Search tips
Search criteria 


Logo of scanLink to Publisher's site
Soc Cogn Affect Neurosci. 2008 December; 3(4): 344–352.
Published online 2008 October 28. doi:  10.1093/scan/nsn040
PMCID: PMC2607056

A neural basis for the effect of candidate appearance on election outcomes


Election outcomes correlate with judgments based on a candidate's visual appearance, suggesting that the attributions viewers make based on appearance, so-called thin-slice judgments, influence voting. Yet, it is not known whether the effect of appearance on voting is more strongly influenced by positive or negative attributions, nor which neural mechanisms subserve this effect. We conducted two independent brain imaging studies to address this question. In Study 1, images of losing candidates elicited greater activation in the insula and ventral anterior cingulate than images of winning candidates. Winning candidates elicited no differential activation at all. This suggests that negative attributions from appearance exert greater influence on voting than do positive. We further tested this hypothesis in Study 2 by asking a separate group of participants to judge which unfamiliar candidate in a pair looked more attractive, competent, deceitful and threatening. When negative attribution processing was enhanced (specifically, under judgment of threat), images of losing candidates again elicited greater activation in the insula and ventral anterior cingulate. Together, these findings support the view that negative attributions play a critical role in mediating the effects of appearance on voter decisions, an effect that may be of special importance when other information is absent.

Keywords: social cognition, faces, trait judgment, voting, political science

We rapidly evaluate others based on their appearance, an effect that has been well demonstrated in social psychology (Hassin and Trope, 2000; Todorov and Uleman, 2003; Willis and Todorov, 2006) and not lost on political scientists (Rosenberg, 1986; Rosenberg et al., 1991). In fact, recent behavioral studies have shown that judgments about candidates’ physical appearance correlate with real election outcomes (Todorov et al., 2005; Ballew and Todorov, 2007), suggesting that information derived from visual appearance alone (so-called thin-slice information) affects voting behavior. In particular, work done by Todorov and coworkers showed that competence judgments made about candidates who ran in 2006 gubernatorial elections picked out the winner at significantly above-chance levels (57 ± 6%, mean ± s.d.) (Ballew and Todorov, 2007). This replicated an earlier study, where the average association between an individual viewer's competence judgments and electoral victory ranged from 53 ± 10% to 59 ± 7% for various types of political office (Todorov et al., 2005). Interestingly, participants made these judgments about politicians with whom they were unfamiliar, and after only very brief exposures to the images (100–2000 ms). These findings are quite remarkable given the amount of information about candidates to which a typical voter is exposed. For example, in the 2006 U.S. midterm elections, candidates and their interest groups spent over a billion dollars on advertising to inform voters of their party affiliation, record, policies and personal qualities (CNN, 2006). While some political scientists view such rich information as the primary driver of voter decisions (Popkin, 1991; Prior, 2005), there are other data suggesting that voters make use of much sparser information (Downs, 1957; Alvarez, 1997).

Given that mere appearance seems to influence voting behavior, this raises the question of what psychological and neural processes might mediate this effect. Here we investigate whether the effect of appearance on voting is more strongly influenced by positive or negative attributions (which might be either implicit or explicit; Galdi et al., 2008). Given the association between appearance-based judgments about competence (a putatively positive trait) and electoral victory (Todorov et al., 2005), one might expect that positive attributions from a politician's appearance dominate in influencing decisions by voters. Yet there are several reasons to hypothesize that negative attributions also play a role. First, one would predict that thin-slice attributions play a larger role in influencing voters who have little knowledge of candidates, and there is strong evidence that, when voters know little about a candidate, perceived negative aspects of a candidate exert a stronger influence than positive aspects on voter turnout (Lau and Pomper, 2001; Martin, 2004; Stevens et al., 2008), party defection, the number of self-reported reasons for voting for or against, and predicting overall positive and negative evaluations of a candidate (for a review, see Lau, 1982). Second, voters make implicit ingroup/outgroup distinctions about candidates (Iyengar and Simon, 2000), and negative information dominates the evaluation of outgroup members (Forgas and Fiedler, 1996). To the extent that candidates occupy a default outgroup position for voters lacking relevant information, as suggested by evidence showing that increasing a candidate's facial similarity to a voter significantly increases that voter's support (Bailenson et al., 2006), negative attributions will dominate voter decisions. Third, negative motives play a significant role in voter decisions, particularly in so-called negative voting, in which a vote for Candidate A is really just a vote against Candidate B (Kernell, 1977; Fiorina and Shepsle, 1989).

There is, thus, reason to hypothesize that candidate appearance affects voting through processes that evaluate both positive as well as negative traits, and that may be both implicit and explicit. To test whether positive or negative attributions play a primary role, and to identify the neural regions involved, we conducted two fMRI studies using images of real political candidates. The first study asked participants to vote in the scanner and the second asked an independent group of participants to make both positive and negative trait judgments. In the first study, we would expect the particular brain structures activated during voting to provide insight into the relative influences of the above factors. Specifically, if voting is more influenced by positive attributions, we would expect vote winners to elicit greater activation of brain structures known to be involved in affective processing. Conversely, if negative attributions dominate, vote losers should elicit the greater activation in these structures. Following up on the results from this first study, our second study used explicit trait judgments (threat, deceitfulness, attractiveness, competence) to further investigate the relative contributions made by positive and negative attributions in linking candidate appearance to voter decisions. We expected that similar neural mechanisms would be revealed, further supporting the findings from the first study.


