Brain correlates of aesthetic expertise: A parametric fMRI study

Brain and Cognition 69 (2009) 306–315

Contents lists available at ScienceDirect

Brain and Cognition

journal homepage: www.elsevier .com/locate /b&c

Brain correlates of aesthetic expertise: A parametric fMRI study

Ulrich Kirk a,b,*, Martin Skov a, Mark Schram Christensen a,c, Niels Nygaard d

a Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital, Hvidovre, Kettegaard Allé 30, DK-2650 Hvidovre, Denmarkb Wellcome Laboratory of Neurobiology, Anatomy Department, University College London, Darwin Building, Gower Street, London WC1E 6BT, UKc Department of Exercise and Sport Sciences, University of Copenhagen, The Panum Institute, Blegdamsvej 3, DK-2200 Copenhagen N, Denmarkd Institute for Architecture and Aesthetics, Aarhus School of Architecture, Norreport 20, DK-8000 Aarhus C, Denmark

a r t i c l e i n f o

Article history:Accepted 1 August 2008Available online 9 September 2008

Keywords:NeuroaestheticsExpertise, orbitofrontal cortexSubcallosal cingulate gyrusArchitectureFaces

0278-2626/$ - see front matter � 2008 Elsevier Inc. Adoi:10.1016/j.bandc.2008.08.004

* Corresponding author. Address: Danish Researchnance, Copenhagen University Hospital, Hvidovre, KHvidovre, Denmark.

E-mail address: [email protected] (U. Kirk).

a b s t r a c t

Several studies have demonstrated that acquired expertise influences aesthetic judgments. In this para-digm we used functional magnetic resonance imaging (fMRI) to study aesthetic judgments of visuallypresented architectural stimuli and control-stimuli (faces) for a group of architects and a group ofnon-architects. This design allowed us to test whether level of expertise modulates neural activity inbrain areas associated with either perceptual processing, memory, or reward processing. We show thatexperts and non-experts recruit bilateral medial orbitofrontal cortex (OFC) and subcallosal cingulategyrus differentially during aesthetic judgment, even in the absence of behavioural aesthetic rating differ-ences between experts and non-experts. By contrast, activity in nucleus accumbens (NAcc) exhibits a dif-ferential response profile compared to OFC and subcallosal cingulate gyrus, suggesting a dissociable rolebetween these regions in the reward processing of expertise. Finally, categorical responses (irrespectiveof aesthetic ratings) resulted in expertise effects in memory-related areas such as hippocampus and pre-cuneus. These results highlight the fact that expertise not only modulates cognitive processing, but alsomodulates the response in reward related brain areas.

� 2008 Elsevier Inc. All rights reserved.

1. Introduction

In psychological models of aesthetic experience it is generallyassumed that art-related expertise influences subjects’ preferencefor works of art (Leder, Belke, Oeberst, & Augustin, 2004). Indeed,a substantial number of behavioural studies have confirmed thatlevel of expertise modulates the aesthetic evaluation of art objects(Eysenck & Castle, 1970; Gordon, 1951/1952, 1956; Hekkert, Peper,& van Wieringen, 1994, Hekkert & van Wieringen, 1996a, 1996b;O’Hare, 1976; Schmidt, McLaughlin, & Leighten, 1989). It is there-fore likely that art experts use different neural processes for deter-mining aesthetic evaluation than non-experts. The question wewish to raise here is whether this putative difference in aestheticevaluation can be detected as a difference in neural activitythrough the use of functional magnetic resonance imaging (fMRI).

It has been shown by imaging experiments that acquired exper-tise is associated with changes in brain structures underlying per-ceptual and memory processes, even on a macro-anatomical scale.For example, in a study using voxel-based morphometry analysis,Maguire and colleagues (2000) found that grey matter volume inthe posterior hippocampus of London taxi drivers is greater than

ll rights reserved.

Centre for Magnetic Reso-ettegaard Allé 30, DK-2650

in age-matched controls, and that the size of this increase corre-lates positively with time spent taxi driving. Furthermore, severalexperiments have demonstrated that musicians, after years ofplaying, respond differently to musical inputs as compared tonon-musicians (for a review, see Schlaug, 2003). For example, ina recent fMRI study, Bangert and colleagues (2006) compared brainactivity in groups of musicians and non-musicians as they pas-sively listened to a piano sequence and found elevated activity inthe musicians in regions of the temporal lobe associated with audi-tory processing, and in frontal regions associated with motorcontrol.

Several neuroimaging studies have investigated cortical areasthat are recruited when subjects make aesthetic evaluations froma variety of stimulus modalities such as paintings (Cela-Condeet al., 2004; Kawabata & Zeki, 2004; Vartanian& Goel, 2004), music(Blood & Zatorre 2001; Blood, Zatorre, Bermudez, & Evans, 1999;Koelsch, Fritz, von Cramon, Müller, & Friederici, 2006; Brown, Mar-tinez, & Parsons, 2004; Menon& Levitin, 2005), faces (Aharon et al.,2001; Nakamura et al., 1998; O’Doherty et al. 2003; Winston,O’Doherty, Kilner, Perrett, & Dolan, 2007) and geometrical figures(Jacobsen, Schubotz, Höfel, & Cramon, 2006). Taken together, thesestudies suggest that the computation of aesthetic preferences forobjects predominantly relies on the activity of cortical and subcor-tical areas implicated in the processing of reward; especially stria-tum, orbitofrontal cortex (OFC) and anterior cingulate cortex (ACC)(for a review, see Skov, in press.) It is therefore important to inves-

U. Kirk et al. / Brain and Cognition 69 (2009) 306–315 307

tigate whether expertise influences aesthetic evaluation throughthe modulation of neural activity in these areas. Since the medialOFC is not only found to correlate with subjective hedonic valuein most of the studies mentioned above, but have also been dem-onstrated to be involved in coding stimulus value of a variety ofother sensory modalities, including taste (O’Doherty et al., 2001;Small, Zatorre, Dagher, Evans, & Jones-Gotman, 2001; Small et al.,2003), olfactory (Anderson et al., 2003; Gottfried, Deichmann, Win-ston, & Dolan, 2002; Rolls, Kringelbach, & de Araujo, 2003), andsomatosensory (Rolls, O’Doherty et al., 2003), we hypothesized thatthis region would reflect a modulation of aesthetic assessmentaccording to level of expertise.

To accomplish this experimental aim, naïve subjects (i.e. sub-jects professing to have no great interest or expertise in art orarchitecture) and expert subjects (i.e. graduate students in archi-tecture and professional architects) were asked to rate the aes-thetic value of a series of images containing both buildings andfaces during an event-related fMRI paradigm (see Fig. 1). Wehypothesized that the expert-specific conditions (i.e. buildingimages) would significantly affect both aesthetic ratings and neuralactivity differentially in the two groups. Since earlier psychometricstudies have found that people in different cultures, and of bothsexes, tend to agree as to which faces are attractive (Langloiset al., 2000), we predicted the two groups’ aesthetic ratings andneural processing would not differentiate for face images.

2. Experimental methods

2.1. Subjects

A total of 24 healthy volunteers (11 experts/13 non-experts; 6female experts/7 female non-experts; experts mean age: 30.8years; age range 26–42 years; non-experts mean age: 27.2 years;age range 22–32 years; all subjects were right-handed) werescanned. We excluded two subjects (both male non-experts) fromthe analysis for clinical reasons. The experts were recruited fromarchitectural offices and schools where they were graduate orpost-graduate students. Non-experts were all undergraduate or

Fig. 1. Experimental paradigm. A fixation cross was shown for 1000 ms followed bystimulus presentation with a duration of 3000 ms in which subjects were instructedto indicate the level of aesthetic appeal by means of button-press on a scale from 5(highest appeal) to 1 (lowest appeal). Examples of stimuli used in the scanningsessions are displayed.

graduate students with no formal education in any art-relatedfield. Written informed consent was obtained from all subjectsand ethical approval (KF-01-131/03) was obtained before theexperiment. All subjects had normal or corrected-to-normal vision,and none had a history of neurological or psychiatric disorders.

