Gray- and White-Matter Anatomy of Absolute Pitch Possessors

Transcription

1 Cerebral Cortex May 2015;25: doi: /cercor/bht334 Advance Access publication December 4, 2013 Gray- and White-Matter Anatomy of Absolute Pitch Possessors Anders Dohn 1,2, Eduardo A. Garza-Villarreal 1,3,4, M. Mallar Chakravarty 5,6,7, Mads Hansen 1,8, Jason P. Lerch 9,10 and Peter Vuust 1,2 1 Center of Functionally Integrative Neuroscience, University of Aarhus, Aarhus 8000, Denmark, 2 The Royal Academy of Music, Aarhus/Aalborg 8000, Denmark, 3 Department of Neurology, Faculty of Medicine and University Hospital Dr. José Eleuterio González, Universidad Autonoma de Nuevo Leon, Monterrey, Nuevo Leon 64460, Mexico, 4 Neuroscience Unit, Center for Research and Development in Health Sciences (CIDICS), Universidad Autonoma de Nuevo Leon, Monterrey, Nuevo Leon 64460, Mexico, 5 Kimel Family Translational Imaging Genetics Research Laboratory, Research Imaging Centre, Centre for Addiction and Mental Health Department of Psychiatry and, 6 Department of Psychiatry and Institute of Biomaterials and Biomedical Engineering, University of Toronto, Toronto, ON M6J 1H4, Canada, 7 Rotman Research Institute, Baycrest, Toronto, ON M6A 2E1, Canada, 8 Department of Psychology and Behavioral Sciences, Aarhus University, Aarhus 8000, Denmark, 9 Program in Neuroscience and Mental Health, The Hospital for Sick Children, Toronto, ON M5G 1X8, Canada and 10 Department of Medical Biophysics, University of Toronto, Toronto, ON M5G 2M9, Canada Address correspondence to Eduardo A. Garza-Villarreal, M.D., Department of Neurology, Faculty of Medicine and University Hospital, Universidad Autonoma de Nuevo Leon, Ave. Gonzalitos s/n col. Mitras Centro, CP 64460, Monterrey, Nuevo Leon, Mexico. eduardo.garzav@uanl.mx Anders Dohn and Eduardo A. Garza-Villarreal contributed equally. Absolute pitch (AP), the ability to identify a musical pitch without a reference, has been examined behaviorally in numerous studies for more than a century, yet only a few studies have examined the neuroanatomical correlates of AP. Here, we used MRI and diffusion tensor imaging to investigate structural differences in brains of musicians with and without AP, by means of whole-brain vertex-wise cortical thickness (CT) analysis and tract-based spatial statistics (TBSS) analysis. APs displayed increased CT in a number of areas including the bilateral superior temporal gyrus (STG), the left inferior frontal gyrus, and the right supramarginal gyrus. Furthermore, we found higher fractional anisotropy in APs within the path of the inferior fronto-occipital fasciculus, the uncinate fasciculus, and the inferior longitudinal fasciculus. The findings in gray matter support previous studies indicating an increased left lateralized posterior STG in APs, yet they differ from previous findings of thinner cortex for a number of areas in APs. Finally, we found a relation between the whitematter results and the CT in the right parahippocampal gyrus. In this study, we present novel findings in AP research that may have implications for the understanding of the neuroanatomical underpinnings of AP ability. Keywords: absolute pitch, cortical thickness, DTI, music, TBSS Introduction The brains of musicians have attracted much attention in recent years as a model of neural plasticity and potentially lifelong development of a specific, highly specialized skill. As an expert level of musicianship requires intense musical practice from childhood and throughout the musical career invoking complex auditory, associative, and motor skills, the hypothesis that musicians brains are anatomically different from nonmusicians has frequently been tested (Schlaug et al. 1995a; Schlaug 2001; Schneider et al. 2002; Sluming et al. 2002; Munte et al. 2002; Gaser and Schlaug 2003; Hutchinson et al. 2003; Jancke 2009). A very small subset of musicians possesses the unique ability called absolute pitch (AP). This ability is defined as the ability to identify a musical pitch absolutely (i.e., without a reference pitch at hand) (Takeuchi and Hulse 1993; Ward 1999) and its prevalence in Western cultures is frequently reported to be around 0.01% (Bachem 1955; Profita and Bidder 1988), whereas the prevalence in Asian populations has been reported to be markedly higher (Gregersen et al. 1999; Deutsch et al. 2006). AP has been investigated in numerous studies for more than a century with regard to behavioral features and limitations (Abraham 1901; Petran 1932; Takeuchi and Hulse 1991; Miyazaki 1993; Dooley and Deutsch 2011) as well as its associated personality traits (Brown et al. 2003; Dohn et al. 2012). AP is generally considered to develop in early childhood as a composite product of early commencement of musical training and conceivably a genetic predisposition (Baharloo et al. 2000; Gregersen et al. 2001; Zatorre 2003a; Miyazaki 2004; Deutsch et al. 2006; Athos et al. 2007) although the etiology of AP still remains a controversial issue. Thus, the AP phenotype is of interest to neuroscientists in that AP possessors constitute a model for investigating cognitive faculties and associated neuroanatomy in the interplay of genetic, anatomical, and environmental factors (Zatorre 2003a, 2003b). With the advent of modern structural scanning techniques, a pioneer morphometric in vivo study on gray-matter (GM) differences revealed that AP musicians (APs), compared with non-ap musicians (non-aps), showed stronger leftward asymmetry of the planum temporale (PT) (Schlaug et al. 1995b), a triangular region posteriorly adjacent to Heschl s gyrus accounting for a part of Wernicke s area. This region is known to be involved in auditory and language processing and yields gross leftward asymmetry in the majority of the general population (Geschwind and Levitsky 1968; Steinmetz 1996). This asymmetry finding has subsequently been replicated using different morphometric methods (Chen et al. 2000; Luders et al. 2004) and Keenan et al. (2001) suggested that a pruning of the right PT may account for the enhanced PT asymmetry in APs. However, other studies did not find a significant difference in asymmetry between APs and non-aps in the PT (Zatorre et al. 1998; Bermudez and Zatorre 2005; Bermudez et al. 2009), but only between APs and a larger group of controls unselected for musical skill (Zatorre et al. 1998) as well as between musicians and nonmusicians (Bermudez et al. 2009). Recently developed techniques of measuring and analyzing GM concentrations have facilitated whole-brain investigations which Bermudez et al. (2009) used to show that AP possessors had thinner cortex in a great number of regions, including frontal and parietal The Author Published by Oxford University Press. All rights reserved. For Permissions, please journals.permissions@oup.com

