A total of 25 active meditation practitioners were accrued through referrals and advertisements in various meditation venues. Three individuals showed macroscopic cerebral abnormalities without clinical significance and were excluded from the study. Our final sample included 22 active meditation practitioners and 22 controls from the International Consortium for Brain Mapping (ICBM) database of normal adults (http://www.loni.ucla.edu/ICBM/Databases/
), matched for gender and age. There were 9 men and 13 women in each group. Age ranged between 30 and 71 years (meditators mean age: 53.00 years [SD: 11.54]; controls mean age: 53.09 years [SD: 11.38]). The maximum allowed age difference within a matched pair was one year. The level of education in meditators and controls was comparable (meditators / controls): 36% / 36% above college level; 45% / 54% college level; 14% / 9% below college level. All subjects were right-handed, except one control subject who was left-handed, where handedness was determined based on self-reports of hand preference for selected activities. All subjects were required to be free of any neurological disorders and gave informed consent according to institutional guidelines (Institutional Review Board of the University of Los Angeles, California [UCLA]).
Years of meditation practice ranged between five and 46 years (mean: 24.18 years [SD: 12.36]), where styles included Zazen, Samatha, Vipassana, and others. Although long-time practices can vary greatly (over time and with respect to the mental exercises performed), more than half of all meditators indicated deep concentration as being an essential part of their practice (63%). About a third of them engaged control of breath (36%), visualization (32%), as well as attention to external and internal stimuli / events (32%). Other elements, however less frequently indicated, included withdrawal of sensory perceptions (14%) and letting go of thoughts (18%). The length of formal meditation ranged from 10 to 90 minutes each session, with the majority of meditators (59%) having sessions daily.
Brain images were acquired on a 1.5-T MRI system (Siemens Sonata) using a 3D T1-weighted sequence (MPRAGE) with the following parameters: TR = 1900 ms; TE = 4.38 ms; flip angle = 15°; 160 contiguous 1 mm sagittal slices; FOV = 256 mm × 256 mm2; matrix size = 256 × 256, voxel size = 1.0 × 1.0 × 1.0 mm.
Voxel-based GM Volume Analysis (Local Approach)
Data were processed and examined using the SPM5 software (Wellcome Department of Imaging Neuroscience Group, London, UK; http://www.fil.ion.ucl.ac.uk/spm
), where we applied VBM standard routines and default parameters implemented in the VBM5 toolbox (http://dbm.neuro.uni-jena.de/vbm.html
). Images were bias field corrected, tissue classified, and registered using linear (12-parameter affine) and non-linear transformations (warping), within the same generative model (Ashburner and Friston, 2005
). Subsequently, analyses were performed on GM segments, which were multiplied by the non-linear components derived from the normalization matrix in order to preserve actual GM values locally (modulated GM volumes). Importantly, GM segments were not multiplied by the linear components of the registration in order to account for individual differences in brain orientation, alignment, and size globally. Finally, the modulated GM volumes were smoothed with a Gaussian kernel of 14 mm full width at half maximum (FWHM). These smoothed modulated GM volumes are hereafter referred to as GM to simplify matters.
Voxel-wise GM differences between active meditators and controls were examined using independent-sample t-tests. In order to avoid possible edge effects between different tissue types, we excluded all voxels with GM values of less than 0.1 (absolute threshold masking). Statistical outcomes were corrected using small volume corrections, applying a 60 mm diameter of a sphere, and family-wise error (FWE) corrections for multiple comparisons. Significant outcomes were restricted to clusters exceeding 693 voxels (spatial extent threshold), in order to decrease the risk of detecting spurious effects due to noise. This spatial extent threshold corresponds to the expected number of voxels per cluster, calculated according to the theory of Gaussian random fields.
Supplemental Voxel-based GM Volume Analysis (Local Approach): Co-varying for Age
Although meditators and controls were carefully matched for age, we conducted an additional VBM analysis comparing meditators against controls (as described above), while co-varying for age. Again, we excluded all voxels with GM values of less than 0.1 (absolute threshold masking). Since these analyses were exploratory, we abstained from applying corrections for multiple comparisons (i.e., outcomes are presented as uncorrected findings at p<0.001) and applied an extent threshold of k=279 (corresponding to the expected number of voxels per cluster, re-calculated according to the new adjusted model).
Total Brain and GM Volume Analysis (Global Approach)
Using the tissue-classified partitions from the VBM analysis (i.e., GM, white matter [WM], and cerebrospinal fluid [CSF]), overall volumes were determined in cm3 as the sum of voxels representing GM+WM+CSF (total brain volume) or GM only (total GM volume). Both (a) total brain volumes and (b) total GM volumes were compared between meditators and control subjects using independent-sample t-tests.
Parcellated Volume Analysis (Regional Approach)
Cortical and subcortical structures were then parcellated using a validated hybrid discriminative / generative model, as detailed elsewhere (Tu et al., 2008
). Briefly, low-level information (i.e., signal intensity and local geometric properties) are used to determine the probability that a voxel belongs to a given structure. High-level information (i.e., global shape information), in connection with local smoothness constraints, are used to enforce the connectivity of each structure and its shape regularity. The applied hybrid model integrates both low- and high-level information into a unified system to reveal a gridface structure (voxel-wise labeling) which explicitly represents the three-dimensional topology of a particular region.
We automatically generated labels for the following regions: (1) left inferior temporal gyrus; (2) right insula; (3) right hippocampus; (4) right superior frontal gyrus; and (5) right middle frontal gyrus. These regions were selected based on outcomes of the two existing morphometric studies comparing brain anatomy between meditators and non-meditators (Lazar et al., 2005
; Holzel et al., 2008
. Brain size-adjusted (i.e., scaled) volumes of these cortical and subcortical structures were determined in cm3
as the sum of voxels belonging to a particular label, followed by comparing regional volume measures between meditators and controls using independent-sample t-tests.
Relationships between local GM and Number of Meditation Years
Finally, in order to compare our findings with others in the literature indicating positive correlations between meditation experience and brain structure, we also set out to examine whether there is an effect of the duration of meditation practice (measured in years) on local GM. For this purpose, we conducted two different analyses: First, we examined voxel-wise correlations between local GM and the number of meditation years in a regression analysis. However, we also speculated that possible positive correlations between meditation experience and GM are likely to be modulated and/or canceled out by changes in brain structure that are shown to occur with age (i.e., older subjects have the longest meditation history but are more prone to brain atrophy (Sowell et al., 2003
)). Unlike previous studies, the current study used a sample with a large age range (30–71 years), where 68 percent of all meditators were older than 50 years. Given that the trajectory of age effects varies considerably across different brain regions (Sowell et al., 2003
), simply integrating age as a co-variate in the statistical model is likely to further bias outcomes.
Thus, we decided to take the alternative approach of splitting the meditation groups based on years of meditation experience, and compared the two groups of meditators against two age-matched control groups using a 2×2 ANOVA. More specifically, one meditation group (Sample A) included individuals with less than 20 years of meditation experience (n=13) and the second meditation group (Sample B) included individuals with more than 20 years of meditation experience (n=9). The separation of these two groups was based on plotting the number of meditation years, where the clustering of the data revealed the 20 years-marker as the main separator between data clouds. The two control groups (Sample A / Sample B) included n=13 / n=9 subjects, respectively. This analysis strategy allowed for the additional examination of interactions between group status (Meditators / Controls) and mediation experience (Sample A / Sample B).
For both approaches, in correspondence with the original VBM analysis (described above), we excluded all voxels with GM values of less than 0.1 and corrected statistical outcomes using small volume corrections as well as family-wise error (FWE) corrections for multiple comparisons.