Grammar learning in older adults is linked to white matter microstructure and functional connectivity

NeuroImage 62 (2012) 1667–1674

Contents lists available at SciVerse ScienceDirect

NeuroImage journal homepage: www.elsevier.com/locate/ynimg

Grammar learning in older adults is linked to white matter microstructure and functional connectivity Daria Antonenko ⁎, 1, Marcus Meinzer 1, Robert Lindenberg, A. Veronica Witte, Agnes Flöel ⁎⁎ Department of Neurology, NeuroCure Cluster of Excellence, and Center for Stroke Research Berlin, Charité-Universitätsmedizin Berlin, Chariteplatz 1, 10117 Berlin, Germany

a r t i c l e

i n f o

Article history: Accepted 27 May 2012 Available online 31 May 2012 Keywords: Diffusion tensor imaging Fractional anisotropy Probabilistic tractography Resting-state functional connectivity Artificial grammar learning

a b s t r a c t Age-related decline in cognitive function has been linked to alterations of white matter and functional brain connectivity. With regard to language, aging has been shown to be associated with impaired syntax processing, but the underlying structural and functional correlates are poorly understood. In the present study, we used an artificial grammar learning (AGL) task to determine the ability to extract grammatical rules from new material in healthy older adults. White matter microstructure and resting-state functional connectivity (FC) of task-relevant brain regions were assessed using multimodal magnetic resonance imaging (MRI). AGL performance correlated positively with fractional anisotropy (FA) underlying left and right Brodmann areas (BA) 44/45 and in tracts originating from left BA 44/45. An inverse relationship was found between task performance and FC of left and right BA 44/45, linking lower performance to stronger inter-hemispheric functional coupling. Our results suggest that white matter microstructure underlying specific prefrontal regions and their functional coupling affect acquisition of syntactic knowledge in the aging brain, offering further insight into mechanisms of functional decline in older adults. © 2012 Elsevier Inc. All rights reserved.

1. Introduction Even though certain aspects of language processing are frequently impaired in healthy older adults (e.g., Meinzer et al., 2009, 2012b; Tyler et al., 2010; Wingfield and Grossman, 2006), very few studies assessed the underlying neural substrates of these impairments. In particular, syntax is a fundamental feature of the human language faculty, important for communication in private and work-related contexts (Kaestle et al., 2001). In the context of the increasing proportion of elderly persons worldwide, age-related decline or preservation of syntactic processes has enormous relevance for individuals and for society. A better understanding of maladaptive or beneficial functional and structural neural changes enables the development of potential mechanism-driven techniques to improve functioning (e.g., using non-invasive brain stimulation, Flöel et al., 2011; Meinzer et al., 2012a). Moreover, language impairments are among the earliest symptoms of age-associated pathological conditions like Alzheimer's disease and its precursors (Bickel et al., 2000; Henry et al., 2004; Murphy et al., 2006; Taler and Phillips, 2008) and are also core symptoms of poststroke language impairments (aphasia; e.g., Meinzer et al., 2011). Given that patients suffering from degenerative diseases and stroke

⁎ Correspondence to: D. Antonenko, Department of Neurology, CharitéUniversitätsmedizin Berlin, Chariteplatz 1, 10117 Berlin, Germany. ⁎⁎ Corresponding author. E-mail addresses: [email protected] (D. Antonenko), agnes.fl[email protected] (A. Flöel). 1 Contributed equally. 1053-8119/$ – see front matter © 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2012.05.074

are usually older adults, the assessment of changes in language processing and their underlying neural substrates in healthy adults are highly relevant as they are the baseline from which these patients depart. Age-related cognitive decline has been linked to structural alterations in cerebral white matter, and to changes in functional brain activity and connectivity patterns (for reviews see Goh, 2011; Madden et al., 2009; Park and Reuter-Lorenz, 2009). Previous studies have linked lower language performance in older adults to deterioration in bilateral white matter microstructure, as measured by fractional anisotropy (FA) using magnetic resonance imaging (MRI)-based diffusion tensor imaging (DTI; Obler et al., 2010; Stamatakis et al., 2011). For example, Obler et al. (2010) found that naming performance was correlated to FA values in left-hemisphere language-relevant brain regions as well as their right-hemisphere homologues. The authors concluded that successful naming skills of older adults may depend on white matter microstructure in brain regions and pathways in both hemispheres. In line with findings in other cognitive domains (Park and ReuterLorenz, 2009) functional MRI (fMRI) studies that used language paradigms to compare local task-related activity patterns in young and older healthy adults frequently found more bilateral processing in the older groups as compared to a unilateral (left-lateralized) pattern in young adults (Meinzer et al., 2012b; Tyler et al., 2010; Wierenga et al., 2008). The behavioral relevance of age-related changes in the lateralization of task-related activity for cognition, and for language in particular, has not yet been conclusively established. Regarding non-linguistic tasks, some studies reported beneficial effects of additional activity in the hemisphere that is not-dominant for the task (e.g., Cabeza, 2002; Park and Reuter-Lorenz, 2009), while others interpreted their findings