2 areas. Jancke et al. (2012) used graph theoretical analysis of cortical thickness (CT) covariations in which they found decreased global interconnectedness in AP musicians but also increased local connectivity in perisylvian areas. In the present study, we investigated CT in APs compared with non-ap musicians. This analysis is performed using the T 1 -weighted sequences at the nodes of a 3D polygonal mesh, where CT is the distance between the pial surface and the white-matter (WM) surface calculated with submillimetric accuracy in every vertex of the surface-based analysis (Lerch and Evans 2005; Winkler et al. 2010). Biologically, CT is the product of the neuronal bodies, dendrites, the synapses, the glia, and vasculature within the cortical layers. Thinning or thickening of the cortex measured by MRI could therefore be related to changes any or all of these elements (Rosas et al. 2002; Gogtay et al. 2004). Further, we investigated WM differences between these groups using tract-based spatial statistics (TBSS). TBSS is a whole-brain voxel-based analysis of WM data by using any diffusion-based metric such as the fractional anisotropy (FA) values. It allows for a more robust group statistical analysis than tractography (Smith et al. 2006). To our knowledge, this is the first study of AP literature that combines GM and WM analyses within the same sample of subjects. Given the aforementioned reports of leftward PT asymmetry, increased bilateral perisylvian cortical connectivity, and thinner frontal and parietal cortices in APs compared with non-aps, we hypothesized that the superior temporal regions, primarily in the left hemisphere, would have a thicker cortex, whereas the dorsolateral prefrontal cortex would have a thinner cortex. Furthermore, we hypothesized higher FA values in temporal WM tracts bilaterally and in the left superior longitudinal fasciculus (SLF). These tracts have been previously implicated in APs in WM studies (Loui et al. 2009, 2011a). Materials and Methods Participants Thirty-five musicians with a mean age of 29 (range = 18 43) were recruited into the study. They consisted of 2 groups: 17 musicians with APs and 18 musicians without APs (non-aps). However, one of the recruited non-aps did not complete the MRI sequence due to claustrophobia and was subsequently excluded from the study. The 2 groups were matched with regard to sex (χ 2 = 0.0, P > 0.2), age (Z = 0.86, P > 0.2), age of onset of musical training (Z = 1.0, P > 0.2), handedness (as assessed using the Edinburgh Handedness Inventory, Oldfield 1971) (Z = 0.1, P > 0.2), number of weekly hours of music practice and performance (Z = 0.1, P > 0.2), and years of musical training (Z = 0.1, P > 0.2). Table 1 summarizes these matching criteria. Furthermore, using the musical ear test (MET), the 2 groups were matched with regard to musical aptitude (Z= 0.9, P > 0.3). Although the participants played different primary instruments (which also can be seen in Table 1), all participants reported familiarity with the piano. All participants were of Caucasian ethnicity and native Danish, and all participants received compensation for being in the study. The participants were primarily recruited through the Royal Academy of Music, Aarhus, and the Music Department at Aarhus University. The study was approved by the local ethics committee (The Central Denmark Region Committees on Biomedical Research Ethics), and written informed consent was obtained from each participant after detailed explanation of the experimental procedure. Procedure The APs were primarily identified through word of mouth and through advertisements at the Danish Royal Academy of Music and the Music Table 1 Information of demography and musical background of the participants Group APs Non-APs Mean (SD) Mean (SD) Number of subjects Sex (male/female) 14/3 14/3 Age 28 (7.5) 30 (6.8) Years of musical training 15 (5.2) 15 (4.7) Weekly hours of music practice and performance 16 (11.6) 16 (9.4) Age of onset of musical training 5.5 (2.0) 6.2 (2.3) Handedness Right-handed Left-handed 3 2 Ambidextrous 2 1 Primary instrument Piano 8 7 Voice 3 3 Guitar 2 3 Other 4 4 Note: Information of demography and musical background of the participants. Aps, musicians with absolute pitch, non-aps, musicians without absolute pitch. Handedness is measured by the Edinburgh Handedness Inventory score (Oldfield 1971) where >80 designates right-handedness, < 80 left-handedness, and those in between were designated ambidextrous. SD, standard deviation. Department at the local university. The non-aps were found subsequently through advertisements and were selected using matching criteria. All participants were tested behaviorally by the same experimenter (the first author). To verify self-reported AP and to make a clear distinction between APs and non-aps, all participants completed an online pitch identification test (PIT) described and provided by Athos et al. (2007) and developed by Baharloo et al. (1998). All imaging data were acquired using the same MRI scanner. After giving informed consent and after the image acquisition, all participants completed the PIT on a laptop with stereo headphones. No participants reported any problems with either the auditory stimuli or the answering procedure in the PIT. Finally, the participants completed a questionnaire regarding age, sex, musical background (primary instrument, age on onset of musical training, etc.), and experience with AP. The non-aps were told that they did not have to answer the questions regarding AP experience. Behavioral Measures The PIT consisted of 80 trials: 40 randomly selected sine wave tones and 40 randomly selected digitized piano tones. The participants were asked to listen to the presented tones and to identify them by responding via an onscreen piano keyboard. The tones had duration of 1 s with an interlude of 2 s between the tones. Four pure tones and 4 piano tones were excluded from the scoring due to their position at the outermost range of the keyboard, resulting in 72 counting trials. Participants were given 1 point for each correct answer and ¾ point for each error of a semitone. We averaged the participants scores in mean pure tones and mean piano tones, and those scoring above a threshold of 36 were designated APs. The mean expected score by chance is with 95% of expected values lying between scores of 8.5 and To make sure both groups have similar musical expertise, all participants completed the MET, a newly developed test designed for measuring musical abilities objectively and quantitatively in both musicians and nonmusicians (Wallentin et al. 2010). This test consists of 104 trials in which participants listen to 2 musical phrases and subsequently judge whether or not they were identical by responding on an answer sheet. The first half of the test is a melodic subtest consisting of 52 pairs of melodic phrases, played with sampled piano sounds, and the other half is a rhythm subtest contains 52 trials with rhythmical phrases, played with wood block sound. Before each subtest, participants are given 2 example trials with feedback. Half of the trials in each session (26) are same trials and half are different trials, with the order randomized in both sessions. The participant s score is the percentage of correct answers out of the 104 trials GM and WM Anatomy of AP Possessors Dohn et al.

