You are on page 1of 64

Emotional modulation



Emotional Modulation of Cognitive Control: Approach-Withdrawal States DoubleDissociate Spatial From Verbal 2-back Task Performance

Jeremy R. Gray Harvard University

in press, Journal of Experimental Psychology: General © APA, please do not reproduce without APA permission

Regular mail: Campus Box 1125 Washington University St Louis, MO 63130 Phone: Fax: Email: 314-935-8547 314-935-4711

Overnight mail: Psychology Building, Rm. 225A Washington University St Louis, MO 63130

Emotional modulation Abstract Emotional states might selectively modulate components of cognitive control. To test this hypothesis, 152 undergraduates (equal numbers male and female) were randomly assigned to watch short videos intended to induce emotional states (approach, neutral,


withdrawal). Each video was followed by a computerized 2-back working memory task (spatial or verbal, equated for difficulty and appearance). Spatial 2-back performance was enhanced by a withdrawal state and impaired by an approach state; the opposite pattern held for verbal performance. The double dissociation held more strongly for participants who made more errors than average across conditions. The results suggest that approach-withdrawal states can have selective influences on components of cognitive control, possibly on a hemispheric basis. They support and extend several frameworks for conceptualizing emotion-cognition interactions.

Emotional modulation Emotional Modulation of Cognitive Control: Approach-Withdrawal States DoubleDissociate Spatial From Verbal 2-back Task Performance


Does being "scared speechless" reveal anything about the influence of emotional states on cognition? Are there related effects on spatial processing--"giddy with joy", perhaps? Although these phenomena are quite rare, more important if less dramatic effects have been well documented. For example, actively constructing a social judgment during an emotional state increases the likelihood that the judgment will be biased, whereas judgments retrieved from memory during an emotional state are relatively immune to such distortion (Forgas, 1995). Indeed, a number of affective factors including emotional states, arousal, mood, stress, trait emotion, and emotional pathology can influence human performance on diverse cognitive tasks (e.g., Brown, Scott, Bench, & Dolan, 1994; Dalgleish & Power, 2000; Forgas, 1995; Gray, 1999; Heller, 1990; Heller & Nitschke, 1998; Humphreys & Revelle, 1984; Isen, 1993; Koelega, 1992; Oaksford, Morris, Grainger, & Williams, 1996; Revelle, 1993). How are we to broadly integrate these results and many others? One possibility derives from the meta-theoretical position that cognition and emotion both function as control systems that regulate cognition and behavior (e.g., as articulated by Carver & Scheier, 1982; 1990; Braver & Cohen, in press; Kosslyn & Koenig, 1992; Simon, 1967). Subsystems that contribute to control and are involved in both emotion and cognition might be critical loci of interaction. The term cognitive control (e.g., Braver & Cohen, in press; Posner & Snyder, 1975) refers inclusively to processes that guide or coordinate flexible information processing, especially in situations that are novel or complex. Cognitive control is not only self-regulatory but often effortful and typically intentional. Components of cognitive control include various forms of working memory, inhibition, attention, and self-monitoring. Cognitive control is not restricted to executive control

Many of these functions depend in part on prefrontal cortex (Fuster. It is clear that some important effects of emotion are mediated by specific and relatively automatic control systems (see e. 1999). Ellis & Ashbrook. 1999). and includes active maintenance. creative problem solving) in terms of slight increases in cortical dopamine (Ashby.g. Tomarken & Keener. 1967).. a demonstration of . 1988.. Emotional states could have non-selective effects on cognitive control.Emotional modulation 4 processes.g. & Turken. Grafman. the model focuses only on positive affect. 1997. 1987. 1999). Emotional states might also differentially influence components of cognitive control. Depue & Collins. Many results broadly consistent with this hypothesis have been obtained in a number of social-cognitive domains (see e. These are all important hypotheses. 1990. Forgas.g. but do not speak to the question of selective effects. as described below.. Holyoak. even opposite effects.g. Goldman-Rakic. 1998). LeDoux.. e. 1988). Smith & Jonides. 1995. Two neuropsychological accounts predict such selective effects (Heller. Darke.. While specifying a specific neurobiological mechanism (cf. A neurocomputational model accounts for the effects of positive affect upon complex cognition (e. thereby influencing all forms of controlled processing in the same manner. Isen.g. & Boller. 1996). having stronger effects on some controlled processes than others. However. Another kind of non-selective effect is motivational: A given emotion may simply increase or decrease the willingness to engage in any effortful task. Emotion might also modulate cognitive control in order to better coordinate overall selfregulation. Emotion could serve as an interrupt signal that redirects cognition to operate upon a new agenda signaled by the shift in emotional-motivational state (Simon. 1995). Attentional or cognitive resource models predict that emotional states of whatever type take up resources that are then no longer available for controlled cognition (e.

Hemispheric . 1997) and facial expression (Ekman. for spatial versus non-spatial information (Baddeley & Hitch. 1998).. attention to novel versus familiar information.. Freisen. initiation. Tomarken & Keener. Selectivity could result from different pools of cognitive resources. Cacioppo. and inhibitory control of cognition and action. Heller & Nitschke. however (e. Klein. for example. 1998). 1995. Tomarken & Keener. & Hatfield. sequencing. it is surprisingly elusive using autonomic measures. and fine versus gross motor control. 1991. Gray. The aim of this article is to test a key prediction of the idea that emotional states can modulate components of cognitive control selectively and adaptively. 1980). e. Selectivity is not a trivial finding. Kosslyn & Koenig. Thus. 1998. 1998). such that some are enhanced and others impaired according to the demands of the emotional state (cf. Berntson. the active maintenance and prioritization of approach versus withdrawal goals (cf. 1995. Tomarken & Keener. The 5 non-selective hypotheses above could be easily extended to incorporate selective effects. Tomarken & Keener. 1974) or for each cerebral hemisphere (Boles & Law. 1998). & Ancoli. Functions modulated by emotional states might include. it is possible without being obviously the case that cognitive control functions might be selectively modulated by emotional states (Heller & Nitschke. 2000b.Emotional modulation selective effects of emotional states on components of cognitive control is lacking. 1998.g. Davidson.g. Heller & Nitschke. Fox. Modulation on a hemispheric basis would be one way to achieve selectivity (Heller & Nitschke. but non-essential or even deleterious during reward seeking (approach-related states). 1993). Emotional states have selective effects on neural activity (on a hemispheric basis. For example. 1998). behavioral inhibition might be more important during conditions of potential threat (which produce withdrawal-related emotional states). 1998.

I present the current account in slightly more detail. the most critical of which was tested. Many of the data come from tasks that are good for localizing . In the 6 rest of the article. 1990. Heller & Nitschke. 1994. followed by three experiments that tested for and support selective influences. subjectively pleasant (positively valenced) emotion tends to facilitate performance on tasks that depend more on the left prefrontal cortex (PFC). 1997. 1998). before it is worth trying to argue for a detailed. Toward a Proto-Theory The point of this section is to raise a logical possibility: that approach and withdrawal emotional states selectively modulate cognitive control functions to coordinate and prioritize among high-level (cognitive) self-regulatory functions. and it has not been demonstrated in previous work. psychologically adaptive account of an emotional modulation of cognitive control. Selectivity is a critical first prediction of this account. Goldberg. 1997. and serves well to organize the extant literature (Heller & Nitschke. 1998). A number of effects of emotion on cognition can be categorized in part given some knowledge of the mediating neural systems on a regional basis: as anterior or posterior within the left or right hemisphere (Heller. Although the account as presented here is speculative and clearly raises far more questions than it answers. & Lovell. 1993. This framework is more nuanced than presented here. Hellige. direct evidence that emotional states are able to have selective effects is needed. However. 1998). Empirically. Smith & Jonides. it leads to a more definite specification of the problem. whereas subjectively unpleasant (negatively valenced) emotion tends to facilitate performance on tasks dependent on right PFC (Heller & Nitschke. and makes numerous predictions. 1999). provides a framework for theoretical investigation. Podell.Emotional modulation differences have been suggested for a number of these functions (see Banich.

These general functions include maintaining the continuity of motivation (including prospective and retrospective memory). such as between-task comparisons.or withdrawal-related emotions bias the ability of prefrontal cortex to organize behavior over time (Fuster. in press). 1990.g. which is more detailed than presented here. Schwarz. consider the example of goal-regulation (see also Gray. An information processing system able to benefit from having goals would need not only a way to maintain them actively. Another perspective (Tomarken & Keener. 1967). 1995. in press). To illustrate the potential of this quite general conception of cognition-emotion relationships.g. 1994. see Ekman & Davidson.. These considerations . Lang. Simon. Tomarken and Keener's hypothesis is exciting. Braver & Cohen. The general idea that emotion helps mediate priorities is widely held (e. 1998) is that approach. 1991. However. and their review of evidence suggesting that depression can be understood as a failure of suppression of interference is supportive. 1990. as the functions of prefrontal cortex). and the shifting of strategy. many of these tasks are not well- 7 characterized in terms of basic information processing (Heller & Nitschke. Tomarken and Keener's hypothesis. is relatively specific about the kinds of emotion (approach.. Gray. 1999).Emotional modulation brain damage on a regional basis. the suppression of interference. Lazarus. which reduces the confidence one can place in further interpretation of the data. but also a way to modulate their influence when situationally inappropriate. 1997). Because they are potentially powerful. Goals can be considered representations that help control behavior and bias how other information is processed (e. withdrawal) and kinds of psychological functions to be prioritized (those proposed by Fuster. they must also be regulated when circumstances change and they would do more harm than good: What is beneficial in one circumstance can have dire consequences in another (Gray. 1997. 1997).

Emotional modulation suggest that a goal management system needs to incorporate mechanisms for both active maintenance and selective regulation. Working memory (Baddeley & Hitch,


1974) is well suited to provide active maintenance of goals or other information that act as top-down constraints on cognition and behavior (Braver & Cohen, in press). Emotional states are well suited for regulation, as dampening or enhancing goals in a way that is consistent with on-going circumstances as they interact with motivation. The active maintenance of goals could be implemented by having hemispheric specialization for approach goals (left) and withdrawal goals (right) (Tomarken & Keener, 1998) to make regulation by emotional states as simple as possible (e.g., via neuromodulators, see General Discussion). These basic considerations hint at the possibility of complementary roles of working memory and emotion in managing goals effectively: Working memory could maintain active goals, and emotional states could regulate active goals based on circumstances, selectively prioritizing approach or withdrawal goals. Other examples could be given of functions that might benefit from selective modulation on a hemispheric basis, but far more conceptual background than can be presented here would be needed to progress from examples to theory (including computational and neurobiological considerations, see Ashby et al., 1999; Gray, in press). Nonetheless, it is a logical possibility that selective effects of emotion on cognitive control could play a functional role in high-level self-regulation. For further progress, a first need is empirical: evidence for (or against) selective effects of emotional states on components of cognitive control. The prediction of the current account is that approach and withdrawal emotional states can enhance or impair performance depending on the particular emotion and cognitive control function involved, and that different emotional states can have opposite effects.

Emotional modulation Evidence for such a double dissociation would argue strongly for a selective effect.


