Ne

Ja

b

c

a

ARRAA

KAAfPV

1

dsmco

aucsscsc

sT

0d

Neuropsychologia 48 (2010) 607–618

Contents lists available at ScienceDirect

Neuropsychologia

journa l homepage: www.e lsev ier .com/ locate /neuropsychologia

eural networks involved in voluntary and involuntary vocal pitch regulation inxperienced singers

ean Mary Zaratea,b,∗, Sean Woodb,c, Robert J. Zatorrea,b

Montréal Neurological Institute, McGill University, 3801 University Street, Montréal, Québec, CanadaInternational Laboratory for Brain, Music, and Sound Research (BRAMS), 1430 Mont-Royal Boulevard West, Montréal, Québec, CanadaDepartment of Computer Science, Université de Montréal, C.P. 6128, Montréal, Québec, Canada

r t i c l e i n f o

rticle history:eceived 9 February 2009eceived in revised form 16 July 2009ccepted 24 October 2009vailable online 6 November 2009

eywords:udio–vocal integrationuditory feedback

MRIitch shift

a b s t r a c t

In an fMRI experiment, we tested experienced singers with singing tasks to investigate neural correlatesof voluntary and involuntary vocal pitch regulation. We shifted the pitch of auditory feedback (±25 or 200cents), and singers either: (1) ignored the shift and maintained their vocal pitch or (2) changed their vocalpitch to compensate for the shift. In our previous study, singers successfully ignored and compensatedfor 200-cent shifts; in the present experiment, we hypothesized that singers would be less able to ignore25-cent shifts, due to a prepotent, corrective pitch-shift response. We expected that voluntary vocal reg-ulation during compensate tasks would recruit the anterior portion of the rostral cingulate zone (RCZa)and posterior superior temporal sulcus (pSTS), as our earlier study reported; however, we predicted thata different network may be engaged during involuntary responses to 25-cent shifts. Singers were lessable to ignore 25-cent shifts than 200-cent shifts, suggesting that pitch-shift responses to small shifts

ocal control are under less voluntary control than responses to larger shifts. While we did not find neural activityspecifically associated with involuntary pitch-shift responses, compensate tasks recruited a functionallyconnected network consisting of RCZa, pSTS, and anterior insula. Analyses of stimulus-modulated func-tional connectivity suggest that pSTS and intraparietal sulcus may monitor auditory feedback to extractpitch-shift direction in 200-cent tasks, but not in 25-cent tasks, which suggests that larger vocal correc-tions are under cortical control. During the compensate tasks, the pSTS may interact with the RCZa and

untar

anterior insula before vol

. Introduction

Electrophysiological, tracer, and lesion studies in animals haveemonstrated that vocalization recruits a constellation of neural

tructures, ranging from motor/premotor cortical areas [i.e., pri-ary motor cortex, supplementary motor area, anterior cingulate

ortex] and subcortical regions (basal ganglia, thalamus) to an arrayf brainstem structures, including periaqueductal gray, substantia

Abbreviations: ACC, anterior cingulate cortex; aINS, anterior insula; aSTG,nterior superior temporal gyrus; BA, Brodmann area; IPL, inferior parietal lob-le; IPS, intraparietal sulcus; M1, primary motor cortex; mid-PMC, mid-premotorortex; PAC, primary auditory cortex; PostC, postcentral gyrus; pre-SMA, pre-upplementary motor area; pSTG, posterior superior temporal gyrus; pSTS, posterioruperior temporal sulcus; PT, planum temporale; RCZa, anterior portion of rostralingulate zone; SMA, supplementary motor area; SMG, supramarginal gyrus; STG,uperior temporal gyrus; STS, superior temporal sulcus; vPMC, ventral premotorortex.∗ Corresponding author at: Montréal Neurological Institute, Cognitive Neuro-

cience Unit, 3801 University Street, Room 276, Montréal, Québec, Canada H3A 2B4.el.: +1 514 398 8519; fax: +1 514 398 1338.

E-mail address: jean.zarate@mail.mcgill.ca (J.M. Zarate).

028-3932/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.oi:10.1016/j.neuropsychologia.2009.10.025

y vocal pitch correction occurs.© 2009 Elsevier Ltd. All rights reserved.

nigra, reticular formation, and motoneuron pools (Jurgens, 2002).Neuroimaging studies have confirmed that many of these regionsare also involved in human vocalization, including speech andvarious singing tasks (Brown, Martinez, Hodges, Fox, & Parsons,2004; Brown, Martinez, & Parsons, 2006; Jeffries, Braun, & Fritz,2003; Kleber, Birbaumer, Veit, Trevorrow, & Lotze, 2007; Ozdemir,Norton, & Schlaug, 2006; Paus, Petrides, Evans, & Meyer, 1993;Perry et al., 1999; Riecker, Ackermann, Wildgruber, Dogil, & Grodd,2000; Schulz, Varga, Jeffires, Ludlow, & Braun, 2005). Sensory feed-back during vocalization not only stems from proprioception fromthe vocal apparatus but also from auditory feedback processed bytemporal lobe regions [e.g., superior temporal gyrus (STG), supe-rior temporal sulcus (STS)], which process vocal sounds, speech,and other auditory stimuli (Belin, Zatorre, & Ahad, 2002; Scott& Johnsrude, 2003). At times, vocal adjustments are necessary ifthere is a mismatch between the intended and actual vocal out-

put or if the environmental tasks change (e.g., noisy background);this vocal regulation requires the integration of vocal motorcontrol and auditory processes (also known as “audio-vocal inte-gration”), but the neural substrates involved in this process are notwell-understood.

6 sycho

gf1(LBeLKpvi(is&&Litcfritt(liwWvssdastbsiCaricNeSZm1Pma(

rnaSbvtd

08 J.M. Zarate et al. / Neurop

Previous behavioral studies have investigated audio–vocal inte-ration underlying vocal pitch regulation by manipulating auditoryeedback, either by adjusting the feedback amplitude (Lombard,911; Siegel & Pick, 1974) or by altering the fundamental frequencyi.e., perceived pitch) of the auditory feedback (Burnett, Freedland,arson, & Hain, 1998; Burnett & Larson, 2002; Burnett, McCurdy, &right, 2008; Donath, Natke, & Kalveram, 2002; Hafke, 2008; Haint al., 2000; Jones & Keough, 2008; Jones & Munhall, 2000, 2005;arson, 1998; Larson, Burnett, & Kiran, 2000; Natke, Donath, &alveram, 2003; Natke & Kalveram, 2001). Such auditory feedbackerturbations often elicit fast, compensatory adjustments in eitherocal amplitude or pitch, such as the Lombard reflex [an increasen vocal amplitude in response to decreased feedback amplitudeLombard, 1911; Siegel & Pick, 1974)] or the pitch-shift response,n which the vocal pitch is quickly adjusted, often in the oppo-ite direction of the feedback shift (Burnett et al., 1998; Burnett

Larson, 2002). In a previous neuroimaging experiment (ZarateZatorre, 2008), we modified the pitch-shift paradigms used by

arson and Burnett to target cortical substrates of audio–vocalntegration. Rather than delivering pitch-shifted feedback for lesshan 1 s as in the Larson/Burnett studies, we maintained a ±200-ent shift in feedback (one whole tone, in musical terminology)or approximately 3 s to increase the likelihood of capturing neu-al activity associated with audio–vocal integration. Subjects werenstructed either to: (1) ignore the pitch-shifted feedback and keepheir vocal output steady, or (2) compensate for the pitch shift, sohat the shifted feedback would sound like the original target notei.e., cancel out the pitch shift in the feedback). We believed theatter task would recruit the brain regions involved in audio–vocalntegration, since subjects needed to monitor auditory feedback

hile regulating their vocal output to cancel out the feedback shift.e tested non-musicians and experienced singers to determine if

ocal training would modify neural activity associated with theseinging tasks. During our “compensate” task, we found two possibleubstrates for audio–vocal integration, each of which was depen-ent on vocal experience: (1) non-musicians showed increasedctivity in the dorsal premotor cortex (dPMC), and (2) experiencedingers showed increased activity in the anterior portion of the ros-ral cingulate zone (RCZa) and posterior STS (pSTS). The dPMC haseen implicated in selecting movements associated with particularensory cues (Chouinard & Paus, 2006; Petrides, 1986), includ-ng auditory–motor interactions (Chen, Penhune, & Zatorre, 2008;hen, Zatorre, & Penhune, 2006; Zatorre, Chen, & Penhune, 2007),nd thus may serve as a basic sensorimotor interface as people,egardless of vocal experience, adjust their vocal output after hear-ng feedback perturbation. In general, the RCZa is implicated inonflict monitoring (Botvinick, Cohen, & Carter, 2004; Botvinick,ystrom, Fissell, Carter, & Cohen, 1999; Carter et al., 1998; Durstont al., 2003; MacDonald, Cohen, Stenger, & Carter, 2000; Picard &trick, 1996, 2001), while the pSTS processes vocal stimuli (Belin,atorre, Lafaille, Ahad, & Pike, 2000; Kriegstein & Giraud, 2004) anday be involved in extracting specific sound features (Celsis et al.,

999; Warren, Scott, Price, & Griffiths, 2006; Warren, Uppenkamp,atterson, & Griffiths, 2003). We proposed that as people undergoore vocal training or experience, the interface between the RCZa

nd pSTS may be increasingly recruited for audio–vocal integrationZarate & Zatorre, 2008).