2.2. Stimulus set

Visual achromatic stimuli belonging to two categories, build-ings and faces, were used as stimulus material. One hundred andsixty-eight building stimuli were selected from various online re-sources. The surrounding of the building image was shaded so thatthe building was in focus for each stimulus. This was accomplishedin Photoshop (version 7.0, Adobe, USA). Any image noticeably dis-torted (e.g., proportion and illumination) by this process was ex-cluded from the stimulus pool. Building stimuli were presentedwith a resolution of 600 pixels in height and varying width witha maximum of 1024 pixels. Prior to scanning the building stimuliwere exposed to an aesthetic judgment scale in a behavioural pilotstudy by a separate cohort of subjects (7 experts/6 non-experts; 3female experts/3 female non-experts; experts mean age 34.3 years;age range 27–44 years; non-experts mean age 29.2 years; agerange 27–30 years). Level of appeal was measured using an aes-thetic rating scale from 1 to 5, where 1 was defined as ‘‘very unap-pealing” and 5 as ‘‘very appealing”. The stimuli conformed to abalanced distribution in the frequency of each rating bin betweenexperts and non-experts. To investigate whether there were differ-ences between the two groups for rating-specific stimuli, i.e.whether there was an image-wise difference between expertsand non-experts for buildings, further analyses were applied. Thebuilding stimuli were selected according to two sub-classes: a for-mal/stylistic sub-classification (‘modernist’ and ‘non-modernist’architecture) and a typological sub-classification (‘private’ and‘public’ architecture). This was done in order to further controlfor a potential skewed preference distribution between the groups;for instance, experts might all prefer modernist and non-expertsmight all prefer non-modernist buildings. However, this poten-tially confounding effect did not amount to significant differencesbetween stimuli sub-classes across groups in subsequent statisticalanalyses (F(7,40) = 1.78; p > .1).

The face database was provided by the Max-Planck Institute forBiological Cybernetics in Tuebingen, Germany. 168 face stimuliwere selected; half of the stimuli were female faces. Stimuli wererated by a separate group of subjects (n = 10/4 females; meanage 28.4 years; age range 26–30 years) for level of appeal in abehavioural pilot study prior to scanning. Level of appeal was ratedusing the same aesthetic rating scale as described above. One hun-dred and sixty-eight faces were selected from the high, middle andlow ends of the appeal ratings in order to obtain a balanced distri-bution. The face stimuli were masked in order to remove hair andwere adjusted to be of equal size and luminance by using Photo-shop (version 7.0, Adobe, USA). The faces were centred in a588 � 600 pixel black background and presented at a screen reso-lution of 1024 � 768 pixels.

2.3. Experimental paradigm

The experimental protocol consisted of an event-related designin which subjects were scanned while being presented with each ofthe 168 face stimuli and the 168 building stimuli in a pseudoran-dom order, making a total of 336 presentations. On each trial, a fix-ation cross was presented for 1000 ms on a grey backgroundfollowed by a stimuli presentation for 3000 ms. Subjects were in-structed to press one of five buttons on a response key-pad withtheir right hand to indicate their aesthetic judgment (1 = veryunappealing, 5 = very appealing). Randomly interspersed with the

308 U. Kirk et al. / Brain and Cognition 69 (2009) 306–315

stimuli presentations were 56 null event trials (grey screen). Totalscanning-time per subject was 26 min. in one session. The longparadigm lasting 26 min could have potentially given rise to fati-gue in the subjects. However, we did not observe any deviationsin behavioural differences when the first part of the scanningwas compared with the last part of the scanning. In particularthe trial to trial judgment variability was assessed as a measureof fatigue, under the assumption that fatigued subjects would tendto evaluate similarly from trial to trial when they were fatigued.Missing trials and the reaction time were also used as indirectmeasures, and these did not change either. Prior to scanning, sub-jects were informed that the study was concerned with investigat-ing aesthetic judgments, but no reference was made to theexperimental aims. After the scanning task was complete, subjectswere presented with the stimuli again, this time outside the scan-ner, where they rated each stimulus for familiarity. Familiarity rat-ings were entered into the design matrix as regressors of nointerest. Stimuli were presented and responses collected using E-prime (Psychology Software Tools, Inc.). The stimuli were back-projected via a LCD projector onto a transparent screen positionedover the subjects’ head and viewed through a tilted mirror fixed tothe head coil.

2.4. fMRI data acquisition

The functional imaging was conducted by using a 3 Tesla scan-ner (Siemens, Magnetom Trio, Erlangen, Germany) to acquire gra-dient T2

* weighted gradient echo (GR) echo planar images (EPI) tomaximize the blood oxygen level-dependent (BOLD) contrast(echo-time, TE = 30 ms; repetition time, TR = 2400 ms; flip angle,FA = 90�). The EPI sequence was optimized in order to reduce signaldrop-out in OFC (Deichmann, Gottfried, Hutton, & Turner, 2003).Each functional image was acquired in an interleaved way, begin-ning with 2nd slice (slice No. 2,4,. . .,40, 1,3,. . .,39) when countedfrom the bottom, comprising 40 axial slices each 3.0 mm thick,consisting of a 64 � 64 matrix with an in-plane resolution of3 � 3 mm. This gave near whole-brain coverage, excluding inferiorparts of the cerebellum. Each session consisted of 654 volumes. Thesubjects’ pulse and respiration were recorded using an MRI-com-patible pulse oximeter, and a respiration belt, both sampled at50 Hz. After the functional scan, a T1 weighted MPRAGE structuralsequence was acquired, using a phased array head coil to providehigh-resolution anatomical detail.

2.5. fMRI data analysis

Image pre-processing and data analysis was performed usingSPM2 (Wellcome Department of Imaging Neuroscience, London,UK). The EPI images were spatially realigned (Friston et al.,1995). This was followed by temporal realignment, which cor-rected for slice-time differences using the middle slice as referenceslice. Images were then normalized to the Montreal NeurologicalInstitute (MNI) EPI-template provided in SPM2. Finally, a spatialfiltering was performed by applying a Gaussian smoothing kernelof 8 mm FWHM (full width at half-maximum).

Following pre-processing a general linear model (GLM) was ap-plied to the time course data, where each event was modelled witha separate single impulse response function time-locked to middlestimulus time and then convolved with the canonical haemody-namic response function (HRF), including its temporal and disper-sion derivatives in order to capture small variations in the onsetand width of the BOLD responses.

A parametric regression analysis was used (Buchel, Holmes,Rees, & Friston, 1998) that allowed us to model on/off, linear andnon-linear haemodynamic responses using orthogonalized polyno-mial expansion functions. This was performed for each of the two

stimulus conditions using subject-specific aesthetic ratings in or-der to model a potential parametric modulation of aesthetic rat-ings. The on/off or 0th order parametric regression analysisallows inferences to be made about variations in the responseacross the two subject groups independent of the aesthetic ratings.First-level analysis was performed on each subject to generate asingle mean parameter corresponding to each term of the polyno-mial expansion. In order to correct for the structured noise inducedby respiration and cardiac pulsation we included RETROICOR (RET-ROspective Image based CORrection method) nuisance covariatesin the design matrix (Glover, Li, & Ress, 2000). These regressorsare a Fourier expansion of the aliased cardiac and respiratory oscil-lations. We included six regressors for respiration and ten regres-sors for cardiac pulsation. We also included 24 regressors thatremove residual movement artefacts with spin history effects,which have been shown to remain even after image realignment(Friston, Williams, Howard, Frackowiak, & Turner, 1996). This setof nuisance regressors have also been shown to reduce inter andintra subject variation significantly (Lund, Nørgaard, Rostrup,Rowe, & Paulson, 2005). Having all four types of nuisance regres-sors in the design improves the assumption of independently andidentically distributed errors (Lund, Madsen, Sidaros, Luo, & Nic-hols, 2006). For the analysis we also applied a high pass filter witha cut-off frequency at 1/128 Hz. This high pass filter removes anytemporal drift that oscillates slower than once every 128 s, and itwill therefore remove slowly varying drift caused by hardwareinstabilities.