Although not a strict requirement, differences should be shown ideally on tasks that are as simple as possible and potentially related to hemispheric differences. Neuropsychological tasks, reasoning, and creative problem solving tasks tend to be too complex for drawing inferences about elementary functions because such tasks permit numerous strategies, such as recoding of verbal material into spatial terms, and potentially draw on a great many abilities (e.g., imagery, memory, and so on). A second need is theoretical: a conceptual basis for selective effects that goes beyond mere logical possibility of the type discussed above. This is by far the more daunting task, in part because there is little empirical constraint. Extant evidence tends to be neurobiological, e.g., related to hemispheric differences (Gray, 2000b; Heller, 1990; Heller & Nitschke, 1998; Tomarken & Keener, 1998) or dopamine systems (Ashby et al., 1999), and so only indirectly informs hypotheses about psychological function. Moreover, there are a number of complex issues involved in understanding hemispheric differences in emotion and cognition (Davidson, 1992; Hellige, 1993) and of relating classes of emotion to neurotransmitter systems (Panksepp, 1998). Notwithstanding the difficulties, the current conception makes a clear-cut prediction that can be rigorously tested: Selective effects of emotional states on cognitive control should be possible. If substantiated, this would make theory development more attractive as an enterprise. Rationale of the General Method If induced emotional states can reveal a double dissociation between two cognitive control functions, it would be strong evidence for selective effects of emotion on cognitive control. Such a dissociation would be still more meaningful if the tasks and types of emotional states are themselves relatively well-understood. I therefore used

Emotional modulation spatial and verbal working memory tasks, and approach-withdrawal emotion. These


tasks and emotions would hint at a hemispheric basis for the dissociation, although this was not tested. Spatial-Verbal Working Memory Working memory refers to the system for the short-term, active maintenance and manipulation of information in the range of seconds (Baddeley & Hitch, 1974). Working memory is a natural choice for a cognitive control function here because of its importance in cognition and the diversity of efforts to specify its component processes (e.g., Baddeley, 1996; Owen, Evans, & Petrides, 1996b; Smith & Jonides, 1999; Smith, Jonides, Marshuetz, & Koeppe, 1998; Waltz et al., 1999). Most relevant for present purposes, it has spatial and verbal subsystems (Baddeley & Hitch, 1974), as shown by behavioral challenge and brain damage (see Baddeley, 1986), functional neuroimaging (see D'Esposito et al., 1998; Smith & Jonides, 1999), and individual differences (Shah & Miyake, 1996). Moreover, considerable functional neuroimaging evidence suggests that some components of working memory are specialized by hemisphere, with verbal active maintenance left lateralized and spatial active maintenance right lateralized in inferior frontal gyrus (D'Esposito et al., 1998; Smith & Jonides, 1999). Distinguishing among working memory, attention, and other aspects of cognitive control is challenging conceptually and empirically (see Braver & Cohen, in press; Miyake & Shah, 1999). Baddeley (1986) drew on a model of attention (Norman & Shallice, 1986) to describe the central executive component of working memory. Moreover, various theorists use the same term with different emphases. Fortunately, the difficulty in demarcating various aspects of cognitive control is not critical for present purposes. The spatial-verbal distinction is conceptually clear and one of the most robust empirical results in the human working memory literature. Moreover,

For example. are not withdrawal related.g. Perlman Lorch. The approach-withdrawal distinction draws empirical support from: clinical populations with emotional pathology (Heller & Nitschke. EEG in neonates (Fox & Davidson. Nitschke. Fox. and so their theoretical roles are less clear than that of states defined in terms of motivation (approach. 1996. The asymmetry concerns subjectively experienced emotion. 1998).. The approach-withdrawal distinction is concerned with goal-directed emotions. 1998. 1995. 2000b. Adolphs. and not post-goal attainment (or non-attainment) emotions (Davidson. The emphases on both function and hemispheric specialization are not essential. 1997). 1986). whereas withdrawal states lead to greater right hemisphere activation (see Davidson. Heller. Borod. 1998). 1982). Sutton & Davidson. Damasio. Gray.. such as satiation.Emotional modulation spatial and verbal tasks can be designed that are well-matched psychometrically. . such as disappointment. and not the perception of emotion in others not the production of emotion (which are typically right lateralized. positive or negative) are not directly functional. 10-month old infants (Davidson & Fox. & Miller. Classes of emotion defined in terms of subjective valence (e. & Nicholas. are not approach related. and some that are subjectively negative. 1995. but could be useful in developing a theoretical account. Considerable evidence suggests hemispheric specialization for anterior areas. 1986. Heller et al. Koff. withdrawal). Approach states lead to greater left hemisphere activation. Tranel. Approach-Withdrawal Emotional States Approach versus withdrawal emotion (Davidson. & Damasio. 1998). some emotions that are subjectively pleasant or positive. 1991) is one of the conceptually clearest and best-validated distinctions between functional classes of experienced emotion. 11 Identifying which more specific aspects of cognitive control are influenced by emotion will clearly be important in further research.

BIS) and trait sensitivity to cues of reward (individual differences in a behavioral activation system. 1982) model of personality. whereas others react little if at all. However. 1997). the asymmetry is found when individual differences in arousal are controlled (Canli. BAS). Kalin. Any given emotion induction is not likely to induce purely approach or withdrawal emotion (if such states exist). Gray's (1970. Desmond. & Gabrieli. Some people react strongly to a given motivating stimulus (reward. 1994) that assess trait sensitivity to cues of threat (i.. To foreshadow the method. the BIS-BAS scales can be used to predict which individuals should experience stronger emotional states and therefore show larger effects of those states on cognitive control. Kalin. Although functional neuroimaging results have been mixed (Canli. Individual differences in trait emotion suggest a methodological advantage to focusing on approach-withdrawal emotion.e. then trait differences .Emotional modulation normal adults. threat). almost certainly not across all participants. A. Sutton & Davidson. Their trait emotional reactivity to cues of reward or threat should selectively amplify the degree to which a given emotion induction induces an approach. Glover. Zhao. for review and evidence). 1995). 1999). BIS and BAS are orthogonal dimensions. Asymmetry in anterior brain activity correlates significantly with the BIS-BAS measures (Harmon-Jones & Allen. If an effect of emotion on cognition depends specifically on approach-withdrawal emotion. Shelton. based on J. EEG in rhesus 12 monkeys (Davidson. Personality measures have been developed (Carver & White. individuals who are more reactive should be more affected. one is not predictive of the other. individual differences in a hypothetical behavioral inhibition system. Larson. People differ in how sensitive they are to cues of reward (approach motivation) and cues of threat (withdrawal motivation) (see Larsen & Ketalaar. & Shelton. 1991. 1998). & Davidson. 1998). 1992.or withdrawal-related state. and adult clinical populations (see Davidson. 1997.

2000a. Heller & Nitschke. I sought to dissociate spatial from verbal working memory. In sum. 11 experiments suggested a dissociation between spatial and verbal working memory performance during emotional states consistent with Heller's regional framework (e. mechanistic inquiry into how normal emotional states influence cognitive control. Moreover. using induced emotion to dissociate spatial from verbal 2-back performance is a test case of a potentially quite general. no studies specifically contrasted spatial with verbal working memory on tasks matched for difficulty. mediated by the strength of induced emotional state. and so motivation or task difficulty may have confounded the results. operationalized in terms of performance on 2-back tasks. A double dissociation would be strong evidence that emotion can have selective influences. Thus in three behavioral experiments. Previous work Previous work hints at a dissociation between spatial and verbal working memory during approach-withdrawal emotional states. I did not test which components of cognitive control necessary for 2-back performance were more influenced. In a recent review (Gray. However. General Method .g.Emotional modulation 13 in approach-withdrawal emotion should predict the magnitude of the effect. 2000a). but focused empirical tests are lacking. Two experiments were not consistent.. 1998). Thus the extant literature (see Gray. 1997. but has not tested for it directly and is methodologically compromised for answering the question. nor test whether hemispheric specialization is the basis for the interaction. 1998) hints at selective effects of emotion on spatial and verbal working memory. Heller & Nitschke. no studies focused on approach-withdrawal emotion specifically.

whereas individuals more sensitive to approach would show larger effects of the approach videos but not the withdrawal videos. None participated more than once. Equal numbers men and women participated. a variant of the Continuous Performance Task (Rosvold. neutral. 2-back tasks. and Scream (Craven. I unobtrusively noted behavioral reactions (e. mean 19. an Australian travelogue G'Day Australia (Queensland Tourist and Travel Corporation. and withdrawal video. overt startle. Sarason. & Hill. and all were randomly assigned to groups within the constraints of counterbalancing. 1987). Two sets of three 9-10 minute video clips were used to induce emotional states. 1985). Set 2 had excerpts from Best of America's Funniest Home Videos (Vin Di Bona Productions. 1990). Bransome. They gave informed consent and were 14 paid for their participation. respectively. 1978 / 1989). To verify the emotion induction. 1996). For the cognitive control task.4. Procedure The key aspect of the protocol was for each participant to watch a short video intended to induce an emotional state.. I asked most of the participants to give selfreport emotion ratings immediately before and after each video (n = 140). Roger and Me (Moore. Mirsky.Emotional modulation Participants in all experiments (total n = 152) were right-handed Harvard undergraduates. Set 1 consisted of excerpts from Candid Camera (Funt. in Experiment 3. 1956) was selected on . laughing) for n = 128. and then perform 100 trials of a computerized 2back task. Emotion induction. respectively. Finally. Each set had an approach. I explicitly tested the prediction that individuals more sensitive to cues of threat would show larger effects of the withdrawal videos but not approach videos. & Beck.g. 1989). Yablans. age range 18-27. and Halloween (Carpenter.