Although we outlined possible substrates for voluntary vocalegulation in this prior study, we did not systematically study theeural correlates of the pitch-shift response itself, which is alsoform of vocal regulation that relies on audio–vocal integration.

ince the pitch-shift response may be more involuntary, it maye governed by different substrates than those outlined above foroluntary vocal regulation. In fact, Burnett et al. (1998) suggestedhat the midbrain periaqueductal gray (PAG) may a possible can-idate for audio–vocal integration during the pitch-shift response,

logia 48 (2010) 607–618

due to its connections and its role in vocalization. Electrical andpharmacological stimulation of the squirrel monkey PAG elicitsvocalization (Dujardin & Jurgens, 2005; Suga & Yajima, 1988), andthe human PAG is active during voiced speech when compared towhispered speech, suggesting that the PAG is involved in motornetworks that produce vocal fold activity (Schulz et al., 2005). ThePAG receives input from a huge array of sensory cortical and sub-cortical regions, including higher order auditory areas (e.g., STS),superior and inferior colliculi, lateral lemniscus, and the nucleusgracilis, which suggests that the PAG may be involved in vocalresponses to external stimuli (Dujardin & Jurgens, 2005). The PAGmay receive information about auditory feedback via the inferiorcolliculus (Huffman & Henson, 1990) or the lateral lemniscus andinitiate a quick, compensatory vocal response to any changes infeedback, such as the Lombard reflex (Nonaka, Takahashi, Enomoto,Katada, & Unno, 1997) or the pitch-shift response.

In our earlier study (Zarate & Zatorre, 2008), we made aninteresting observation—during the ignore task, we saw pitch-shift responses only in the non-musicians; we therefore concludedthat vocal training must have helped singers suppress pitch-shiftresponses when asked to ignore a large, 200-cent shift. Giventhat only singers suppressed pitch-shift responses when ignoringlarge pitch perturbations and generally produced more uniformbehavioral results than non-musicians in our previous experi-ment, in the current study, we investigated the neural correlatesof audio–vocal integration during both small pitch-shift responsesand larger, intended vocal adjustments only in experienced singers.In the present experiment, singers performed the same ignore andcompensate tasks from our first experiment, but we utilized twodifferent shift magnitudes: 200-cent and 25-cent pitch shifts. Sinceour previous experiment has already shown that singers can suc-cessfully ignore and compensate for a 200-cent shift, we expectedthat the response magnitudes between these tasks would be sig-nificantly different. In contrast, given that pitch-shift responsesare better suited to fully correct for smaller pitch perturbationsthan larger ones (Liu & Larson, 2007), and hence are thought tobe under more automatic control, we hypothesized that singerswould be less able to suppress pitch-shift responses to 25-centshifts than to 200-cent shifts; thus, we did not expect significant dif-ferences in response magnitudes for ignoring and compensating forthis smaller shift. We predicted that the brain regions that singersrecruited for ignoring and compensating for the large shift would besimilar to those reported in our prior experiment (Zarate & Zatorre,2008). However, during the 25-cent tasks, we hypothesized thatnot only similar regions would be recruited as in the large-shifttasks, but that the PAG would also be specifically recruited duringelicited pitch-shift responses in the ignore task.

2. Materials and methods

2.1. Subjects

A total of 13 healthy subjects were recruited from the McGill University com-munity and surroundings areas. All subjects (mean age = 23 ± 3.93 years old) wereright-handed, had normal hearing, and were devoid of neurological or psycho-logical disorders and contraindications for functional magnetic resonance imaging(fMRI) techniques. All subjects gave informed consent to participate in this study,in accordance with procedures approved by the Research Ethics Committees of theMcConnell Brain Imaging Centre and the Montréal Neurological Institute. Three sub-jects were withdrawn from the study due to problems performing the tasks, andanother subject was excluded for moving excessively during the scanning session.The remaining nine subjects (three male), all categorized as experienced singers,had an average of 11 years (±4.28 years) of formal vocal training and/or experience,were currently practicing or performing at the time of the study, and did not partic-ipate in our previous experiment (Zarate & Zatorre, 2008). According to self-report,

none of the subjects possessed absolute pitch.

2.2. Equipment

During familiarization sessions, subjects sat in front of a lab computer screenand were given a microphone (Røde NT5, Silverwater, Australia) and a pair of

sycho

hamMOahtcptmttjpdIt(a(

2

cpbmtes(savtttsfuipawactDs

TstohiF&s(vai±rmtaar

2

lpw

J.M. Zarate et al. / Neurop

eadphones (Sennheiser HD 280 PRO, Wedemark, Germany) through which alluditory stimuli were delivered. During scanning sessions, the subjects were givenagnetic-resonance (MR) compatible headphones (MR-Confon Peltor Optimex,agdeburg, Germany) and an MR-compatible microphone (FOM-Optimic 2155,ptoacoustics, Or-Yehuda, Israel). All visual cues were back-projected onto a screent the subjects’ feet, and subjects viewed the screen via a mirror attached to theead coil. For both sessions, the microphone was connected to a mixer to amplifyhe voice signal before it was sent to a VoiceOne digital signal processor (TC Heli-on Vocal Technologies, Westlake Village, CA, USA). During the entire experiment,ink noise was delivered through the headphones to reduce bone conduction, sohat the manipulated vocal signal from the digital signal processor would be the

ain source of auditory feedback to the subjects. All auditory stimuli (pink noise,arget vocal waves, and auditory feedback) were delivered to the headphones viahe mixer, and all volume levels were adjusted to comfortable levels for each sub-ect. Pink noise was delivered at an average of 68.3 dB SPL A, while the target waveresentation was presented at an average of 4.1 dB SPL above the pink noise. Theelivery of target waves, visual prompts to cue subjects for singing, and Musical

nstrument Digital Interface (MIDI) system-exclusive messages to control the digi-al signal processor were all controlled by Media Control Functions (MCF) softwareDigiVox, Montréal, Canada). Auditory feedback (via the digital signal processor) andll vocalizations were digitally recorded onto a Marantz PMD-670 digital recorderMarantz Professional, Itasca, IL, USA).

.3. Experimental paradigm

During the familiarization session, subjects practiced singing tasks and controlonditions to prepare for the fMRI scanning session. For all singing tasks, we firstresented a target note and then used a visual cue to prompt subjects to sing the noteack using the syllable /a/. All subjects were trained to sing with minimal mouthovement to reduce movement artifacts in the fMRI session. They were instructed

o keep their jaws slightly open and lips closed, so that at the beginning and end ofvery sung note, only their lips, but not their jaws, moved. Each singing task was pre-ented in blocks of five trials, with the same 2-s target note for each trial [176.99 Hz∼F3) for males, 355.03 (∼F4) for females]. In one task, after hearing the target note,ubjects were cued to sing the note for 4 s (“simple singing”). During pitch-shift tasks,pproximately 1 s after the onset of singing (shift onset range: 1000–1500 ms), theoice was shifted either up or down by 200 cents (one whole tone) or 25 cents viahe digital signal processor and remained shifted until the end of the trial. For theserials, subjects were instructed to make a different response in each of two distinctasks: (1) ignore the shifted feedback and keep the vocal output as steady as pos-ible on the original note (“IGN”), or (2) correct the shifted feedback so that theeedback sounded like the target note (“COMP”). We maintained the feedback shiftntil the end of the trial to increase the probability of finding brain regions involved

n vocal pitch regulation with fMRI techniques. Two control conditions were alsoresented: (1) a condition with only pink noise playing in the background, used tossess “baseline” cortical activity in the MR scanner; and (2) a perception condition,hich presented a target note that subjects did not have to sing back, thus serving

s an auditory control for all singing tasks in the scanner. In both of these controlonditions, subjects were visually cued to breathe out normally, rather than sing;herefore, these conditions also served as a respiratory control for the singing tasks.uring familiarization, the subjects went through four experimental runs with all

inging tasks and control conditions included in each run.A few days after the familiarization session, each subject was tested in a Siemens

rio 3T MR scanner. While in the scanner, subjects were exposed to all of theinging tasks and control conditions presented in the familiarization session. Prioro functional scanning, a high-resolution (voxel = 1 mm3) T1-weighted scan wasbtained for anatomical localization. During the two functional runs, one whole-ead frame of forty contiguous T2*-weighted images were acquired in an ascending,

nterleaved fashion (TE = 60 ms, TR = 10.3 s, 64 × 64 matrix, voxel size = 3.5 mm3,OV = 224 mm2). We utilized a sparse-sampling design (Belin, Zatorre, Hoge, Evans,Pike, 1999)—tasks were performed during the silent periods between scan acqui-

itions to: (1) prevent scanner noise from interfering with the auditory stimuli and2) reduce any effect of movement due to vocalization, since scanning occurred afterocalizations were completed. We also used cardiac-triggered gating to minimizeny pulsatile artifacts in subcortical structures (Guimaraes et al., 1998). Relative tim-ngs between scan acquisitions and tasks were systematically varied or “jittered” by500 ms to maximize the likelihood of obtaining the peak of the hemodynamic

esponse for each task (Belin et al., 1999). Each subject went through two experi-ental runs in the scanner. At the end of the scanning session, the simple singing

ask was presented a total of 20 times, while each pitch-shift task was presentedtotal of 40 times (20 trials for each pitch-shift direction); one brain image was

cquired per trial. The order of all singing tasks and control conditions within eachun was counterbalanced across subjects.

.4. Behavioral analyses

We automated the statistics extraction process using the Python programminganguage in conjunction with de Cheveigné’s Matlab implementation of the YINitch extractor (de Cheveigné & Kawahara, 2002). The individual vocalization filesere first extracted from all subjects’ recordings, each of which spanned the dura-

logia 48 (2010) 607–618 609

tion of each experimental session. To facilitate vocalization extraction, the presentedtarget notes (and auditory feedback via the digital signal processor) were recordedonly in the right channel of a stereo recording, while the left channel received inputonly from the microphone directly (i.e., raw vocal output). By subtracting the enve-lope of the left channel from that of the right, we isolated the target notes fromthe rest of the signal. The beginning of each individual vocalization file was definedas the end of the preceding target. The end of the individual vocalization file wasthe point in time when the signal became silent after having risen above half of itsmaximum amplitude, or onset of the next target, whichever came first. The isolatedindividual vocalization files were then exported as 44.1 kHz audio files.