Study 1: simulated voting

In the first study, participants voted in a simulated election. Participants viewed grayscale pictures of 100 pairs of unfamiliar real politicians, one Republican and one Democrat, who competed in the 2006 U.S. midterm elections (Figure 1A; see Methods for details). Participants saw the image of each of the two candidates in sequence, for only 1 s, separated by a blank screen, and after a delay were asked to cast their vote. All the data were collected before the 2006 election, ensuring that participants could not have been influenced in any way by the real election outcomes.

Fig. 1
Trial design of the two studies. Each experiment showed participants grayscale images of real politicians, one at a time, separated by a variable interstimulus interval. For both experiments, stimuli are depicted as they appeared on the screen. (A) Study ...

In order to increase the sensitivity of our fMRI analyses, we limited our analyses to voxels within brain areas already known to be associated with the evaluation of facial appearance and affective processing. We produced region of interest (ROI) masks using the Automated Anatomical Labeling Toolbox for SPM (Tzourio-Mazoyer et al., 2002). These ROIs are as follows: (i) the bilateral temporal lobes, including the fusiform gyrus (associated with gaze and face processing; Kanwisher and Yovel, 2006); (ii) the bilateral caudate (associated with positive evaluation of faces; Kim et al., 2007); (iii) the bilateral putamen (associated with reward-based processing; O’Doherty et al., 2004); (iv) the bilateral gyrus rectus (associated with positive evaluation of faces; Kim et al., 2007); (v) the bilateral orbitofrontal cortex (associated with positive evaluation of faces; Gottfried et al., 2003; Kim et al., 2007); (vi) the bilateral insula [associated with perception of lack of trustworthiness (Winston et al., 2002), and pain perception (Ploghaus et al., 1999; Salomons et al., 2007), but also with positive evaluation of faces (Kim et al., 2007)]; (vii) the bilateral amygdala (associated mostly with negative facial attribution; Winston et al., 2002); and (viii) the bilateral anterior cingulate (associated with social rejection; Eisenberger et al., 2003; Somerville et al., 2006). We report only those clusters surviving FWE correction at P < 0.05, as determined by Monte Carlo simulation using AlphaSim in AFNI (Xiong et al., 1995; Cox, 1996) (see Supporting Information).

We analyzed the fMRI data for this first study by using the voting data provided by the participants in our experiment. We estimated a general linear model in which the appearance of an individual candidate's image was modulated by its vote share in the simulated election, a variable that we refer to as lab vote share (see Methods for details). We found that positive lab vote share elicited no significant activation in any ROIs. In fact, positive lab vote share did not result in any significant activation anywhere, even with a whole-brain analysis. In contrast, negative lab vote share (i.e. election loss) elicited robust, statistically significant activations in bilateral insula [222 voxels (48, −3, −9), Z = 3.81; 179 voxels (−45, 12, 9); Z = 3.96; Supporting Information Table S1; Figure 2A], and bilateral anterior cingulate cortex [239 voxels (3, 33, 9), Z = 3.73; Supporting Information Table S1; Figure 3A]. Thus, these regions are increasingly engaged by viewing candidates with larger margins of electoral loss.

Fig. 2
Activation in the insula/parainsula correlates with election loss in both studies. (A) Study 1: activation during the simulated voting study in bilateral insula [blue circles; left insula (−45, 12, 9) and right insula (48, −3, −9)] ...
Fig. 3
Activation in ventral anterior cingulate cortex correlates with election loss in both studies. (A) Study 1: activation during the simulated voting study in bilateral anterior cingulate [blue circle; (3, 33, 9) and (−9, 21, 24)] was negatively ...

The lack of any significant brain activations elicited by viewing winners in our simulated vote, coupled with the robust activations elicited by viewing losers, suggests that negative attributions from appearance may play a predominant role in mediating how appearance influences voting.

Study 2: candidate trait judgments

In our second study, we sought to further investigate the relative contributions to this effect made by positive and negative attributions (which might be either implicit or explicit), as well as the underlying neural structures we had found in the first study. In order to selectively enhance processing of one attribution over another, we asked participants to make overt judgments in the scanner. To best investigate the effects of candidate appearance alone, we maximized thin-slice conditions with a rapid stimulus presentation. We used a subset of the same images previously shown to elicit an association between real-electoral outcome and judgments of competence (Todorov et al., 2005).

Study participants made social judgments based on the images of real, but unfamiliar, political candidates who ran against one another in the 2000, 2002 and 2004 U.S. Congressional elections, mainly in the House of Representatives. Participants made binary judgments about 30 pairs of candidates, one Republican and one Democrat (in randomized order), on two putatively positive traits, attractiveness (Attr) and competence (Comp), and two putatively negative traits, public deceitfulness (Dect) and personal threat (Thrt), in four separate scanning sessions (see Methods). Each trial consisted of a protocol that has been previously used to investigate face preferences (Kim et al., 2007), in which the two images in a pair of candidates were presented sequentially, for only 30 ms each (unmasked), one alternating with the other, until the participant pushed a button to indicate which of the two faces showed more of the trait being judged (Figure 1B).