471): encoding.. temporal ordering (to identify which item was two previous).. recoding of spatial information into verbal terms (or vice versa) is generally more difficult than doing the task as intended. Pilot work established display parameters at which spatial and verbal versions were well matched for difficulty at a group level for errors and RT. following each video). letters appearing in locations) and are instructed to indicate. the spatial and non-spatial versions can have exactly the same visual appearance and response demands. whether the current item matches the one seen two trials previously. press the D key). Jonides et al.g. Third. guessing). RT on target trials in which a correct response is made contribute to the measure of average RT for that session. inhibition (of currently irrelevant items). the tasks strongly constrain how people typically engage them. Reaction time (RT) and accuracy are recorded for each trial. In the 2-back task (Braver et al. First. Each participant is instructed to press one of two computer keys on each trial to indicate whether the item on the current trial matches the item shown two trials ago (same. storage. not the stimuli or mode of responding. they are relatively well understood. Given the time constraints of 3 s per trial. press the S key) or not (different. and response execution. as assessing processing speed when RT is the least likely to be contaminated by spurious influences (e. for 15 each item.g. (1997) describe the 2-back task as requiring seven elementary operations (p. and there are 100 trials per session (i. While recoding is still possible.Emotional modulation methodological grounds. participants view a series of items (e.. . Second.e. The 2back tasks have several strengths for present purposes. rehearsal.. it would work against showing a differential influence of emotion on task performance. the instructions vary. 1997). matching (of current to stored). Based on a conceptual analysis with converging evidence from patterns of brain activation. Participants in the experiments reported here also found the tasks to be of equal difficulty.

major axis = 7. 1993). j. Flatt. The same sequence of letters-in-locations was used for the spatial and verbal versions.8 cm between square centers on the vertical. ignoring the identity of the letter. with no more than two targets in a row. 30% of trials are targets (S) and 70% are nontargets (D). To ensure a uniform pace of the task. Instruction and 12 trials (repeated if necessary) were given prior to any video to eliminate confusion the participants may have had about the task. The square can appear in one of six possible locations around the center of the screen (roughly following an iso-acuity ellipse: minor axis = 5. The letters appear in black 16 against a white background. The spatial instructions are to press S or D depending only on the location of the box on the screen. The verbal instructions are to press S or D depending on the identity of the letter inside the box. m) presented one per trial using PsyScope (Cohen. l. but without becoming practiced at the task. k. Each participant was given explicit instruction about the task.2 cm horizontal). d. Each square with letters is shown for 500 ms against a background of jumbled letters. appears as a nontarget as close to equally often as possible. & Provost. The familiarization session was intended to enable participants to proceed as quickly as possible from the end of the video to engaging in the task. both on the computer screen and orally by the experimenter during a brief familiarization session. ignoring the location. the task progresses with no delay to the next trial. g. MacWhinney. Each item (letter or position) occurs as a target equally often. . followed by a 2500 ms period of the background alone. An omission is recorded and scored as an error.4 cm square. and is otherwise presented at random. f. if the participant does not respond within the 500 ms + 2500 ms time period.Emotional modulation The task has 10 letters (b. c. Multiple copies of the letter for a given trial are shown in 48 point Geneva font inside a 5. with the 17" monitor resolution set at 832 x 624 pixels. h.

and so the same task can impose different demands on executive function. That is.05.Emotional modulation Analyses Individual differences in working memory capacity are thought to reflect 17 differences in controlled processing (see Engle. the degree of controlled processing required will depend in part on the individual. because they often no longer tax executive function (see Barkley. it should be greatest in magnitude for those performing below average. it should be stronger in individuals finding the 2-back tasks more challenging. In the work reported here. That is. . For all statistical tests. First. a given task will impose different cognitive demands at different ages. 1994). alpha was set to . Participants were divided into two groups based on performance: above average (low-error group) and below average (high-error group).. Although the tasks are same across participants. & Tuholski. when adults are acquiring cognitive skills. individuals can differ in executive abilities (here as a function of age). 1990). 1997. more effort and controlled attention are required initially than subsequently during skilled or practiced performance (Raichle et al. For both age and practice. substantial individual variation in 2-back performance was found. and so the same task can impose a different demand. Second. 1999). Tasks appropriate for studying executive function at one age are often inappropriate at an older age. Kane. p values are 2-tailed. Diamond. not the task alone. In pilot work. analyses were performed to both control for and capitalize on these individual differences. individuals can differ in ability (here as a function of practice). in child development. the cognitive control required depends in part on the individual and situation. If the predicted Emotion x Task interaction is specific to cognitive control. Two examples may help make this part of the design more clear. Thus if the effect of emotion is specific to controlled processing.

anxious) described the current emotional state. spatial 2-back performance should be better in a withdrawal state than an approach state. For those giving ratings. 2-back task (spatial. Self-focused attention to an emotional state while giving the ratings could also dampen the emotion. Only the first set of videos was used. whereas the opposite should hold for verbal 2-back performance. 1998). taken as indices of approach emotion. Method The participants (n = 24). The target terms were amused. energetic. neutral-approach- . video order (withdrawal-neutral-approach. calm. 2-back tasks. sad. As suggested by reviews of the extant literature (Gray. The instructions were to indicate how strongly each of eight emotion terms (bored. verbal). high arousal. angry. so that the likely transient effects of emotion would be stronger. withdrawal emotion. self-report ratings were collected before and after each video to assess emotional state. The factors counterbalanced between-subjects were: sex. anxious. This was indicated by making a mark on a 10 cm horizontal line provided for each term to indicate the current level.Emotional modulation Experiment 1 Experiment 1 tested the key prediction that spatial and verbal controlled processing would be influenced selectively by approach and withdrawal emotional 18 states. low arousal. happy. respectively. 1974). Omitting the ratings for the participants was intended to make the time between the end of the video and the beginning of the 2-back task as short as possible. and videos are as described in the general method. which was subsequently scored in mm from 0 (not at all) to 100 (extremely) (Bond & Lader. amused. Heller & Nitschke. For half of the participants. each video was preceded and followed by the emotion self-report ratings. 2000a. and low interest. and bored. energetic. calm. Design.

for the high-error group it was significantly so. The contrast supported the prediction in the high-error group for errors. (To focus on the key interaction.044. It was also significant within the higherror group alone. low error) were defined based on a mean split of errors averaged across all three 2-back sessions. p = . and did so three times following each of the three videos. the effect of spatial-verbal task on the withdrawalapproach difference scores is equivalent to an Emotion x Task interaction. Results For all participants the crossover interaction was in the predicted direction.042. approach-withdrawal-neutral). It was not present in the low-error group. albeit not significantly. t(14) = +2. p = . 7) = 7.69.Emotional modulation withdrawal. Each participant performed only the spatial version or only the verbal version.87.03. shown in Figure 1. and the results are consistent with this prediction. t(26) = 0. consistent with the contrast analyses. The withdrawal minus approach differences in RT and errors were the dependent measures. Across all participants.21. was tested using Contrast 1 (Table 1). In a test of the difference . 19 Two groups (high error. 2-back performance after the neutral video was not included here.033. it is not necessary for showing the dissociation. p = . F(1. p = . this Emotion x Task effect was upheld significantly for errors. p = . using difference scores to collapse the repeated measures. The data can also be examined in terms of a relative effect. but not within the low-error group alone. which are summarized in Table 1.86. The Emotion x Task interaction should emerge more strongly in the high-error group if it is specific to cognitive control.) In an ANOVA (spatial-verbal and high-low error groups between subjects).17. F(1. 13) = 0. and ratings/no ratings. with no speed-accuracy trade-off. 20) = 4. with no speed-accuracy tradeoff. F(1.03. The reliability of the pattern across the condition means.

5.e. The spatial and verbal tasks were well matched. the High-Low Error x ApproachWithdrawal x Spatial-Verbal interaction was marginal.80. videos within subjects). Average errors and RT over all three video conditions were taken as the summary measures of difficulty. this interaction was in the same direction. energetic.50. There were no reliable differences between the tasks for errors or RT. F(1. in the direction expected (a larger Emotion x Task interaction within the high-error group than the low-error group). ps > . p = .. p = . 20) = 3.18. The videos induced the intended emotional states. the effect of ratings/no-ratings was not reliable. and ratings/no-ratings between subjects. In a mixed model ANOVA (spatial-verbal. Anxiety (withdrawal) ratings were higher after the withdrawal video but not after the approach video. 11) = .Emotional modulation between the high-error and low-error groups. suggesting equal arousal despite opposite valence. Of note. The Emotion x Task interaction of the participants giving ratings did not differ from those who did give ratings. 11) = 0. higherror/low-error.94. calm. there were no differences in either calm or energetic (low or high arousal. respectively) between the two emotional videos. Control Analyses 20 Task difficulty. . based on self-reported changes in emotion.45.065. suggesting heightened interest in the emotion conditions. approach-withdrawal within subjects). Emotion manipulation. The scores for bored (low interest) were significantly and similarly decreased (i.01. F(1. 16) = 1. Other. Potential spatial-verbal differences were tested by mixed model ANOVAs (spatial-verbal task between subjects. F(1. Amused (approach) ratings were higher after the approach video but not the withdrawal video. F(1. summarized in Table 2. all ps > . less bored) for both emotional videos but unchanged for the neutral video. for RT.

the plausibility of a number of alternative accounts is greatly reduced. whereas verbal performance was enhanced by an approach state and impaired by a withdrawal state. The double dissociation also means that an arousing but emotionally neutral condition was not required in the design. neutral performance was typically intermediate. given the self-report ratings showing similar levels of arousal for the oppositely valenced emotion inductions. Motivational differences between the emotion conditions are not plausible on this basis. The neutral condition is not required to show the dissociation.55). the crossover interaction held in absolute terms. Discussion Spatial performance was enhanced by a withdrawal state and impaired by an approach state. Thus the experiment provides evidence that components of cognitive control can be dissociated by induced emotional states. Because the Emotion x Task interaction was a double dissociation. This means that the double dissociation--itself a strong result--is even more straightforward to interpret. identical in appearance). the effect size of the Emotion x Task interaction was half the size in the group giving ratings (r = .27) than not giving ratings (r = .Emotional modulation Nonetheless. In particular.1 A second aim was to establish that the key effect depends on approach . yet it is additionally informative. The effect is not attributable to differential arousal. For 21 participants making more errors than average. The double dissociation held in a direct comparison between well-matched tasks (equated for difficulty. there was no main effect of the neutral versus emotion conditions. Experiment 2 A first aim of Experiment 2 was to replicate the findings in Experiment 1 in a large sample.

video order (emotional-neutral-neutral-emotional. 1977). Eysenck. withdrawal). The between-subjects factors were sex. verbal) crossed by video type (emotional. These data are presented as Experiment 3 for greater continuity. The within-subjects factors were task (spatial. followed by a 2-back task. 22 Each participant filled out a consent form and the State-Trait Anxiety Inventory (STAI) (Spielberger. & Barrett. All participants saw the two neutral videos. After the last 2-back task and a final emotion ratings. The measures of self-reported emotional state were obtained in part to validate the emotion induction and in part to pilot-test items for a rating scale sensitive to approach-withdrawal emotional states. for the other half they were the approach videos. to allow within-subjects comparisons of the influence of an emotion condition against a neutral condition for both spatial and verbal tasks. with 11 terms per scale. form Y. Method The participants (n = 128) and tasks are described in the General Method. Only the data for the terms amused (approach). anxious (withdrawal). 1994) and the EPQ-R (Eysenck. 1985). participants completed the BIS-BAS scale (Carver & White. These data were collected using a procedure very similar to that described in Experiment 1. two emotional and two neutral. valence of the emotional video (approach. Each participant saw four videos. For half of the participants.Emotional modulation and withdrawal emotion rather than some other aspect of the videos. short form. task order (spatial-spatial-verbal-verbal. and content are reported here. neutral-emotionalemotional-neutral). the 2 emotional videos were the withdrawal videos. Each participant then had the familiarization trials. Each video was preceded and followed by a brief emotion rating scale. and four repetitions of video-plus-task. They were debriefed and paid for their participation. neutral). verbal-verbal-spatial- .