1668

D. Antonenko et al. / NeuroImage 62 (2012) 1667–1674

as the result of inefficient recruitment of specialized brain regions in the dominant hemisphere, disinhibition of non-specialized networks or enhanced demands placed on top-down control processes (e.g., Fling et al., 2011; Li and Lindenberger, 1999; Park and Reuter-Lorenz, 2009; Rajah and D'Esposito, 2005). In the language domain, beneficial effects of bilateral processing have primarily been reported for relatively easy tasks (Tyler et al., 2010; Wierenga et al., 2008) while it was associated with reduced performance during more challenging tasks (Meinzer et al., 2009, 2012b; Peelle et al., 2010). This would be in line with the assumption that control processes in left and/or right prefrontal areas can effectively compensate for structural damage or inefficient recruitment of specialized task-relevant neural populations to a certain degree (Davis et al., 2008; Park and Reuter-Lorenz, 2009). However, such “scaffolding” networks may be less efficient and prone to errors and at a higher level of challenge this may result in lower performance (cf. Park and Reuter-Lorenz, 2009, pp. 185–86). More recently, researchers have also begun to explore changes in functional connectivity (FC) to assess potential age-associated changes in coordinated brain activity at the network level (Andrews-Hanna et al., 2007; Chen et al., 2009; Davis et al., 2011). In the language domain, however, only one recent study addressed cortical agerelated changes at the functional network level. In this study, Peelle et al. (2010) found that reduced connectivity (i.e., co-variation between brain areas) within a mostly left-hemispheric syntax-processing network in older compared to young adults was associated with slower reaction times during a syntactic comprehension task. To study the acquisition of novel syntactic knowledge, which is an important aspect of natural language learning (Forkstam et al., 2006; Petersson et al., 2004), the artificial grammar learning (AGL) task has proven to be effective (de Vries et al., 2008, 2009; Kürten et al., 2012). In this task, subjects are asked to memorize letter strings without being aware of their inherent grammatical structure (implicit acquisition period; Reber, 1967). Afterwards participants are presented with novel sequences and asked to decide whether those are in line with the implicitly learned syntactic structure. Structural and functional brain imaging as well as brain stimulation studies confirmed a crucial involvement of left Brodmann areas (BA) 44/45 in young adults during AGL paradigms (de Vries et al., 2009; Flöel et al., 2009; Forkstam et al., 2006; Petersson et al., 2004; Uddén et al., 2008). Forkstam et al. (2006) suggested a specific role of the left BA 44/45 in extraction of syntactic rules. AGL performance can further be divided in chunkbased (explicit; more superficial) and rule-based (implicit) learning performance. In chunk-based learning, participants learn fragments or so-called chunks of the training items, such as letter pairs or triplets (Perruchet and Pacteau, 1990). In rule-based learning, implicit learning of the rules underlying the artificial grammar, such as an ‘M’ may be followed by a ‘V’ after a variable number of ‘S’ (see Fig. 1 and Reber, 1967), have to be encoded in order to successfully classify novel letter strings as grammatical or non-grammatical. Notably, in young subjects, rule-based AGL performance was correlated with