As one would expect if participants made meaningful judgments, positive trait judgments were positively correlated (Attr and Comp, r = 0.39, P = 0.002), those for negative traits were positively correlated (Thrt and Dect, r = 0.61, P < 0.0001) and those between Comp and Thrt were negatively correlated (Comp and Thrt, r = −0.39, P = 0.002). No other statistically reliable relationships were seen in the behavioral data alone.

The relationship between our behavioral data regarding competence judgments and real-world electoral outcome was in line with the published findings we reviewed above (Todorov et al., 2005; Ballew and Todorov, 2007): we found that our participants were above chance in judging winners of real elections as more competent [55 ± 9%, t(15) = 2.15, P < 0.05], with the same average individual accuracy as seen in the prior studies. When we examined the majority group competence judgments, comparing candidates who were characterized as competent by a majority of our participants with those who had won elections, the association trended positively [55%, χ2(1) = 1.00, P > 0.1, against an expected 50%] but did not reach statistical significance. This is likely due to our fMRI-scale sample size, as sample sizes of 40 or more are generally required to achieve reliability on this particular measure for competence judgments (Todorov et al., 2005).

Consistent with the robust effect of election loss we found in our first study, a novel behavioral finding from the second study was that the strongest association between election outcome and trait judgments was seen for personal threat judgments. Majority group personal threat judgments corresponded to election loss 65% of the time [χ2(1) = 9.00, P < 0.05], and average individual accuracy was also above chance [57 ± 10%, t(15) = 2.65, P < 0.05]. In fact, the association between majority personal threat judgments and election outcome was stronger than that for competence [χ2(1) = 4.04, P < 0.05] and public deceitfulness [χ2(1) = 9.00, P < 0.05], although not reliably different from attractiveness [χ2(1) = 2.01, P > 0.1]. In addition, only the association between personal threat judgments and election loss survived in a multiple binomial regression model relating all four social judgments to the election outcomes (beta = 1.5, P = 0.03, r2 = 0.1, P = 0.01).

As previously reported (Todorov et al., 2005), the association between election outcome and attractiveness judgments was not statistically different from chance for the average individual [46 ± 10%, t(15) = −1.46, P > 0.1] or for majority group judgments [58% correspondence with election loss, χ2(1) = 2.56, P > 0.1]. Judgments of public deceitfulness across individuals [49 ± 9%, t(15) = −0.34, P > 0.1] and group majority judgments [50%, (χ2(1) = 0, P > 0.1] also did not differ from chance in associating with election outcome.

Our behavioral findings from the second study are, thus, consistent with what we inferred from our first study: there appears to be a primary role for negative attributions in mediating the effect of candidate appearance on election outcome. Interestingly, this may be especially the case for attributions that affect one's personal welfare (i.e. the personal threat judgments viewers made). Given these behavioral findings, and given that our first study revealed no activations for winners, we focus subsequent imaging analyses on the condition of personal threat (see Supporting Information for more detail). This allows us to determine if loser-elicited activations seen in our first study are also seen here for candidates who lost real elections, under conditions where we aimed to most enhance negative attribution (i.e. in judging personal threat from a smiling politician).

We analyzed the fMRI data from our second study by estimating a general linear model in which separate regressors were formed for the first onset of images based on whether those candidates had won or lost in real elections and on whether those candidates were the majority choice with respect to reflecting a particular trait (see Methods for details). We contrasted the parameter estimates obtained in response to the pictures of candidates who had won and those who had lost real elections. We report significant activations surviving FWE-corrected thresholding at P < 0.05 (see Supporting Information).

Again consistent with Study 1, we found no significant activations in our regions of interest for candidates who won real elections (see Supporting Information for complete details). Instead, we found that candidates who lost real elections, compared to those who won, elicited greater activation in the insula/parainsula [18 voxels (45, 0, −15), t = 4.80; Figure 2B and C; Table S2B.] and in the ventral anterior cingulate cortex [24 voxels (15, 39, 0); t = 4.02; (9,45,6), t = 3.90; Figure 3B; Table S2B.]. These locations are within the regions we found for Study 1 and further support the idea that negative attribution is primary in mediating the effects of candidate appearance on voter decisions, and itself is mediated by a network of structures that include the insula/parainsula and ventral anterior cingulate regions. Further evidence for this interpretation comes from the observation that losers elicited an increase in activation in the insula, while winners elicited a decrease (Figure 2C).

