Pratto & John, 1991) or visual fixation (Fiske, 1980) as dependent variables. The adaptive and evolutionary advantages of this bias are obvious: The consequences of a dangerous or injurious event are often much more dramatic than the conse- quences of ignoring or reacting slowly to neutral or even appetitive stimuli (e.g., Ekman, 1992 Ohman, �� 1992). The combination of high spatial and high temporal resolution may provide information on a third important issue: which areas are activated in each subprocess within each type of attention to emotional stimulation. Data exist indicating that, in the VAC alone, emotional stimuli elicit activity in more than 15 sources or ���clusters��� along with those activated by all (emotional and neutral) visual stimuli (see, e.g., Lang et al., 1998). It is important to investigate which subprocess of attention they reflect (if any) by studying their temporal characteristics. This issue may be closely related, first, to time-related neg- ativity bias, since circuitry that computes urgent utility (e.g., avoidance of a dangerous event) often diverges from the circuitry subserving deep identification and discrimination (Shizgal, 1999). Secondly, it may also be related to the current debate on whether the evaluative systems (and the neural substrates) involved in the processing of positive and negative stimuli are inde- pendent or overlapping. Emotional content of stimula- tion has been traditionally assumed to be defined by two dimensions, valence (whose extremes are positive- neg- ative) and arousal (calming- arousing). Initially proposed by Russell (1979), the ���valence arousal��� or ���circumplex��� model considers that both positive and negative events are processed (and trigger emotional reactions) in the same, and single, valence-related evaluative circuit. The idea that positive and negative situations activate differ- ent evaluative systems (associated, at least in part, with separate neural mechanisms) has recently been de- fended (Cacioppo & Gardner, 1999 Davidson & Irwin, 1999 Tellegen, Watson, & Clark, 1999). This perspective has received experimental support (Northoff et al., 2000 Lane, Reiman, Ahern, Schwartz, & Davidson, 1997 Le- Doux, 1995). The use of a methodology with both high temporal and relatively high spatial resolution may facilitate the exploration of spatial differences even in the rapid and transient attentional subprocesses, whose involvement in this positive- negative ���spatial independ- ence��� has not yet been defined. The Present Study Electroencephalographic (EEG) recording of event-re- lated brain potentials (ERPs) constitutes a powerful and necessary tool for exploring these important issues concerning the interaction of attention and emotion. Apart from high temporal resolution, ERPs are able to provide acceptable spatial information if a dense and wide distribution of electrodes is used and subsequent recordings are analyzed through source localization 1110 Journal of Cognitive Neuroscience Volume 13, Number 8

algorithms. To the best of our knowledge, and surpris- ingly, ERP data on the interaction between emotion and visual attention have not been provided to date. There- fore, the main goal of the present study was to remedy this situation. A wide conception of this interaction should take into account several forms of attention. In this regard, the first specific objective was to explore the two types of attention previously mentioned: expect- ancy-related and input processing-related attention. The second objective was to study the temporal character- istics of both types of attention, and particularly those associated with the positive or negative valence of the emotional stimulation. Finally, the third objective was to explore the possibility that subprocesses within each type of attention present, to some extent, different brain sources. Expectance is usually elicited in ERP research through the experimental paradigm employed to evoke the contingent negative variation (CNV). In its traditional form, this paradigm (that, due to its timing character- istics, cannot be employed in hemodynamic studies) requires the presentation, in each trial, of a warning signal (S1) and subsequently of an ���imperative��� stimulus (S2) (Walter, Cooper, Aldridge, McCallum, & Winter, 1964). The onset of S2 requires a rapid response from subjects (e.g., to press a button as rapidly as possible in order to terminate S2). During the interval between S1 and S2, in which the subject is alert, the CNV appears, terminating just after S2 onset. CNV amplitude relates directly to expectancy-related attention to S2 (McCal- lum, 1988). If S2 consists of a visual image, and if its visual exploration is required by the task, ERP elicited by this second stimulus is also able to reflect input process- ing-related attention. Four emotional categories were studied in the present experiment, in which an S1 (hereafter labeled ���cue���)- S2 (���target���) type task was employed: arousing-positive (A+), arousing-negative (A ), neutral (N) and relaxing (R). These categories are intended to cover the main combinations of positivity, negativity, and arousal in order to study their influence on attention (as explained elsewhere [Lang et al., 1997], certain combinations, such as R , are virtually impossible). Cue stimuli were simple white-on-black schematic line drawings or symbols. As described in detail in the Methods section, each symbol was identified in the instructions with a particular object or motif (e.g., insect, building, etc.). Cue presentation consisted of two drawings or symbols belonging to the same emotional category, one above the other (see Figure 12). The ideographic-type structure of cues en- sured that the stimuli were physically homogeneous, with no significant differences in their complexity, color, brightness, contrast, or size. Also, this type of visual cue was preferred to other types, such as words, because they provided the pictorial nature of the target stimuli (consisting of photographs, as explained below), thus facilitating the involvement of similar processing systems in response to both cue and target stimuli. Moreover, simple ideographic stimuli would be less prone to having deep intrinsic emotional significance, at least by comparison with photographs or realistic drawings. A color photograph (target), was presented 1750 msec later (see Figure 11). Photographs always represented one of the motifs symbolized by the two drawings of the cue presentation. Participants were told to identify verbally which of the two drawings of the cue corre- sponded to the target (verbal response was given 2 sec after target onset). The instructions did not explicitly establish that the experiment dealt with emotional reactions, and, as explained, the cue indicated only in an implicit way the emotional category of the following target. This strategy had two advantages. First, it helped to avoid a situation whereby participants considered that some of the stimuli were more important for the task than others (e.g., emotional stimuli more important than neutral ones), and thus to avoid a relevance-for- task effect, often described in previous studies (e.g., Duncan-Johnson & Donchin, 1977: the stimuli on which the task focuses tend to elicit the highest amplitudes in certain endogenous ERP components). Second, homog- enization of cognitive demands makes it easier to reach more solid conclusions about the emotion-related cau- sation of possible attentional differences. In general, tasks in which target information is pre- dicted (100% correctly announced) by a cue have a characteristic that is very interesting and relevant to the present study: They allow for the random presenta- tion of different stimuli minimizing the effect of the orienting response, at least in its relation to novelty or nonexpectance of stimulation. Since only active, selec- tive attention was of interest in the present study, the orienting response could represent an important inter- ference for posttarget ERPs. Moreover, in natural envi- ronments, situations in which subjects perceive some cues or signals that trigger affective-related expectance and prepare them for possible forthcoming emotional events are relatively frequent. These situations have as yet scarcely been investigated in relation to brain activity. The study of input processing-related attention required that only attended stimuli (cues and targets) were analyzed. To ensure this, recordings obtained in trials in which subjects responded erroneously were elimina- ted. It was also necessary to avoid emotional categories being heterogeneous with respect to the level of atten- tion needed to carry out the task correctly: differences should be due to the affective content of each category, but not to its difficulty with respect to the task. Control analyses were planned in order to discard, if necessary, those categories differing in this aspect from the rest. Our first hypothesis is that expectancy-related and input processing-related phases of the task, which, as explained, associate with different types of attention, will be characterized by the activation of different brain areas. The former should associate with the activation Carretie �� et al. 1111