The statistical parametric maps were entered into a second-le-vel, random effects analysis (RFX) accounting for the between sub-ject variance. Experts and non-experts were treated as separategroups in an ANOVA model using the beta-estimates of the twogroups and the two stimuli conditions for the linear and the qua-dratic expansions. Equal variance was not assumed, thus SPM2’soptions for non-sphericity correction was applied (Glaser& Friston,2004).

Using t-contrasts allowed us to test for correlations of the fMRIBOLD signal and the parameters of interest performed as on/off,linear and non-linear parametric modulations, respectively. Re-ported p-values were set at a threshold of p < .001, uncorrected,unless otherwise stated. In order to correct for multiple compari-sons in the medial orbitofrontal cortex, a region in which activationwas predicted on the basis of our a priori hypothesis, we usedsmall volume corrections (SVC) (Worsley et al., 1996) constrainingour analysis to this region using a sphere with a 10 mm radius. Weused the coordinates reported in Kawabata and Zeki (2004) formedial OFC. Before using SVC, we transformed coordinates givenby Kawabata and Zeki (2004) from Talairach space to MNI space(http://www.mrc-cbu.cam.ac.uk). The coordinates of all activationsare reported in MNI space.

3. Results

3.1. Behavioural results

We first inspected the two groups’ behavioural responses, i.e.aesthetic ratings, collected during scanning (see Fig. 2). A two-way ANOVA with two factor levels (buildings, faces) and groups(experts, non-experts) revealed significant differences betweenstimulus conditions (F(1,10) = 54.42; p < 2 � 10�7) (see Fig. 2A),but no significant differences between groups (F(1,10) = 1.89;p > .1). Furthermore, no significant interactions between stimulusconditions and groups was observed (F(1,10) = 1.44; p > .2). Thesame analysis was applied to the reaction-time data (RT) collectedduring scanning. However no significant differences were foundbetween groups (F(1,10) = 0.45; p > .5). Analysing the mean ratings

Fig. 2. Behavioural responses collected during scanning. (A) Mean aesthetic ratings for the two stimulus conditions and both subject groups. The mean rating for buildingstimuli for experts was 3.29 (SD = 0.21) and for non-experts 3.28 (SD = 0.32). For face stimuli the mean rating for experts was 2.7 (SD = 0.25) and for non-experts 2.79(SD = 0.33). (B) Mean reaction times (RT) for stimulus conditions and subject groups. Examination of RTs revealed that average RT for building stimuli for experts was 2003 ms(SD = 357.4) and for non-experts 2011 ms. (SD = 562.4). The average response time for face stimuli for experts was 1810 ms (SD = 256.6) and for non-experts 1696 ms(SD = 440.4). Response latencies between groups and stimulus conditions did not differ significantly in a one-way ANOVA (F(3,40) = 1.48; p < .23). (C) Distribution of ratingsacross the five rating bins for building stimuli, where each rating bin is bears a scale from 5 (high appeal) to 1 (low appeal) (x-axis) and the response frequency across subjectsin percent is shown (y-axis). Error bars indicate SD. (D) Distribution of ratings across the five rating bins for face stimuli. Error bars indicate SD.

U. Kirk et al. / Brain and Cognition 69 (2009) 306–315 309

of the two stimulus classes, these results suggest that experts andnon-experts did not display significantly different behaviour inmaking an aesthetic judgment of either faces or buildings. We nextinspected whether experts and non-experts differed in the fre-quency of each rating bin (see Fig. 2C and D), as such a differencemight not be reflected when inspecting the mean aesthetic ratings(see Fig. 2A). Moreover, such a potential difference in the distribu-tion of ratings could also have consequences for interpreting thefMRI results, since a different distribution in the frequency of rat-ing bins between groups could potentially account for differencesin the linear fits between the two groups. No significant differenceswere found between groups in the frequency of each rating bin forface stimuli (see Fig. 2D). For building stimuli (see Fig. 2C) no sig-nificant differences between groups was observed for rating bin 4and 5 (reflecting high appeal) and bin 1 and 2 (reflecting low ap-peal). However, the middle rating bin resulted in a significant dif-ference between groups (two sample t = 3.02; df = 10; p < .01).Finally, we looked at the variance of RTs between the two groups.Although we found no significant differences between the twogroups in mean ratings and RTs, it was possible that a differencebetween groups would be reflected in greater variance in RTs be-tween the groups. However, we found no such difference in vari-ance across groups (F(1,4) = 0.73; p > .4), rating bin (F(1,4) = 0.24;

p > .9), or stimulus type (F(1,4) = 2.25; p > .1). Likewise, we ob-served no significant interactions (group � rating, group � stimu-lus, and group � rating � stimulus).

3.2. fMRI results

3.2.1. Correlation between the BOLD signal and linear aesthetic ratingsTo test whether architectural expertise modulated brain activ-

ity associated with making aesthetic judgments a 1st order para-metric regression model using the subject-specific behaviouralresponses was applied. We defined the expertise effect as the inter-action between the two subject groups and the two stimulusconditions [Expert_Build–Expert_Faces] � [Non-Expert_Build–Non-Expert_Faces]. This analysis allowed us to focus on voxelsfor which the difference between the responses for the two stimu-lus conditions varied across the two subject groups. The interactionrevealed significant activations in bilateral subcallosal cingulategyrus (�4, 30, 2; z = 4.47; p < .05, corrected for multiple compari-sons using false discovery rate, FDR; 14, 40, �2; z = 4.32; p < .05,FDR) (see Fig. 3). When we performed small volume corrections(SVC) we observed significant activity in bilateral medial OFC(�8, 30, �20; z = 3.83; p < .05, FDR, SVC; 6, 34, �16; z = 3.40;p < .05, FDR, SVC). These significant interactions were further

Fig. 3. The figure shows areas where the BOLD signal correlates with a linear fit for the interaction [Expert_Build–Expert_Faces] � [Non-Expert_Build–Non-Expert_Faces]. Theupper left panel shows activation in left subcallosal cingulate gyrus. The upper right panel shows parameter estimates for subcallosal cingulate gyrus (�4, 30, 2), where the x-axis reflects the experimental conditions and the y-axis shows BOLD signal changes. The lower left panel shows activation in left medial OFC and the lower right panel showsparameter estimates from the hottest voxels in medial OFC (�8, 30, �20). Activations are overlaid on sagittal sections of the canonical SPM structural image. Activations aredisplayed at p < .001, uncorrected. Error bars indicate 90% confidence interval.

310 U. Kirk et al. / Brain and Cognition 69 (2009) 306–315

investigated by examining contrasts of parameter estimates. Fig. 3describes the subject-averaged parameter estimates from the peakactivations at the group level in subcallosal cingulate gyrus andmedial OFC, which show the marked difference between expertsand non-experts for building stimuli. Experts had greater activa-tion in subcallosal cingulate gyrus and medial OFC when makingaesthetic judgments of buildings, while activation levels for thesame voxels for face stimuli in both groups were essentially bal-anced, suggesting that the observed interaction effects depend onacquired expertise of architecture. Interestingly, the subcallosalcingulate gyrus correlated positively with aesthetic judgmentsfor experts while for non-experts it was negatively correlated.However, both groups correlated positively in OFC, while onlythe activation level was significant for experts compared to non-experts.

A potentially confounding effect in this paradigm would havebeen introduced if experts had any prior knowledge of the buildingstimuli, so that the cortical differences between groups reflectedrecognition effects rather than aesthetic judgment, specifically inmedial OFC (Frey& Petrides, 2002). To control for this we regressedfamiliarity data (collected post-scanning for building stimuli) ontobrain activity. SVC was applied, constraining our analysis to themedial OFC activation (�8, 30, �22 and 6, 34, �16). However, thisanalysis did not produce any supra-threshold voxels at p < .001,uncorrected (not shown), indicating that familiarity effects didnot contribute to the results in medial OFC. The likely explanationfor this is simply that only very few buildings were recognized bythe experts, and hence, any effects of familiarity would be mod-elled out by the applied high pass filter.