To link our two studies directly, we first chose a region in the right insula from Study 1 (with simulated voting) and queried this region [a 10 mm radius sphere centered on the peak voxel (48, −3, −9)] with respect to the contrast effects seen in Study 2 (with real voting). Consistent with Study 1, we found in Study 2 that the contrast of loser > winner, under the condition of threat judgment, resulted in a significantly enhanced activation in this region [t(15) = 1.87, P < 0.05; Figure 2D]. However, we found no significant effect under any of the other judgment conditions (P > 0.1). Similarly, we chose a region in the right anterior cingulate from Study 1 and queried this region [a 10 mm radius sphere centered on the peak voxel (3, 33, 9)] with respect to the contrast effects seen in Study 2. Again, we found that the contrast of loser > winner resulted in a significantly enhanced activation in the region, for threat judgment only [t(15) = 1.98, P < 0.05; Figure 3C; all other conditions, P > 0.1, except for attractiveness, which shows a near-significant effect of winner > loser, t(15) = 1.71, P = 0.054]. Thus, brain regions identified in Study 1 showing a differential sensitivity for images of election losers, compared to winners, show this same sensitivity in Study 2 under conditions in which negative attributions are putatively enhanced, and this time for real election outcomes. This is further evidence that negative attributions are primary in mediating the effect of appearance on voting.

While analyses in Study 2 focused primarily on the threat judgment condition, it is important to note that we observed activations under the other three conditions also (see Supporting Information for details). The direction of the significant election contrasts (loser > winner, or winner > loser) was in line with what one would expect given the valence of the social judgment condition. Thus, both threat and deceit conditions produced activations primarily for loser > winner, whereas attractiveness and competence conditions produced activations primarily for winner > loser. We interpret these data to show that either positive or negative attributions can be enhanced with a sufficiently valenced social judgment context, while negative attributions are primary under the context of voting, particularly when there is a lack of other information about the candidates.


The activation patterns in the insula and anterior cingulate are similar between our two studies, especially considering that they (i) involved different groups of participants, (ii) used different images of political candidates, (iii) used different tasks, and (iv) used different measures of electoral outcome (i.e. simulated and real). In both studies, the activations were elicited by images of candidates who had lost in an election, simulated or real, consistent with the notion that the activations reflect processing in negative attribution. Taken together, the studies suggest that elicitation of negative emotional processes may predominate in mediating the connection between candidate appearance and voting behavior. This interpretation of the data is based on several observations. First, winners of our simulated election elicited no activations in any brain region, while losers elicited robust activation in both the insula and ventral anterior cingulate. While both of these regions have been shown to be sensitive to positive as well as negative aspects of appearance, under various conditions, our interpretation is that here they were responding to negative aspects since they were strongly activated by candidates who lost. This is consistent with literature associating the insula/parainsula (Coan et al., 2006; Lamm et al., 2007) and the ventral anterior cingulate (Eisenberger et al., 2003; Somerville et al., 2006) with the processing of negatively valenced emotions in social situations. The insula is an area known to mediate interoceptive processing and feelings (Craig, 2002), such as sensations of pain or internal discomfort (Singer et al., 2004; Coan et al., 2006), and the right ventral anterior cingulate is implicated in panic attacks (Eser et al., in press), fear (Williams et al., 2006; Bryant et al., 2007) and uncontrollable pain (Salomons et al., 2007). These regions, including the parainsula (Stefanacci and Amaral, 2002), are also known to connect strongly with the amygdala (Amaral et al., 1992), a structure known to play a key role in negative affect associated with faces. Second, judgments of personal threat were most robustly correlated with election outcomes. Finally, under conditions likely to enhance negative attributions (i.e. examining faces for personal threat), we again saw that candidates who won elections elicited no activations in our regions of interest, while those who lost elections elicited greater activation in the insula and ventral anterior cingulate.

These findings are all the more surprising given that nearly all of our politicians were smiling (92% in Study 1; 100% in Study 2) and none showed any overt negative facial expressions. To our knowledge, this is the first demonstration that images with expressions that are overtly positive can nonetheless drive brain activations related to negative evaluations. It is also the first demonstration of a link between voter decisions and brain activations in people making social judgments.

It is important to qualify the findings in this paper in several ways. First, although the weight of our findings suggests a preferential role for negative attributions from candidate appearance, we did see some areas associated with positive emotional processing in Study 2. However, the behavioral judgments under these conditions (competence and attractiveness; cf. Supporting Information) were not as robustly correlated with real electoral outcomes. Thus, we do not wish to rule out that positive attributions may contribute to the effect of appearance on voter decision making, but this effect may be small compared to the effect of negative attributions. It is also possible that positive attributions are simply more variable across individuals than negative attributions, thus diluting their group effect. Nevertheless, our findings support a model in which the contribution made by negative attributions predominates when voters make decisions based on limited information, in line with findings from political science (Lau and Pomper, 2001; Martin, 2004; Stevens et al., 2008).

A fundamental question in politics is the extent to which voters’ decisions are driven by positive motives, which induce them to vote for candidates that they like, or by negative ones, which induce them to vote for the candidate that they do not dislike (i.e. negative voting). As detailed above, there is evidence that negative motives play a role, if not an exclusive one, in voters’ decisions (Kernell, 1977; Lau, 1985; Fiorina and Shepsle, 1989). The results from our two studies suggest that political ‘intangibles’, such as a candidate's appearance, might also work primarily via negative motives. This raises a final question about the nature of those intangibles: what is it about a person's appearance that signals negative traits and influences election loss? Future studies with considerably larger stimulus sets, and with experimental manipulations of facial features, will be required to address this question.