. the emotion inductions were weaker and the tasks were more practiced (ps < . The reason is that the data were analyzed by focused contrast (for greater specificity) rather than factorial ANOVA (used in Experiment 3 and control analyses). set 2-set 2-set 1-set 1). For RT. the first 32 participants were tested before the next 32 were begun).e. i. For errors.) Note that Contrast 1 does not recapitulate the experimental design (factorial. (There was a tradeoff in the low-error group.32) and less strongly within the low-error group (r = +. A mean split on % errors across all 2-back sessions was used to define two groups based on performance (high-error.e..e. Interpretation of the data from all participants. effects due to serial position within the four repeated measures) were present and would work against showing an Emotion x Task interaction. repeated measures) in two ways.15). The reason is that two strong order effects (i. First. the High group alone.24). the two neutral groups were combined to give a more stable estimate of performance in the neutral condition. with 4 participants per cell. only the data from the first 2-back session were used in this analysis. Thus there were 32 cells. And second. the same trend was present. Results The data are shown in Figure 2 and summarized in Table 3. and video version order (set 1-set 1-set 2-set 2. see Order effects under Control analyses). This pattern held more strongly within the high-error group (r = +. The participants in the two Neutral groups . which compromises interpreting their performance in isolation.. as tested by Contrast 1 (Table 1).0001. and High versus Low groups are not compromised. there was no speed-accuracy tradeoff across all participants or within the high-error group.Emotional modulation 23 spatial). low-error). Random assignment was done within blocks of 32 (i. the crossover interaction was reliably present across all participants (effect size r = +. In later sessions.

as shown by a focused contrast on the data shown in Figure 2 A. with the data fourth-root transformed to acceptably equate the variance between groups. which is not problematic. The neutral groups are not required for showing a double dissociation. age.59. p =.53.. The spatial-neutral group made fewer errors (8. verbal = +1.4%).17. and the contrasts were computed accordingly (Rosenthal & Rosnow.26). as tested meta-analytically. F(2.89.41. emotional arousal might somehow influence spatial and verbal tasks differentially. p = . the effect size for the high-error group (r = +. For errors.0096. but depends as well on both the participants and the conditions under which the task is performed (e. 116) = 4. there was no difference between the tasks for errors. Using lambdas for spatial = -1. t(122) = 1. One might think that performance in the neutral condition is somehow a better measure of task difficulty.) Control Analyses Task difficulty. or for RT. In fact. and so the difference is of no real concern.Emotional modulation should not differ during the first 2-back session because their protocols diverged only upon presentation of the second (non-neutral) video.0091. for RT the same trend held albeit not significantly. For errors. 1985). the 3 (Emotion) x 2 (Task) x 2 (Error group) interaction was significant. Z = +2. However. p = . difficulty is not a property of a task. t(122) = .24.82. This means there are unequal n 24 per group. (An ANOVA with high-low error group as a factor also supports this conclusion. in the direction expected and with no speedaccuracy tradeoff. As a hypothetical example. Z = +0.6% errors) than the verbal-neutral group (12. difficulty in the neutral condition is unreliable as a guide to difficulty in the emotion conditions.59. p = . and so conclude that the tasks are not well matched. The spatial and verbal 2-back tasks were well matched for these participants. p = .g.21) was significantly larger than the effect size for the low-error group (r = -. even if they . practice).

with ps < . Participants made the most errors on the 2-back tasks in the first session. 9. . there was a main effect of session for errors. The videos were less effective at inducing emotional states if they came later in the session order. and content.0001. there were significant effects of session for all three. Difficulty needs to be matched in the emotion conditions and it clearly was.Emotional modulation were equated in neutral conditions.1. and .0001. No participant laughed during both withdrawal videos.27. Discussion This experiment found further evidence for a double dissociation between spatial-verbal 2-back performance under induced emotional states. and 3. In three mixed model ANOVAs (session within subjects. anxious. 12. with the three dependent variables of change scores for amused. Six 25 participants of 64 (9%) gasped or startled overtly at some point during both withdrawal videos (at least once per video). Emotion Manipulation. and RT. replicating the results of Experiment 1 in a large sample. Fs(3. anxious.2. 381) = 36. suggesting a systematically less effective emotion induction. The mean change scores for amused.2 Order effects. video order between). F(3. The videos induced emotional states as intended. F(3. mean errors in the first session was 11.3. nor startled during either approach video. 381) = 12.4.0001.8%. and content declined monotonically over the course of the four sessions. The self-report ratings of emotional state differed strongly by video type and in the directions expected. and 8. ps < . as assessed by the ratings.2%. respectively. Averaging over video and task conditions.02. The cross-over interaction held in absolute .0%. versus 8.5% in subsequent sessions. 372) = 16. 37 of 64 (57%) laughed audibly at some point during both approach videos (at least once per video). In a repeated measures ANOVA (session within subjects).

Davidson. Tomarken. the BIS-BAS scales (Carver & White. The strength of the resulting emotional states should differ systematically across participants as a function of trait reactivity. Those making fewer errors showed no effect and perhaps an opposite one (a speed-accuracy tradeoff makes it uncertain). 1990). Using individual differences in this way gives a strong test in a large sample with well-validated measures. suggesting specificity to cognitive control. Further. & Henriques. The counterintuitive prediction that spatial performance would be enhanced by a withdrawal state and impaired by an approach state was uniformly upheld. as comparisons to other work suggest (Sobotka. Those most reactive should show the largest influences of the videos on spatial versus verbal performance. & Senulis. It held significantly more strongly for participants making more errors than average than for those making fewer errors. a strategy that capitalizes on individual variation has much to recommend it. rather than other correlates of watching the videos. trait emotion measures are available. 1994) consist of self-report questions that target sensitivity to cues of threat and cues of reward. . Davidson. the influence of the videos should be specific to approachwithdrawal emotion. For approachwithdrawal. The logic is that some participants should be more reactive to specific classes of emotional stimuli than other participants. The BIS-BAS measures correlate significantly with the prefrontal EEG asymmetry (Harmon-Jones & Allen. Trait scales that meet the present requirements are available. Because well-validated. The videos do serve as an emotional challenge and plausibly involve some degree of approach and withdrawal emotion.Emotional modulation 26 terms across all participants. 1992. Experiment 3 The aim of this experiment was to show that the influence of the videos on 2back performance is attributable to emotional states.

1991). The prediction is that BIS-BAS measures should be associated with the Emotion x Task interaction. was selected to assess neuroticism and extraversion. 1985). Watson & Clark. The two factors are associated with both reactivity to affective stimuli and the level of typical affect (Costa & McCrae. Larsen & Ketalaar. 1980. the EPQ-R (Eysenck et al.Emotional modulation 27 1997. The other individuals (high-high. neuroticism and extraversion are appropriate measures of trait emotion against which to compare the BIS-BAS measures in terms of their effectiveness at predicting the Emotion x Task interaction. That is. 1997). and highBAS/low-BIS individuals (high reactivity to cues of reward but not threat) have greater left activity. 1984). To show specificity of the effect to approach-withdrawal emotion. high-BIS/low-BAS individuals (high reactivity to cues of threat but not reward) have greater right hemisphere activity. low-low) are intermediate in terms of hemispheric asymmetry. Sutton. the strategy was to contrast the 2-back performance of participants who are simultaneously above the mean on BIS and below the mean on BAS (denoted high BIS/low BAS) against those . short form. For this reason. Almost all factor-analytic assessments of personality from self-report measures find dimensions associated with negative and positive emotionality (Costa & McCrae. For a comparison trait-emotion scale.. Sutton & Davidson. 1998. Extraverts tend to experience positive affect more often and to react more positively to pleasant events. and more strongly than a comparison trait measure that is less specific at targeting approach-withdrawal reactivity. Neurotics tend to experience negative affect more frequently and to react more negatively to unpleasant events. 1980. The relation should be more evident after an emotional provocation than a control (neutral) condition if these individual differences are acting as selective amplifiers of approachwithdrawal emotional states. & Ketalaar. Gross.

High and low trait groups were defined by mean splits on each of the four measures (BIS. the same participants were categorized instead on the basis of mean splits on the EPQ-R trait measures. The prediction is that the high BIS/low BAS group should show the Emotion x Task 28 interaction in response to withdrawal videos. the high BAS/low BIS group should show the Emotion x Task interaction in response to the approach videos. The specific pattern of spatial versus verbal 2-back performance within trait groups was tested using a focused contrast on groups defined by trait emotion scores. and extraversion). This effect should be stronger after an emotion induction than a neutral condition.e. and other). For neuroticism-extraversion. yet not strongly in response to the withdrawal videos. as three trait groups (high neuroticism / low extraversion. neuroticism.Emotional modulation who are below the mean on BIS and above the mean on BAS (i. there may well be an association with 2-back performance. The strength of the emotion x task interaction found in Experiment 2 was indeed predicted by BIS-BAS scores. high BAS/low BIS). which are emotion related but not as specific to approach-withdrawal emotion. yet not strongly in response to approach videos. BAS. withdrawal) x 3 Trait (high BIS/low BAS.. Conversely. Method All data were collected in Experiment 2. high BAS/low BIS. For neuroticism-extraversion. and the other groups should be intermediate. Results BIS-BAS. Six groups were defined: 2 Video (approach. high BAS / low BIS should show better verbal than spatial performance but only after an approach video. Errors and RT are summarized in Figure 3 A . but a less specific one. and other). high extraversion / low neuroticism. The prediction is that: high BIS / low BAS should show better spatial than verbal 2-back performance but only after a withdrawal video.

or the emotion-neutral condition (Table 5). shown in Figure 3 B. the contrast is in the same direction as for BIS-BAS. For BIS/BAS. which mostly confirm the selectivity in differential strength of the induced emotions. F(1.Emotional modulation as spatial-verbal difference scores within Video x Trait groups for participants in the high-error group. A guiding assumption was that the high and low trait groups would respond differently to the same videos. the contrast was not significant for either the emotiononly or emotion-neutral comparison. but not significant for either the emotion-only condition. summarized in Table 5. which they did. Emotional reactivity. 60) = 8. anxious. Contrast 2 was applied to the data for groups defined by high/low neuroticism x high/low extraversion. The mean change scores for amused. but also when controlling for baseline ability by subtracting each subject's neutral performance for each task (emotion-neutral analyses). Thus neuroticismextraversion scores were not reliably associated with the specific pattern although the trends were in the same direction. 29 Table 4. Control Analyses Several relevant control analyses are reported in Experiment 2.52. Differences in reactivity were in the expected directions. Within the high-error group. The reliabilities of these associations were tested using Contrast 2. as shown by post-video minus pre-video change scores on the self-reported emotion measures. Neuroticism-extraversion groups did not show a reliable association with 2-back performance. . the interactions of Video (approach. and content were submitted to ANOVAs. Across all participants. withdrawal) x Trait group (high BIS/low BAS. the full set of analyses are summarized in Table 5. Neuroticism-Extraversion. The same pattern held across all participants. The pattern was reliable not only after the emotional videos (emotion-only analyses). high BAS/low BIS) were significant and in the direction expected for anxious.