When we performed the inverse interaction, which shouldhighlight areas exhibiting an elevated parametric response tobuildings over faces in non-experts compared to experts, [Non-Expert_Build–Non-Expert_Faces]� [Expert_Build–Expert_Faces], we

found that no areas correlated significantly with such a responseprofile (p < .001, uncorrected).

In order to formally identify possible common brain areas be-tween the two subject groups that scaled linearly with preferenceresponses for building stimuli we performed a conjunction be-tween experts and non-experts. This analysis did not reveal anysignificant voxels (p < .001, uncorrected), supporting our hypothe-sis that processes involved in aesthetic evaluation are differentiallymodulated in the two groups.

3.2.2. Correlation between the BOLD signal and non-linear aestheticratings

We furthermore modelled 2nd order polynomial expansions ofthe subject-specific aesthetic judgments. This analysis makes itpossible to seek evidence for brain activity that correlates signifi-cantly with a positive 2nd order non-linear response profile, whichhas the form of a u-shaped function (where responses are maximalfor appealing and unappealing stimuli compared to neutrally ratedstimuli) to account for additional variance not captured by the lin-ear 1st order term. An interaction analysis using this non-linearorder term [Non-Expert_Build–Non-Expert_Faces]� [Expert_Build–Expert_Faces] and [Expert_Build–Expert_Faces]� [Non-Expert_-Build–Non-Expert_Faces] did not produce any significant activity(p < .001, uncorrected). In order to search for common areas with anon-linear response profile for building stimuli regardless of groupwe performed a conjunction analysis. One region in the ventral stri-atum, namely the left nucleus accumbens (NAcc) (�10, 10, �4;z = 4.94; p < .008, FDR), and also a small cluster in the left anteriorthalamus (�14, �4, 12; z = 4.75; p < .008, FDR), were both significantin the conjunction analysis. In order to further investigate the role ofNAcc and the anterior thalamus we employed a conjunction analysisincluding both stimulus conditions and both groups, but the resultdid not meet a corrected threshold. We therefore applied SVC using

Fig. 5. The upper panel display activation of the left hippocampus from theinteraction [Expert_Build–Expert_Faces] � [Non-Expert_Build–Non-Expert_Faces]using the zero-order parametric analysis that display voxels activated irrespectiveof aesthetic rating. The lower panel shows parameter estimates for the lefthippocampus, where the x-axis reflects the experimental conditions, and the y-axisshows BOLD signal changes. Activation is displayed at p < .001, uncorrected. Errorbars indicate 90% confidence interval.

Table 1Summary of interaction effects for the parametric regression analysis irrespective ofaesthetic ratings

Brain region MNI coordinates z Score Number of voxels

[Expert_Build–Expert_Faces] � [Non-Expert_Build–Non-Expert_Faces]R hippocampus 38, �28, �8 3.60 18L hippocampus �26, �14, �14 3.29 17

U. Kirk et al. / Brain and Cognition 69 (2009) 306–315 311

the clusters from the building-specific conjunction, and found thatleft NAcc (�10, 8, �4; z = 3.42; p < .05, FDR, SVC) was significantlymore active in both stimuli conditions in both groups (see Fig. 4).The activation in left anterior thalamus did not survive SVC. These re-sults suggest that left NAcc plays a role in encoding high and low aes-thetic values that is not modulated by expertise or stimulus modality.

3.2.3. Correlation between the BOLD signal and expertise irrespectiveof aesthetic ratings

Finally, in order to identify voxels that responded differentiallyin the two groups per se—i.e., irrespective of aesthetic rating—aninteraction analysis using regressors from a zero-order parametricregression analysis was conducted. An interaction analysis [Ex-pert_Build–Expert_Faces] � [Non-Expert_Build–Non-Expert_Faces]showed distinct specificity for buildings compared to faces inexperts relative to non-experts in bilateral hippocampus, left pre-cuneus and cerebellum (see Fig. 5 and Table 1).

In the converse interaction [Non-Expert_Build–Non-Expert_Fa-ces] � [Expert_Build–Expert_Faces] we observed significant activa-tions in bilateral calcarine gyrus bilateral and fusiform gyruslocated adjacent to the collateral sulcus and inferior lingual gyrusposterior to the parahippocampal gyrus (see Fig. 6 and Table 1).

We furthermore conducted a conjunction analysis using thebuilding-specific main effects for both groups [Expert_Build–Ex-pert_Faces] and [Non-Expert_Build–Non-Expert_Faces]. We ob-served bilateral activation of the parahippocampal place area(PPA) [30, �42, �14; �30, �46, �10, FDR] (see Supplementarymaterial) that has been found to respond selectively to houses,landscapes and other environmental sceneries (Epstein& Kanwish-er, 1998).

4. Discussion

The present experiment extends other studies of expertise tosuggest that acquired expertise not only impacts on cognitiveand perceptual systems (Bangert et al., 2006; Maguire et al.,

Fig. 4. The upper panel shows activation in left NAcc using the 2nd order non-linearterm. The lower panel shows the parameter estimates in both groups and for bothstimulus conditions in left NAcc (�10, 8, �4) where the x-axis reflects theexperimental conditions consisting of both groups and both stimulus conditions,and the y-axis shows BOLD signal changes. Activations are overlaid on sagittal,coronal and axial sections of the canonical SPM structural image. Activation isdisplayed at p < .008, FDR). Error bars indicate 90% confidence interval.

�34, �16, �20L precuneus �6, �54, 22 3.61 37R cerebellum 16, �62, �16 3.57 26

26, �68, �24 3.44 13

[Non-Expert_Build–Non-Expert_Faces] � [Expert_Build–Expert_Faces]R fusiform gyrus 32, �54, �4 3.35 9L fusiform gyrus �30, �56, 0 3.73 42R calcarine gyrus 18, �58, 16 3.59 21L calcarine gyrus �18, �64, 14 3.79 30R pons 16, �18, �30 4.41 69

Activations are shown at (p < .001, uncorrected). L, left hemisphere; R, righthemisphere.

2000), but also modulates the response of brain areas associatedwith the processing of reward. However, the processing of rewardhas been linked to several brain areas, including the ventral teg-mental area, ventral striatum, amygdala and OFC (for a review,see McClure, York, & Montague, 2004), and our results show thatonly parts of this system are modulated by expertise during aes-thetic judgment. In contrast to expertise effects observed in OFCand subcallosal cingulate gyrus, we found that activity in left NAccwas elevated in both groups and stimuli conditions in response toappealing and non-appealing stimuli.

The response profile of medial OFC in both groups exhibited apositive linear correlation with aesthetic ratings. However, whencompared to each other the increase was significantly higher inthe experts than in the non-experts. The fact that the medial partof OFC shows sensitivity to the magnitude of aesthetic value is inaccordance with studies on reward processing showing that therelative reward value of stimuli is reflected by the amplitude of

Fig. 6. The figure displays the interaction [Non-Expert_Build–Non-Expert_Fa-ces] � [Expert_Build–Expert_Faces] from the zero-order parametric analysis thatshows brain areas activated, irrespective of the actual aesthetic ratings. The upperpanel displays a bilateral activation of the fusiform gyrus on slices where z-coordinates for the three slices are ascending from �6, �4 and �2, respectively.Evident on the slices are the collateral sulcus just lateral to the activation in thefusiform gyrus. The lower panel displays the corresponding parameter estimates inthe right fusiform gyrus. The x-axis reflects the experimental conditions. The y-axisshows BOLD signal changes. Activations are displayed at p < .001, uncorrected. Errorbars indicate 90% confidence interval.