Study 1: simulated vote

Participants. Twenty-four participants (seven female, aged 18–38) participated in the study. Participants had no history of neurological or psychiatric illness and were not on psychotropic medications. Participants had no previous knowledge of any of the political candidates whose images were used in the study, and reported no recognition of any of the politicians. All procedures were approved by the Institutional Review Board at the California Institute of Technology, and participant consent was obtained according to the Declaration of Helsinki.

Stimuli. Stimuli consisted of 200 grayscale images of political candidates who ran in the real 2006 U.S. midterm elections for either the Senate (60 images), the House of Representatives (74 images), or Governor (66 images). The stimuli were collected from the candidates’ campaign Web sites and other Internet sources. An electoral pair consisted of two images of candidates, one Republican and one Democrat, who ran against one another in the real election. Due to the racial and gender composition of the candidates, 70 of the 100 pairs were of male politicians, and 88 of 100 pairs involved two Caucasian politicians. An independent observer classified 92% of the images as ‘smiling’. In 57% of the pairs, both candidates were frontal facing; in the rest at least one was facing to the side. Except for transforming color images into a gray scale, the stimuli were not modified. Images were presented using video goggles (Resonance Technologies Inc.; The stimulus presentation and response recording was controlled by Cogent 2000 (Wellcome Department of Imaging Neuroscience;

The study was conducted in the month before the 2006 election. An effort was made to avoid pairs in which one of the candidates (e.g. Hillary Clinton) had national prominence or participated in a California election, and familiarity ratings collected from all of the participants after the scanning task verified the stimuli were unfamiliar. On a scale of 1 (completely unfamiliar) to 7 (very familiar), the mean rating was 1.65 (s.d. = 1.03) for the Democratic candidates and 1.62 (s.d. = 0.98) for the Republican candidates.

Procedure. Participants were instructed that they would be asked to vote for real political candidates who were running against each other in the upcoming midterm election. In particular, they were asked to decide who they would be more likely to vote for given that the only information that they had about the politicians were their portraits.

Each trial consisted of three events (Figure 1A). First, a picture of one of the candidates was centrally presented for 1 s. Second, after a blank screen of length 1–10 s (uniform distribution), the picture of the other candidate in the pair was presented for 1 s. Third, after another blank screen of length 1–10 s, the pictures of both candidates were presented side by side. At this point, participants were asked to cast their vote by pressing either the left or the right button. They had a maximum of 2 s to make a decision. Participants made a response within this time frame in 100% of the trials. Trials were separated by a 1–10 s blank screen. The order of presentation of the candidates as well as their position in the final screen was fully randomized between participants.

Neuroimaging data acquisition. Imaging data were collected on a Siemens 3.0-T Trio MRI scanner. Whole-brain, high-resolution (1 × 1 × 1 mm3) T1-weighted images were collected for each participant and coregistered with the mean functional, T2*-weighted images. For the fMRI data, we collected gradient-echo T2*-weighted echoplanar images with blood oxygenation level-dependent contrast using an interleaved, ascending image sequence (parameters: TR = 2.75 s, TE = 30 ms, field of view = 192 mm, 44 slices at 3 mm thick, 64 × 64 voxels, resulting in a voxel size of 3 × 3 × 3 mm3). We used a tilted acquisition, at 30° relative to the anterior commissure–posterior commissure line in order to achieve good signal in both the orbitofrontal cortex and subcortical regions. In addition, we used an eight-channel phased array coil that yields a 40% signal increase in OFC over the standard head coil.

Neuroimaging data analysis. All data analysis was performed using SPM5. We discarded the first five EPI images to allow for signal equilibration, applied slice-timing correction (centered at TR/2), realigned all volumes, spatially normalized a standard EPI template with a resampled isotropic voxel size of 3 × 3 × 3 mm3, spatially smoothed the data using a Gaussian kernel with FWHM of 8 mm and applied intensity normalization and high-pass temporal filtering (filter width 128 s).

We constructed regressors corresponding to the onsets of the images and additionally used six motion regressors. We used the lab vote share as a parametric modulator of the image onset regressor. The lab vote share is the fraction of our participant group who voted for the politician, and the negative lab vote share is simply this fraction subtracted from 1. In our first analysis, we used the positive lab vote share as a parametric modulator, and in our second analysis, we used the negative lab vote share as a parametric modulator. We applied linear contrasts within each participant, and took this to the random effects level using t-tests. To select statistically significant clusters, we applied an FWE-corrected threshold of P < 0.05 (see Supporting Information for details).

Study 2: trait judgments

Participants. Twenty-two Caucasian women (aged 20–35) participated in the study (note: none of these individuals participated in Study 1). At the time of the experiment, they were registered to vote and had voted in one or more of the following national elections: 2000, 2002 and/or 2004. Participants had no history of neurological or psychiatric illness and were not on psychoactive medications. Participants had no prior knowledge of any of the political candidates whose images were used in the study and reported no recognition of any of the politicians. Neuroimaging data from six participants were rejected due to excessive motion. The behavioral data of the 16 participants included in the study did not differ significantly from those of six participants rejected for excessive motion. All procedures were approved by the Institutional Review Board at the California Institute of Technology.