p = . One possibility is that the term amused might not be specific to approachrelated emotion. p = . 60) = 1. all ps < . Of note. BIS-BAS scores were significantly and specifically associated with differential spatial-verbal 2-back performance after the emotion induction.31. F(1. Discussion The specificity of the Emotion x Task interaction to BIS-BAS groups was striking. Thus. p = . F(1. In addition to differential reactivity. BIS and neuroticism were positively correlated with STAI-S. neuroticism. A measure of state anxiety (STAI-S) was obtained on entering the laboratory before any emotion induction or task.37. verbal 2-back performance was worse than spatial for those participants who were above the mean on BIS and below the mean on BAS. with others intermediate.40. As expected from random assignment. 1998). the group seeing withdrawal videos did not differ from those seeing approach videos on BIS. p = .Emotional modulation 30 p = . 60) = 6.58. After the withdrawal videos.81. content. The opposite held for those above the mean on BAS and below the mean on BIS.04. Trait measures. the high BIS/low BAS participants viewing the withdrawal videos had the strongest differential reactivity (anxiety).0049. F(1. although not for amused. This was shown within-subjects by . or amused. The relation is a stronger result than simply an association with trait measures.05.. As summarized in Table 6.013. 69) = 0. 69) = 1. because it was not explained by differential ability at baseline.10. p = . For neuroticism-extraversion.81. 122) < 1. trait groups should show different levels of on-going affect (Gross et al.30. and these participants showed the strongest effects on spatialverbal 2-back performance.18. all Fs(1. and content. 69) = 0. which they did. whereas BAS was negatively correlated with STAI-S. F(1. F(1.73. BAS. ps > . or extraversion. the Video x Trait group interactions were in the same directions but not significantly so for anxious.

the results are also consistent with a further aspect of the theoretical account: modulation on a hemispheric basis.Emotional modulation 31 subtracting neutral performance. General Discussion Performance on psychometrically matched spatial and verbal 2-back tasks was influenced oppositely by induced approach and withdrawal emotional states. It both supports and extends current frameworks of emotion-cognition interactions (Ashby et al. and there are at least two interpretations. The overall pattern was that an approach state tended to impair spatial performance and improve verbal. whereas a withdrawal state tended to improve spatial performance and impair verbal. Gray. The double dissociation is strong evidence for a selective effect of emotion on components of cognitive control. Under either interpretation. 1994). 1999). 1999. the specificity to BIS-BAS was gratifying.. see Rusting & Larsen. then they should reveal dissociations between tasks that depend if only in part on opposite hemispheres. A. The BIS-BAS scales could simply be more accurate measures of the same underlying traits (for evidence. Considerable neuroimaging evidence suggests that the active maintenance of verbal information . at least for approach-withdrawal emotion. 1997. Conversely. and so on. Beyond simply showing a selective effect of emotion on cognitive control. it could be argued that the BIS-BAS dimensions are in the same space as neuroticismextraversion but rotated within it and so measure essentially two different aspects of personality (J. If approach-withdrawal states modulate cognitive control functions selectively by cerebral hemisphere. 1998). Heller & Nitschke. removing performance differences attributable to not only traits but also sex. Thus the results strongly suggest specificity to emotional states. The relation did not hold as well for neuroticism-extraversion. Tomarken & Keener.

Other work suggests that some of the pairwise dissociations are reliable (for reviews see Brown et al. Bromocriptine. but with some speed-accuracy tradeoff. within each experiment many of the possible tests of pairwise differences between the six conditions would not have been statistically reliable (in Figure 2 A. the only pairwise test having p < . n = 128). 1982. supplementary motor and premotor areas). D’Esposito. The contrasts show only that the overall pattern is reliable. 1998). Finding ways to make the effect larger would facilitate further investigation.24 (Experiment 2. 1995. 1999). There is also considerable evidence for hemispheric specialization in approach-withdrawal emotion (Davidson.. 32 There are several important yet less obvious points to note about what the results do and do not show. Eysenck & Calvo. 2000b. The Emotion x Task interaction was strongest for those performing below average. Heller & Nitschke. influenced executive-task . The modest effect size partly reflects large individual differences in both emotional reactivity and task performance. 2000a.Emotional modulation depends more on left PFC (Broca's area. The overall pattern was reliable across experiments as tested by focused contrasts. 1997). whereas the active maintenance of spatial information depends more on right PFC (premotor areas) (Smith & Jonides. a dopamine receptor agonist. However. 1997). 1994. Modulation of cognitive control on a hemispheric basis is a parsimonious explanation for the current results. Participants performing above average showed a significant effect in the opposite direction for errors. taken in the context of physiological evidence. Gray. Gray. Interestingly.05 was verbal approach vs. Sutton & Davidson. Eysenck. withdrawal for RT). some effects of dopaminergic drugs on cognitive control depend on individual differences in working memory capacity (Kimberg. & Farah. 1997. 1992. The most stable estimate of the effect size was r = .

D'Esposito et al. This finding could be relevant here because of work suggesting a link 33 between positive emotion and dopamine (Ashby et al.. If so. 1974. It is also possible that the participants performing better than average were able to exert compensatory control. Moreover. An obvious direction for future research is to investigate the effects of approachwithdrawal emotional states on other components of cognitive control.. While working memory is clearly a strong candidate for a more specific locus of the effect.Emotional modulation performance oppositely (a crossover interaction) for high versus low capacity individuals. 1998. including the coordination of dual task performance. 1999). 1999). components of executive processing. more work is needed to establish this possibility. 1999). Solomon & Corbit. If this pattern is reliable. it is potentially explicable in terms of reciprocal inhibition between approach-withdrawal systems (cf. and clearly tax cognitive control.. strategy . A final nuance is that the emotional states appeared to bias both spatial and verbal performance away from neutral. individual differences in performance on working memory tasks are thought to reflect differences in controlled processing (Engle et al. 1974). Thus the data suggest that the effect is specific to cognitive control. and so showed little effect of emotion on the behavioral measures. rather than merely improving performance in what would be a hemisphere-congruent manner. Depue & Collins. The 2-back tasks are widely used to activate working memory networks in neuroimaging studies. Smith & Jonides. The analyses using individual differences in 2-back task performance suggest but do not unambiguously implicate working memory. Is the effect truly specific to working memory. they might show a cost on a secondary task. 1999. as suggested but not demonstrated by the 2-back data? Further components include: active maintenance versus executive processing (Baddeley & Hitch.

and so on. selective attention. It would therefore be interesting if the . it is unlikely to describe all the data. regardless of valence.. The most simple prediction of the current account is that cognitive control functions that show hemispheric specialization in PFC will also show selective modulation by approach and withdrawal states. Posner & Peterson. Liddle.. whereas easy tasks do not (Erber & Tesser. Gardner. and memory monitoring (Baddeley. & Grafman. just as dual task methods. 1999). 1996. An influence of cognitive control on emotion is likely. Hager. i. 1992). 1997) and cognition. 1944). sustained attention and vigilance (Pardo. subgoal maintenance and execution (Koechlin. Tasks rarely target a single cognitive process. & Heron. Effortful cognitive tasks return an induced mood to baseline. inhibitory control. have provided insight into cognitive control. sequencing) should be more accurate or faster during approach-related states than withdrawal-related states. Normal emotional states might have unique strengths and applications as a tool for investigating cognitive control.Emotional modulation 34 switching. Friston. & Raichle.g.. Even if such a simple account is largely correct. brain damage. & Berntson. Fox. pharmacological challenge. 1991. 1996. Another source of complexity is that approach and withdrawal systems are not simply mirror images of each other in functional terms (see Cacioppo. Panzer. That is. 1990). sequencing (temporal ordering). a reverse effect. Roberts. Robbins. & Frackowiak.3 given their strong psychological validity and potential for functionally specific influences on self-regulation. normal emotional states might also prove useful. Basso. 1999. 1989. Functions that depend more on left PFC (e. Pietrini. such as intentional memory encoding (Kelley et al.. and there could be conflicting effects of emotion on different task components. Hyder et al. 1998). and intentional or "willed" action (Frith. 1994). 1991. 1996).e. Shallice & Burgess. the regulation of prepotent or dominant responses (Diamond & Goldman-Rakic. Miller.

and other components for withdrawal behavior and spatial working memory. the spatial and verbal tasks could be interpreted as behavioral probes of prefrontal subsystems that also support the active maintenance of approach-withdrawal goals (cf. but to regulate the systems responsible for maintaining active approachwithdrawal goals.Emotional modulation 35 effectiveness of a cognitive control task at normalizing mood depended on the task and mood. It seems likely that goals would require active maintenance. Yet an association of verbal working memory with approach emotion and spatial working memory with withdrawal emotion could be coincidental (e. That active maintenance but not manipulation shows hemispheric specialization (Smith & Jonides. and the functional organization of the brain. Toward a Theoretical Basis Theorizing will need to proceed cautiously. Tomarken & Keener.g. That neural networks for sequencing are more left lateralized .. For example. emotion. with the idea of goal-regulation as a conceptual thread running through cognition. Under this account. The current results are consistent. approach and withdrawal states could have opposite effects on 2back performance not to regulate the active maintenance of spatial and verbal information. perhaps simply an uninteresting consequence of co-lateralization). whereas sustained attention might be required more for withdrawal and spatial controlled processing. 1998). to speculate. Conversely. at least. Another possible reason for the empirical association of enhanced verbal 2-back performance with approach-related states is that some components of cognitive control could be more important for both approach behavior and verbal working memory. sequencing is plausibly required more for approach and verbal controlled processing. 1999) fits with cognitive considerations concerning the critical computations for goal management. but not a high degree of on-going manipulation.

Conversely. DA might mediate the effects of positive emotion on cognition (see Ashby et al. and leads to phasic activity responses that are stimulus-specific (Aston-Jones. Given strong biochemical and physiological similarities between the DA and NE. Computational models can account quantitatively for the effects of DA on working memory both at the level of neural systems (Braver & Cohen. & Gunturkun. 2000. & Meyerson. like DA. & Evans. To speculate. 1984). in press) and single neurons (Durstewitz.Emotional modulation 36 (Banich.. pharmacologic administration of NE agonists leads to increased anxiety or panic at high doses (Gray. 1996a) and for sustained attention more right lateralized (Cabeza & Nyberg. and yohimbine facilitates the acoustic startle reflex in humans (Morgan et al. Valentino. Foote. Diffuse effects that can nonetheless be segregated on a gross physical basis would have the twin advantages of being able to modulate a number of component functions at the same time (within a hemisphere). 1983. NE and DA systems are functionally similar at the synaptic level yet anatomically segregated. 1994. 1983). Petrides. with NE showing . NE might partially mediate some aspects of withdrawal-related emotion. Pardo et al. Zuckerman. Van Bockstaele. 1999. 1997.. dopamine (DA) and norepinephrine (NE). exerts a modulatory effect at both inhibitory and excitatory synapses. Lateralization of components of cognitive control could serve as a simple way to make a physiological regulatory mechanism as easy to implement as possible. Owen. it is possible that they exert exhibit similar patterns of activation dynamics and similar neuromodulatory effects on cognitive control functions in PFC. 1999). Owen. Doyon. 1993). NE. Bloom. For example. & Aston-Jones. the diffuse projecting catecholamines. Kelc. have neuromodulatory functions plausibly suited for such a role. Davis. 1991) in PFC is consistent with this possibility. & Tsaltas. yet also be able to have specific effects (between hemispheres). 1999).. Depue & Collins.