3 The behavioral data from the groups were assessed for normality with the Kolmogorov Smirnov test, which revealed violations of normality assumptions. Accordingly, we used the nonparametric Mann Whitney U test to test for differences in PIT and MET scores (as well as matching criteria) between the groups. Image Acquisition Images were acquired using a 3T GE Signa EXCITE MR system (Milwaukee, WI, USA) with an eight-channel Invivo head coil (Invivo, Gainesville, FL, USA). The study consisted in the acquisition of a highresolution T 1 3D volume and a diffusion tensor imaging (DTI) sequence. The T 1 3D acquisition consisted of a T 1 SPGR sequence, with TI/TR/TE = 750/6.6/2.8 ms, FOV mm 2, 256-by-256 image matrix, 146 contiguous slices with a thickness of 1.2 mm with no gap. Resulting voxel size was mm 3. DTI data were acquired performed with a double spin-echo single-shot EPI sequence. The diffusion-encoding scheme consisted of 26 directions isotropically distributed in space, using a b-factor of 1000 s/mm 2. In addition, 6 b =0s/mm 2 images were acquired. The maximum gradient strength was kept at 36 mt/m. Fortysix slice locations of slice thickness 3.0 mm (0 mm gap) were acquired, using 240 mm FOV in a 128-by-128 image matrix. TR/TE = 12500/88 ms. Two repeated scans were performed, with a total DTI acquisition time of 14 min. Resulting voxel size was mm 3. Cortical Thickness Preprocessing and Analysis CT was estimated using the CIVET processing pipeline (version ; Montreal Neurological Institute). All T 1 -weighted images were first linearly aligned to the ICBM 152 average template using a 9-parameter transformation (3 translations, rotations, and scales) (Collins et al. 1994) and preprocessed to minimize the effects of intensity nonuniformity (Sled et al. 1998). Images were then classified in GM and WM and cerebrospinal fluid (Zijdenbos et al. 2002). The hemispheres were then modeled as GM and WM surfaces using a deformable model strategy that generates 4 separate surfaces defined by vertices each (Kim et al. 2005). CT was derived between homologous vertices on GM and WM surfaces were derived using the t-link metric and subsequently blurred with a 20-mm surface-based diffusion kernel (for subsequent statistical analyses) (Lerch and Evans 2005). Normalizing for head or brain volume has little relationship to CT, and this risks introducing noise into the analyses; thus, native-space thicknesses were used in all analyses reported (Yasser et al. 2005; Sowell et al. 2007). Homology across the population was achieved using through nonlinear surface-based normalization that utilizes a midsurface (between pial and WM surfaces) (Robbins et al. 2004). This normalization uses a novel depth-potential function (Boucher et al. 2009) that fits each subject to a minimally biased surface-based template (Lyttleton et al. 2007). All vertex-wise analyses were performed in the RMINC package ( All surface-based analyses were corrected for multiple comparisons using the false discovery rate (FDR) (Genovese et al. 2002). All vertexwise statistics were carried out using a general linear model (GLM) that included age, sex, and years of musical training in the model. We also ran a GLM with the tbss_cluster (see TBSS analysis) to investigate its association to CT. DTI Preprocessing The DTI images were eddy current effects corrected by taking the first volume of the first sequence as a reference using FSL s Diffusion Toolbox (Smith et al. 2004). Afterward, we obtained FA images from each subject. In this analysis, the mean FA images were created by fitting a tensor model to the raw diffusion data using FDT, and then brain-extracted using BET (Smith 2002). All subjects FA data were then aligned into standard space using the nonlinear registration tool FNIRT (Rueckert et al. 1999; Andersson et al. 2007a, 2007b). TBSS Analysis Voxel-wise statistical analysis of the FA data was carried out using TBSS (Smith et al. 2004, 2006). The mean FA image was created and thinned to create the mean FA skeleton that represents the centers of all tracts common to the group. The mean FA skeleton was further thresholded by a FA value of 0.2 to exclude peripheral tracts where there was significant intersubject variability and/or partial volume effects with gray matter. Each subject s aligned FA data were then projected onto this skeleton and the resulting data fed into voxel-wise crosssubject statistics. To identify FA differences between APs and controls, the skeletonized FA data were fed into the voxel-wise statistics analysis which is based on nonparametric approach utilizing randomization. The statistical analysis was performed by the FSL randomize program using 5000 random permutations. Two contrasts were estimated: APs greater than controls and controls greater than APs. As FA can be also influenced by age, sex, handedness, and onset age of musical training, these variables were entered into the analysis as covariates to ensure that any observed difference of FA between groups was independent. Threshold-free cluster enhancement (TFCE) (Smith and Nichols 2009), was used to obtain the significant differences between 2 groups at P < 0.01, after accounting for multiple comparisons by using family-wise error (FWE). From the results of voxel-wise group comparisons, the skeletal regions showing significant intergroup differences were located and labeled anatomically by mapping the FWE-corrected statistical map of P < 0.01 to the Johns Hopkins University (JHU)-ICBM- DTI-81 WM labels atlas (Mori et