Stimuli. Stimuli consisted of 60 grayscale images of smiling political candidates who ran in real U.S. elections for the House of Representatives or Senate in either 2000, 2002 or 2004 (30 pairs of opponents). The stimuli comprised a subset of those used in the 2005 study by Todorov et al. (2005), and were selected by three of the experimenters so that both images in an electoral pair (i) were frontal facing, (ii) were of the same gender and ethnicity and (iii) had clear, approximately central presentation of faces that were of approximately the same size. An electoral pair consisted of two images of candidates, one Republican and one Democrat, who actually ran against one another in a real election. Due to the racial/ethnic and gender composition of the original image library, all stimuli were of Caucasian politicians, and 8 of the 30 pairs were of female politicians. Stimuli were preprocessed to normalize overall image intensity while maintaining good image quality, across all 60 images. All images were presented centrally, via an LCD projector and a rear-projection screen, onto a mirror attached to the MRI head coil, approximately 10 inches from a participant's eyes. Stimuli subtended approximately 8° of visual angle. Stimulus control and response recording used the Psychophysics Toolbox v2.54 (Brainard, 1997; Pelli, 1997) in Matlab (the Mathworks, Natick, MA). A pilot behavioral study confirmed that the social judgments made about our selected stimuli were representative of the entire set of face stimuli from which they were drawn (the entire set used by Todorov et al. (2005).

Procedure. Participants were instructed that they would be asked to make judgments about real political candidates who ran against one another in real elections. They were told that they would only be given the images of the politicians to inform their judgments. Image order was counterbalanced across participants. Participants made judgments about candidates’ attractiveness (Attr), competence (Comp), public deceitfulness (Dect) and personal threat (Thrt) in four separate scanning sessions. For threat judgments, participants were asked which candidate in a pair looked more likely to act in a physically threatening manner toward them (i.e. personal threat). For attractiveness judgments, participants were asked which candidate looked more physically attractive to them (i.e. personal attractiveness). For competence judgments, participants were asked which candidate looked more competent to hold national congressional office. For public deceitfulness, participants were asked which candidate looked more likely to lie to the voters (i.e. public deceit). Each session took approximately 9 min to complete. Of 16 participants, six made personal threat decisions prior to making other decisions, while the remainder made personal threat decisions after making decisions about other attributes. There were no effects of block order.

Each trial in a decision block consisted of the sequential presentation of two images in an electoral pair, image A then image B, until a participant entered a decision about the pair via a button press (Figure 1B). This follows a protocol we have used successfully in prior studies of face preference (Kim et al., 2007). An A/B cycle on a given trial proceeded as follows: (i) central presentation of a fixation rectangle that surrounded the area in which an image was to appear; (ii) after 4–6 s, a 30 ms display of image A surrounded by the fixation box, accompanied by a small black dot in the lower left corner (indicating that this was image A); and (iii) after 3–4 s, a 30 ms display of image B surrounded by the fixation box, accompanied by a small black dot in the lower right corner (indicating that this was image B). Cycles were separated by 4–6 s and continued until a participant entered a button press or until 30 s had elapsed, whichever came first (no participant ever took the 30 s). Participants were asked to attend overtly to the space inside the rectangle in preparation for a candidate image. We used eyetracking (MRI-compatible Long-Range Optics Model, Applied Science Laboratories, Bedford, MA) to ensure that participants were looking at the stimuli. Trials that required just one A/B cycle for participants to make their judgment are referred to as one-cycle trials, those that required two cycles are two-cycle trials, and so on.

Behavioral data analysis. Participants primarily took two cycles to decide [76 ± 2% of trials were two-cycle trials (mean ± s.d. across 16 participants, all judgment conditions)]. There was not enough data from other types of trials to conduct a full random effects analysis, so only data from two-cycle trials were examined.

Correlations between social judgments. For each candidate and for each judgment, we first calculated the judgment share, which was just the proportion of participants who decided that candidate was more threatening, attractive, competent or deceitful. Using these values, we calculated Pearson correlation coefficients between the different social judgments that participants made.

Correspondence between social judgments and electoral outcome. We conducted two types of analyses, as done for previous studies (Todorov et al., 2005). To determine average individual association between judgments and real electoral outcome, we calculated the percent agreement between each participant's social judgments and electoral outcome, and then averaged across participants. A simple t-test allowed us to determine whether the mean individual association differed significantly from chance (50%). To determine the association between majority group judgment and real electoral outcome, we first counted which politician in a pair had the most participants naming them as attractive, competent, deceitful and threatening. We then calculated the percent agreement between this social judgment outcome and the signed electoral outcome (for instance, agreement between who was judged to look more competent and real-world election winning; or who was judged to look more threatening and real-world election losing). A χ2 test finally determined whether majority associations were different from chance (50%).