1984. Wittling. 1991). and neural connectivity (Oke. The neuromodulatory effects of DA on working memory are a promising avenue for integrative research. 1997. 1998). 1993). & Glick. 1978). involving hemispheric lesions (Robinson. there have been relatively few investigations of how emotional states modulate PFC functions. 1974). and so there are not many constraints on such accounts. such a substrate must support differential effects of approach-withdrawal emotion on spatial and verbal 2-back tasks. The present results suggest two. mechanistic framework (Braver & Cohen. Keller.Emotional modulation greater posterior and dorsal connections and DA showing greater connectivity in 37 ventral anterior regions. Individual differences are likely to be important (Kimberg et al. 1995). 1999). Constraints on models of PFC Viable models of PFC must accommodate both cognitive and emotional functions within a unified. Fitzgerald. 1998. Finally. Summary . However. the NE and DA systems are mutually inhibitory within the PFC (Tassin. but comes from a range of both human and animal studies. in press. & Adams. Medford. Revelle. The evidence for laterality could be stronger.. It is also intriguing that NE appears more right-lateralized and DA more left-lateralized (for reviews see Tucker & Williamson. Damasio. Second. a substrate critical for 2-back performance is needed that can be influenced by approachwithdrawal emotion. 1985). consistent with opponent processing in emotion (Solomon & Corbit. Keller. First. Metcalfe & Mischel. cf. measurement of neurotransmitter metabolites under emotional conditions (Carlson. Integrating the two constraints would be facilitated by understanding more specifically the mechanisms of cognitive control that are modulated by approach-withdrawal emotion.

The results support and extend several neuropsychological frameworks for cognition-emotion interactions. emotional states can selectively modulate components of cognitive control. In other words.Emotional modulation Approach-withdrawal emotional states can double dissociate performance on spatial and verbal 2-back tasks. and constrain models of PFC function. 38 .

G. R. Neuropsychology. J. Working memory. E.. 529-550. H. Quarterly Journal of Experimental Psychology. A. Aston-Jones. The psychology of learning and motivation (Vol. A. A. 76787687. & Nicholas.. C. A. 5-28. G. Banich. Cortical systems for the recognition of emotion in facial expressions. & Lader. R. and PTSD: Neurobiological and clinical parallels. Boles. & Meyerson. J.. & Hitch. Koff. pp. Baddeley. A. 47-89). U. H.. M. R. D.. 24. A. (1997). R... (1974). Exploring the central executive. DC: American Psychiatric Press. Ashby. New York: Guilford.Emotional modulation References 39 Adolphs. A. Baddeley. Van Bockstaele.). & Law.. 8. D. M. (1999). British Journal of Medical Psychiatry. 169180. In G. F. T.. stress. In M. 47.. Boston: Houghton Mifflin. (1986). A simultaneous task comparison of differentiated and undifferentiated hemispheric resource theories. A. Working memory. . M. J. (1994). Baddeley. (1997).. 211-218. Tranel. A neuropsychological theory of positive affect and its influence on cognition. E. & Turken. 49(A). New York: Academic Press. (1986). Valentino. Journal of Experimental Psychology: Human Perception and Performance. G. New York: Oxford. Bower (Ed. Damasio. 106. The expression and perception of facial emotion in brain-damaged patients. M. Journal of Neuroscience. ADHD and the nature of self-control. 204-215. 24. (1996). (1974). Perlman Lorch. & Damasio. Barkley. B. A. Isen. (1998). Locus coeruleus. (1996). Neuropsychologia. Catecholamine function in post traumatic stress disorder: Emerging concepts . Psychological Review. Murburg (Ed.. M. M. The use of analog scales in rating subjective feelings. M. Washington.). Borod. 16.. T. Bond.. B.

Neuroimage. Jonides.. T. E. 201-207.).. Cabeza. Monsell & J. Berntson.. W. J. Braver.. Bench. W. Carlson. Desmond. S. Klein. D. & Glick. (1999)... & Cohen. A parametric study of prefrontal cortex involvement in human working memory. Nystrom. 9(14). & Hatfield. 12. G. S. Cacioppo. 76.). New York: Guilford Press. Jr. 119-142).. D. G. (1991). 3233-3239. The psychophysiology of emotion. (2000). MA: MIT Press. C. Canli. G. C... T. J. (1993). Cohen. E. Smith. (1994). Lewis & J. (1999). E. E. & Nyberg. 313-318. 24(4). J. D.. In M. (in press). W. (1998). L. R. J. Side and region dependent changes in dopamine activation with various durations of restraint stress. Glover. The affect system has parallel and integrative processing components: Form follows function. G. The Neuroscientist.. Journal of Cognitive Neuroscience. R. Gardner. Brain Research. Scott. Cacioppo.. & Berntson. & Noll. J.. 829-847. (1997). Psychological Medicine. L. R. Cognitive function in depression: its relationship to the presence and severity of intellectual decline. Hemispheric asymmetry for emotional stimuli detected with fMRI.. D.. Z. & Gabrieli. Neuroreport. T. T.Emotional modulation Braver. 49-62. J.. Keller. D.. T. 5(4). G. J. L. Hemispheric asymmetry in the experience of emotion: A perspective from functional imaging. Imaging cognition II: An empirical review of 275 PET and fMRI studies. J.. 839-855. N.. . On the control of control: The role of dopamine in regulating prefrontal function and working memory. In S. Handbook of emotions (pp. D. L. J. R. Driver (Eds.. L. J.. S. M. & Dolan. Haviland (Eds.. 40 Brown. G. Journal of Personality and Social Psychology. 1-47. 550(2). Cambridge. T. E. 5(1). Attention and Performance XVIII .. Canli. J. Zhao. C. Fitzgerald.

(2000).. Brain Research: Cognitive Brain Research. D'Esposito. Carver.. and health psychology. K. 1-13. & White. D. (1978 / 1989). MacWhinney. Emotion in the perspective of an integrated nervous system. (1998)... Behavior Research Methods. (Eds.. T. 19-35. and affective responses to impending reward and punishment: The BIS/BAS scales. 668-678. Carver. Zarahn. & Scheier. & McCrae. Control theory: A useful conceptual framework for personality-social. 7(1). Craven. J. 97(1). Functional MRI studies of spatial and nonspatial working memory. M. S. D... Shin. 38. Burbank. M. F. Dalgleish. New York: John Wiley & Sons. Journal of Personality and Social Psychology. S. R. Halloween [Film]. J.. Psychological Bulletin. & Computers... (1990). 92(1). G. (1993). & Lease. 25. 111-135.. Journal of Personality and Social Psychology. T. Carver. D. Aguirre. A. Ballard. CA: Dimension Home Video. J. (1998). & Provost.. Los Angeles: Heron Communications. & Power. . Damasio. T. B. & Hill. clinical. (Original work produced 1978). 67. Influence of extraversion and neuroticism on subjective well-being: Happy and unhappy people. R. 257-271. I. R. J. Flatt. M. F.. (1994). 26(2-3). J. Yablans.. Scream [Film]. Brain Research: Brain Research Reviews. C. Handbook of cognition and emotion. E. 83-86.Emotional modulation 41 Carpenter. (1980). P. 319-333. R. M. Origins and functions of positive and negative affect: A control-process view. C. behavioral activation.). C. Behavioral inhibition. Instruments. K. (1996).. & Scheier. Cohen. W. M. S. Psychological Review. Costa. PsyScope: An interactive graphic system for designing and controlling experiments in the psychology laboratory using Macintosh computers. (1982).

S. Biological Psychiatry. Cognition and Emotion. . Journal of Personality and Social Psychology. Diamond. Davidson. Brain asymmetry (pp. & Shelton. N. 115-138. A. R. 32. Diamond. 307-330. (1990). Depue. P. R. (1992). 22. E. Neurobiology of the structure of personality: Dopamine. The development and neural bases of memory functions... (1988). & Fox. F. & Collins. R. Experimental Brain Research. N.). 438-451. 12351237. Science. (1999). MA: MIT Press. J. Annals of the New York Academy of Sciences. & Goldman-Rakic. (1992). Cambridge. J. Davidson. P. Cognition and Emotion. Affective style and affective disorders: Perspectives from affective neuroscience. 267-317. In R. J. Comparison of human infants and rhesus monkeys on Piaget's A B task: Evidence for dependence on dorsolateral prefrontal cortex. Behavioral and Brain Sciences. Davidson & K. Kalin. 24-40. (1989). S.Emotional modulation Darke.. 361-387). Cerebral asymmetry. 7. H. Lateralized effects of diazepam on frontal brain electrical asymmetries in rhesus monkeys. 12(3). Hugdahl (Eds. J. Davidson. A. S. H. Davidson. and extraversion. 491-569. Cerebral asymmetry and emotion: Conceptual and methodological conundrums. 608. Effects of anxiety on inferential reasoning task performance. R. (1995). 74. as indexed by the A-not-B and delayed response tasks. 218. in human infants and infant monkeys. (1998). emotion. R. and affective style. Asymmetrical brain activity discriminates between positive and negative affective stimuli in human infants. (1982). R. A. facilitation of incentive motivation... 55(3). J. 42 Davidson. A. 499-505. J.

G.. 6(6)... (1982). M. R.). D. 339-359. New York: Oxford. .. M.). Eysenck. & Tuholski. In K. (1999). (1999). New York: Springer-Verlag. and social behavior (pp. J. R. Task effort and the regulation of mood: The absorbtion hypothesis. Journal of Neuroscience. (1985). Personality and Individual Differences. Attention and arousal: Cognition and performance. & Ashbrook.. P. W. J. 28. Foote. W. H. & Gunturkun. A. W.. 844914. Forgas (Eds. E. (1983).. & Calvo. Eysenck.. M. M. (1980).. V.. W. Kane. cognition. The nature of emotion. Ekman. S. A neurocomputational theory of the dopaminergic modulation of working memory functions. P. A revised version of the psychoticism scale. P. 21-29. Ellis. Bloom. Engle. Eysenck. O. W. 2807-2822. Physiological Reviews. P. 63(3). 1125-1134. & Tesser. W. (1994). & Davidson. & Barrett. L. H. Eysenck. Erber. R.Emotional modulation 43 Durstewitz. Cognition and Emotion. NY: Hogrefe. (1988). Kelc.. Freisen. Shah (Eds. (Eds. Journal of Personality & Social Psychology. Resource allocation model of the effects of depressed mood states on memory.). 25-43). Facial signs of emotional experience. Fielder & J. (1992). Models of working memory . Miyake & P. 6. A. Lewiston. In A. Anxiety and performance: The processing efficiency theory. Ekman. G. & Aston-Jones. M. S. J. Affect. 39(1-sup-6). (1992). and functions of the prefrontal cortex. Individual differences in working memory capacity and what they tell us about controlled attention. B. Journal of Experimental Social Psychology.. S... F. 19(7). 409-434. general fluid intelligence. G. & Ancoli. S. Nucleus locus ceruleus: New evidence of anatomical and physiological specificity. New York: Cambridge University Press.