Once the individual vocalization files were extracted, YIN was used to calculatefundamental frequency (f0), signal power, and aperiodicity every 32 samples [result-ing in a frame rate of 1378.125 Hz, i.e. (44.1/32) kHz]. Within each file, the beginningand end of a given vocalization were defined as the first and last frames for whichthe signal power is greater than 5% of the maximum and the aperiodicity is below0.1. Since YIN normally calculates f0 in octaves relative to 440 Hz, we modified thecode to determine f0 relative to the frequencies of our target waves and then multiplyeach value by 1200 to convert to cents (one octave equals 1200 cents); subsequently,the mean f0 was calculated for each vocalization. For pitch-shift vocalizations, wedefined a pitch-shift window of 100 ms (50 ms before and after the programmedshift time) to account for the digital signal processor’s variability in pitch-shift deliv-ery. Therefore, the pre-shift mean included f0 values from the beginning of eachvocalization to the pitch-shift window. The post-shift mean consisted of f0 valuesfrom the last second of each vocalization. We then subtracted the pre-shift meanfrom the post-shift mean to calculate average response magnitude for each pitch-shift vocalization. For each subject, all mean f0 values and response magnitudes werefirst calculated within each trial and then averaged across all trials within each taskin a specific shift direction (e.g., ignore, shifted up 200 cents; compensate, shifteddown 25 cents).

In our previous experiment, we found no significant differences in the behavioralresults between the familiarization and fMRI sessions, so we present only the fMRIsession results in this paper. In one set of analyses, we used the average responsemagnitude as the dependent variable for pitch-shift tasks. After separating tasks byshift direction (i.e., up or down), the pitch-shift results were analyzed with two-wayrepeated-measures analyses of variance (ANOVAs, instruction by shift magnitude).In a second set of analyses, we converted response magnitudes to percentages of theshift magnitude by dividing the absolute response magnitude in each pitch-shift trialby the absolute pitch-shift magnitude and multiplying each value by 100; this helpedus determine how much correction was produced either by pitch-shift responses orvoluntary vocal pitch changes. The percent response magnitudes were analyzedusing a three-way ANOVA (instruction by shift magnitude by shift direction). TheScheffé test was used for all post hoc analyses.

2.5. fMRI analyses

To correct for motion artifacts, all blood-oxygen-level-dependent (BOLD)images from both functional runs were realigned with the fourth frame of the firstrun using the AFNI software (Cox, 1996). To increase the signal-to-noise ratio ofthe imaging data, the images were spatially smoothed with an 8-mm full-widthat half-maximum (fwhm) isotropic Gaussian kernel. Prior to analysis, the first fourframes were excluded from further analyses to remove T1-saturation effects; theseframes were acquired either during practice singing trials or presented instruc-tions. For each subject, we conducted our image analyses in a similar fashion tothat described in our first paper (Zarate & Zatorre, 2008), using fMRISTAT, whichinvolves a set of four Matlab functions that utilize the general linear model for anal-yses (Worsley et al., 2002). The motion-correction parameters obtained with theAFNI software were used as covariates in fMRISTAT to further account for motionartifacts in the imaging results. Before group statistical maps for each contrast ofinterest were generated, in-house software was used to non-linearly transform eachsubject’s anatomical and functional images into standardized MNI/ICBM stereotaxiccoordinate space, using the non-linearly transformed, symmetric MNI/ICBM 152template (Collins, Neelin, Peters, & Evans, 1994; Mazziotta et al., 2001; Talaraich& Tournoux, 1988). The program stat summary assessed the threshold for signifi-cance by selecting the minimum among the values given by a Bonferroni correction,random field theory, and the discrete local maximum to account for multiple com-parisons (Worsley, 2005). The threshold for a significant peak was t = 4.9 at p = 0.05,using a whole-brain search volume. We report peaks of neural activity if their voxelor cluster p-values are less than 0.05. While some peaks did not meet the criticalthreshold, they fell within regions previously reported in our earlier study (Zarate& Zatorre, 2008). For these a priori regions, we corrected the threshold for smallvolumes and report peaks if their corrected voxel or cluster p-values are 0.05 orless.

In our analyses of functional connectivity, the general linear model was fittedto account for the neural activity due to a stimulus (e.g., any singing task). Then,

the remaining or residual activity within a specific voxel (the “seed” voxel) wasregressed on the activity within the rest of the brain (on a voxel-by-voxel basis) todetermine where activity significantly covaries with the activity at that seed voxel,without the effect of a stimulus (Friston, 1994; Worsley, Charil, Lerch, & Evans, 2005).We also performed analyses of stimulus-modulated functional connectivity, whichassessed how the connectivity is affected by the stimulus or task of interest (Friston

610 J.M. Zarate et al. / Neuropsychologia 48 (2010) 607–618

F -shiftew onsesw

er

ibtti

naEo(d(2

2

avf

3

3

usmprlbCpb

abrp2vlpap

2008), including bilateral primary and secondary auditory areas,bilateral sensorimotor mouth regions, bilateral supplementarymotor areas (SMA), right ventral premotor cortex, left thalamus,left lateral globus pallidus, and bilateral medial geniculate nuclei(Supplementary Table S1).

Fig. 2. Percent response magnitudes (±S.E.) for IGN and COMP tasks collapsed across

ig. 1. Average response magnitudes (±S.E.) for tasks with auditory feedback pitchere significantly larger than in the other tasks (marked with *, ps < 0.001). The respith +, p < 0.07).

t al., 1997). Using stat summary, the critical t-thresholds for connectivity analysesanged from 5.00 to 5.08 (all ps = 0.05, corrected for multiple comparisons).

To perform conjunction analyses with two contrasts of interest, we utilized ann-house tool called mincmath to find the minimum t-statistic at each voxel acrossoth contrast images. The conjunction results were then tested against the “conjunc-ion null hypothesis”, which entailed using the critical t-values for just one contrast,o determine whether there was significant neural activity in certain brain regionsn both contrasts (Nichols, Brett, Andersson, Wager, & Poline, 2005).

The locations of peak neural activity or connectivity were classified using: (1)euroanatomical atlases (Duvernoy, 1991; Talaraich & Tournoux, 1988); (2) prob-bilistic maps or profiles for the Heschl’s Gyrus (Penhune, Zatorre, MacDonald, &vans, 1996), planum temporale (Westbury, Zatorre, & Evans, 1999), mouth regionf the sensorimotor cortex (Fox et al., 2001), inferior frontal gyrus pars opercularisTomaiuolo et al., 1999), and basal ganglia (Ahsan et al., 2007); and (3) locationsefined by previous reports or reviews on the medial frontal and cingulate areasPicard & Strick, 1996, 2001) and subdivisions of the premotor cortex (Chen et al.,008).

.6. Data exclusions

For behavioral analyses, 28 out of 1620 fMRI recordings were excluded fromnalyses due to equipment failure, subject-performance error, or problems withocalization extraction. For fMRI analyses, 106 out of 2160 frames were excludedrom analyses due to equipment failure or performance errors.

. Results

.1. Behavioral results

The two-way ANOVAs performed separately on downward- andpward-shifted tasks gave similar results. Both ANOVAs revealedignificant two-way interactions between instruction and shiftagnitude [down: F(1,8) = 504.85, p < 0.001; up: F(1,8) = 306.22,< 0.001]. Scheffé post hoc tests determined that as expected, the

esponses to compensating for a 200-cent shift (COMP200c) werearger than responses to ignoring the 200-cent shift (IGN200c) andoth 25-cent tasks (COMP25c and IGN25c; Fig. 1, ps < 0.001). TheOMP25c responses were larger than IGN200c responses (Fig. 1;< 0.001 for downward pitch-shift, p < 0.07 for upward pitch-shift)ut were not significantly different from IGN25c responses.

While there were no significant differences between IGN200cnd IGN25c response magnitudes, the IGN200c responses inoth directions were closer to 0-cents magnitude than IGN25cesponses, suggesting that singers were more capable of sup-ressing prepotent pitch-shift responses to 200-cent shifts than to5-cent shifts. To test for this directly, we converted the absolute

alues of all response magnitudes to percentages of the abso-ute pitch-shift magnitude. The three-way ANOVA performed onercent response magnitudes revealed a significant two-way inter-ction between instruction and shift magnitude [F(1,8) = 21.86,< 0.01], and post hoc tests determined that percent response

d (a) downwards and (b) upwards. In both directions, the responses to COMP200cto COMP25c were larger than those in IGN200c (marked with !, p < 0.001; marked

magnitudes in IGN200c, IGN25c, COMP200c, and COMP25c weresignificantly different from each other (Fig. 2, all ps < 0.05), withthe exception of COMP200c and IGN25c (p > 0.1). Therefore, singersproduced significantly larger pitch-shift responses while ignoringa 25-cent shift than a 200-cent shift. While singers did not correctfully for the 200-cent shift (87.66% correction), they overcom-pensated for the 25-cent shift (112.67% correction). Additionally,since percent response magnitudes were significantly differentbetween both 25-cent tasks, this overcompensation suggests thatthe response magnitudes in the COMP25c task cannot be solelyattributed to involuntary pitch-shift responses, but rather thatsingers voluntarily attempted to correct for the small perturbation.

3.2. fMRI results

3.2.1. Basic functional network for simple singingSimple singing, when contrasted with perception, recruited a

functional network similar to that seen in our and others’ previousexperiments (Kleber et al., 2007; Perry et al., 1999; Zarate & Zatorre,

shift direction. The percent response magnitude for IGN200c was smaller than in allother tasks (marked with *, ps < 0.001). Similarly, the percent response magnitudefor COMP25c was larger than responses in other tasks (marked with !, ps < 0.05). Thepercent response magnitudes for COMP200c and IGN25c were both significantly dif-ferent than IGN200c and COMP25c (marked with #, ps < 0.001) but were not differentfrom each other.

J.M. Zarate et al. / Neuropsychologia 48 (2010) 607–618 611

Table 1Functional networks associated with ignoring pitch-shifted feedback.