al. 2008) and JHU-WM Tractography Atlas in MNI space (Mori et al. 2005; Wakana et al. 2007; Hua et al. 2008). The TBSS results were inflated using tbss_fill only for visualization purposes. Coordinates are shown in MNI space. DTI Individual Analysis In order to investigate the WM fasciculi related to the differences observed in the TBSS analysis, we performed individual tractography. For this, we transformed the TBSS results back to native space in order to extract the region-of-interest (ROI) mask from the cluster showing significant group FA differences. To do this, the significant skeleton-space cluster voxels are projected back from its position on the mean skeleton to the nearby position at the center of the nearest tract in the subject s FA image in standard space. Then, this point is warped back into the subject s native space by inverting the nonlinear transformation mapping each FA image to MNI space. After this transformation, we ended up with the individual ROI derived from the significant cluster (tbss_cluster), aligned in space to the individual FA image. From the tbss_cluster, we calculated a mean area FA (mafa) value for each participant, by averaging the FA in all voxels in each tbss_cluster. We then prepared the corrected DTI images for deterministic tractography in each subject using Diffusion Toolkit (Wang et al. 2007), a software toolbox which provides precise diffusion imaging analysis and visualization capabilities (Granziera et al. 2009). Diffusion tensor estimation was performed using the linear least-squares fitting method (Wang et al. 2007). The raw data were not smoothed or sharpened prior to reconstruction. Deterministic tractography was subsequently performed in TrackVis software using the fiber assignment by continuous tracking algorithm (Mori et al. 1999; Xue et al. 1999; Mori and van Zijl 2002). Only fibers with lengths of >10 mm were included. Results Behavioral Data The APs had a mean PIT score of 60.3 (SD = 10.7) whereas the non-aps had a mean PIT score of 14.5 (SD = 4.0). This difference was found to be statistically significant (Z = 5.0, P < 0.001) indicating that the 2 groups were clearly segregated with regard to their ability to identify and label pitch. The APs had a mean MET score of 87.0 (SD = 5.9), whereas the non-aps had a mean MET score of 84.5 (SD = 6.9). This difference was found not to be statistically significant (Z = 0.9, P > 0.3) indicating that the 2 groups were at similar levels of musical aptitude. Cerebral Cortex May 2015, V 25 N

4 Imaging data Cortical Thickness All significant findings were in the direction of a thicker cortex in the AP group than the non-ap group. These findings included the bilateral superior temporal gyrus (STG) (BA 41, 42, 22), the left inferior frontal gyrus (lifg) (BA 44, 45), the right supramarginal gyrus (rsg), the right parahippocampal gyrus (rpg), the bilateral anterior cingulate cortex (ACC) and the subcallosal cingulate gyrus (SCG) (BA 24, 35 32), medial part of Table 2 Cortical thickness t-value Stereotaxic coordinates (MNI space) x y z Left hemisphere Superior temporal gyrus BA BA BA Inferior frontal gyrus (BA 44, 45) Anterior cingulate cortex Subcallosal cingulate gyrus (BA 24, 32, 35) Lingual gyrus Fusiform gyrus Postcentral gyrus Right hemisphere Superior temporal gyrus BA BA Supramarginal gyrus Parahippocampal gyrus Anterior cingulate cortex Subcallosal cingulate gyrus (BA 24, 32, 35) Superior frontal gyrus Lingual gyrus Postcentral gyrus Precentral gyrus Stereotaxic coordinates of the areas of GM differences. BA, Brodmann areas. the right superior frontal gyrus (rsfg) (BA 6), left fusiform gyrus (lfg), bilateral lingual gyrus (LG), bilateral postcentral gyrus (PosG), and precentral gyrus (PreG) (BA 4) (Table 2 and Fig. 1). There were no significant peaks in the non-ap > AP direction. WM Analysis The TBSS analysis (AP > non-ap) showed that APs had higher FA compared with the matched controls (P < 0.01; TFCEcorrected) in a single significant cluster (Fig. 2), located in the right temporal lobe s subgyral white matter ( peak voxel: x = 39, y = 16, z = 11; t = 4.1); specifically, within the path of the inferior fronto-occipital fasciculus and the inferior longitudinal fasciculus (ILF) according to the JHU ICBM-DTI-81 White-Matter Labels Atlas (Mori et al. 2005). The contrast non-ap > AP did not yield any significant results after correction for multiple comparisons. The tractography showed similar fiber bundles in all participants, derived from the tbss_cluster (Supplementary Fig. 1). The tracts were compared with the Fiber Tract-based Atlas of Human White-Matter Anatomy (Wakana et al. 2004), where we could confirm that most of the fiber bundles corresponded to association fibers: the inferior fronto-occipital fasciculus, the ILF but also the uncinate fasciculus. Cortical Thickness and TBSS Examining the mean FA value of the significant WM cluster against the cortical thickness across the whole brain revealed a significant association in a single cluster in the right parahippocampal gyrus (BA 36) (Fig. 3). Discussion Here, using CT and TBSS analysis we show that the GM and WM of musicians with AP differ from those of musicians without AP. We found increased CT in APs compared with Figure 1. T-statistic maps showing the cortical thickness findings of the contrast of APs > non-aps. Graphs are only for exemplification purpose. All images are corrected for multiple comparisons using FDR (see Materials and Methods) GM and WM Anatomy of AP Possessors Dohn et al.