Neuroimaging data acquisition. This was as for Study 1, except that TR = 2 s, and each EPI image had 34 slices at 4 mm thick.

Neuroimaging data analysis. All data preprocessing was done in SPM2 and analyses were conducted with SPM5 (Wellcome Department of Imaging Neuroscience, London, UK; (Friston et al., 2005). Preprocessing was identical to Study 1, except that we used a spatial smoothing filter of 6 mm FWHM.

We combined all four judgment conditions in one design matrix and proceeded in three steps. In the first step, we estimated a general linear model with AR(1). Second, we calculated first-level contrasts of group chosen (GC) vs group unchosen (GU) for each of the four judgment choices, and real-world election winner (RW) vs loser (RL), again for each of the four judgment conditions, as well as interactions between conditions. Finally, for each of these first-level contrasts, we calculated a second-level random effects contrast using a one sample t-test. To select statistically significant clusters, we applied an FWE-corrected threshold of P < 0.05 (see Supporting Information for details). Our analysis focused on the first cycle of the 2-cycle trials. We did this for two reasons: to maintain consistency with our first study and to maximize thin-slice conditions.

To examine second-level effects in Study 2 using activation ROIs from Study 1, we used the RFXPLOT toolbox for SPM5 ( We formed a mask to select all voxels inside a sphere (10 mm radius) for a given ROI from Study 1 [right insula: around (48, −3, −9); right anterior cingulate: around (3, 33, 9)]. We applied this mask to the individual first-level contrast images from participants in Study 2, selecting all voxels inside the mask (both sub- and suprathreshold) and calculated a mean ROI contrast for each participant. We used t-tests to determine whether the group mean contrasts were significantly different from zero.


Supplementary data are available at SCAN online.


The authors wish to thank Jan Glaescher for helpful discussions and for the use of rfxplot. This research was financed in part by a grant from the Gordon and Betty Moore Foundation, and with support from the National Science Foundation (grant number SES-0134618 to A.R.) and from the National Institutes of Health (to R.A. and M.L.S.).