(a) IGN200c ∩ IGN25c (b) IGN200c–IGN25c

Left Right Left Right

x y z t x y z t x y z t x y z t

Auditory STG 66 −16 10 4.4STS 62 −20 2 4.2pSTS 58 −36 8 3.5Planum temporale −58 −22 10 4.5 52 −20 6 3.7 −60 −44 20 4.4

Frontal BA 6/44 52 10 30 4.8

Parietal Supramarginal gyrus −52 −46 22 4.1

R n (a) sw orrect

3f

tacrrjrottasFwrS

Fcec

Angular gyrus

egions of peak neural activity during the IGN200c and IGN25c singing tasks. Sectioith more activity during IGN200c than during IGN25c. All peak/cluster ps ≤ 0.05, c

.2.2. Additional brain regions involved in ignoring pitch-shiftedeedback

Since we had no specific hypotheses about the direction ofhe pitch shift, we combined the imaging results for each taskcross both shift directions. When singers ignored either a 200-ent or a 25-cent shift, they recruited a similar network ofegions in addition to the basic network for singing (specificegions in each task are listed in Supplementary Table S2). Con-unction analyses between IGN200c and IGN25c determined thatight Brodmann area (BA) 6/44 (ventral premotor cortex and parspercularis of the inferior frontal gyrus) and bilateral planumemporale were recruited for both tasks (Table 1a, Fig. 3a). A con-rast between these tasks showed that IGN200c required morectivity in right STG, STS and pSTS, and left planum temporale,

upramarginal gyrus, and angular gyrus than IGN25c (Table 1b,ig. 3a), but no regions showed significantly increased activityhen IGN25c was contrasted with IGN200c, since overall neu-

al activity was weaker during IGN25c (Supplementary Table2).

ig. 3. Brain regions associated with pitch-shift tasks when compared to simple singinompared with simple singing. Right: Auditory areas and supramarginal gyrus displayngaged a shared functional network when compared to simple singing. Right: Extensontrasted with COMP25c. All peak/cluster ps ≤ 0.05, corrected. Refer to legend for abbre

−58 −56 20 3.5

hows the shared regions between both IGN tasks, while section (b) displays regionsed. Refer to legend for abbreviations.

3.2.3. Additional brain regions involved in compensating forpitch-shifted feedback

As singers corrected for either the 200-cent or the 25-cent shift,they displayed similar patterns of increased neural activity (specificregions recruited during each task are shown in SupplementaryTable S3). Conjunction analyses between COMP200c and COMP25cshowed a common network with increased activity in bilateralBA 6/44, anterior insulae, pre-SMA, right RCZa, bilateral mid-premotor cortex (mid-PMC), intraparietal sulci, and supramarginalgyri, and right STS and planum temporale (Table 2a, Fig. 3b). A con-trast between both tasks revealed more activity within bilateralplanum temporale, STG, STS, and right pSTS during COMP200c thanCOMP25c (Table 2b, Fig. 3b), which is similar to the increased audi-tory cortical activity observed in the contrast between IGN200c and

IGN25c.

3.2.4. Functional connectivity during pitch-shift tasksFor connectivity analyses in IGN tasks, we chose a seed voxel

in the right pSTS since this region displayed more activity in

g. (a) Left: The shared network of regions recruited during both IGN tasks whened more activity during IGN200c than during IGN25c. (b) Left: Both COMP tasksive increases in auditory cortical activity were associated with COMP200c whenviations.

612 J.M. Zarate et al. / Neuropsychologia 48 (2010) 607–618

Table 2Functional networks associated with compensating for pitch-shifted feedback.

(a) COMP200c ∩ COMP25c (b) COMP200c–COMP25c

Left Right Left Right

x y z t x y z t x y z t x y z t

Auditory aSTG −58 −4 0 4.1 56 4 −2 4.6STG −58 −24 6 3.9 68 −16 8 4.6STS 64 −28 4 3.9 −66 −18 2 3.3 64 −18 4 4.1pSTS 52 −42 12 4.1Planum temporale 66 −22 10 4.1 −62 −34 14 4.7 64 −28 16 5.5

Motor RCZa (ACC BA 32) 2 20 40 3.8Pre-SMA −4 4 54 5.7 2 18 44 3.8Mid-PMC −48 0 42 4.6 44 0 46 5.0

Multimodal Anterior insula −34 22 2 5.4 32 22 2 6.2

Frontal BA 6/44 −52 8 30 4.4 52 10 24 7.6Inf. frontal (BA 44) −40 16 26 3.4

Parietal Intraparietal sulcus −42 −40 46 4.5 40 −44 48 4.4−36

R ectionr r ps ≤

e(etrtStpsw

FwsdatR

Supramarginal gyrus −36 −48 48 4.3 58

egions of peak neural activity during the COMP200c and COMP25c singing tasks. Segions with more activity during COMP200c than during COMP25c. All peak/cluste

xperienced singers than non-musicians in our first experimentZarate & Zatorre, 2008) and was also active during IGN200c in thisxperiment. Table 3 and Fig. 4a show a vast network of regionshat are functionally connected to right pSTS, including auditoryegions, motor and premotor regions, insulae, BA 44, postcen-ral gyri, inferior parietal lobule, and various subcortical regions.

timulus-modulated functional connectivity analyses determinedhat the IGN200 task modulated the connectivity between rightSTS and right intraparietal sulcus, bilateral postcentral gyri, rightensorimotor cortex, and a few regions along the posterior medialall, when compared to the effect of simple singing (Table 3,

ig. 4. Functional and stimulus-modulated functional connectivity in IGN200c and COMith a right pSTS seed voxel (MNI/ICBM152 world coordinates 54, −42, 12; all voxel p

pecifically enhanced connectivity of the right pSTS seed voxel with right intraparietalifferent overlap patterns between three connectivity maps during COMP200c tasks, gennterior insula (32, 22, 2); all voxel ps ≤ 0.001, uncorrected. The Venn diagram above depiask specifically enhanced connectivity between the right pSTS voxel and bilateral intraparefer to legend for abbreviations.

46 4.3

(a) shows the shared regions between both COMP tasks, while section (b) displays0.05, corrected. Refer to legend for abbreviations.

Fig. 4b). Analyses of stimulus-modulated functional connectivityrevealed no significant differences in task-modulated connectiv-ity between IGN25c and simple singing or between IGN25c andIGN200c.

For all connectivity analyses in COMP tasks, we chose seed vox-els in the right pSTS and right RCZa, since these regions were

more active in experienced singers during COMP200c than in non-musicians in our previous experiment (Zarate & Zatorre, 2008). Wealso chose a seed voxel in the right anterior insula, since this regionwas originally part of our hypothesized network for audio–vocalintegration and was significantly active in all COMP tasks in this

P200c tasks. (a) The functional connectivity map during IGN200c tasks, generateds ≤ 0.001, uncorrected). (b) When compared to simple singing, the IGN200c tasksulcus and bilateral postcentral gyri. All peak/cluster ps ≤ 0.05, corrected. (c) Theerated with seed voxels in right pSTS (54, −42, 12), right RCZa (2, 20, 40), and rightcts the color legend used to show overlap in connectivity maps. (d) The COMP200cietal sulci when compared with simple singing. All peak/cluster ps ≤ 0.05, corrected.

J.M. Zarate et al. / Neuropsychologia 48 (2010) 607–618 613

Table 3Connectivity associated with ignoring pitch-shifted feedback.

Functional connectivity Stimulus-modulated connectivity

R pSTS R pSTS

x y z t x y z t

Auditory Left PAC −38 −30 12 4.1Left pSTG −56 −54 18 4.5Left STS −56 −30 6 5.2Right STS 46 −30 0 5.1Left pSTS −60 −56 20 4.6Left planum temporale −58 −44 20 7.3Right planum temporale 36 −32 20 5.0

Motor Right ACC—BA 32 (RCZa) 10 12 40 3.9Left SMA −8 −12 66 4.0Right SMA 10 −6 54 5.9Left pre-SMA −10 2 58 3.3Right pre-SMA 8 2 58 7.2Right M1 40 −2 50 4.6Right mid-PMC 44 0 40 4.3Left vPMC −58 12 10 4.2Right vPMC 50 4 32 5.0Left subcentral −48 −2 10 3.8Right subcentral 50 −2 8 4.0Right central/rolandic operculum 42 2 16 3.9

Multimodal Right anterior insula 30 16 8 3.7Left posterior insula −34 −26 10 3.5Right posterior insula 32 −26 20 4.8Right BA 6/44 42 12 28 4.6Right sensorimotor cortex 30 −22 46 4.7Right posterior cingulate 14 −24 40 3.2Left paracentral lobule −12 −32 62 3.7Right paracentral lobule 8 −30 54 3.8

Frontal Left inferior frontal—BA 44 −52 12 6 5.5

Parietal Left postcentral −40 −42 58 5.6 −10 −34 68 5.2Right postcentral 30 −32 50 4.6Right postcentral (opercular) 50 −20 18 4.2Right IPL 62 −26 20 6.0Right IPS 28 −40 38 5.3Right parietal operculum 50 −22 16 4.5

Thalamus Right thalamus 12 −12 8 4.1

Basal Ganglia Left putamen −28 −10 0 3.8Right putamen 30 −10 −2 4.6Left lateral globus pallidus −26 −18 2 3.9Right lateral globus pallidus 22 −16 10 6.7

L ht pSTc ging.

emtahrtumanrl(rtccbC

eft: Brain regions whose activity is significantly correlated to activity within rigonnectivity with the right pSTS during the IGN200c task compared with simple sin

xperiment. Table 4 (and Supplementary Table S4) shows thatost of the regions recruited during COMP200c are also func-

ionally connected to each other, with the exceptions of the RCZand anterior insula seed voxels with pSTS. The pSTS seed voxel,owever, is functionally connected with both the RCZa and ante-ior insula on a subthreshold level. In fact, Fig. 4c demonstrateshe overlap between connectivity maps (all thresholded at t = 3.17,ncorrected p = 0.001), and all three seed voxels overlap in theedial motor regions, bilateral BA 6/44 and anterior to mid-insulae,

nd left planum temporale. In stimulus-modulated functional con-ectivity analyses, we found that the connectivity between theight pSTS and bilateral intraparietal sulci was significantly modu-ated by COMP200c when compared to the effect of simple singingTable 4, Fig. 4d). Interestingly, this is similar to the connectivityesults for IGN200c—the IGN200c task also significantly enhanced

he connection between right pSTS and right intraparietal sul-us. As seen in the IGN tasks, analyses of stimulus-modulatedonnectivity revealed no significant differences in connectivityetween COMP25c and simple singing or between COMP25c andOMP200c.