5 Figure 2. Axial view of several slices of the template brain (bottom-up) showing the TBSS significant cluster (mafa) AP > non-aps (age, sex, handedness, and onset age of musical training as covariates). The mean FA skeleton from the TBSS is in green. The cluster was inflated using tbss_fill for visualization purposes only. R, right; L, left, z, z-plane in mm (MNI). Figure 3. Association between the individual WMfrom the TBSS analysis and cortical thickness (CT). Here, we show the medial view of the right hemisphere, where the WM cluster is related to a single CT cluster in the right parahippocampal gyrus (left). The scatter plot shows the relation between the variables (right), with the groups in different colors. AP, absolute pitch possessors, non-ap, controls. non-aps in several areas of the cortex. Furthermore, the APs were found to have higher FA in the WM specifically situated in the temporal lobe, within the path of the association fibers. Finally, the higher FA in the temporal lobe was associated with a higher CT in the rpg. Differences in WM as measured with DTI between APs and non-aps have only sparsely been assessed, revealing a leftward asymmetry in the FA of the SLF in AP possessors (Oechslin et al. 2010) and higher FA values in WM pathways connecting the STG with the middle temporal gyrus (MTG) bilaterally (Loui et al. 2011a). Taken together, these structural GM and WM findings point relatively concordantly toward regions around the sylvian fissure and in the temporal lobe. However, it should be noted that a considerable proportion of them made use of predefined ROIs within these particular perisylvian and temporal regions and may have missed differences elsewhere in the brain. A common yet unproven hypothesis states that AP possessors possess an internal pitch template from which they reference and retrieve musical pitch (Ward 1963, 1999; Levitin 2004; Levitin and Rogers 2005). This long-term memory pitch template could be putatively encoded by the auditory system (Bidelman et al. 2011) and stored in hippocampus (Teki et al. 2012), yet additional brain areas including the visual system may also contribute to generating this template (e.g., by reading music) (Huang et al. 2010). Hence, AP ability may form a complex structural network that involves a number of brain areas related to basic cognitive functions. The tremendous rapidness in pitch identification of a genuine AP possessor (cf. Miyazaki 1990) suggests that the process of referencing the auditory input to a possible internal pitch template may be related to association fibers that interconnect specialized brain areas (Loui et al. 2011a). Nevertheless, it is still not yet clear if this internal pitch template exists for pitch identification in the AP ability. In our study, we found areas already mentioned in other structural and functional AP studies in the past; however, we also found a wide range of areas with greater CT of the APs that seem to be related to basic cerebral processing, such as memory, motor, somatosensory, auditory, visual, and association, as well as the complex processing such as language and attention. In our analysis, we attempted to control for many known sources of variation of cortical thickness and FA to investigate subject differences based solely in the AP ability. However, it is not clear if all our GM significant peaks are related to the AP ability itself. This could only be confirmed by further cross-sectional and, specially, longitudinal Cerebral Cortex May 2015, V 25 N

6 studies. We will now discuss the GM and WM findings as related to the AP ability. Gray Matter The GM analysis showed increased cortical thickness in the bilateral STG with a left lateralization, an area known to be related to auditory and frequency processing. This is consistent with previous anatomical findings of (Schlaug et al. 1995b) in AP musicians and leans on previously shown left lateralization suggested to exhibit AP ability. We also found higher cortical thickness in the lifg, which in musicians has been related to complex music listening (Levitin and Menon 2003), harmony processing (Maess et al. 2001; Garza Villarreal et al. 2011), and complex rhythms (Vuust et al. 2006). In APs, however, the IFG has only been related to pitch naming expertise (Wilson et al. 2009) and with pitch memory in fmri studies (Schulze et al. 2009), suggesting perhaps a verbal component in the pitch labeling. The thicker cortex in the right SG has not been shown in other AP studies. Nevertheless, in fmri studies in musicians, SG has been related to increased activation during pitch memory tasks compared with nonmusicians (Gaab et al. 2003a; Gaab and Schlaug 2003). We found increased cortical thickness in the lfg and the bilateral LG, both not previously found in APs. Studies have shown that these areas are both involved in visual memory processing (Slotnick and Schacter 2006) and semantic processing (Heath et al. 2012; Rama et al. 2012), strongly suggesting a verbal component in pitch labeling as mentioned before. Or perhaps, we could speculate that the designation of musical pitch labels to an auditory frequency may include a semantic type of component when referencing a pitch stimulus to a proposed internal memory pitch template. We also observed thicker cortex in the ACC. The ACC is a brain area, as Paus (2001) writes, where a regulatory network related to the brainstem nuclei, interacts with an executive network. ACC is often associated with error detection (e.g., in Stroop tasks) (Carter et al. 1998; Carter et al. 2000; Bush et al. 2000; Botvinick et al. 2001) and found to be involved in working memory tasks with musical chords (Pallesen et al. 2010) as well as in musical improvization (Berkowitz and Ansari 2008). Previous studies have not directly studied a functional or structural relation between the ACC and the AP ability, so it is an open question if this area is related to the AP ability itself. However, in musicians, neural activity in ACC increases during enhanced working memory performance (Pallesen et al. 2010). AP possession includes the capability to detect incongruity between an auditory pitch and a visual note (e.g., a choir s first note) in which the ACC could very well play