  • Alvarez RM. Information and Elections. Ann Arbor, MI: University of Michigan; 1997. pp. 109–56.
  • Amaral DG, Price JL, Pitkanen A, Carmichael ST. The anatomical organization of the primate amygdaloid complex. In: Aggleton JP, editor. The Amygdala: Neurobiological Aspects of Emotion, Memory, and Mental Dysfunction. New York: Wiley-Liss; 1992. pp. 1–67.
  • Bailenson JN, Garland P, Iyengar S, Yee N. Transformed facial similarity as a political cue: a preliminary investigation. Political Psychology. 2006;27:373–85.
  • Ballew CC, Todorov A. Predicting political elections from rapid and unreflective face decisions. Proceedings of the National Academy of Science. 2007;104:17948–53. [PubMed]
  • Brainard DH. The Psychophysics Toolbox. Spatial Vision. 1997;10:433–6. [PubMed]
  • Bryant RA, Felmingham K, Kemp A, et al. Amygdala and ventral anterior cingulate activation predicts treatment response to cognitive behaviour therapy for post-traumatic stress disorder. Psychological Medicine. 2007;38:556–61. [PubMed]
  • CNN. Record amount spent on campaign ads in 2006. 2006.
  • Coan JA, Schaefer HS, Davidson RJ. Lending a hand: social regulation of the neural response to threat. Psychological Science. 2006;17:1032–9. [PubMed]
  • Cox RW. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Computers and Biomedical Research. 1996;29:162–173. [PubMed]
  • Craig AD. How do you feel? Interoception: the sense of the physiological condition of the body. Natural Reviews. Neuroscience. 2002;3:655–66. [PubMed]
  • Downs A. An Economic Theory of Democracy. New York: Harper; 1957. pp. 207–59.
  • Eisenberger NI, Lieberman MD, Williams KD. Does rejection hurt? An FMRI study of social exclusion. Science. 2003;302:290–2. [PubMed]
  • Eser D, Leicht G, Lutz J, et al. Human Brain Mapping. Functional neuroanatomy of CCK-4-induced panic attacks in healthy volunteers. [Epub ahead of print; 19 December 2007] (in press)
  • Fiorina MP, Shepsle KA. Is negative voting an artifact? American Journal of Political Science. 1989;33:423–39.
  • Forgas JP, Fiedler K. Us and them: mood effects on intergroup discrimination. Journal of Personality and Social Psychology. 1996;70:28–40.
  • Friston KJ, Williams SC, Howard R, Frackowiak RSJ, Turner R. 2005. Online resource, Wellcome Trust Centre for Neuroimaging.
  • Galdi S, Arcuri L, Gawronski B. Automatic mental associations predict future choices of undecided decision-makers. Science. 2008;321:1100–02. [PubMed]
  • Gottfried JA, O'Doherty J, Dolan RJ. Encoding predictive reward value in human amygdala and orbitofrontal cortex. Science. 2003;301:1104–07. [PubMed]
  • Hassin R, Trope Y. Facing faces: studies on the cognitive aspects of physiognomy. Journal of Personality and Social Psychology. 2000;78:837–52. [PubMed]
  • Iyengar S, Simon AF. New perspectives and evidence on political communication and campaign effects. Annual Review of Psychology. 2000;51:149–69. [PubMed]
  • Kanwisher N, Yovel G. The fusiform face area: a cortical region specialized for the perception of faces. Philosophical Transactions of the Royal Society of London B: Biological Sciences. 2006;361:2109–28. [PMC free article] [PubMed]
  • Kernell S. Presidential popularity and negative voting. American Political Science Review. 1977;71:44–66.
  • Kim H, Adolphs R, O'Doherty JP, Shimojo S. Temporal isolation of neural processes underlying face preference decisions. Proceedings of the National Academy of Sciences of the USA. 2007;104:18253–58. [PubMed]
  • Lamm C, Batson CD, Decety J. The neural substrate of human empathy: effects of perspective-taking and cognitive appraisal. Journal of Cognitive Neuroscience. 2007;19:42–58. [PubMed]
  • Lau RR. Negativity in political perception. Political Behavior. 1982;4:353–78.
  • Lau RR. Two explanations for negativity effects in political behavior. American Journal of Political Science. 1985;29:119–138.
  • Lau RR, Pomper GM. Effects of negative campaigning on turnout in U.S. Senate elections, 1988-1998. Journal of Politics. 2001;63:804–19.
  • Martin PS. Inside the black box of negative campaign effects: three reasons why negative campaigns mobilize. Political Psychology. 2004;25:545–562.
  • O'Doherty J, Dayan P, Schultz J, Deichmann R, Friston KJ, Dolan RJ. Dissociable roles of ventral and dorsal striatum in instrumental conditioning. Science. 2004;304:452–4. [PubMed]
  • Pelli DG. The VideoToolbox software for visual psychophysics: transforming numbers into movies. Spatial Vision. 1997;10:437–42. [PubMed]
  • Ploghaus A, Tracey I, Gati JS, et al. Dissociating pain from its anticipation in the human brain. Science. 1999;284:1979–81. [PubMed]
  • Popkin S. The Reasoning Voter: Communication and Persuasions in Presidential Campaigns. Chicago: University of Chicago Press; 1991. pp. 7–43.
  • Prior M. American Journal of Political Science. Vol. 49. 2005. News vs. entertainment: how increasing media choice widens gaps in political knowledge and turnout. p. 577.
  • Rosenberg SW. The image and the vote: the effect of candidate presentation on voter preferences. American Journal of Political Science. 1986;30:108–27.
  • Rosenberg SW, Kahn S, Tran T, Le M.-T. Creating a political image: shaping appearance and manipulating the vote. Political Behavior. 1991;13:345–67.
  • Salomons TV, Johnstone T, Backonja MM, Shackman AJ, Davidson RJ. Individual differences in the effects of perceived controllability on pain perception: critical role of the prefrontal cortex. Journal of Cognitive Neuroscience. 2007;19:993–1003. [PubMed]
  • Singer T, Seymour B, O'Doherty J, Kaube H, Dolan RJ, Frith CD. Empathy for pain involves the affective but not sensory components of pain. Science. 2004;303:1157–62. [PubMed]
  • Somerville LH, Heatherton TF, Kelley WM. Anterior cingulate cortex responds differentially to expectancy violation and social rejection. Nature Neuroscience. 2006;9:1007–8. [PubMed]
  • Stefanacci L, Amaral DG. Some observations on cortical inputs to the macaque monkey amygdala: an anterograde tracing study. Journal of Comparative Neurology. 2002;451:301–23. [PubMed]
  • Stevens D, Sullivan J, Allen B, Alger D. What's good for the goose is bad for the gander: negative political advertising, partisanship, and turnout. Journal of Politics. 2008;70:527–41.
  • Todorov A, Mandisodza AN, Goren A, Hall CC. Inferences of competence from faces predict election outcomes. Science. 2005;308:1623–26. [PubMed]
  • Todorov A, Uleman JS. The efficiency of binding spontaneous trait inferences to actors' faces. Journal of Experimental Social Psychology. 2003;39:549–62.
  • Tzourio-Mazoyer N, Landeau B, Papathanassiou D, et al. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage. 2002;15:273–89. [PubMed]
  • Williams LM, Liddell BJ, Kemp AH, et al. Amygdala-prefrontal dissociation of subliminal and supraliminal fear. Human Brain Mapping. 2006;27:652–61. [PubMed]
  • Willis J, Todorov A. First impressions: making up your mind after a 100-ms exposure to a face. Psychological Science. 2006;17:592–8. [PubMed]
  • Winston JS, Strange BA, O'Doherty J, Dolan RJ. Automatic and intentional brain responses during evaluation of trustworthiness of faces. Nature Neuroscience. 2002;5:277–83. [PubMed]
  • Xiong J, Gao J.-H, Lancaster JL, Fox PT. Clustered analysis for functional MRI activation studies of the human brain. Human Brain Mapping. 1995;3:287–301.

Articles from Social Cognitive and Affective Neuroscience are provided here courtesy of Oxford University Press