S (54, −42, 12) during IGN200c trials. Right: Brain regions displaying enhancedAll peak/cluster ps ≤ 0.05, corrected. Refer to legend for abbreviations.

4. Discussion

4.1. Behavioral results

As seen in our previous experiment (Zarate & Zatorre, 2008),singers were capable of both ignoring and compensating for thelarge shift. However, as predicted, pitch-shift responses to smallfeedback perturbations were not easily suppressed in the ignoretask, as demonstrated by a larger percent response magnitude dur-ing IGN25c than during IGN200c. For all behavioral analyses, weanalyzed the last second of each vocalization, which correspondsto a late pitch-shift response that begins 300 ms after the pitchshift and is subject to voluntary control (Burnett et al., 1998; Hainet al., 2000)—including our ignore and compensate instructions.Even though singers were instructed to keep their vocal output

steady in the IGN25c task, their late pitch-shift response alreadymay have been influenced by the more automatic early pitch-shiftresponse, which is elicited 100–150 ms after the pitch shift (Burnettet al., 1998; Hain et al., 2000). Similar pitch-shift aftereffects thatoccur much later than the early pitch-shift response have been

614 J.M. Zarate et al. / Neuropsychologia 48 (2010) 607–618

Table 4Connectivity associated with compensating for pitch-shifted feedback.

Functional connectivity Stimulus-modulated connectivity

R pSTS R pSTS

x y z t x y z t

Auditory Left PAC −46 −22 14 3.9Right PAC 52 −14 8 4.9Right STG 58 −8 8 5.9Left pSTG −58 −40 14 5.4Left STS −50 −30 0 4.5Right STS 62 −20 2 5.6Left pSTS −60 −48 14 5.7Left planum temporale −60 −38 12 5.0Right planum temporale 62 −22 4 5.7

Motor Left ACC—BA 24 −4 4 30 4.2Right ACC—BA 24 2 −8 32 4.1Left ACC—BA 32 (RCZa) −6 8 40 4.9Right ACC—BA 32 (RCZa) 8 14 32 4.2Right SMA 2 0 50 4.1Right pre-SMA 12 8 54 4.2Left M1 −52 −4 38 4.0Right M1 56 −4 34 4.9Left vPMC −56 8 20 4.9Right vPMC 48 2 38 4.0Left subcentral −52 −6 22 3.9Right subcentral 60 −10 18 4.5

Multimodal Left posterior insula −32 −26 16 4.8Right posterior insula 32 −28 18 4.7Left BA 6/44 −46 10 32 4.8Right BA 6/44 44 10 32 7.1

Frontal Left inferior frontal—BA 44 −44 16 18 4.4

Parietal Left postcentral −48 −20 18 3.8Right postcentral 62 −22 26 5.4Left IPL −60 −24 38 4.3Right IPL 58 −26 34 5.1Left IPS −30 −54 40 4.2 −38 −52 56 4.2Right IPS 32 −46 40 4.5 32 −44 44 5.1Left supramarginal −54 −38 30 5.0Right supramarginal 60 −36 30 4.1

Thalamus Left thalamus −10 −18 6 5.3Right thalamus 8 −14 6 5.9

Basal ganglia Left putamen −26 −24 14 4.3Right putamen 30 10 4 4.9Left lateral globus pallidus −20 −4 2 4.5Right lateral globus pallidus 26 −12 2 3.4

L TS (54w eak/cl

rqcaebvpmmt

4

Ctsp

eft: Brain regions whose activity is significantly correlated to activity within right pSith the right pSTS during the COMP200c task compared with simple singing. All p

eported—even after the shifted feedback was turned off, subse-uent vocalizations with normal auditory feedback still showed aompensatory adjustment from the original vocal pitch (Donath etl., 2002; Jones & Keough, 2008; Jones & Munhall, 2000, 2005; Natket al., 2003). The compensatory early pitch-shift response helps sta-ilize the vocal motor system and correct for minor errors duringocalization (Burnett et al., 1998; Liu & Larson, 2007). For smalleritch perturbations, we suggest that early pitch-shift responsesay be more robust and thus influence late pitch-shift responses,aking them less amenable to voluntary control during IGN25c

han during IGN200c.

.2. fMRI results

The functional networks recruited during IGN200c andOMP200c, when compared to simple singing, were similar tohose reported in our previous paper, despite variability fromlightly different experimental protocols (e.g., subject pool sam-ling, different magnetic field strengths, the use of different

, −42, 12) during COMP200c. Right: Brain regions displaying enhanced connectivityuster ps ≤ 0.05, corrected. Refer to legend for abbreviations.

methods and templates for resampling brain images into stereo-taxic space).

Although singers were more successful at ignoring a large shiftthan a small shift, we still found that both IGN tasks engaged a simi-lar functional network for maintaining vocal output in the presenceof altered auditory feedback, including bilateral planum temporaleand right BA 6/44; we also determined that these regions were func-tionally connected with each other and additional cortical regions.A contrast between the two tasks showed that IGN200c recruitedmore activity within various auditory areas (including right pSTS)and left supramarginal gyrus than IGN25c. The left supramarginalgyrus has been recently associated with pitch memory (Gaab, Gaser,& Schlaug, 2006; Gaab, Gaser, Zaehle, Jancke, & Schlaug, 2003) andtherefore may be recruited to keep the original target note in mind

as singers maintain their vocal output on that pitch and ignore thelarge 200-cent pitch shift.

While singers undercorrected for the 200-cent shift and over-corrected for the 25-cent shift, they recruited a similar functionalnetwork for voluntary vocal adjustments in both COMP tasks,

sycho

iastwptt

4f

amscttpefKsotepstpf2oaZomCccccwctrGdppmia

4

ttt1ciaMBa

J.M. Zarate et al. / Neurop

ncluding right STS and planum temporale, right RCZa, bilateralnterior insulae and BA 6/44, and bilateral intraparietal sulci andupramarginal gyri, regardless of the shift magnitude. We foundhat during the COMP200c task, most of the regions within this net-ork were functionally connected with each other, particularly theSTS, RCZa, and anterior insula. When we compared the two COMPasks, we found that COMP200c required more activity within audi-ory regions, including right pSTS, than COMP25c.

.2.1. Posterior STS: a possible substrate for monitoring auditoryeedback

In both 200-cent tasks, we found increased auditory corticalctivity when compared to 25-cent tasks. Recent fMRI experi-ents have also reported that larger pitch changes in auditory

timuli engaged more auditory cortical activity than smaller pitchhanges (Hyde, Peretz, & Zatorre, 2008; Rinne et al., 2007). Whilehis enhancement of auditory cortical activity may be attributedo the salience of larger pitch changes, we propose that the rightlanum temporale, which is involved in pitch processing (Hydet al., 2008), and the right pSTS, which extracts particular soundeatures from vocal stimuli (Belin et al., 2000; Celsis et al., 1999;riegstein & Giraud, 2004; Warren et al., 2006, 2003), are alsopecifically recruited as singers monitor their auditory feedback inur 200-cent tasks. In support of the pSTS’s proposed role, we notehat connectivity between the right pSTS and intraparietal sulci wasnhanced in both COMP200c and IGN200c tasks, compared to sim-le singing. Rinne et al. (2007) also found that the intraparietalulcus was recruited in response to larger pitch shifts in the audi-ory discrimination task. The cortex within the intraparietal sulcuslays a role in somatosensory and visuo-spatial transformationsor motor tasks (Astafiev et al., 2003; Grefkes, Ritzl, Zilles, & Fink,004; Tanabe, Kato, Miyauchi, Hayashi, & Yanagida, 2005), and inur previous paper, we proposed that the intraparietal sulcus maylso be involved in frequency-related transformations (Zarate &atorre, 2008); the intraparietal sulcus’ involvement in these typesf operations is further demonstrated by its recruitment duringusical transposition tasks (Foster & Zatorre, in press). Thus for the

OMP200c tasks, the pSTS may interact with the intraparietal sul-us to extract the pitch-shift direction to prepare the ensuing vocalorrection in the proper direction. During the IGN200c tasks, theonnectivity between the right pSTS and bilateral somatosensoryortex was also enhanced. The frequency information extractedithin the pSTS and further processed within the intraparietal sul-

us may be combined with somatosensory information to maintainhe current vocal output and ensure that it does not change inesponse to the pitch-shifted feedback (see Kleber, Veit, Birbaumer,ruzelier, & Lotze, in press). Since functional connectivity analysesetermined that similar regions have correlated activity with rightSTS in both COMP200c and IGN200c, we speculate that if a com-ensatory pitch-shift response occurred during IGN200c, singersay then utilize the rest of the functionally connected network,

ncluding RCZa and anterior insula, to readjust their vocal outputnd correct the pitch-shift response.