a role. The parahippocampal gyrus (PG) is a brain area related to memory encoding (van Strien et al. 2009; Bergmann et al. 2012; Hirshhorn et al. 2012) and recalling verbal experiences (Wagner et al. 1998). It is interesting that the APs were found to have a thicker cortex in this area due to a proposed, yet unproven, memory pitch template that APs may possess (Levitin 1994; Ward 1999). It has also been shown that the PG is important for action-sound representation (the prediction of which sound an action will produce), an important feature of the AP ability (Petrini et al. 2011). Furthermore, the PG has been related to affective musical perception of unpleasantness (Gosselin et al. 2006; Khalfa et al. 2008). The implication for AP is most likely related to the memory component of the ability; however, it is not clear if there is an affective component. The SCG is the portion of the cingulate gyrus lying ventral to the corpus callosum, from the anterior boundary of the genu to the rostrum. It has been shown to be an important node in a network that includes cortical structures, the limbic system, thalamus, hypothalamus, and brainstem nuclei, including BA 24, 25, and 32 (Hamani et al. 2011). BA 24 and 32 have been related to affective and cognitive motor functions, whereas BA 25 has neural connections to nucleus accumbens, amygdala, and periacqueductal gray and has been related to visceromotor control. The ACC and SCG are known to be crucial for primate vocalization, which is used to express primarily internal emotional states (Devinsky et al. 1995). Although it is not straightforward as to what this relationship may mean for the AP ability, we can speculate that it could relate to the correct vocalization of a pitch, musical chord, or note without an external reference. Primary somatosensory (PreG) and motor (PosG) cortices were also thicker in APs than non-aps. However, as the APs and non-aps were matched and controlled with regard to age of onset of musical training, years of musical training, and current amount of musical activity, this result may not be related to differences musical practice, but perhaps a type of sensory-motor specialization. The PreG and PosG include BA 6 which has been previously associated with AP ability (Zatorre et al. 1998; Bermudez et al. 2009), and motor BA 4 (Calvo- Merino et al. 2005; Savini et al. 2012; Takeuchi et al. 2012). The PosG is functionally more activated in APs than non-aps during music listening (Loui et al. 2012). The right superior frontal gyrus (rsfg) has been related to switching between distinct cognitive tasks (Cutini et al. 2008), learning, and working memory performance (Nestor et al. 2008; Vasic et al. 2008). Finally, the IFG has also motor functions for language production (Grabski et al. 2012) and singing (Ozdemir et al. 2006). Hence, it seems there are clear differences in several motor areas between APs and non-aps in this study. Even though in our study, we controlled for effects of musical expertise and onset age of musical practice it is not clear if these differences in motor areas are purely related to the AP ability. In a recent study, Bermudez et al. (2009) showed that APs had less CT in the areas they found to be significantly different. Surprisingly, although we used the same analysis methods and software, all our GM findings showed a thicker cortex in the APs compared with the non-aps, contrary to our expectations and to the findings in the Bermudez study. The main explanation could be that, in our study, we focused on the AP ability instead of musicianship, by reducing the variation due to musicianship and with a slightly higher sample size. Also, in our study, the AP test (the PIT) was highly conservative and found clear-cut differences between APs and non-aps, whereas in the Bermudez s study they did not find a clear difference, therefore they chose the strongest and weakest performers based on their in-house AP test. AP classification is key to understand the AP ability as well as CT differences, and until there is a standardized AP test, the results may not be consistent between studies. Another possible explanation of the discrepancy in CT may be found in the highly differing sex distributions. After several reports of substantial sex effects in morphometric studies (Amunts et al. 2000; Nopoulos et al. 2000; Good et al. 2001), later VBM studies of musicians only included male participants (Sluming et al. 2002; Gaser and Schlaug 2003). Moreover, a previous VBM study on APs detected a sex effect (Luders et al. 2004) and 1384 GM and WM Anatomy of AP Possessors Dohn et al.

7 various neuroanatomical studies on musicians have found a number of structural differences between male musicians and male nonmusicians whereas no similar significant results were found between female groups (Schlaug 2001; Hutchinson et al. 2003; Lee et al. 2003) which was substantiated by a functional study (Gaab et al. 2003b). Although our groups of participants were accurately matched with regard to sex it should be noted, however, that an unusual large proportion of the participants in our study were male (28 of 34), whereas the large majority of the participants in the study by Bermudez et al. (2009) were female (47 of 71). In summary, although our CT analysis was similar to the Bermudez study, the method, the sample, and the hypotheses may account for the differences in CT results, which should be studied further. White Matter We found one single significant cluster within the right temporal lobe with higher FA in APs, specifically within the path of the right inferior fronto-occipital fasciculus (IFOF), the uncinate fasciculus (UF), and the ILF. This finding is unique in the AP literature in that the only 2 previous DTI studies on AP focused on the SLF (Oechslin et al. 2010) and the tracts between the STG and the MTG (Loui et al. 2011a) respectively, both finding a leftward asymmetry. In spite of this evidence, we found a rightward asymmetry in the TBSS analysis. There could be several explanations for this result. The TBSS is a broad whole-brain analysis of FA, whereas tractography is constrained on particular ROIs. In the TBSS analysis, we correct for multiple comparisons for the whole brain, which