white matter microstructure underlying left but not right BA 44/45 and tracts originating from this area (Flöel et al., 2009). Furthermore, a significant age-related decline in performance has been shown for chunk-based but not rule-based learning elements representing explicit learning during AGL (Kürten et al., 2012; Meulemans and Van der Linden, 1997). The underlying neural correlates of inter-individual variability in older adults, however, have not been thoroughly scrutinized so far. In particular, previous studies have not combined structural and functional imaging to provide a comprehensive assessment of age-related changes in syntax processing. Therefore, in the present study we assessed AGL performance, as well as white matter microstructure and inter-hemispheric resting-state FC in 20 healthy older adults. Based on the study by Flöel et al. (2009) that investigated white matter correlates of AGL in young adults, we hypothesized that higher AGL performance, particularly for chunk-based learning, would be associated with higher FA values in left BA 44/45 and in white matter pathways originating from this region in older adults as well. However, as a number of studies implicated right frontal involvement in language processing in older adults (Obler et al., 2010; Stamatakis et al., 2011), we also expected to find an association of performance with white matter underlying right BA 44/45. Moreover, functional imaging and brain stimulation studies suggested that suppression of contralateral activity can benefit language and motor task performance (Schäfer et al., 2012; Thiel et al., 2006). In particular, in the context of highly lateralized functions like syntax processing (Friederici, 2011), performance may depend on successful inhibition of contralateral brain regions that are not involved in the respective task (Netz et al., 1995). Therefore, we also hypothesized that enhanced FC of specialized left-hemisphere regions (i.e., BA 44/45) with their right-hemisphere homologues, possibly mediated by reduced transcallosal or inter-hemispheric inhibition (Fling et al., 2011), would be associated with decreased behavioral performance. 2. Materials and methods 2.1. Participants Twenty healthy elderly German speakers participated in the study (ten women; mean±SD age: 69.9±3.0 years, range: 64–76; years of education 16.3 ±3.3 years, range: 11–22). Subjects were recruited from a local database of healthy older adults who underwent an extensive neuropsychological testing to assure normal cognitive functioning (working memory: digit span forward and backward (Wechsler, 1987), executive functioning: phonemic [letter S] and semantic [category animals] based fluency, and Trail-Making Test A and B (Consortium to Establish a Registry for Alzheimer's Disease (CERAD)-Plus; www.memoryclinic.ch); verbal memory: Verbal Learning Memory Task (VLMT) (Helmstaedter et al., 2001)). Performance scores in neuropsychological tests were within age-corrected norms, see Table 1 for details. All participants were righthanded (Edinburgh Handedness Inventory (Oldfield, 1971)) and had no history of medical or neurological disorders. Data of one participant were not included in the analyses because he did not understand the task instruction (resulting in a d-primeb 0, see below). Data of two participants had to be excluded either from DTI (n= 1) or resting state analyses (n= 1) due to severe movement artifacts, resulting in a total of 18 participants in each analysis and 17 in the linear regression model (see below). The study was approved by the local ethics committee. All subjects gave written informed consent and were reimbursed for their participation. 2.2. AGL task

Fig. 1. Graphical representation of the finite state grammar used to generate the “grammatical” letter strings.

The AGL task has been described in detail elsewhere (de Vries et al., 2009; Flöel et al., 2009; Kürten et al., 2012). In short, the task comprised two steps: (1) an acquisition part during which an artificial grammar

D. Antonenko et al. / NeuroImage 62 (2012) 1667–1674 Table 1 Baseline assessment.

Digit span forward Digit span backward Phonemic fluency (S) Semantic fluency (animals) TMT-A (s) TMT-B (s) VLMT: learning VLMT: immediate recall VLMT: delayed recall

Mean ± SD

Range

7.4 ± 1.6 6.2 ± 2.0 17.6 ± 5.0 27.2 ± 5.3 32.6 ± 9.8 75.0 ± 27.2 54.0 ± 10.0 10.1 ± 3.2 10.8 ± 3.2

4–11 2–10 9–25 18–39 23.6–64.7 44.5–168.7 35–70 4–15 5–15

VLMT learning includes the sum of all five learning trials with the same 15 words (resulting in a maximum score of 75 words). TMT-A/B: Trail-Making Test Version A/ B; VLMT: Verbal Learning and Memory Test.