312 U. Kirk et al. / Brain and Cognition 69 (2009) 306–315

neural activity in OFC (Kringelbach, 2005; Tremblay & Schultz,1999). For instance, in studies comparing subjects ingesting foodin states of hunger and satiety a contrast of these two states revealsdifferent neural responses in OFC, indicating that OFC neuronscodes reward aspects of a stimulus rather than sensory aspects(Kringelbach, O’Doherty, Rolls, & Andrews, 2003). Furthermore,several studies suggest that the medial aspect of the human OFCrepresents the hedonic attributes involved in preference judg-ments of various stimulus types (Aharon et al., 2001; Andersonet al., 2003; Blood et al., 1999; Gottfried et al., 2002; Kawabataet al., 2004; O’Doherty et al., 2001; O’Doherty et al., 2003; Rolls,Kringelbach, O’Doherty, Rolls, & Andrews, 2003; Rolls, O’Doherty,et al., 2003; Small et al., 2001, 2003). The implication of these pre-vious findings for the present results is that medial OFC may be en-gaged under conditions where behavioural decision making basedon stimulus reward value is required (Bechara, Damasio, & Dama-sio, 2000; Wallis, 2007). Recently, several studies have demon-strated that medial OFC responses can be modulated bytop-down information such as knowledge of the price of a wine(Plassmann et al., 2008), brand information (McClure, York et al.,2004), and visual word descriptors influence preference for odours(de Araujo, Rolls, Velazco, Margot, & Cayeux, 2005). The novelty ofour results is that the representation of stimulus value, or possiblyintrinsic motivation, in medial OFC varies with expertise level tosuch an extent that the experts displayed higher activation to thebuilding stimuli than the non-experts, but not to the control-stim-uli, i.e. faces.

In contrast to OFC, voxels in the subcallosal part of the anteriorcingulate gyrus responded inversely to high and low ratings in theexperts compared to non-experts. This result supports the growingrecognition that anterior cingulate gyrus and OFC contribute dis-tinct component processes to decision making (Rushworth, Beh-rens, Rudebeck, & Walton, 2007). The anterior aspect of thecingulate gyrus forms an anatomical interface between the OFCand premotor cortex. Since it also receives afferents from subcorti-

cal dopaminergic neurons and inputs from dorsolateral prefrontalcortex, it has been suggested that the anterior cingulate integratesthe affective drive and action strategies for the purpose of selectingappropriate motor responses, i.e. making decisions how to act(Paus, 2001). Another possibility might be that subcallosal cingu-late gyrus activity reflects the subjects’ monitoring of their ownemotional state (Ochsner et al., 2004), whereby appealing buildingsare more arousing to the experts than to the non-experts. Indeed,the subcallosal aspect of the cingulate gyrus has previously beenimplicated in such forms of emotional processing, including the re-call of happy autobiographical memories (Lane, Reinman, Ahern,Schwartz, & Davidson, 1997) and attending to emotionally stimu-lating words (Maddock, Garrett, & Buonocore, 2003). Interestingly,decreasing musical dissonance, associated with an elevated experi-ence of pleasure (Blood et al., 1999) and passive listening to unfa-miliar, pleasant musical compared to a rest condition (Brown et al.,2004) has also been shown to produce enhanced activity in subcal-losal cingulate gyrus. Finally, as pupillometry data was not re-corded in the present study we are unable to assess whether ornot such effects would have detected differences between thetwo groups. Indeed there is evidence for involvement of the subcal-losal cingulate in generating and monitoring autonomic interocep-tive states (Critchley, 2004).

It is notable that while activity in OFC and the subcallosal cin-gulate gyrus were sensitive to the level of expertise, the behav-ioural responses did not parallel this difference in neuralactivation between the two groups. As a number of previousbehavioural studies have found an effect of expertise on aestheticevaluation to be robust, our result was somewhat surprising(Eysenck & Castle, 1970; Gordon, 1951/1952; Gordon, 1956;O’Hare, 1976; Schmidt et al., 1989). One possible reason for thisdiscrepancy between our study and earlier ones could be the meth-od of comparison used in our study. It is conceivable that compar-ing means of rating is not sufficiently sensitive to detectdifferences in experts’ and non-experts’ ratings. Experts and non-experts are known to respond differentially to dimensions of stim-ulus qualities such as craftsmanship and quality (Hekkert & vanWieringen, 1996b), or chromatic versus achromatic versions ofpaintings (Hekkert & van Wieringen, 1996a). It is conceivable thatour set of buildings stimuli lack perceptual properties that system-atically influence the differential aesthetic assessment of expertsand non-experts. On the other hand, the fact that we did not ob-serve a behavioural difference in the two groups’ responses tothe building stimuli strengthens the neural result. If both a differ-ence in behaviour and neural activity between the two groups hadbeen observed, the difference in neural activity might have beenconfounded by the differences in behavioural responses. In thepresent situation where the two groups differ in neural activitybut not in behaviour we can be sure that what we observe is a truegroup difference.

Our finding of a positive bivalent response in the ventral stria-tum, specifically the left NAcc, in both subject groups and to bothstimuli types replicates and extends previous findings that NAccand OFC play different functional roles in reward processing (Knut-son, Fong, Adams, Varner, & Hommer, 2001; O’Doherty et al., 2002;O’Doherty et al., 2003; Tremblay& Schultz, 1999; Watanabe, 1999).Whereas the OFC is thought to process reward outcomes, the NAccis generally believed to subserve the prediction of reward and tocompute the variance between reward expectation and the actualreward (for reviews, see Knutson & Cooper, 2005; Montague, Hy-man, & Cohen, 2004). Although earlier results have found non-lin-ear response profiles in the NAcc as discussed below, to ourknowledge there has been no previous descriptions of NAcc activ-ity with negative aesthetic ratings. Electrophysiological recordingsin animals have demonstrated that NAcc neurons increase theirresponse to positive reward prediction errors (situations that are

U. Kirk et al. / Brain and Cognition 69 (2009) 306–315 313

better than expected) and decrease their response to negative pre-diction errors (Apicella, Ljungberg, Scarnati, & Schultz, 1991;Schultz, Apicella, Ljungberg, Romo, & Scarnati, 1993). Neuroimag-ing work has subsequently replicated these findings (e.g., O’Doher-ty et al., 2006; Seymour, Daw, Dayan, Singer, & Dolan, 2007; Spiceret al., 2007). However, recently another hypothesis has been putforth, suggesting that activations of the ventral striatum are notsensitive to errors in prediction, but rather encode salience dimen-sions of the stimulus. This idea heralds from imaging studies wherethe ventral striatum has been shown to correlate with predictionerrors regardless of valence (Jensen et al., 2007; Zink, Martin-Skur-ski, Chappelow, & Berns, 2004). A recent fMRI study by Cooper andKnutson (2008), though, suggests that the ventral striatum maycompute the interaction of valence and salience, depending uponthe context wherein motivational behaviour takes place. Directlycomparing degrees of valence and salience, Cooper and Knutsonfound that both the valence and the salience of anticipated incen-tives correlated with NAcc activation. More specifically, in thisstudy when outcomes were uncertain and salience high, NAcc acti-vation increased for anticipated loss and gain, whereas NAcc acti-vation increased for anticipated gain and decreased foranticipated loss when outcomes were certain and salience low. Inour study the experimental set-up might be seen as similar to thefirst situation (uncertain outcome and high salience). However, weneither manipulated the salience of the pictures nor the relationbetween anticipation and reward, so this remains conjecture. Sinceit is possible that the subjects covertly anticipated the reward out-come of the upcoming stimuli based on the recently transpiredjudgment act, we cannot rule out the possibility that the NAcc acti-vation reflects prediction error signalling.