constrains our significant differences to areas that show the most differences in FA. Therefore, the differences in left SLF, STF, and MTG found by Oechslin et al. and Loui et al. may not appear in the TBSS analysis due to low effect size or the multiple comparisons. Also, our sample size was slightly higher than both these studies, thereby perhaps improving power to detect significant differences between groups. We also controlled for the covariates age, sex, handedness, and onset age of musical training that could account for variation in FA not related to the AP ability itself. Finally, our study as well as Oechslin s and Lui s differs between each other in the PIT to determine AP. Our AP determination was highly conservative, hence scoring AP by chance was very low. In summary, the rightward asymmetry we found results from our particular sample, AP determination and analysis, therefore this and other studies cannot be generalized until higher sample sizes are studied and a standarized AP determination is implemented. The ILF interconnects visual association areas of the occipital lobe with lateral and medial anterior temporal lobe regions. Lateral branches pass to the inferior, middle, and STG in the right hemisphere whereas medial branches pass to the uncus and PG (Catani et al. 2002; Wakana et al. 2004). The IFOF is a large and long association bundle of fibers that interconnects frontal (including Broca s and adjacent areas) and occipital lobes but it also contains fibers that connect to the posterior part of the parietal and temporal lobe. In fact, it contains fibers that connect the auditory cortex with the prefrontal cortex (Kier et al. 2004). The UF interconnects the frontal and temporal lobes through the temporal stem and constitutes the ventral route between frontal language areas (i.e., BA 44, 45, and 47) and the posterior STG, including Wernicke s area (Kubicki et al. 2002; Kier et al. 2004; Parker et al. 2005). Hence, these are all complex association fibers that interconnect the temporal lobe with other lobes within the same hemisphere, particularly the ventral part of the prefrontal cortex. This finding suggests that these WM pathways in the right hemisphere are involved in AP and interestingly, these pathways connect to the right PG where we also found to reveal thicker cortex in APs and related pitch identification proficiency. Previous studies not related to AP ability have shown that higher FA values in frontotemporal WM pathways correlates with increased cortical thickness in language areas in the left hemisphere (Phillips et al. 2011); however, a recent study on the WM integrity associated with performance in a pitchbased grammar learning task has revealed that brain structures subserving pitch-based learning are right lateralized (Loui et al. 2011b) in normal individuals. Therefore, our finding corroborates the importance of the right WM pathways in AP ability. A comparison of the mafa in this cluster against CT across the whole brain revealed a significant correlation in the PG. This area has not previously been assessed in relation to AP; however, since the unique faculty of labeling, a single musical tone may require a certain memory encoding for designating musical pitch labels, it is possible that the PG may play a role in the genesis of AP. Structural Network Integration The GM results showed a left lateralized STG, IFG, LG, and FG, and a right lateralized SG, ACC, SCG, PreG, PosG, and PG. The WM results showed higher FA in the right temporal lobe related to long-range association tracts and that, in turn, may be related to the right PG. Together these results suggest that the AP ability may be bilaterally and widely distributed. Although it is difficult to explain the relation between the left lateralized areas and the right FA values, this may reflect that each hemisphere contributes differently to perform the same task. As mentioned earlier, the STG is an area consistently found in AP studies that, in connectomics terms, could possibly act as a main central hub for the AP ability, directing the structural and/or functional connectivity between structures (Van Dijik et al. 2010). Overall, there is a need for larger scale studies regarding the AP ability as a fine-tuned structurally and functionally distributed cognitive process. In conclusion, we here show that the brains of closely matched, homogeneous groups of musicians with and without AP differ in both GM and WM. Some differences in CT are in similar areas to those of previous structural and functional studies. However, we also found areas not particularly related to the AP ability. As well, we showed that APs have higher FA values in the right temporal lobe, within the path of association tracts. This suggests that the AP ability may be associated with an anatomically efficient network, in turn associated with musical expertise in general that seems to be highly specialized in the AP population. However, to uncover the origin of the structural differences in APs versus non-aps (i.e., whether the structures represent a predisposition for AP or whether they simply reflect the brain s plasticity through musical activity and employment of AP ability), longitudinal studies on children prior to acquiring AP should be conducted. Supplementary Material Supplementary can be found at: org/. Cerebral Cortex May 2015, V 25 N