1. CT: Vestron Video. & Davidson. 863-872. (1987). & Boller. S. Podell. P.. Behavioural Research and Therapy. pp. J. J.). In V. R.). & S. Mood and judgment: The affect infusion model (AIM). B. (1995). Goldberg. F. (3rd ed. & Frackowiak. New York: Oxford. P. Psychological Bulletin. K. (1994). (1985).. New York: The New York Academy of Sciences. 24(3). 8. Neuropsychologia.Emotional modulation Forgas. Lateralization of frontal lobe functions and cognitive novelty.. 117(1). Taste-elicited changes in facial signs of emotion and the asymmetry of brain electrical activity in human newborns. (1986). Gray. The prefrontal cortex. Grafman... A. M. & Lovell. Friston. A. sec. (1991). N. (1995). Proceedings of the Royal Society of London B. (1970). 244(1311). Fox. 46. J. 769). F. Willed action 44 and the prefrontal cortex in man: a study with PET. C. J. If it's not left. 417-422. Goldman-Rakic. 1. N.. The psychophysiological basis of introversion-extraversion. Structure and functions of the human prefrontal cortex. R. S. Journal of Neuropsychiatry and Clinical Neurosciences. it's right: Electroencephalographic asymmetry and the development of emotion. New York: Lippincott-Raven. Funt. F. R.. 373-417). 6(4). Liddle. Geiger (Eds. J. P. Frith. Stamford. (Vol. K. Candid Camera [Film]. . D. (Eds. 241-246. A. 371-378.). Circuitry of primate prefrontal cortex and regulation of behavior by representational memory. (1991). Fox. Handbook of physiology (Vol. (1997). Fuster. K. Mountcastle. A. Plum. 39-66. 249-266. E. American Psychologist.. Holyoak. J.

S. K. Cognition and Emotion. J. Gray. 329-331).. 4(2). J. Brain systems that mediate both emotion and cognition. R. To appear in S. The nature of emotion (pp.). A. Oaksford (Eds. In P.. and anxiety. Gray. Psychological and physiological relations between anxiety and impulsiveness. A. 24. NJ: Erlbaum. Gray. R. J. A. impulsivity. (2000a).. New York: Oxford. J..Emotional modulation 45 Gray. Relations between affect and personality: Support for the affect-level and affect-reactivity views. J. Behavioral activation sensitivity and resting frontal EEG asymmetry: Covariation of putative indicators related to risk for mood disorders. Davidson (Eds. . Manuscript submitted for publication. J. Gray. Oxford: Oxford University Press. R. Journal of Abnormal Psychology. R. Moore & M. Davis. Emotional modulation of working memory subsystems: A neurocomputational hypothesis. Personality dimensions and emotion systems. E. Gray. (1994). 25. Harmon-Jones. E. S. R. A bias toward short-term thinking in threat-related negative emotional states. Effects of emotion on verbal and spatial cognition: A review. Gross. J. Gray. C. J. A. J. (1983). 65-75. The biological bases of sensation seeking. 106(1). 279-288.) Emotional cognition.. (1990). Sutton. (1997). Personality and Social Psychology Bulletin. & Allen. & Tsaltas. In M. Hillsdale. (1999). J. (2000b). (1998). J. T.. 269-288. Owen. Manuscript in preparation. 159-163. (in press). Personality and Social Psychology Bulletin. N. Zuckerman (Ed. & Ketalaar. John Benjamins. Hemispheric differences in experienced emotion: Prosocialselfish or approach-withdrawal? Comment on Buck (1999). Gray. J. (1982). Ekman & R. The neuropsychology of anxiety: An enquiry into the functions of the septo-hippocampal system.).

"Willed action": a functional MRI study of the human prefrontal cortex during a sensorimotor task. Trabasso (Eds. NJ: Erlbaum. C. W. 91(2). Proceedings of the National Academy of Science U S A. W. Cognition and Emotion.. 12. & Miller. Regional brain activity in emotion: A framework for understanding cognition in depression.. B. 11. Phelps. Heller. Personality. W. Lateralization in emotion and emotional disorders.). 26-32. New York: Guilford. (1998). Blamire. A. motivation. F. A. J. The neuropsychology of emotion: Developmental patterns and implications for psychopathology. R. and performance: A theory of the relationship between individual differences and information processing. B. Haviland (Eds. Stein. (1993). Humphreys. Hyder. Psychological and biological approaches to emotion (pp. B. Wiggins. S. S. Isen. J. The puzzle of regional brain activity in depression and anxiety: The importance of subtypes and comorbidity. Cognition and Emotion. & Nitschke. L. 637661. MA: Harvard University Press. M. A. Lewis & J. Heller. A. (1997). Current Directions in Psychological Science. Hemispheric asymmetry. 6989-6994. Cambridge.. In N.. Labar. (1997)...). G. . J. J. & Shulman. M. & Nitschke. 7(7). 261-277). B. Leventhal. Hillsdale. Hellige. J. (1984).. K. G. In M. Psychological Review. Positive affect and decision making. Nitschke. W. B. 167-211). (1990).. M. 94(13).. & T. (1998). 153-184. E. 421-447. & Revelle. Heller.Emotional modulation 46 Heller.. W. Handbook of emotion (pp. (1993).

R. cortisol. L. E. M.. Koechlin.. Smith. D’Esposito. (1991). F. J. S.. T.Emotional modulation 47 Jonides. Ollinger. A. Extraversion and vigilance performance: 30 years of inconsistencies.. (1995). & Koeppe. E. E. Z. M. 112(2).. E. 50(5). & Grafman.. Lang. The emotional brain. Buckner. C. Nature. E. Kosslyn. Koelega.. Cohen. J. Kalin. and behavior associated with fearful temperament in rhesus monkeys. J. N. M. E. K. Pietrini. M. Snyder. 462-475.. (1998). Akbudak. E. Journal of Personality and Social Psychology. A. 239-258. Larson.. J. Verbal working memory load affects regional brain activation as measured by PET. Kosslyn. R. R. Shelton. J. D. (1997). Miezin. E. Journal of Cognitive Neuroscience. 286-292. E. The emotion probe: Studies of motivation and attention. E. 445-484). Wet mind. S. New York: Simon and Schuster. Asymmetric frontal brain activity.. 3581-3585. H. Kelley.. Emotion and adaptation. W. J.. Epilogue: Wet Mind in 1995. M. (1992). J. E. & Koenig.. M. 372-385. G. The role of the anterior prefrontal cortex in human cognition. American Psychologist. O. S. & Ketalaar. S. Psychological Bulletin.. S. P. M. (1992). E. B. (1998). 927-936. 20(5). 148-151. (1995). M. .. R. & Farah. 9. J. Kimberg.. Minoshima.. Y. R... H. New York: Free Press. Raichle.. Neuroreport. S. P. Panzer. Lauber. Wet mind (pp. Neuron. H... 61. Larsen.. S. O.. 112. E. Personality and susceptibility to positive and negative emotional states.. (1996). Lazarus. & Petersen. Conturo. New York: Free Press. S. (1991). 399. LeDoux. Awh. Behavioral Neuroscience. N. & Davidson. 132-140.. (1997). (1999). 8. & Koenig. Basso. Effects of bromocriptine on human subjects depend on working memory capacity. Schumacher. New York: Oxford.. J. McDermott.. Hemispheric specialization in human dorsal frontal cortex and medial temporal lobe for verbal and nonverbal memory encoding.

V. European Journal of Neuroscience. R. A. Evidence for a two-stage model of spatial working memory processing within the lateral frontal cortex: a positron emission tomography study. & Charney. C. Oke. A. McV. (1978). J.. 8(2). Yohimbine-facilitated acoustic startle reflex in humans. Psychological Review. 431-465). 22(2). J. and central executive processes.. Grainger. Doyon. and Cognition. C.. & Adams. M... A hot/cool-system analysis of delay of gratification: Dynamics of willpower. Consciousness and selfregulation (Vol.. A. 353-364. Keller. reasoning. 31-38.. Hunt (Ed. E. C. Evans. Petrides. Davis. 476-492. Norman. 1-18). M. (1999). A. Burbank. Experimental studies of conflict. & Petrides.Emotional modulation Metcalfe.. P. Models of working memory. In J. 48 Morgan. & Evans. In G. New York: Plenum. 200. F. Owen. (1986)... Roger & Me [Film]. (1996). & Shallice. I. New York: Oxford University Press. Panksepp. 342-346. M.).). H. D. Moore. 6(1). (1996b). (Eds. A. New York: Ronald. A. 4. Owen. B. Planning and spatial working memory: A positron emission tomography study in humans. W. Medford.. M. New York: Cambridge University Press. M. S. A.. A. S. (1989).. Affective neuroscience. (1996a). D. pp. 110. Shapiro (Eds. Lateralization of norepinephrine in the human thalamus. 1141-1143.). & Williams. Grillon.. M. Attention to action: Willed and automatic control of behavior. Journal of Experimental Psychology: Learning. J. C. R. pp. & Mischel. (1998). M.. Schwartz & D. Morris.. (1944). M.. J. J. 106. Science. (1993). CA: Warner Brothers. G. (1999).. M.. E. Mood. 3-19. Oaksford. Personality and the behavior disorders (Vol. Southwick. & Shah. Psychopharmacology. Krystal. . Cerebral Cortex. 1. T. Memory. N. Miller. Miyake.

. Pardo. W. & Rosnow. G. Localization of a human system for sustained attention by positron emission tomography. 61-64. R. Raichle.. Contrast analysis. Prefrontal cognitive processes: Working memory and inhibition in the antisaccade task. I. In A. Annual Review of Neuroscience. C. Practice-related changes in human brain functional anatomy during nonmotor learning. awareness. Videen. B. & Peterson. Journal of Experimental Psychology: General. T. R. Hollywood.. P. E.. 135156). (1987). L. The attentional system of the human brain. L. and control (pp. New York. Robbins. Philosophical Transactions of the Royal Society of London. R. 49 Posner.. & Petersen. E. S. A. G'day Australia: Like nothing else on Earth [Film]. J. T. CA: Paramount Pictures... Hager. Hillsdale: Erlbaum. Fox. & Heron. 8-26. (1975). & Snyder. Solso (Ed. Robinson. I. V. J.. J. Fox.. Roberts. & Raichle. E. (1993). R. (1985). Information processing and cognition (pp. (1991).. M. (1994). M. Revelle.Emotional modulation Pardo. J. New York: Oxford. 346-373). 374-393. . M. 349. Lateralized behavioral and neurochemical consequences of unilateral brain injury in rats. NY: Academic Press. Posner. Individual differences in personality and motivation: 'noncognitive' determinants of cognitive performance.. T. 351(1346). New York: Cambridge University Press. In R. M. Queensland Tourist and Travel Corporation. MacLeod. (1990). R. (1996). C. 13. Cerebral Cortex. Attention and cognitive control.). (1985).. V. 123(4). O. Nature. W.). S. D. R. Attention: Selection.. Weiskrantz (Eds. E. 4(1). P. 1463-1470. Rosenthal. Baddeley & L. M. L. Dissociating executive functions of the prefrontal cortex. T. Cerebral lateralization in nonhuman species (pp. (1994). 25-42. A. Fiez. 55-85).