.2.2. The role of the insula in audio–vocal integrationIn our previous experiment, we originally hypothesized that

he anterior insula played a role in audio–vocal integration forhree reasons: (1) this region has reciprocal connections with audi-ory areas and the anterior cingulate cortex (Mesulam & Mufson,982; Mufson & Mesulam, 1982); (2) the anterior insula’s cytoar-hitecture and projections make this region more amenable to

ntegrating auditory input with other modalities, including visualnd vocal motor systems (Ackermann & Riecker, 2004; Bamiou,usiek, & Luxon, 2003; Bushara, Grafman, & Hallett, 2001; Lewis,

eauchamp, & DeYoe, 2000; Rivier & Clarke, 1997); and (3) thenterior insula may be involved specifically in audio–vocal integra-

logia 48 (2010) 607–618 615

tion since its activity is enhanced during overt speech and singingwhen compared with covert or internal vocalization (Riecker et al.,2000). Although singers displayed increased activity in the ante-rior insula during the compensate task in our prior study, thisactivity did not survive the group comparison between singersand non-musicians (Zarate & Zatorre, 2008). The anterior insulamay have been recruited to a much lower, subthreshold level innon-musicians, and since the insula was only weakly active insingers, the group contrast did not show any significant differencesin insular activity. Accordingly, we did not report the insula as anexperience-dependent substrate for audio–vocal integration. In thepresent experiment, however, the anterior insula was one of themost strongly recruited regions during both COMP tasks, so wecannot dismiss its possible role in audio–vocal integration.

While anatomical studies report that the insula shares con-nections with auditory regions and the anterior cingulate cortex(Mesulam & Mufson, 1982; Mufson & Mesulam, 1982), most audi-tory regions have connections with the mid-dorsal and posteriorinsula (Augustine, 1996); this is supported by our functional con-nectivity results. Yet, the anterior insula may still receive input fromauditory regions via intra-insular connections (Augustine, 1996)and via higher order auditory areas, as demonstrated by a weakcorrelation in activity between the anterior insula and planum tem-porale in this experiment. Additionally, Augustine (1996) reportedthat the anterior insula shares connections with BA 24; our over-lapping connectivity maps support this and also demonstrate thatthe anterior insula is functionally connected with BA 32. Together,these cingulate areas are classified as the rostral cingulate zone(RCZ), which can be subdivided into a posterior portion (RCZp) andan anterior portion, RCZa (Picard & Strick, 1996, 2001). The RCZp isassociated with voluntary response selection, including speech andsinging (Paus et al., 1993; Picard & Strick, 1996, 2001). The RCZa isinvolved in conflict monitoring in a variety of contexts (Botvinicket al., 2004, 1999; Carter et al., 1998; Durston et al., 2003) and maybe recruited in our COMP tasks due to the conflict between theintended note and altered auditory feedback. In summary, the ante-rior insula may be classified as a higher order association area dueto its projections and cortical architecture (Rivier & Clarke, 1997).With its connections with auditory regions and the anterior cin-gulate cortex, the anterior insula may modulate vocalizations byintegrating auditory processes (via pSTS) with conflict monitoring(via RCZa) and vocal output selection (via RCZp) and thus contributeto audio–vocal integration.

4.2.3. The role of BA 6/44 during vocal pitch regulationIn all pitch-shift tasks, regardless of the shift magnitude, the

right ventral premotor cortex (vPMC; BA 6) and the right parsopercularis of the inferior frontal gyrus (BA 44) were significantlyactive when compared to simple singing. Both of these regionsare associated with vocal motor planning and control (Binkofski& Buccino, 2004, 2006)—the vPMC is implicated in singing andovert speech-related tasks (Ghosh, Tourville, & Guenther, 2008;Nakamura, Dehaene, Jobert, Le, & Kouider, 2007; Perry et al.,1999), whereas lesions or electrical stimulation in pars operculariscan impair or arrest speech (Jurgens, 2002; Quinones-Hinojosa,Ojemann, Sanai, Dillon, & Berger, 2003). The vPMC is also involvedin transforming spatial information into a motor response (Fogassiet al., 2001; Rizzolatti, Fogassi, & Gallese, 2002); spatial informationmay stem from the intraparietal sulcus, since a recent diffusion-weighted imaging tractography study reported that the vPMChas connections with this particular area (Rushworth, Behrens,

& Johansen-Berg, 2006). During our pitch-shift tasks, we proposethat the pitch-shift direction may be encoded by the intraparietalsulcus, and that information is transmitted to the vPMC, whichis co-activated with the pars opercularis. This co-activation maybe attributed to both cytoarchitectural similarities between the

6 sycho

rceRotp

4

siptUctnstpsnosci

cteesooesaasfmrip

5

nuiZapataAersaEP

16 J.M. Zarate et al. / Neurop

egions (Binkofski & Buccino, 2004, 2006) and high probabilisticonnections between the vPMC and pars opercularis (Rushwortht al., 2006). Since BA 44 and premotor areas, including vPMC andCZp, interact with the primary motor cortex and other regionsf the vocal motor system (Jurgens, 2002), BA 6/44 may also con-ribute to executing the correct voluntary vocal response in ouritch-shift tasks.

.2.4. Investigating the involuntary pitch-shift responseOne of the goals of this experiment was to determine the neural

ubstrates of the involuntary pitch-shift response. Since the PAGs implicated in initiating vocal responses to external stimuli (e.g.,itch-shifted feedback, Dujardin & Jurgens, 2005), we hypothesizedhat this region may play a crucial role in the pitch-shift response.nfortunately, we did not find increased neural activity specifi-ally associated with the pitch-shift responses during the IGN25cask, nor did we find any regions with significantly modulated con-ectivity due to 25-cent singing tasks when compared to simpleinging or either of the 200-cent tasks. However, it is indicativehat the increased functional connectivity between pSTS and intra-arietal sulcus was only found in the two 200-cent tasks relative toimple singing, and not in either of the 25-cent tasks. This findingeeds to be confirmed, but it would be broadly compatible with theverall concept we propose in this paper—responses to 200-centhifts are under greater voluntary control than responses to 25-ent shifts, and in turn, this is the consequence of greater functionalnteractions between the two cortical regions.

We suggest that the imaging results reported in this paper mayoincide with the late pitch-shift response. As previously discussed,his late component may have been influenced by the more robustarly component during the 25-cent tasks. Since the latency of thearly pitch-shift response is only 100–150 ms, while the sparse-ampling fMRI paradigm captures neural activity only on the orderf seconds, the temporal resolution of our fMRI protocol hinderedur ability to capture the neural substrate of this response. Futurexperiments designed to capture the neural correlates of the pitch-hift response may require a continuous acquisition sequence tonalyze the hemodynamic response function time delay in corticalnd subcortical regions (see Brass & von Cramon, 2002), but thecanner noise may interfere with vocal production and auditoryeedback manipulation. Alternatively, EEG/ERP or MEG methods

ay complement our fMRI studies, since they have greater tempo-al resolution than fMRI methods and may reveal crucial temporalnformation about the interaction between regions governing theitch-shift response.

. Conclusion

In this experiment, we tested experienced singers to investigateeural correlates of voluntary (via the COMP tasks) and invol-ntary vocal pitch regulation (i.e., elicited pitch-shift responses

n the IGN25c task). As seen in our first experiment (Zarate &atorre, 2008), experienced singers were capable of correcting fornd ignoring a 200-cent pitch shift. While singers almost com-letely suppressed pitch-shift responses in IGN200c, they were lessble to suppress pitch-shift responses during IGN25c; this suggestshat pitch-shift responses to smaller shifts may be more robustnd under less voluntary control than responses to larger shifts.lthough we could not verify the specific neural substrates gov-rning audio–vocal integration during the involuntary pitch-shift

esponse, we confirmed that the previously hypothesized sub-trates of audio–vocal integration, the anterior cingulate cortex,uditory cortex, and the insula (see Ackermann & Riecker, 2004;liades & Wang, 2003; Muller-Preuss, Newman, & Jurgens, 1980;erry et al., 1999; Riecker et al., 2000; Zarate & Zatorre, 2008), are

logia 48 (2010) 607–618

involved in audio–vocal integration. More specifically, subdivisionsof these regions, namely the RCZa, pSTS, and anterior insula arerecruited as experienced singers voluntarily adjust their vocal pitchduring the COMP tasks, regardless of the pitch-shift magnitude.Importantly, the stimulus-modulated connectivity results suggestthat the pSTS is specifically involved in monitoring auditory feed-back, and via connections with the intraparietal sulcus, encodes thedirection of pitch shifts in our 200-cent tasks, whereas this seemsnot to be the case in the 25-cent tasks. Connectivity analyses alsoindicate that the pSTS is functionally connected with the anteriorcingulate cortex and the insula. Thus, the pSTS may be involved ina network that routes pitch-shift information to the RCZa eitherdirectly through its shared connections with the anterior cingu-late cortex or indirectly via its connections with the insula. TheRCZa may register cognitive conflict due to the mismatch betweenshifted auditory feedback and the intended vocal output, and sub-sequently initiate proper vocal pitch correction via its connectionswith the RCZp and the rest of the vocal motor system.

Broadly speaking, the functional connectivity results observedin our experienced singers resemble our previous findings of func-tional connectivity between auditory cortex, anterior cingulatecortex, and insula in both non-musicians and experienced singers(Zarate & Zatorre, 2008). Although this network was not specifi-cally recruited by non-musicians during COMP200c tasks in thatstudy, the fact that non-musicians possess this functionally con-nected network suggests that they have the potential to engage thisnetwork for audio–vocal integration during voluntary vocal pitchregulation if they undergo vocal training and practice.

Acknowledgments

We gratefully acknowledge Michael Petrides, Marc Schön-wiesner, and the McConnell Brain Imaging Centre staff for theirassistance. This work was supported by grants from the Cana-dian Institutes of Health Research (CIHR; RJZ), the Eileen PetersMcGill Majors Fellowship (JMZ), and the Centre for InterdisciplinaryResearch in Music Media and Technology (CIRMMT; JMZ, SW).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.neuropsychologia.2009.10.025.

References

Ackermann, H., & Riecker, A. (2004). The contribution of the insula to motor aspectsof speech production: A review and a hypothesis. Brain and Language, 89,320–328.