The zero-order parametric regression analysis identified brainregions showing a typological response to the two stimulus con-ditions irrespective of aesthetic ratings. Hence, this analysis de-tects differences in the two groups’ response to the stimulusmaterial beyond those differences specifically related to makingan aesthetic judgment. Inspection of the parameter estimates insignificant voxels from the interaction showed that experts hadsignificantly more activity in hippocampus, precuneus and cere-bellum relative to non-experts for expert-stimuli (buildings) com-pared to control-stimuli (faces). Precuneus is often reported toplay a role in integrating the current input with prior establishedknowledge (Fletcher et al., 1995; Maguire, Frith, & Morris, 1999)and in episodic memory retrieval (Krause et al., 1999). Our datasuggests that the demand on the precuneus is higher in expertswhen perceiving expert-stimuli, as building conditions presum-ably depend more on connections between retrieved informationand prior knowledge for this group relative to non-experts. Thehippocampus has also been consistently implicated in episodicmemories (Brown & Aggleton, 2001; Eichenbaum, Schoenbaum,Young, & Bunsey, 1996; Eichenbaum, Yonelinas, & Ranganath,2007). Hippocampus activation has been associated with condi-tions where subjects correctly recollect contextual informationcompared to conditions where they do not (Cansino, Maquet, Do-lan, & Rugg, 2002). Our findings support these data and suggestthat the hippocampus and precuneus may be selectively engagedduring memory retrieval in experts. As we have ruled out famil-iarity effects in the data, experts may have attempted to organizenew information into a framework of prior knowledge and usethis information to guide and bias aesthetic judgments. The pos-sibility that the hippocampus and the precuneus are specificallyinvolved in biasing preference judgments based on recruitmentof episodic memory has also been suggested by other studies(Jacobsen et al., 2006; McClure et al., 2004). Specifically, a charac-teristic of episodic memories in the present study might be in-creased encoding of associations in experts (Eichenbaum et al.,2007) responding to expert-stimuli, which is dissociable from

stimuli recognized based on familiarity (Eldridge, Knowlton, Fur-manski, Bookheimer, & Engel, 2000).

For the converse interaction analysis we found no activation inareas involved in episodic memory formation. However, we foundactivity in regions of the visual cortex and in the ventral temporalcortex, such as the calcarine gyrus and in the fusiform gyrus. Thebilateral activation of the fusiform gyrus is interesting as this re-gion, although distinctly demarcated by the collateral sulcus, isanatomically closely located to an area straddling the anteriorend of lingual gyrus which has been deemed the location of abuilding-sensitive region (Aguirre, Zarahn, & D’Esposito, 1998).Further evidence for building selectivity shows that the medialportion of the fusiform gyrus, including the collateral sulcus, dem-onstrates greater fMRI signal change in response to buildings ascompared to faces and chairs (Ishai, Ungerleider, Martin, Schouten,& Haxby, 1999). There is disagreement about the exact anatomicallocation of a building-sensitive region, but it seems to include bothinferior lingual gyrus and the fusiform gyrus surrounding the col-lateral sulcus. Our activation, clearly located in the fusiform gyrus,is, however, distinct from the face-sensitive region within the fusi-form gyrus (Kanwisher, McDermott, & Chun, 1997), which is lo-cated inferior and lateral to our building-sensitive voxels. This isfurthermore evidenced by the parameter estimates in the build-ing-sensitive region of the fusiform gyrus, where it is shown thatthis area is unresponsive to face stimuli in both groups (seeFig. 6). The region we observed in the fusiform gyrus is located justadjacent to the parahippocampal gyrus. Several neuroimagingstudies have demonstrated that the posterior portion of the para-hippocampal gyrus is involved in the representation of large-scaleplaces and scenes (Epstein & Kanwisher, 1998; Maguire, Frith, Bur-gess, Donnett, & O’Keefe, 1998). It is noteworthy that we observedactivity in the PPA in the conjunction analysis between [build-ings > faces] for experts and non-experts. Evident in this conjunc-tion is also activation, beyond the PPA, of the entire ventraltemporal cortex and visual cortex (see Supplementary material),suggesting that the response to buildings is not restricted to the re-gion that responds maximally to that object category located in thefusiform gyrus. However, this effect may also be driven by differ-ences in visual stimulation across the two stimulus conditions, be-cause building trials were presented with varying pixel width. Thefact that building stimuli activate the entire ventral temporalcortex, albeit to varying degrees, suggests, in agreement with otherreports (Ishai et al., 1999), that the representation of buildings inthis portion of the cortex may be feature-based rather than build-ing-sensitive per se. Such an interpretation may account for the dif-ferential activation of the fusiform gyrus between experts and non-experts evident in the interaction analysis. The demand on thisportion of the fusiform gyrus is higher for non-experts comparedto experts presumably due to experts’ recruitment of episodicmemory, whereas non-experts are more sensitive to specific per-ceptual features of building stimuli, which finds further supportby the relative stronger activation in the calcarine gyrus in non-ex-perts relative to experts.

In conclusion, we have demonstrated that expertise modulatesbrain areas to both aesthetic processing and to cognitive or typo-logical processing irrespective of aesthetic ratings. Specifically,our new discovery is that the representation of stimulus value inmedial OFC and bilateral subcallosal cingulate gyrus is modulatedby expertise. We found that only some regions associated with theprocessing of reward are modulated by expertise (OFC, subcallosalcingulate gyrus), while activity in NAcc was typical of both expertsand non-experts, suggesting that these regions play different rolesin reward processing. Furthermore, we have demonstrated that ex-perts and non-experts differ in their neural response to expertisestimuli per se, irrespective of aesthetic ratings. This typological re-sponse was observed bilaterally in the hippocampus and precu-

314 U. Kirk et al. / Brain and Cognition 69 (2009) 306–315

neus, and suggests that experts may integrate current input into aframework of prior knowledge and use this information to orga-nize aesthetic judgments.

Acknowledgments

We thank Prof. S. Zeki, Dr. O.J. Hulme and Dr. T. Lund for helpfuldiscussions. Prof. C. Frith, Dr. M. Self, Dr. V. Cardin and Dr. T. Ramsøyprovided useful comments on the manuscript. P. Neckelmann pre-pared the stimulus material. U. Kirk was supported by a Ph.D. schol-arship from the Danish Medical Research Council; M. Skov wassupported by Hvidovre Hospital’s research foundation; M.S. Chris-tensen was supported by a Ph.D. scholarship from the Faculty of Sci-ence, University of Copenhagen; N. Nygaard was supported by aPh.D. scholarship by the Danish Research Council for the Humani-ties. The MR-scanner was donated by the Simon Spies Foundation.

Appendix A. Supplementary data

The figure displays the building conjunction from the zero-or-der parametric analysis derived from the building-specific maineffects for both groups [Expert_Build–Expert_Faces] and [Non-Ex-pert_Build–Non-Expert_Faces]. Glass-brain activation is displayedat FDR-corrected threshold. Supplementary data associated withthis article can be found, in the online version, at doi:10.1016/j.bandc.2008.08.004.

References

Aguirre, G. K., Zarahn, E., & D’Esposito, M. (1998). An area within the human ventralcortex sensitive to ‘building’ stimuli: Evidence and implications. Neuron, 21,373–383.

Aharon, I., Etcoff, N., Ariely, D., Chabris, C. F., O’Connor, E., & Breiter, H. C. (2001).Beautiful faces have variable reward value: fMRI and behavioural evidence.Neuron, 32, 537–551.

Anderson, A. K., Christoff, K., Stappen, I., Panitz, D., Ghahremani, D. G., Glover, G.,et al. (2003). Dissociated neural representations of intensity and valence inhuman olfaction. Nature Neuroscience, 6, 196–202.

Apicella, P., Ljungberg, T., Scarnati, E., & Schultz, W. (1991). Responses to reward indorsal and ventral striatum. Experimental Brain Research, 85, 491–500.

Bangert, M., Peschel, T., Schlaug, G., Rotte, M., Drescher, D., & Hinrichs, H. (2006).Shared networks for auditory and motor processing in professional pianists:Evidence from fMRI conjunction. Neuroimage, 30, 917–926.

Bechara, A., Damasio, H., & Damasio, A. R. (2000). Decision making and theorbitalfrontal cortex. Cerebral Cortex, 10, 295–307.

Blood, A. J., & Zatorre, R. J. (2001). Intensely pleasurable responses to musiccorrelate with activity in brain regions implicated in reward and emotion.Proceedings of the National Academy of Sciences of the United States of America, 98,11818–11823.