was learned implicitly, and (2) a classification part that assessed performance. Stimuli consisted of consonant letter strings (4–12 letters; mean± SD: 10.7 ± 1.5), generated from the finite state grammar as depicted in Fig. 1, using consonants from the alphabet (M, S, V, R, X). Importantly, the finite-state grammar not only includes learning of short fragments (chunks or local dependencies of elements), but also comprises rule learning (long-distance dependencies) that requires the abstraction of grammatical structure (Forkstam et al., 2006; Poletiek, 2002). Examples of valid letter strings are MSSV, VXVRXRM, and MVRXRRM, derived by progressing from node to node through valid transitions (the arrows), from start to end (Reber, 1967). Invalid letter strings include for example MXRS, VRMMSXR, or VXRSSMV. In the following, letter strings that followed the artificial grammar rules (valid) will be referred to as “grammatical strings” and those that did not (invalid) as “non-grammatical strings”. During the acquisition part (~35 min), that was introduced to the subjects as a working memory task, 100 grammatical strings were presented on a computer screen for 5 s using Presentation® (Neurobehavioral Systems, Albany, CA). Participants were instructed to encode the strings during the presentation and to recall them (by re-typing) as accurately as possible in a self-paced manner after they had disappeared. Only correctly re-typed sequences (i.e., all letters of the string in correct order) counted as correct response. In the subsequent classification task (~15 min), 100 different grammatical and 100 non-grammatical strings were presented and subjects had to classify the presented strings by a button press as either “grammatical” or “non-grammatical” according to the implicitly acquired rules. Grammatical strings in this classification task can further be differentiated with respect to their superficial familiarity (chunk strength: relative frequency of equal bi- or trigrams of letters in the strings) relative to the learned strings in the acquisition task. Thus, separate performance measures could afterwards be calculated for high and low chunk strength items reflecting a more superficial or chunk-based (with explicit elements) versus rule-based (implicit) grammar learning ability (Forkstam et al., 2006; Lieberman et al., 2004). The task generated four response categories: hits (i.e., correct classifications), correct rejections, false alarms, and misses. D-prime measures were calculated (z-transformed hit rate minus z-transformed false alarm rate) to index general AGL performance (d-prime; primary outcome measure) as well as specific performances for items with high (d-primeHIGH) and low chunk strengths (d-primeLOW). For all correlations of AGL performance with imaging parameters, nonparametric statistical tests (Spearman rho's (rs) two-tailed correlation coefficients and permutation testing) were conducted. 2.3. MRI data acquisition All subjects underwent MRI in a 3 T Siemens Magnetom Trio Scanner with a 12-channel head coil. We obtained fMRI scans at rest

1669

(TR = 2300 ms, TE= 30 ms, 34 slices, voxel size of 3.0 × 3.0 × 4.0 mm3, flip angle = 90°) and diffusion-weighted images using a spin-echo EPI sequence (TR = 7500 ms, TE= 86 ms, 61 axial slices, voxel size of 2.3 × 2.3× 2.3 mm3; 64 directions with a b-value of 1000 s/mm 2 and one b0). High-resolution T1-weighted MPRAGE images (TR = 1900 ms, TE = 2.52 ms, 192 sagittal slices, voxel-size of 1.0 ×1.0 × 1.0 mm3, flip angle= 9°) were acquired in order to detect potential structural abnormalities and to facilitate normalization of the functional scans. 2.4. DTI analyses FSL (www.fmrib.ox.ac.uk/fsl) was used for image pre-processing which included correction for eddy currents and head motion using affine registration (Jenkinson and Smith, 2001), extraction of nonbrain tissue (Smith et al., 2004), calculation of a probability distribution of fiber directions for each voxel while allowing estimates of two directions per voxel (Behrens et al., 2007). 2.4.1. Region of interest (ROI) definition The Juelich Histological Atlas was used to define ROIs. As this atlas provides probabilistic maps for BA 44 and BA 45, these were combined (or merged) and binarized in order to create ROIs of left and right BA 44/ 45. Specifically, voxels surpassing a probability of 0.5 were deemed as belonging within the ROI and assigned a value of 1, whereas voxels below the 0.5 probability threshold were considered outside the ROI mask and assigned 0. As part of the perisylvian language system, a control ROI within the inferior parietal lobe (IPL) was chosen. This area was found to be activated during AGL performance (Forkstam et al., 2006). Left V1 (BA 17) served as a language-unrelated control ROI. In order to extract inter-hemispheric FA, an additional ROI was defined in the anterior corpus callosum and thresholded at the probability of 0.5, as the microstructure of anterior callosal fibers was used as an indicator for inter-hemispheric structural connectivity in various studies (for a review see Fling et al., 2011). The ROIs were transformed to match the space of each subject's diffusion data (“native space”) using affine registration (Jenkinson and Smith, 2001) in order to extract FA values for further analysis. 2.4.2. Tract-based spatial statistics (TBSS) The TBSS method aims at testing for local correlations between voxel-wise FA values and (in this case) behavioral data (Smith et al., 2006). Instead of considering the whole brain white matter, the procedure computes individual white matter skeletons (centers of white matter tracts). Voxel-wise correlations between FA values and d-prime measures were performed in order to obtain clusters associated with AGL performance while controlling for age-related effects (i.e., regressing out the covariate “age”). Permutation-based statistical analyses were conducted using the program “randomise” implemented in FSL. This program allows modelling and statistical inferences in the context of the standard General Linear Model (GLM; Nichols and Holmes, 2002). Threshold-free cluster enhancement (TFCE; threshold for significant voxels: p b 0.05) was used as an alternative to clusterbased thresholding to enhance cluster-like structures in an image without having to define an initial cluster-forming threshold or carry out a large amount of data smoothing (Smith and Nichols, 2009). The analysis was restricted to voxels of the skeleton within BA 44/45 (equivalent to Flöel et al., 2009). For visualization, significant clusters were “thickened” (by convolving the thresholded statistical images by a 3 mm sphere) and overlaid onto the group mean FA skeleton. 2.4.3. Probabilistic tractography Tractography was performed from seed voxels that were specified as follows: The probabilistic maps of left BA 44/45 (see above) were thresholded at the probability of 0.5. The thresholded maps were then overlaid onto the individual white matter skeletons obtained