876-882. (1996). J. Higgins (Eds. Handbook of motivation (Vol. Personality and Individual Differences. (1999). & Burgess. 95(3). Marshuetz. J.Emotional modulation 50 Rosvold. Sorrentino & E. pp. 351(1346). 367372. & Koeppe. H. A. P. Storage and executive processes in the frontal lobes. Motivational and emotional controls of cognition. J. Rusting. 343-350. N. M. Simon.. Journal of Experimental Psychology: General. R.. 527-561).. 1405-1411. P. S. 74. 2. Smith. R.. T. Sarason. & Beck. J. (1998). Davidson. 83(4). Components of verbal working memory: Evidence from neuroimaging. (1992). (1996). & Miyake. Clarifying Gray's theory of personality: A response to Pickering.. H. A. Smith. 26. 283. A. Proceedings of the National Academy of Science USA. L. 20. S. E. & Jonides. Corr and Gray. Psychological Review. A continuous performance test of brain damage. I. Science. Philosophical Transactions of the Royal Society of London B. A. Bransome. 236-247. & Larsen.. J.. E. 29-39. 4-27.). Schwarz. Anterior brain electrical asymmetries in response to reward and punishment. 1657-1661. E. A. In R. The separability of working memory resources for spatial thinking and language processing: An individual differences approach. Electroencephalography and Clinical Neurophysiology. (1956).. . Journal of Consulting Psychology. R. E. C.. (1999). H. (1990). E. Shah. T.. L. Mirsky. D. Jonides. & Senulis. Shallice. C. E. Sobotka. F... (1967). New York: Guilford. Happy but mindless? Mood effects on problem solving and persuasion. The domain of supervisory processes and temporal organization of behaviour.. 125(1).

. J. B. Tassin. An opponent-process theory of motivation: I. J. A. 81.. B. C. C. 119-145. Boone. K. & Miller. A system for relational reasoning in human prefrontal cortex. 42. J. 8. A. Norepinephrine-dopamine interactions in the prefrontal cortex and the ventral tegmental area: relevance to mental diseases. Advanced Pharmacology. Watson.. Thomas. Tomarken. (1997). & Davidson. Vin Di Bona Productions. 204-210. Holyoak. D. & Corbit. J. & Williamson. Mishkin. (1998). D. Sutton. A.. Asymmetric neural control systems in human self-regulation. M. B. D. 10. L. Psychological Bulletin. Davidson. P. Psychological Review. (1990). (1999). 465-490. B. P. & Clark. Psychological Review. 185-215. (1974). R. 712-716. Negative affectivity: The disposition to experience aversive emotional states.. (1977). 119-125. 791-801. Palo Alto. & Henriques. Tomarken. A. R. J. Tucker. K. . A. Cognition and Emotion. Knowlton.Emotional modulation 51 Solomon. Best of America's Funniest Home Videos [Film]: CBS / Fox Videos.. & Keener. Spielberger. J. Psychological Science. de Menezes Santos... Psychological Science. D. J. Temporal dynamics of affect. R. Journal Personality and Social Psychology. CA: Mind Garden. 91(2). S. 12. (1984). A. J. 387-420. F. M. Waltz.. S. (1984).. Resting frontal brain asymmetry predicts affective responses to films. (1990). 59(4). K.. (1998). J. State trait anxiety inventory (Form Y). Prefrontal brain asymmetry: A biological substrate of the behavioral approach and inhibition systems. 96.. Frontal brain asymmetry and depression: A self-regulatory perspective. L... D. J. L. R.

W. Brain asymmetry (pp. In R. (1984). 413-471.). (1995). Davidson & K. MA: MIT Press. Zuckerman. J. Hugdahl (Eds. 305-357). Brain asymmetry in the control of autonomic-physiologic activity. M. Sensation seeking: A comparative approach to a human trait. 52 . Behavior and Brain Sciences. Cambridge.Emotional modulation Wittling. 7.

and Ken Nakayama for their helpful comments on previous versions of the manuscript. Richard J. whom I thank for his support. Gray. Washington University). guidance. Larsen. Daniel Simons. St Louis.wustl. I also thank Todd S. Braver. Correspondence concerning this article should be addressed to the author at Campus Box 1125. Kosslyn. . Jerome Kagan. MO 63130.Emotional modulation Author Note 53 Jeremy R. Psychology Department (now at the Department of Psychology. and Michelle Leichtman. Washington University. I thank Annya Hernandez and Tracy Thall for their superb assistance in testing some of the participants. or sent by internet to jgray@artsci. The work reported here was carried out as part of my doctoral dissertation with Stephen M. Randy J. McNally.

details available on request.Emotional modulation Footnotes 1 54 The Approach-Withdrawal x Spatial-Verbal task interaction replicated in an experiment not reported here with n = 32 using the same tasks and videos. . 2 Change scores were computed as post-video level minus pre-video level and used as dependent variables in ANOVAs (approach-withdrawal group betweensubjects). 3 I thank Todd Braver for this insight.

Experiment 1 Spatial Withdrawal Contrast 1: All RT Errors High error RT Errors 897 12.8 970 9. df = 14). with MSEs from repeated-measures ANOVAs.5 942 7 -1 Neutral 0 Approach +1 Withdrawal +1 Verbal Neutral 0 Approach -1 55 MSE 12391 9.48 Note. df = 44) and the High-error group (n = 9.44 16500 8.7 933 7.8 971 6. within-subjects) x 2 (Task. for All participants (n = 24.6 993 14.1 919 8. between-subjects) conditions.6 930 16.9 960 7. .0 924 11.Emotional modulation Tables Table 1 Contrast 1 and 2-back Performance. Mean RT (ms) and errors (%) are given for the 3 (Video.6 893 13.5 983 12.

3 -14. Experiment 1 Video Term Amused (approach) Anxious (withdrawal) Bored (low interest) Calm (low arousal) Energetic (high arousal) Approach +24. Emotion ratings are mean change scores.3 Neutral +20. A positive change score indicates an increase after watching the video (e. more anxious after the withdrawal video than before it)..2 Withdrawal +24.Emotional modulation Table 2 Self-reported Emotion Ratings. post-video minus pre-video (possible range = -100 to +100). .g.5 -20.05 from a paired t-test against zero (no change).3 -20.9 56 Note. shown if p < .

4 12.8 Low-error group RT: Spatial Verbal 841 1016 982 1056 930 932 +1.9 18.1 8.8 +2.7 High-error group RT: Spatial Verbal Errors: Spatial Verbal 14.Emotional modulation Table 3 2-back Performance.18 +.24 59680 .8 12.007 +.15 47382 1007 1227 1115 1196 1268 1069 +1. Experiment 2 Video Withdrawal Neutral Approach t p Contrast 1 r MSE 57 All participants RT: Spatial Verbal Errors: Spatial Verbal 11.34 .5 12.7 24.18 +.9 +2.036 +.32 65947 913 1095 1011 1113 1099 975 +0.4 17.21 75.6 18.03 65.35 .16 .4 11.74 .38 .74 +.6 12.

36 (. the high-error group (n = 46. Mean RT (ms) and errors (%) are given for the 3 Video (between-subjects) x 2 Task (between-subjects) conditions.0 8. df = 122). Method and Results).Emotional modulation Errors: Spatial Verbal 9.4 8. and effect size r are from Contrast 1.4 7. for all participants (n = 128.26 19. df = 40).5 58 Note.021) -.0 4. with MSEs from ANOVAs. t. df = 76). and the low-error group (n = 82.3 7. The contrast does not recapitulate the design (see Experiment 2. p. .5 -2.

E = extraversion. a negative lambda means the opposite. N = neuroticism.e. High = above the mean on a trait measure. verbal performance better than spatial).. . and a lambda of zero means little difference. Other = high / high or low / low. A positive lambda means that the 2-back spatial minus verbal difference score is predicted to be positive for errors and RT (i.Emotional modulation Table 4 Contrast 2 Withdrawal BIS-BAS: High BAS / low BIS EPQ-R: High E / low N Lambda: 0 -1 Other Other High BIS / low BAS High N / low E -2 High BAS / low BIS High E / low N +2 +1 Other Approach Other 59 High BIS / low BAS High N / low E 0 Note. low = below the mean.

44 a +0.19 a +0.02 .04 -0.004 +2.Emotional modulation Table 5 BIS-BAS Scores Predict the Emotion x Task Interaction. A positive t value indicates that the predicted direction of spatial .77 .73 +0.80 +0.005 .neutral RT Emotion only Emotion . .19 +1.07 p All t(122) p 60 Neuroticism .neutral +2.03 +1.59 +0.31 .14 +0. Experiment 3 BIS-BAS High-error t(40) Errors Emotion only Emotion .03 +1.07 a +1. a The corresponding data are shown in Figure 3.95 a +3.63 +2.extraversion High-error t(40) p All t(122) p - - Note.64 +0.verbal 2-back performance within emotion conditions was upheld as assessed by Contrast 2 (Table 4).

1 (9.12 (3. r BAS Neuroticism 61 Extraversion Note.34) 36.35 +.03) 41.90 (5. Experiment 3 Range Trait measure BIS BAS Neuroticism Extraversion STAI-S STAI-T 11-28 29-52 0-12 0-12 20-67 21-68 20. n = 128.174.4 (9.18 -.34) 8.54 (4.05 +.Trait scale.18 +.06 +.Emotional modulation Table 6 Self-Reported Trait Emotion.06) 33.05. . If |r| > .14 -.64 -.1) Mean (SD) BIS -.34 -.45 -.14) 5. STAI-S = State/Trait Anxiety Inventory .11 (3.21 +.25 -.State scale. STAI-T = State/Trait Anxiety Inventory .26 Simple correlations. p < .66 -.04 +.

shown separately for the high-error and low-error groups. Spatial and verbal 2-back performance. A positive value means performance for that group was worse after the withdrawal video.Emotional modulation 62 200 100 spatial verbal prediction Withdrawal-approach difference: RT (ms) Errors (%) 0 -100 5 0 -5 Low error High error Figure 1. . The dependent variable is a difference score: mean (S.E.) RT and errors after the withdrawal video minus after the approach video. Experiment 1.

Method and Results). and (B) the high-error group only (n = 46). Experiment 2. The display does not recapitulate the experimental design (see Experiment 2.E.Emotional modulation 63 A: 1500 All participants withdrawal neutral approach Contrast 1 B: High-error group RT (ms) Errors (%) 1000 500 0 30 20 10 0 Spatial Verbal Spatial Verbal Figure 2.) RT and errors within the 3 (Video) x 2 (Task) conditions. Spatial and verbal 2-back performance by video condition. Mean (S. for (A) all participants (n = 128). .

The dependent variable is the mean (S.E. or (B) their EPQ-R neuroticism (N) and extraversion (E) scores. . see Table 5 for a summary of related analyses. Participants are classified in terms of (A) their BIS-BAS scores.Emotional modulation 64 A: 200 BIS-BAS scores high-BAS / low-BIS o t h e r BIS-BAS high-BIS / low-BAS Contrast 2 B: EPQ-R scores high-E / low-N other E / N high-N / low-E Spatial-verbal difference: RT (ms) Errors (%) 0 -200 10 0 -10 Withdrawal Approach Withdrawal Approach Figure 3. Spatial-verbal 2-back performance by trait group.) spatial-verbal difference score in the emotion condition within 3 (Trait) x 2 (Video) groups. shown for participants making more errors than average. A positive difference score means verbal performance was better than spatial. Experiment 3.