FUNCTIONAL NEUROANATOMY OF EMOTION 343 areas may represent top-down modulatory effects on 2000).Thus,the amygdala may be more responsible for the visual processing stream.particularly in relation to processing of externally-cued perceptual emotional the amygdala.which was also found to be specinc to 1999),and less visual in on.Anat 100 7:Wh my8gaalaext maral t 1998:Rauch:Shin tal 2000:Dan al.2000).Another potential issue is that many mem- of this top-down proc essing may be the selective atten- ories recalled are likely to involve more than one emo- tion that drives the modulation of visual processing tion (George et al., )and thereby may reduce (Corbetta et al,1993).Anatomical studies has shown tistical power letect sub e amyg that the primate amygdala receives substantial input dal Aggle 1980 are on et a by recalling pe ersonal events in a controued environ ment like a PET scanner is clearly phenomenologically function (ton down propessing) the different from the spontaneous and direct experience eions after finding hypothesized that extrastriate regions have functional al relationship betv nd h behavio uestoAlshn veen chan duction of emotions by recall or imagery may be more 1996 35 Visual indue and the an dala The cult ng sin bepora 9aaem28pcoye96P 1996). =12.93.P CAVEATS AND LIMITATIONS OF REVIEW has a special Meta-analysis methodologies in g ral and this re view in particular.have a n r of limitations worth e Bec the occinital c discussing.First.in the search for potential associa- rtex. dala is positioned to modulate visual input.pased on tions between emotion and brain activation,authors emotional significance.at a variety of levels along this may be biased in reporting certain activations. visual processing stream.Hence.as discussed earlier. ala the r activations may be primarily for pro- n ir on wit tivation studies of emotion.studies may under-report 2001 th ser aad地ar mine sue nding such associa- ally stimuli may be in the serviced of assigr a meta-an IS S emo tional significance to sensory. particularly visual, y in exa inputs (Simpson et al., 2000)0r that because of its only in the ev ocative stimulus.induction method. and modulatory influences. the amygdala may facilitate emotional tasks employed but also in their statistical cant al mn power,in the criteria used in defining significant re- and in metho than h may al e y in the environme the a dala is well no sitioned to alert us to visual threat following percep- s throt the ub- tion by the occipital cortex. traction method,which produced all the activation foci Interestingly.very few recall-driven emotion activa- examined in this review.The subtraction method is tion studi es engaed the mth (at 7%frequency) nited because it does not identify all the regions that me auth have hypou ed in certain emotion/ all and refe nditionareas may represent top-down modulatory effects on the visual processing stream, particularly in relation to the amygdala, which was also found to be specific to visual induction. Anatomically, projections from limbic regions like the amygdala extend to all the processing steps in the ventral visual stream, including the pri￾mary visual cortex (Amaral et al., 1984). A component of this top-down processing may be the selective atten￾tion that drives the modulation of visual processing (Corbetta et al., 1993). Anatomical studies has shown that the primate amygdala receives substantial input from temporal visual-association areas (Iwai et al., 1987, Aggleton et al., 1980). Both Morris et al. (1998a) and Sprengelmeyer et al. (1998) proposed a neuro￾modulatory function (top-down processing) of the amygdala on extrastriate cortical regions after finding an inverse correlation of amygdala and fusiform activ￾ity in response to fearful faces. Morris et al. (1998) hypothesized that extrastriate regions have functional interactions with the amygdala, evidenced by other facial emotion processing studies (George et al., 1993; Adolphs et al., 1996). 3.5 Visual induction and the amygdala. The visual induction method also preferentially activated the amygdala, over recall and auditory-generated emotion. Fifty percent of visual induction studies reported acti￾vation of the amygdala, compared to 7% and 0% of recall and auditory inductions, respectively (X2 12.93, P 0.002). From this, one can propose that the amygdala has a specialized role in processing visually relevant emotional cues, signalling fear, aversiveness, or sa￾lience. Because of its projections to virtually all levels of visual processing in the occipital cortex, the amyg￾dala is positioned to modulate visual input, based on emotional significance, at a variety of levels along this visual processing stream. Hence, as discussed earlier, the amygdalar activations may be primarily for pro￾cessing affective information within the visual process￾ing stream in service of imparting danger warnings (Davis and Whalen, 2001). Similarly, others have sug￾gested that the amygdalar response to visual emotion￾ally stimuli may be in the serviced of assigning emo￾tional significance to sensory, particularly visual, inputs (Simpson et al., 2000) or that because of its modulatory influences, the amygdala may facilitate processing of salient, adaptively significant visual in￾puts (Morris et al., 1998). Given that humans rely on vision, more than hearing or olfaction, to evaluate changes in the environment, the amygdala is well po￾sitioned to alert us to visual threat following percep￾tion by the occipital cortex. Interestingly, very few recall-driven emotion activa￾tion studies engaged the amygdala (at 7% frequency). Some authors have hypothesized that the amygdala is less engaged by recalled emotions than for visually￾evocative emotions (Reiman et al., 1997; Damasio et al., 2000). Thus, the amygdala may be more responsible for processing of externally-cued perceptual emotional stimuli (Reiman et al., 1997; Teasdale 1999), and less involved in internally generated recollection or imag￾ery of those stimuli (Reiman et al., 1997; Whalen et al., 1998; Rauch et al., 1999; Shin et al., 2000; Damasio et al., 2000). Another potential issue is that many mem￾ories recalled are likely to involve more than one emo￾tion (George et al., 1995), and thereby may reduce statistical power to detect subtle changes in the amyg￾dalar response, which may also be selective for certain emotions. The relationship of self-generated emotion by recalling personal events in a controlled environ￾ment like a PET scanner is clearly phenomenologically different from the spontaneous and direct experience that emotion in a natural setting (Mayberg et al., 1997). Another potential explanation of the infre￾quency of amygdala activation in emotional recall may be due to the temporal relationship between changes in blood flow and the behavior in question. Also, the in￾duction of emotions by recall or imagery may be more time-consuming than more direct perceptual induc￾tion, making differences in amygdalar responses diffi- cult to detect in PET studies due to limited temporal resolution, or the time-consuming process may be sus￾ceptible to known habituation effects (Breiter et al., 1996). CAVEATS AND LIMITATIONS OF REVIEW Meta-analysis methodologies in general, and this re￾view in particular, have a number of limitations worth discussing. First, in the search for potential associa￾tions between emotion and brain activation, authors may be biased in reporting certain activations. For example, because the associations between fear and the amygdala and between sadness and the subcallosal anterior cingulate cortex have gained attention in ac￾tivation studies of emotion, studies may under-report failures to find the expected activations, and conse￾quently inflate the probability of finding such associa￾tions in a meta-analysis. Second, there is substantial difficulty in examining results from non-uniform ex￾periments. Emotion activation studies differ widely not only in the evocative stimulus, induction method, and emotional tasks employed but also in their statistical power, in the criteria used in defining significant re￾sults, and in methods of preprocessing imaging data. Such factors may affect the results considerably and add to the diverse activation patterns throughout the entire brain. Third, there are limitations of the sub￾traction method, which produced all the activation foci examined in this review. The subtraction method is limited because it does not identify all the regions that are involved in certain emotion/emotional task but only those that show a significant difference between the target and reference condition. Consequently, the re￾FUNCTIONAL NEUROANATOMY OF EMOTION 343