1670

D. Antonenko et al. / NeuroImage 62 (2012) 1667–1674

from TBSS. Only voxels common to both the probabilistic map and skeleton were used as seed masks for probabilistic tractography in order to assure that seed masks included only voxels belonging to brain white matter (Flöel et al., 2009). To obtain the righthemisphere seed, the mirror image of the left-hemisphere seed was used, so that the number of seed voxels was equal allowing a comparison of generated tracts. A multi-fiber model was fit to the diffusion data at each voxel in order to trace fibers through regions of fiber crossing or complexity (Behrens et al., 2007). The model calculates probability distributions of fiber directions for each voxel while allowing estimates of two directions per voxel. Probabilistic tractography was carried out by propagating 5000 samples through the estimated probability distributions from each seed voxel. In the output connectivity distribution, values at all brain voxels represent the number of samples passing through any given voxel and thus the global connectivity between that voxel and the seed voxels. The resulting streamlines were constrained to voxels with more than 3% of the individual mean connectivity in order to remove spurious connections. Thresholded pathways were then binarized in each subject to obtain a mask for each subject's tracts. Mean FA was extracted from thresholded tracts in order to compute correlations with AGL performance. 2.5. Resting-state fMRI analyses Resting-state FC was assessed using customized processing scripts developed from FreeSurfer (http://surfer.nmr.mgh.harvard.edu), AFNI (http://afni.nimh.nih.gov/afni), and FSL software packages, based on the 1000 Functional Connectomes Project (www.nitrc.org/projects/ fcon_1000) (Biswal et al., 2010). Pre-processing of individual anatomical scans included brain extraction, segmentation into different tissue types, and bias field correction. Pre-processing of individual functional scans comprised motion correction, spatial smoothing with a 6-mm fullwidth-at-half-maximum (FWHM) Gaussian kernel, temporal filtering (0.01–0.1 Hz), and de-trending. The functional scans were co-registered with the anatomical image using affine registrations. Nuisance signals