8 Funding The work was supported by the Royal Academy of Music, Aarhus/Aalborg, Denmark, the Ministry of Culture, Denmark, and The Danish National Research Foundation s Grant to the Center of Functionally Integrative Neuroscience (CFIN). Notes We thank all the musicians with and without absolute pitch for participating in this study. We also wish to thank Dora Zeidler, Ryan Sangill, and Michael Geneser for help with MR data acquisition, Torben Ellegaard Lund for help with data handling, Jesper Frandsen and Luis Concha-Loyola for help with the DTI analysis, and Mikkel Wallentin for help with writing the paper. Conflict of Interest: None declared. References Abraham O Das absolute Tonbewußtsein. Sammelbände der Internationalen Musikgesellschaft. 3:1 86. Amunts K, Jäncke L, Mohlberg H, Steinmetz H, Zilles K Interhemispheric asymmetry of the human motor cortex related to handedness and gender. Neuropsychologia. 38: Andersson JLR, Jenkinson M, Smith S. 2007a. Non-Linear Optimisation. FMRIB Technical Report TR07JA1. Available from: fmrib.ox.ac.uk/analysis/techrep. Andersson JLR, Jenkinson M, Smith S. 2007b. Non-Linear Registration, aka Spatial Normalisation. FMRIB Technical Report TR07JA2 Available from: Athos EA, Levinson B, Kistler A, Zemansky J, Bostrom A, Freimer N, Gitschier J Dichotomy and perceptual distortions in absolute pitch ability. Proc Natl Acad Sci USA. 104: Bachem A Absolute pitch. J Acoust Soc Am. 27: Baharloo S, Johnston PA, Service SK, Gitschier J, Freimer NB Absolute pitch: an approach for identification of genetic and nongenetic components. Am J Hum Genet. 62: Baharloo S, Service SK, Risch N, Gitschier J, Freimer NB Familial aggregation of absolute pitch. Am J Hum Genet. 67: Bergmann HC, Rijpkema M, Fernandez G, Kessels RP Distinct neural correlates of associative working memory and long-term memory encoding in the medial temporal lobe. Neuroimage. 63: Berkowitz AL, Ansari D Generation of novel motor sequences: the neural correlates of musical improvisation. Neuroimage. 41: Bermudez P, Lerch JP, Evans AC, Zatorre RJ Neuroanatomical correlates of musicianship as revealed by cortical thickness and voxel-based morphometry. Cereb Cortex. 19: Bermudez P, Zatorre RJ Differences in gray matter between musicians and nonmusicians. Ann N Y Acad Sci. 1060: Bidelman GM, Gandour JT, Krishnan A Musicians and tonelanguage speakers share enhanced brainstem encoding but not perceptual benefits for musical pitch. Brain Cogn. 77:1 10. Botvinick MM, Braver TS, Barch DM, Carter CS, Cohen JD Conflict monitoring and cognitive control. Psychol Rev. 108: Boucher M, Whitesides S, Evans A Depth potential function for folding pattern representation, registration and analysis. Med Image Anal. 13: Brown WA, Cammuso K, Sachs H, Winklosky B, Mullane J, Bernier R, Svenson S, Arin D, Rosen-Sheidley B, Folstein SE Autismrelated language, personality, and cognition in people with absolute pitch: results of a preliminary study. J Autism Dev Disord. 33: Bush G, Luu P, Posner MI Cognitive and emotional influences in anterior cingulate cortex. Trends Cogn Sci. 4: Calvo-Merino B, Glaser DE, Grezes J, Passingham RE, Haggard P Action observation and acquired motor skills: an fmri study with expert dancers. Cereb Cortex. 15: Carter CS, Braver TS, Barch DM, Botvinick MM, Noll D, Cohen JD Anterior cingulate cortex, error detection, and the online monitoring of performance. Science. 280: Carter CS, Macdonald AM, Botvinick M, Ross LL, Stenger VA, Noll D, Cohen JD Parsing executive processes: strategic vs. evaluative functions of the anterior cingulate cortex. Proc Natl Acad Sci USA. 97: Catani M, Howard RJ, Pajevic S, Jones DK Virtual in vivo interactive dissection of white matter fasciculi in the human brain. Neuroimage. 17: Chen C, Halpern A, Bly B, Edelman R, Schlaug G Planum temporale asymmetry and absolute pitch. Neuroimage. 11:S114. Collins DL, Neelin P, Peters TM, Evans AC Automatic 3D intersubject registration of MR volumetric data in standardized Talairach space. J Comput Assist Tomogr. 18: Cutini S, Scatturin P, Menon E, Bisiacchi PS, Gamberini L, Zorzi M, Dell Acqua R Selective activation of the superior frontal gyrus in task-switching: an event-related fnirs study. Neuroimage. 42: Deutsch D, Henthorn T, Marvin E, Xu H Absolute pitch among American and Chinese conservatory students: prevalence differences, and evidence for a speech-related critical period. J Acoust Soc Am. 119: Devinsky O, Morrell MJ, Vogt BA Contributions of anterior cingulate cortex to behaviour. Brain. 118: Dohn A, Garza-Villarreal EA, Heaton P, Vuust P Do musicians with perfect pitch have more autism traits than musicians without perfect pitch? An empirical study. PLoS One. 7:e Dooley K, Deutsch D Absolute pitch correlates with high performance on interval naming tasks. J Acoust Soc Am. 130: Gaab N, Gaser C, Zaehle T, Jancke L, Schlaug G. 2003a. Functional anatomy of pitch memory an fmri study with sparse temporal sampling. Neuroimage. 19: Gaab N, Keenan JP, Schlaug G. 2003b. The effects of gender on the neural substrates of pitch memory. J Cogn Neurosci. 15: Gaab N, Schlaug G The effect of musicianship on pitch memory in performance matched groups. Neuroreport. 14: Garza Villarreal EA, Brattico E, Leino S, +ÿstergaard L, Vuust P Distinct neural responses to chord violations: a multiple source analysis study. Brain Res. 1389: Gaser C, Schlaug G Brain structures differ between musicians and non-musicians. J Neurosci. 23: Genovese CR, Lazar NA, Nichols T Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage. 15: Geschwind N, Levitsky W Human brain: left-right asymmetries in temporal speech region. Science. 161: Gogtay N, Giedd JN, Lusk L, Hayashi KM, Greenstein D, Vaituzis AC, Nugent TF, Herman DH, Clasen LS, Toga AW et al Dynamic mapping of human cortical development during childhood through early adulthood. Proc Natl Acad Sci USA. 101: Good CD, Johnsrude I, Ashburner J, Henson RNA, Friston KJ, Frackowiak RSJ Cerebral asymmetry and the effects of sex and handedness on brain structure: a voxel-based morphometric analysis of 465 normal adult human brains. Neuroimage. 14: Gosselin N, Samson S, Adolphs R, Noulhiane M, Roy M, Hasboun D, Baulac M, Peretz I Emotional responses to unpleasant music correlates with damage to the parahippocampal cortex. Brain. 129: Grabski K, Lamalle L, Vilain C, Schwartz JL, Vallee N, Tropres I, Baciu M, Le Bas JF, Sato M Functional MRI assessment of orofacial articulators: neural correlates of lip, jaw, larynx, and tongue movements. Hum Brain Mapp. 33: Granziera C, Schmahmann JD, Hadjikhani N, Meyer H, Meuli R, Wedeen V, Krueger G Diffusion spectrum imaging shows the structural basis of functional cerebellar circuits in the human cerebellum in vivo. PLoS One. 4(4):e5101. Gregersen PK, Kowalsky E, Kohn N, Marvin EW Absolute pitch: prevalence, ethnic variation, and estimation of the genetic component. Am J Hum Genet. 65: Gregersen PK, Kowalsky E, Kohn N, Marvin EW Early childhood music education and predisposition to absolute pitch: teasing apart genes and environment. Am J Med Genet. 98: GM and WM Anatomy of AP Possessors Dohn et al.