Blood, A. J., Zatorre, R. J., Bermudez, P., & Evans, A. C. (1999). Emotional responses topleasant and unpleasant music correlate with activity in paralimbic brainregions. Nature Neuroscience, 2, 382–387.

Brown, M. W., & Aggleton, J. P. (2001). Recognition memory: What are the roles ofthe perirhinal cortex and hippocampus? Nature Review Neuroscience, 2, 51–61.

Brown, S., Martinez, M. J., & Parsons, L. M. (2004). Passive music listeningspontaneously engages limbic and paralimbic systems. NeuroReport, 15,2033–2037.

Buchel, C., Holmes, A. P., Rees, G., & Friston, K. J. (1998). Characterizing stimulusresponse functions using nonlinear regressors in parametric fMRI experiments.Neuroimage, 8, 140–148.

Cansino, S., Maquet, P., Dolan, R. J., & Rugg, M. D. (2002). Brain activity underlyingencoding and retrieval of source memory. Cerebral Cortex, 12, 1048–1056.

Cela-Conde, C. J., Marty, G., Maestú, F., Ortiz, T., Munar, E., Fernández, A., et al.(2004). Activation of the prefrontal cortex in the human visual aestheticperception. Proceedings of the National Academy of Sciences of the United States ofAmerica, 101, 6321–6325.

Cooper, J. C., & Knutson, B. (2008). Valence and salience contribute to nucleusaccumbens activation. Neuroimage, 39, 538–547.

Critchley, H. D. (2004). The human cortex responds to an interoceptive challenge.Proceedings of the National Academy of Sciences of the United States of America,101, 6333–6334.

de Araujo, I. E., Rolls, E. T., Velazco, M. I., Margot, C., & Cayeux, I. (2005). Cognitivemodulation of olfactory processing. Neuron, 46, 671–679.

Deichmann, R., Gottfried, J. A., Hutton, C., & Turner, R. (2003). Optimized EPI forfMRI studies of the orbitofrontal cortex. Neuroimage, 19, 430–441.

Eichenbaum, H., Schoenbaum, G., Young, B., & Bunsey, M. (1996). Functionalorganization of the hippocampal memory system. Proceedings of the

National Academy of Sciences of the United States of America, 93,13500–13507.

Eichenbaum, H., Yonelinas, A. P., & Ranganath, C. (2007). The medial temporal lobeand recognition memory. Annual Reviews of Neuroscience, 21, 123–152.

Eldridge, L. L., Knowlton, B. J., Furmanski, C. S., Bookheimer, S. Y., & Engel, S. A.(2000). Remembering episodes: A selective role for the hippocampus duringretrieval. Nature Neuroscience, 11, 1148–1152.

Epstein, R., & Kanwisher, N. (1998). A cortical representation of the local visualenvironment. Nature, 392, 598–601.

Eysenck, H. J., & Castle, M. (1970). Training in art as a factor in the determination ofpreference judgments for polygons. British Journal of Psychology, 61, 65–81.

Fletcher, P. C., Frith, C. D., Grasby, P. M., Shallice, T., Frackowiak, R. S. J., & Dolan, R. J.(1995). Brain systems for encoding and retrieval of auditory-verbal memory: Anin vivo study in humans. Brain, 118, 401–416.

Frey, S., & Petrides, M. (2002). Orbitofrontal cortex and memory formation. Neuron,36, 171–176.

Friston, K. J., Ashburner, J., Frith, C. D., Poline, J. B., Heather, J. D., & Frackowiak, R. S. J.(1995). Spatial registration and normalization of images. Human Brain Mapping,2, 165–189.

Friston, K. J., Williams, S., Howard, R., Frackowiak, R. S. J., & Turner, R. (1996).Movement-related effects in fMRI time-series. Magnetic Resonance in Medicine,35, 346–355.

Glaser, D., & Friston, K. J. (2004). Variance components. In R. S. J. Frackowiak, K. J.Friston, C. D. Frith, R. J. Dolan, C. J. Price, & S. Zeki, et al. (Eds.), Human brainfunction (pp. 781–792). Elsevier Academic Press.

Glover, G. H., Li, T. Q., & Ress, D. (2000). Image-based method for retrospectivecorrection of physiological motion effects in fMRI: RETROICOR. MagneticResonance in Medicine, 44, 162–167.

Gordon, D. A. (1951/1952). Methodology in the study of art evaluation. Journal ofAesthetics and Art Criticism, 10, 338–352.

Gordon, D. A. (1956). Individual differences in the evaluation of art and the nature ofart standards. Journal of Educational Research, 50, 17–30.

Gottfried, J. A., Deichmann, R., Winston, J. S., & Dolan, R. J. (2002). Functionalheterogeneity in human olfactory cortex: An event-related functional magneticresonance imaging study. Journal of Neuroscience, 22, 10819–10828.

Hekkert, P., Peper, L. E., & van Wieringen, P. C. W. (1994). The effect of verbalinstruction and artistic background on the aesthetic judgment of rectangles.Empirical Studies of the Arts, 12, 185–203.

Hekkert, P., & van Wieringen, P. C. W. (1996a). The impact of level of expertise onthe evaluation of original and altered versions of post-impressionistic paintings.Acta Psychologia, 94, 117–131.

Hekkert, P., & van Wieringen, P. C. W. (1996b). Beauty in the eye of expert andnonexpert beholders: A study in the appraisal of art. American Journal ofPsychology, 109, 389–407.

Ishai, A., Ungerleider, L. G., Martin, A., Schouten, J. L., & Haxby, J. V. (1999).Distributed representation of objects in the human visual pathway. Proceedingsof the National Academy of Sciences of the United States of America, 96,9379–9384.

Jacobsen, T., Schubotz, R. I., Höfel, L., & Cramon, D. Y. (2006). Brain correlates ofaesthetic judgment of beauty. Neuroimage, 29, 276–285.

Jensen, J., Smith, A. J., Willeit, M., Crawley, A. P., Mikulis, D. J., Vitcu, I., et al. (2007).Separate brain regions code for salience vs valance during reward prediction inhumans. Human Brain Mapping, 28, 294–302.

Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The fusiform face area: Amodule in human extrastriate cortex specialized for face perception. Journal ofNeuroscience, 17, 4302–4311.

Kawabata, H., & Zeki, S. (2004). Neural correlates of beauty. Journal ofNeurophysiology, 91, 1699–1705.

Knutson, B., & Cooper, J. C. (2005). Functional magnetic resonance imaging ofreward prediction. Current Opinion of Neurobiology, 18, 411–417.

Knutson, B., Fong, G. W., Adams, C. M., Varner, J. L., & Hommer, D. (2001).Dissociation of reward anticipation and outcome with event-related fMRI.NeuroReport, 12, 3683–3687.

Koelsch, S., Fritz, T., von Cramon, D. Y., Müller, K., & Friederici, A. D. (2006). Investigatingemotion with music: An fMRi study. Human Brain Mapping, 27, 239–250.

Krause, B. J., Schmidt, D., Mottaghy, F. M., Taylor, J., Halsband, U., Herzog, H., et al.(1999). Episodic retrieval activates the precuneus irrespective of the imagerycontent of word pair associates: A PET-study. Brain, 122, 255–263.

Kringelbach, M. L. (2005). The human orbitofrontal cortex: Linking reward tohedonic experience. Nature Reviews Neuroscience, 6, 691–702.

Kringelbach, M. L., O’Doherty, J., Rolls, E. T., & Andrews, C. (2003). Activation of thehuman orbitofrontal cortex to a liquid food stimulus is correlated with itssubjective pleasantness. Cerebral Cortex, 13, 1064–1071.

Lane, R. D., Reinman, E. M., Ahern, G. L., Schwartz, G. E., & Davidson, R. J. (1997).Neuroanatomical correlates of happiness, sadness, and disgust. The AmericanJournal of Psychiatry, 154, 926–933.