Ahsan, R. L., Allom, R., Gousias, I. S., Habib, H., Turkheimer, F. E., Free, S., et al. (2007).Volumes, spatial extents and a probabilistic atlas of the human basal ganglia andthalamus. NeuroImage, 38, 261–270.

Astafiev, S. V., Shulman, G. L., Stanley, C. M., Snyder, A. Z., Van, E., & Corbetta, M.(2003). Functional organization of human intraparietal and frontal cortex forattending, looking, and pointing. Journal of Neuroscience, 23, 4689–4699.

Augustine, J. R. (1996). Circuitry and functional aspects of the insular lobe in primatesincluding humans. Brain Research Brain Research Reviews, 22, 229–244.

Bamiou, D. E., Musiek, F. E., & Luxon, L. M. (2003). The insula (Island of Reil) andits role in auditory processing, Literature review. Brain Research Brain ResearchReviews, 42, 143–154.

Belin, P., Zatorre, R. J., & Ahad, P. (2002). Human temporal-lobe response to vocalsounds. Brain Research Cognitive Brain Research, 13, 17–26.

Belin, P., Zatorre, R. J., Hoge, R., Evans, A. C., & Pike, B. (1999). Event-related fMRI ofthe auditory cortex. NeuroImage, 10, 417–429.

Belin, P., Zatorre, R. J., Lafaille, P., Ahad, P., & Pike, B. (2000). Voice-selective areas inhuman auditory cortex. Nature, 403, 309–312.

Binkofski, F., & Buccino, G. (2004). Motor functions of the Broca’s region. Brain and

Language, 89, 362–369.

Binkofski, F., & Buccino, G. (2006). The role of ventral premotor cortex in actionexecution and action understanding. Journal of Physiology (Paris), 99, 396–405.

Botvinick, M. M., Cohen, J. D., & Carter, C. S. (2004). Conflict monitoringand anterior cingulate cortex: An update. Trends in Cognitive Sciences, 8,539–546.

sycho

B

B

B

B

B

B

B

B

C

C

C

C

C

C

C

d

D

D

D

D

E

F

F

F

F

F

G

G

G

G

G

H

H

J.M. Zarate et al. / Neurop

otvinick, M., Nystrom, L. E., Fissell, K., Carter, C. S., & Cohen, J. D. (1999). Conflictmonitoring versus selection-for-action in anterior cingulate cortex. Nature, 402,179–181.

rass, M., & von Cramon, D. Y. (2002). The role of the frontal cortex in task prepara-tion. Cerebral Cortex, 12, 908–914.

rown, S., Martinez, M. J., Hodges, D. A., Fox, P. T., & Parsons, L. M. (2004). The songsystem of the human brain. Brain Research Cognitive Brain Research, 20, 363–375.

rown, S., Martinez, M. J., & Parsons, L. M. (2006). Music and language side by sidein the brain: A PET study of the generation of melodies and sentences. EuropeanJournal of Neuroscience, 23, 2791–2803.

urnett, T. A., Freedland, M. B., Larson, C. R., & Hain, T. C. (1998). Voice F0 responsesto manipulations in pitch feedback. Journal of the Acoustical Society of America,103, 3153–3161.

urnett, T. A., & Larson, C. (2002). Early pitch-shift response is active in both steadyand dynamic voice pitch control. Journal of the Acoustical Society of America, 112,1058–1063.

urnett, T. A., McCurdy, K. E., & Bright, J. C. (2008). Reflexive and volitional voicefundamental frequency responses to an anticipated feedback pitch error. Exper-imental Brain Research, 191, 341–351.

ushara, K. O., Grafman, J., & Hallett, M. (2001). Neural correlates of auditory-visualstimulus onset asynchrony detection. Journal of Neuroscience, 21, 300–304.

arter, C. S., Braver, T. S., Barch, D. M., Botvinick, M. M., Noll, D., & Cohen, J. D.(1998). Anterior cingulate cortex, error detection, and the online monitoringof performance. Science, 280, 747–749.

elsis, P., Boulanouar, K., Doyon, B., Ranjeva, J. P., Berry, I., Nespoulous, J. L., et al.(1999). Differential fMRI responses in the left posterior superior temporal gyrusand left supramarginal gyrus to habituation and change detection in syllablesand tones. NeuroImage, 9, 135–144.

hen, J. L., Penhune, V. B., & Zatorre, R. J. (2008). Listening to musical rhythms recruitsmotor regions of the brain. Cerebral Cortex, 18, 2844–2854.

hen, J. L., Zatorre, R. J., & Penhune, V. B. (2006). Interactions between auditory anddorsal premotor cortex during synchronization to musical rhythms. NeuroImage,32, 1771–1781.

houinard, P. A., & Paus, T. (2006). The primary motor and premotor areas of thehuman cerebral cortex. Neuroscientist, 12, 143–152.

ollins, D. L., Neelin, P., Peters, T. M., & Evans, A. C. (1994). Automatic 3D intersubjectregistration of MR volumetric data in standardized Talaraich space. Journal ofComputer Assisted Tomography, 18, 192–205.

ox, R. W. (1996). AFNI: Software for analysis and visualization of functional mag-netic resonance neuroimages. Computers and Biomedical Research, 29, 162–173.

e Cheveigné, A., & Kawahara, H. (2002). YIN, a fundamental frequency estimator forspeech and music. Journal of the Acoustical Society of America, 111, 1917–1930.

onath, T. M, Natke, U., & Kalveram, K. T. (2002). Effects of frequency-shifted auditoryfeedback on voice F0 contours in syllables. Journal of the Acoustical Society ofAmerica, 111, 357–366.

ujardin, E., & Jurgens, U. (2005). Afferents of vocalization-controlling periaqueduc-tal regions in the squirrel monkey. Brain Research, 1034, 114–131.

urston, S., Davidson, M. C., Thomas, K. M., Worden, M. S., Tottenham, N., Martinez,A., et al. (2003). Parametric manipulation of conflict and response competitionusing rapid mixed-trial event-related fMRI. NeuroImage, 20, 2135–2141.

uvernoy, H. M. (1991). The human brain: Surface, three-dimensional sectionalanatomy and MRI. Wien, New York: Springer–Verlag.

liades, S. J., & Wang, X. (2003). Sensory-motor interaction in the primate audi-tory cortex during self-initiated vocalizations. Journal of Neurophysiology, 89,2194–2207.

ogassi, L., Gallese, V., Buccino, G., Craighero, L., Fadiga, L., & Rizzolatti, G. (2001).Cortical mechanism for the visual guidance of hand grasping movements in themonkey: A reversible inactivation study. Brain, 124, 571–586.

oster, N. E. V., & Zatorre, R. J. (in press). A role for the intraparietal sulcus in trans-forming musical information. Cerebral Cortex.

ox, P. T., Huang, A., Parsons, L. M., Xiong, J.-H., Zamarippa, F., Rainey, L., et al. (2001).Location-probability profiles for the mouth region of human primary motor-sensory cortex: Model and validation. NeuroImage, 13, 196–209.

riston, K. J. (1994). Functional and effective connectivity in neuroimaging: A syn-thesis. Human Brain Mapping, 2, 56–78.

riston, K. J., Buechel, C., Fink, G. R., Morris, J., Rolls, E., & Dolan, R. J. (1997). Psy-chophysiological and modulatory interactions in neuroimaging. NeuroImage, 6,218–229.

aab, N., Gaser, C., & Schlaug, G. (2006). Improvement-related functional plasticityfollowing pitch memory training. NeuroImage, 31, 255–263.

aab, N., Gaser, C., Zaehle, T., Jancke, L., & Schlaug, G. (2003). Functional anatomy ofpitch memory—An fMRI study with sparse temporal sampling. NeuroImage, 19,1417–1426.

hosh, S. S., Tourville, J. A., & Guenther, F. H. (2008). A neuroimaging study of pre-motor lateralization and cerebellar involvement in the production of phonemesand syllables. Journal of Speech Language, and Hearing Research, 51, 1183–1202.

refkes, C., Ritzl, A., Zilles, K., & Fink, G. R. (2004). Human medial intrapari-etal cortex subserves visuomotor coordinate transformation. NeuroImage, 23,1494–1506.

uimaraes, A. R., Melcher, J. R., Talavage, T. M., Baker, J. R., Ledden, P., Rosen, B.

R., et al. (1998). Imaging subcortical auditory activity in humans. Human BrainMapping, 6, 33–41.

afke, H. Z. (2008). Nonconscious control of fundamental voice frequency. Journalof the Acoustical Society of America, 123, 273–278.

ain, T. C., Burnett, T. A., Kiran, S., Larson, C. R., Singh, S., & Kenney, M. K. (2000).Instructing subjects to make a voluntary response reveals the presence of

logia 48 (2010) 607–618 617

two components to the audio–vocal reflex. Experimental Brain Research, 130,133–141.

Huffman, R. F., & Henson, O. W., Jr. (1990). The descending auditory pathway andacousticomotor systems: Connections with the inferior colliculus. Brain ResearchBrain Research Reviews, 15, 295–323.

Hyde, K. L., Peretz, I., & Zatorre, R. J. (2008). Evidence for the role of the right auditorycortex in fine pitch resolution. Neuropsychologia, 46, 632–639.

Jeffries, K. J., Braun, A. R., & Fritz, J. B. (2003). Words in melody: An H215O PET study

of brain activation during singing and speaking. NeuroReport, 14, 749–754.Jones, J. A., & Keough, D. (2008). Auditory-motor mapping for pitch control in singers

and nonsingers. Experimental Brain Research, 190, 279–287.Jones, J. A., & Munhall, K. G. (2000). Perceptual calibration of F0 production: Evi-

dence from feedback perturbation. Journal of the Acoustical Society of America,108, 1246–1251.

Jones, J. A., & Munhall, K. G. (2005). Remapping auditory-motor representations invoice production. Current Biology, 15, 1768–1772.

