Review TRENDS in Cognitive Science Vol9 No.5 May2005 245 important for appraising the aversive or rew and appraisal systems are strikingly similar to those ompared w are 39 sho ate 2nneheoy hich sti tas myg. acc NA 【4 rd type of cognitive regu insula and cingulate 40 ways in which simple stimul ciations are d r expectation formed and Altho the pre syst stimuli are perceived as painful when participants expec avoidance of aversive stimuli 42,exti action of classically ,an thi rostral cingulate regions [461,which might be involved in depend upon interactions between similar cognitiv pain aftect and cognitive expectations about pain, A third approach has directly contrasted top-dow PFC,OFC and/or ACC have been observed consistently, generated by beliets al a stimulus with ing impair. this issue by asking participants either to look at aversiv the appraisal side however,findings have less or to th 818 an dala activa observed in both conditions. Ho ver on top-down extinction.and during reversal learning both striatal 60 8geerao act AC PFC and MPF and amygdala vation h been observed, wit aversive annraisal of an oth wise innocuous image npared with currently reinforced stimuli62.Thes repancies across studies might chan e has also b een studied in the rent Critica a ha as nvo C.O by cingulate contro general.studies have found that rea raisal of neg about,and interpretations of,stimuli,or (ii)by ing to emotion orsal ACC and systems tha late new e onses with stimuli. These and decre trol that in ap raisal systems such as the mygdala or insula in cingulate systems [8,9.The consiste acy of these e oal of reappraisal here ngs (relative to o fact systems recruitedacross studies,however,which migh the use of stimuli that generate strong emotional (see below.and 18.51). said qu ons remain about when and how specific The second type of controlled n is implicate ontrol d appraisa systems int including worki an raisal svstems the tent to which dhavre ho th participante tha different strategies modulate appraisal systems in differ ent ways amygdala and pain- elated cingulate,insula and thalamic Towards a functional architecture of cognitive control of com (ii)incr on of of this re control,includir al cing ate cortex and dorsal and studies whose results can help to elucidate the function yet clo interactions betwee imaging studies of either or important for appraising the aversive or rewarding (as compared with neutral) properties of stimuli. Thus, anticipating a painful shock [37,42,43], heat [38] or injection [39] activates cingulate, insula and amygdala; anticipating pleasant or aversive tastes activates amyg￾dala, nucleus accumbens (NAcc) and/or OFC [44]; and anticipating monetary reward activates NAcc, amygdala, insula and cingulate [40]. The second approach has examined how expectations about how a stimulus might feel influence neural responses to it. Studies have shown that nonpainful stimuli are perceived as painful when participants expect pain, and that this expectation leads to activation of midcingulate regions [45] as well as medial temporal and rostral cingulate regions [46], which might be involved in pain affect and cognitive expectations about pain, respectively. A third approach has directly contrasted top-down responses generated by beliefs about a stimulus with bottom-up responses driven by direct perception of aversive stimuli. To date, only one study has addressed this issue by asking participants either to look at aversive images (bottom-up) or to think about neutral images in negative ways (top-down). Amygdala activation was observed in both conditions. However, only top-down generation activated ACC, LPFC and MPFC systems [17], which might be involved in cognitively generating an aversive appraisal of an otherwise innocuous image. Controlled regulation The use of cognitive change to regulate an existing or ongoing emotional response has also been studied in the context of three different forms of higher cognition and learning. The first type of cognitive regulation is known as reappraisal, and involves reinterpreting the meaning of a stimulus to change one’s emotional response to it [47]. In general, studies have found that reappraisal of negative emotion activates dorsal ACC and PFC systems that support the selection and application of reappraisal strategies, and decreases, increases or maintains activity in appraisal systems such as the amygdala or insula in accordance with the goal of reappraisal [48–54]. There has been variability in the precise prefrontal and appraisal systems recruited across studies, however, which might be attributable to differences in the nature of the stimuli used and the goal or content of reappraisal strategies (see below, and [18,51]). The second type of controlled regulation is implicated in placebo responses to situations that involve no active drug compounds that could impact appraisal systems. Two studies have shown that if participants believe that placebo creams or drugs blunt pain, then painful stimuli elicit less pain and produce (i) decreased activation of amygdala and pain-related cingulate, insula and thalamic regions in combination, with (ii) increased activation of lateral and medial prefrontal regions related to cognitive control, including rostral cingulate cortex and dorsal and right ventral LPFC [55–57]. Although the precise nature of the cognitive processes mediating placebo effects is not yet clear, placebo-related interactions between prefrontal and appraisal systems are strikingly similar to those supporting reappraisal, suggesting that placebo effects are mediated by the active maintenance of beliefs about placebo compounds that change the way in which stimuli are appraised [57]. The third type of cognitive regulation builds on animal models of emotion regulation (e.g. [6,7]) by examining the ways in which simple stimulus–reinforcer associations are formed and altered. Although the precise systems recruited and the nature of interactions among them have differed across studies and paradigms, instrumental avoidance of aversive stimuli [42], extinction of classically conditioned fear responses [58,59] and reversal of stimu￾lus–reward associations [60–63] have been shown to depend upon interactions between similar cognitive control and emotional appraisal systems. On the control side, findings of activation in ventral lateral and medial PFC, OFC and/or ACC have been observed consistently, supported by neuropsychological studies showing impair￾ments of reversal learning in patients with lesions of ventral and orbital but not dorsolateral PFC [64,65]. On the appraisal side, however, findings have been less consistent. For example, amygdala activation has been reported to either decrease [59] or increase [58] during extinction, and during reversal learning both striatal [60] and amygdala activation have been observed, with separate regions of the amygdala tracking previously as compared with currently reinforced stimuli [62]. These discrepancies across studies might be connected with differences in stimulus characteristics, and also how emotional associations are learned and altered. Critical summary In general, studies of cognitive change have shown con￾sistently that emotional appraisal systems can be modu￾lated by PFC, OFC and cingulate control systems activated either (i) by high-level expectations for beliefs about, and interpretations of, stimuli, or (ii) by learning to associate new emotional responses with stimuli. These findings are strikingly similar to control dynamics observed for ‘cold’ forms of control that involve prefrontal and cingulate systems [8,9]. The consistency of these findings (relative to inconsistent results for studies of attentional control) might be attributable to two factors: the use of stimuli that generate strong emotional responses and the use of regulatory strategies that clearly and strongly engage regulatory processes. That being said, questions remain about when and how specific control and appraisal systems interact, including working out exactly why specific control strategies recruit specific control systems and determining the extent to which different strategies modulate appraisal systems in differ￾ent ways. Towards a functional architecture of cognitive control of emotion The goal of this review was to evaluate recent imaging studies whose results can help to elucidate the functional architecture underlying the cognitive control of emotion. Work using animal models of affective learning and imaging studies of either cognitive control or emotional Review TRENDS in Cognitive Sciences Vol.9 No.5 May 2005 245 www.sciencedirect.com