(e.g., motion parameters, white matter, cerebrospinal fluid, and global signal) were removed from the functional signal by multiple regressions. Normalization of inter-individual images for group analyses were performed using linear and non-linear registrations to a scanner-specific template, generated from 116 age-matched (mean ±SD: 63.6± 6.5) healthy older adults (58 women) at 2 mm isotropic voxel size. 2.5.1. Seed ROI selection and seed-based correlations After pre-processing, seed-based correlations were calculated using the standardized residuals of the time series. Individual interhemispheric FC strength was assessed between left and right BA 44/ 45, and intra-hemispheric FC between BA 44/45 and IPLs. As a control, FC was assessed between the left BA 44/45 and left BA 17. ROIs were defined as spheres with a 10-mm radius around the maxima of the respective probability maps of the Juelich Histological Atlas, coregistered to the study-specific template. The resulting Pearson's r correlation coefficient maps were Fisher's z transformed to obtain normal distribution (method similar to Carter et al., 2010) in order to correlate these measures of FC strength with AGL performance. 2.5.2. Linear regression model In order to obtain an overall statistical model for AGL performance, the variables ‘age’, ‘FA of left BA 44/45’, ‘FA of right BAs 44/45’, ‘FA extracted from tracts originating in left BA 44/45’, and ‘FC between left and right BA 44/45’ were entered into a backward stepwise linear regression model. 3. Results 3.1. AGL task During the acquisition phase, subjects correctly re-typed 15.26 ± 11.53 (mean ± SD) of 100 letter strings after they had disappeared. Performance during the acquisition task of older adults in our study was comparable to that of older adults in the study of Kürten et al. (2012) that used the same task. In the classification task, d-prime

Fig. 2. A: correlation of FA values extracted from left and right BA 44/45 with AGL performance (d-prime). B: regions of the white matter skeleton within BA 44/45 (blue) where FA values correlated with AGL performance (d-prime; red-yellow). Coordinates in mm of the MNI space. FA: fractional anisotropy. BA: Brodmann areas. L: left, R: right.

D. Antonenko et al. / NeuroImage 62 (2012) 1667–1674

1671

Table 2 Clusters of the white matter skeleton within left and right BA 44/45 where FA values showed significant positive correlations with AGL performance.

d-prime Left BA 44/45

Right BA 44/45

d-primeHIGH Left BA 44/45

Right BA 44/45

d-primeLOW Right BA 44/45

Cluster

k

Peak p-value

MNI coordinates x, y, z

1 2 3 4 5 6 7 8 9 10 11 12 13

229 77 34 16 14 8 7 7 113 44 24 11 7

0.001 0.005 0.011 0.019 0.019 0.028 0.031 0.037 0.002 0.006 0.021 0.010 0.034

− 43, 11, 13 − 29, 22, 18 − 35, 36, 12 − 51, 16, 23 − 33, 11, 39 − 41, 13, 37 − 41, 4, 31 − 36, 10, 34 49, 2, 14 51, 18, 17 52, 4, 30 44, 15, 34 47, 10, 5

1 2 3 4 5 6 7 8 9 10 11

301 211 129 86 67 30 123 54 36 11 11

b0.001 b0.001 0.001 b0.001 0.004 0.024 0.002 0.014 0.014 0.005 0.027

− 43, 10, 12 − 31, 22, 20 − 42, 6, 29 − 34, 37, 15 − 52, 5, 26 − 52, 16, 20 49, 2, 15 44, 5, 9 33, 25, 17 43, 16, 34 48, 22, 17

1

16

0.017

51, 19, 18

Only clusters with a minimum extent of 6 voxels were included. k =cluster size (# voxels). BA: Brodmann areas.

measures indicated that participants successfully acquired the implicit grammar (d-prime: 0.64 ± 0.37; d-primeHIGH: 0.79 ± 0.54; dprimeLOW: 0.53 ± 0.47). AGL performance (during the classification task) correlated with age for d-prime (rs = −0.50, p = 0.03) and dprimeHIGH (rs = −0.45, p = 0.05), but not for d-primeLOW. There were no correlations of d-prime measures with performance on neuropsychological tests. Post-experimental debriefing revealed that subjects were not aware of the underlying implicit learning during the acquisition task, leading to a subjective experience of “guessing” in the classification task.

3.2. DTI analysis 3.2.1. ROI analysis Mean FA of left as well as right BA 44/45 correlated with AGL performance (d-prime: rs = 0.64, p b 0.01 for left, and rs = 0.50, p = 0.03 for right, Fig. 2A). Mean FA of the left IPL showed a trend for a correlation with d-prime (rs = 0.44, p = 0.07), whereas mean FA of the right IPL as well as the left V1 did not correlate with d-prime measures. Correlations of FA values with AGL performance were mainly driven by d-primeHIGH (rs = 0.66, p b 0.01 for left BA 44/45; rs = 0.43; p = 0.07 for right BA 44/45), correlations with d-primeLOW were not significant. The level of statistical significance was not adjusted for the number of ROIs.