Jurgens, U. (2002). Neural pathways underlying vocal control. Neuroscience andBiobehavioral Reviews, 26, 235–258.

Kleber, B., Birbaumer, N., Veit, R., Trevorrow, T., & Lotze, M. (2007). Overt and imag-ined singing of an Italian aria. NeuroImage, 36, 889–900.

Kleber, B., Veit, R., Birbaumer, N., Gruzelier, J., & Lotze, M. (in press). The brain ofopera singers: Experience-dependent changes in functional activation. CerebralCortex.

Kriegstein, K. V., & Giraud, A. L. (2004). Distinct functional substrates along the rightsuperior temporal sulcus for the processing of voices. NeuroImage, 22, 948–955.

Larson, C. R. (1998). Cross-modality influences in speech motor control: The use ofpitch shifting for the study of F0 control. Journal of Communication Disorders, 31,489–502.

Larson, C. R., Burnett, T. A., & Kiran, S. (2000). Effects of pitch-shift velocity on voiceF0 response. Journal of the Acoustical Society of America, 107, 559–564.

Lewis, J. W., Beauchamp, M. S., & DeYoe, E. A. (2000). A comparison of visual and audi-tory motion processing in human cerebral cortex. Cerebral Cortex, 10, 873–888.

Liu, H., & Larson, C. R. (2007). Effects of perturbation magnitude and voice F0 level onthe pitch-shift reflex. Journal of the Acoustical Society of America, 122, 3671–3677.

Lombard, E. (1911). Le signe de l’elevation de la voix. Annales maladies oreille larynxnez pharynx, 37, 101–119.

MacDonald, A. W., III, Cohen, J. D., Stenger, V. A., & Carter, C. S. (2000). Dissociatingthe role of the dorsolateral prefrontal and anterior cingulate cortex in cognitivecontrol. Science, 288, 1835–1838.

Mazziotta, J., Toga, A., Evans, A., Fox, P., Lancaster, J., Zilles, K., et al. (2001). Aprobabilistic atlas and reference system for the human brain: International Con-sortium for Brain Mapping (ICBM). Philosophical Transactions of the Royal SocietyB: Biological Sciences, 356, 1293–1322.

Mesulam, M. M., & Mufson, E. J. (1982). Insula of the old world monkey. III: Efferentcortical output and comments on function. Journal of Comparative Neurology,212, 38–52.

Mufson, E. J., & Mesulam, M. M. (1982). Insula of the old world monkey. II: Afferentcortical input and comments on the claustrum. Journal of Comparative Neurology,212, 23–37.

Muller-Preuss, P., Newman, J. D., & Jurgens, U. (1980). Anatomical and physiologi-cal evidence for a relationship between the ‘cingular’ vocalization area and theauditory cortex in the squirrel monkey. Brain Research, 202, 307–315.

Nakamura, K., Dehaene, S., Jobert, A., Le, B. D., & Kouider, S. (2007). Task-specificchange of unconscious neural priming in the cerebral language network. Pro-ceedings of the National Academy of Sciences of the United States of America, 104,19643–19648.

Natke, U., Donath, T. M., & Kalveram, K. T. (2003). Control of voice fundamentalfrequency in speaking versus singing. Journal of the Acoustical Society of America,113, 1587–1593.

Natke, U., & Kalveram, K. T. (2001). Effects of frequency-shifted auditory feedbackon fundamental frequency of long stressed and unstressed syllables. Journal ofSpeech Language, and Hearing Research, 44, 577–584.

Nichols, T., Brett, M., Andersson, J., Wager, T., & Poline, J. B. (2005). Valid conjunctioninference with the minimum statistic. NeuroImage, 25, 653–660.

Nonaka, S., Takahashi, R., Enomoto, K., Katada, A., & Unno, T. (1997). Lombard reflexduring PAG-induced vocalization in decerebrate cats. Neuroscience Research, 29,283–289.

Ozdemir, E., Norton, A., & Schlaug, G. (2006). Shared and distinct neural correlatesof singing and speaking. NeuroImage, 33, 628–635.

Paus, T., Petrides, M., Evans, A. C., & Meyer, E. (1993). Role of the human anteriorcingulate cortex in the control of oculomotor, manual, and speech responses: Apositron emission tomography study. Journal of Neurophysiology, 70, 453–469.

Penhune, V. B, Zatorre, R. J., MacDonald, J. D., & Evans, A. C. (1996). Interhemisphericanatomical differences in human primary auditory cortex: Probabilistic map-ping and volume measurement from magnetic resonance scans. Cerebral Cortex,6, 661–672.

Perry, D. W., Zatorre, R. J., Petrides, M., Alivisatos, B., Meyer, E., & Evans, A. C.(1999). Localization of cerebral activity during simple singing. NeuroReport, 10,3979–3984.

Petrides, M. (1986). The effect of periarcuate lesions in the monkey on the perfor-

mance of symmetrically and asymmetrically reinforced visual and auditory go,no-go tasks. Journal of Neuroscience, 6, 2054–2063.

Picard, N, & Strick, P. L. (1996). Motor areas of the medial wall: A review of theirlocation and functional activation. Cerebral Cortex, 6, 342–353.

Picard, N., & Strick, P. L. (2001). Imaging the premotor areas. Current Opinion inNeurobiology, 11, 663–672.

6 sycho

Q

R

R

R

R

R

S

S

S

S

18 J.M. Zarate et al. / Neurop

uinones-Hinojosa, A., Ojemann, S. G., Sanai, N., Dillon, W. P., & Berger, M. S. (2003).Preoperative correlation of intraoperative cortical mapping with magnetic res-onance imaging landmarks to predict localization of the Broca area. Journal ofNeurosurgery, 99, 311–318.

iecker, A., Ackermann, H., Wildgruber, D., Dogil, G., & Grodd, W. (2000). Oppositehemispheric lateralization effects during speaking and singing at motor cortex,insula and cerebellum. NeuroReport, 11, 1997–2000.

inne, T., Kirjavainen, S., Salonen, O., Degerman, A., Kang, X., Woods, D. L., et al.(2007). Distributed cortical networks for focused auditory attention and dis-traction. Neuroscience Letters, 416, 247–251.

ivier, F., & Clarke, S. (1997). Cytochrome oxidase, acetylcholinesterase, and NADPH-diaphorase staining in human supratemporal and insular cortex: Evidence formultiple auditory areas. NeuroImage, 6, 288–304.

izzolatti, G., Fogassi, L., & Gallese, V. (2002). Motor and cognitive functions of theventral premotor cortex. Current Opinion in Neurobiology, 12, 149–154.

ushworth, M. F., Behrens, T. E., & Johansen-Berg, H. (2006). Connection patternsdistinguish 3 regions of human parietal cortex. Cerebral Cortex, 16, 1418–1430.

chulz, G. M., Varga, M., Jeffires, K., Ludlow, C. L., & Braun, A. R. (2005). Functionalneuroanatomy of human vocalization: An H215O PET study. Cerebral Cortex, 15,1835–1847.

cott, S. K., & Johnsrude, I. S. (2003). The neuroanatomical and functional organiza-tion of speech perception. Trends in Neurosciences, 26, 100–107.

iegel, G. M., & Pick, H. L., Jr. (1974). Auditory feedback in the regulation of voice.Journal of the Acoustical Society of America, 56, 1618–1624.

uga, N., & Yajima, Y. (1988). Auditory-vocal integration in the midbrain of the mus-tached bat: Periaqueductal gray and reticular formation. In J. D. Newman (Ed.),The physiological control of mammalian vocalization (pp. 87–107). New York:Plenum Press.

logia 48 (2010) 607–618

Talaraich, J., & Tournoux, P. (1988). Co-planar stereotaxic atlas of the human brain (3rded.). New York: Thieme Medical.

Tanabe, H. C., Kato, M., Miyauchi, S., Hayashi, S., & Yanagida, T. (2005). The sen-sorimotor transformation of cross-modal spatial information in the anteriorintraparietal sulcus as revealed by functional MRI. Brain Research Cognitive BrainResearch, 22, 385–396.

Tomaiuolo, F., MacDonald, J. D., Caramanos, Z., Posner, G., Chiavaras, M., Evans, A.C., et al. (1999). Morphology, morphometry and probability mapping of the parsopercularis of the inferior frontal gyrus: An in vivo MRI analysis. European Journalof Neuroscience, 11, 3033–3046.

Warren, J. D., Scott, S. K., Price, C. J., & Griffiths, T. D. (2006). Human brain mechanismsfor the early analysis of voices. NeuroImage, 31, 1389–1397.

Warren, J. D., Uppenkamp, S., Patterson, R. D., & Griffiths, T. D. (2003). Separatingpitch chroma and pitch height in the human brain. Proceedings of the NationalAcademy of Sciences of the United States of America, 100, 10038–10042.

Westbury, C. F., Zatorre, R. J., & Evans, A. C. (1999). Quantifying variability in theplanum temporale: A probability map. Cerebral Cortex, 9, 392–405.

Worsley, K. J. (2005). An improved theoretical P value for SPMs based on discretelocal maxima. NeuroImage, 28, 1056–1062.

Worsley, K. J., Charil, A., Lerch, J., & Evans, A. C. (2005). Connectivity of anatomical andfunctional MRI data. In 2005 International Joint Conference on Neural Networks

Worsley, K. J., Liao, C., Aston, J., Petre, V., Duncan, G. H., Morales, F., et al. (2002). A

general statistical analysis for fMRI data. NeuroImage, 15, 1–15.

Zarate, J. M., & Zatorre, R. J. (2008). Experience-dependent neural substrates involvedin vocal pitch regulation during singing. NeuroImage, 40, 1871–1887.

Zatorre, R. J., Chen, J. L., & Penhune, V. B. (2007). When the brain plays music:Auditory-motor interactions in music perception and production. NatureReviews Neuroscience, 8, 547–558.