Langlois, J. H., Kalakanis, L., Rubenstein, A. J., Larson, A., Hallam, M., & Smoot, M.(2000). Maxims or myths of beauty? A meta-analytic and theoretical review.Psychological Bulletin, 126, 390–423.

Leder, H., Belke, B., Oeberst, A., & Augustin, D. (2004). A model of aestheticappreciation and aesthetic judgments. British Journal of Psychology, 95, 489–508.

Lund, T. E., Madsen, K. H., Sidaros, K., Luo, W. L., & Nichols, T. E. (2006). Non-whitenoise in fMRI: Does modeling have an impact? Neuroimage, 29, 54–66.

Lund, T. E., Nørgaard, M. D., Rostrup, E., Rowe, J. B., & Paulson, O. B. (2005). Motion oractivity: Their role in intra- and inter-subject variation in fMRI. Neuroimage, 26,960–964.

U. Kirk et al. / Brain and Cognition 69 (2009) 306–315 315

Maddock, R. J., Garrett, A. S., & Buonocore, M. H. (2003). Posterior cingulate cortexactivation by emotional words: fMRI evidence from a valence decision task.Human Brain Mapping, 18, 30–41.

Maguire, E. A., Frith, C. D., Burgess, N., Donnett, J. G., & O’Keefe, J. (1998). Knowingwhere things are: Parahippocampal involvement in encoding object locations invirtual large-scale space. Journal of Cognitive Neuroscience, 10, 61–76.

Maguire, E. A., Frith, C. D., & Morris, R. G. M. (1999). The functional neuroanatomy ofcomprehension and memory: The importance of prior knowledge. Brain, 122,1839–1850.

Maguire, E. A., Gadian, D. G., Johnsrude, I. S., Good, C. D., Ashburner, J., Frackowiak, R.S., et al. (2000). Navigation-related structural change in the hippocampi of taxidrivers. Proceedings of the National Academy of Sciences of the United States ofAmerica, 97, 4414–4416.

McClure, S. M., Li, J., Tomlin, D., Cypert, K. S., Montague, L. M., & Montague, P. R.(2004). Neural correlates of behavioral preference for culturally familiar drinks.Neuron, 44, 379–387.

McClure, S. M., York, M. K., & Montague, P. R. (2004). The neural substrates ofreward processing in humans: The modern role of fMRI. Neuroscientist, 10,260–268.

Menon, V., & Levitin, D. J. (2005). The rewards of music listening: Response andphysiological connectivity of the mesolimbic system. Neuroimage, 28, 175–184.

Montague, P. R., Hyman, S., & Cohen, J. D. (2004). Computational roles for dopaminein behavioural control. Nature, 431, 760–767.

Nakamura, K., Kawashima, R., Nagumo, S., Ito, K., Sugiura, M., Kato, T., et al. (1998).Neuroanatomical correlates of the assessment of facial attractiveness.NeuroReport, 9, 753–757.

Ochsner, K. N., Knierim, K., Ludlow, D. H., Hanelin, J., Ramachandran, T., Glover, G.,et al. (2004). Reflecting upon feelings: An fMRI study of neural systemssupporting the attribution of emotion to self and other. Journal of CognitiveNeuroscience, 16, 1746–1772.

O’Doherty, J. P., Buchanan, T. W., Seymour, B., & Dolan, R. J. (2006). Predictive neuralcoding of reward preference involves dissociable responses in human ventralmidbrain and ventral striatum. Neuron, 49, 157–166.

O’Doherty, J. P., Deichmann, R., Critchley, H., & Dolan, R. J. (2002). Neural responsesduring anticipation of a primary taste reward. Neuron, 33, 815–926.

O’Doherty, J., Rolls, E. T., Francis, S., Bowtell, R., & McGlone, F. (2001). Representationof pleasant and aversive taste in the human brain. Journal of Neurophysiology, 85,1315–1321.

O’Doherty, J., Winston, J., Critchley, H., Perrett, D., Burt, D. M., & Dolan, R. J. (2003).Beauty in a smile; the role of medial orbitiofrontal cortex in facialattractiveness. Neuropsychologia, 41, 147–155.

O’Hare, D. P. A. (1976). Individual differences in perceived similarity and preferencefor visual art: A multidimensional scaling analysis. Perception and Psychophysics,20, 445–452.

Paus, T. (2001). Primate anterior cingulate cortex: Where motor control, drive andcognition interface. Nature Reviews Neuroscience, 2, 417–424.

Plassmann, H., O’Doherty, J., Shiv, B., & Rangel, A. (2008). Marketing actions canmodulate neural representations of experienced pleasantness. Proceedings of theNational Academy of Sciences of the United States of America, 105, 1050–1054.

Rolls, E. T., Kringelbach, M. L., & de Araujo, I. E. (2003). Different representations ofpleasant and unpleasant odours in the human brain. European Journal ofNeuroscience, 18, 695–703.

Rolls, E. T., O’Doherty, J., Kringelbach, M. L., Francis, S., Bowtell, R., & McGlone, F.(2003). Representations of pleasant and painful touch in the humanorbitofrontal and cingulate cortices. Cerebral Cortex, 13, 308–317.

Rushworth, M. F., Behrens, T. E., Rudebeck, P. H., & Walton, M. E. (2007). Contrastingroles for cingulate and orbitofrontal cortex in decisions and social behaviour.Trends in Cognitive Science, 14, 168–176.

Schlaug, G. (2003). The brain of musicians. In I. Peretz & R. Zatorre (Eds.), Thecognitive neuroscience of music (pp. 366–381). Oxford: Oxford University Press.

Schmidt, J. A., McLaughlin, J. P., & Leighten, J. (1989). Novice strategies forunderstanding paintings. Applied Cognitive Psychology, 3, 65–72.

Schultz, W., Apicella, P., Ljungberg, T., Romo, R., & Scarnati, E. (1993). Reward-related activity in the monkey striatum and substantia nigra. Progress in BrainResearch, 99, 227–235.

Seymour, B., Daw, N., Dayan, P., Singer, T., & Dolan, R. (2007). Differential encodingof losses and gains in the human striatum. Journal of Neuroscience, 27,4826–4831.

Skov, M. (in press). The pleasure of art. In M. Kringelbach & K. Berridge (Eds.),Pleasures of the brain. Oxford: Oxford University Press.

Small, D. M., Gregory, M. D., Mak, Y. E., Gitelman, D., Mesulam, M. M., & Parrish, T.(2003). Dissociation of neural representation of intensity and affectivevaluation in human gestation. Neuron, 39, 701–711.

Small, D. M., Zatorre, R. J., Dagher, A., Evans, A. C., & Jones-Gotman, M. (2001).Changes in brain activity related to eating chocolate. Brain, 124, 1720–1733.

Spicer, J., Galvan, A., Hare, T. A., Voss, H., Glover, G., & Casey, B. J. (2007). Sensitivityof the nucleus accumbens to violations in expectation of reward. Neuroimage,34, 455–461.

Tremblay, L., & Schultz, W. (1999). Relative reward preference in primateorbitofrontal cortex. Nature, 398, 704–708.

Vartanian, O., & Goel, V. (2004). Neuroanatomical correlates of aesthetic preferencefor paintings. NeuroReport, 15, 893–897.

Wallis, J. D. (2007). Orbitofrontal cortex and its contribution to decision-making.Annual Reviews of Neuroscience, 30, 31–56.

Watanabe, M. (1999). Attraction is relative not absolute. Nature, 398, 661–662.Winston, J. S., O’Doherty, J., Kilner, J. M., Perrett, D. I., & Dolan, R. J. (2007). Brain

systems for assessing facial attractiveness. Neuropsychologia, 45, 195–206.Worsley, K. J., Marrett, S., Neelin, P., Vandal, A. C., Friston, K., & Evans, A. C. (1996). A

unified statistical approach for determining significant signals in images ofcerebral activation. Human Brain Mapping, 4, 58–73.

Zink, C. F., Martin-Skurski, M. E., Chappelow, J. C., & Berns, G. S. (2004). Humanstriatal responses to monetary reward depend on saliency. Neuron, 42, 509–517.