3.2.2. TBSS Considering an extent threshold of at least six voxels (see Flöel et al., 2009), several clusters within left and right BA 44/45 correlated with AGL performance (d-prime; Fig. 2B and Table 2). Similar clusters within left and right BA 44/45 emerged for d-primeHIGH, but only one within right BA 44/45 for d-primeLOW (Table 2).

Fig. 3. A: correlation of FA values extracted from individual tracts and AGL performance (d-prime). B: probability maps of tracts originating in voxels of the white matter skeleton within left BA 44/45; to generate the canonical image, individual tracts from all subjects were normalized, converted to binary images and then summed; color coding reflects probability of voxels to be present in 50% (dark blue) to 100% of subjects (light blue). Coordinates in mm of the MNI space. FA: fractional anisotropy. L: left, R: right.

3.2.3. Probabilistic tractography White matter pathways originating in BA 44/45 showed a comparable anatomical connectivity pattern with regions supporting languagerelated functions as previous studies (Anwander et al., 2007; Catani et al., 2005; Frey et al., 2008; Friederici, 2009). Pathway volumes were significantly larger for tracts originating from left compared to right BA 44/45 (mean± SD # voxels left: 3983.6 ± 1680.7; right: 3076.5 ± 1787.5; t = 2.48, p = 0.024) which is in line with previous studies showing asymmetries in language-related pathways (Nucifora et al., 2005; Powell et al., 2006). FA values extracted from the pathways however did not differ. Visual inspection of individual pathways revealed a similar pattern of structural connectivity across subjects and hemispheres. BA 44/45 was connected to dorsolateral and medial frontal (including the precentral area/pre-supplementary motor area [preSMA]), inferior parietal (via the superior longitudinal fasciculus [SLF]), and superior temporal (via the extreme capsule fiber system [ECFS] and insular cortex) cortical regions. Furthermore, subcortical gray matter such as the basal ganglia and thalamus were also found to be connected with left and right BA 44/45. Both left- and right-hemisphere pathways

Fig. 4. Correlation of FC between the seeds within left and right BA 44/45 and AGL performance (d-prime). FC: functional connectivity. BA: Brodmann areas. L: left, R: right.

1672

D. Antonenko et al. / NeuroImage 62 (2012) 1667–1674

crossed the midline through the anterior callosal body. In the pathway originating from left BA 44/45, inter-hemispheric connections appeared to be more pronounced with fibers extending to contralateral medial frontal and subcortical regions. In order to assess the behavioral relevance of inter-individual differences of FA in white matter tracts originating within BA 44/45, we correlated FA values extracted from these tracts with AGL performance. Tract-specific FA values of streamlines originating in left BA 44/45 correlated with AGL performance (d-prime: rs = 0.53, p = 0.03; Fig. 3A), whereas FA values of tracts that were reconstructed using seeds in right BA 44/45 did not correlate with AGL performance. No correlations were observed for FA values extracted from left and right tracts with d-primeHIGH or d-primeLOW. Fig. 3B shows the probability map of the white matter tract originating from left BA 44/45 that was generated by normalizing (to standard space) and merging the binarised individual pathways. Accordingly, voxel values represent the number of subjects in whom the pathway is present. 3.3. Resting-state FC FC between left and right BA 44/45 showed a significant negative correlation with AGL performance (d-prime: rs = −0.50, p = 0.04; Fig. 4; both d-primeHIGH and d-primeLOW:rs = −0.46, p = 0.06). FC between left or right BA 44/45 and ipsilateral IPL as well as between left BA 44/45 and ipsilateral V1 showed no significant correlations with any d-prime measure. FC between left and right BA 44/45 showed no significant correlation with inter-hemispheric FA. 3.4. Linear regression model The variables ‘FA of left BA 44/45’ and ‘FC between left and right BA 44/45’ remained in the model (beta/p-values: 0.57/b0.01 and −0.40/ 0.05) which explained 51.3% (adjusted R 2 = 44.3%) of the variance in d-prime. When choosing d-primeHIGH as dependent variable, the two predictors ‘FA of left BA 44/45’ and ‘FC between left and right BA 44/ 45’ (beta/p-values: 0.95/b0.01, and −0.45/b0.01) remained significant in the model, explaining 79.3% (adjusted R2 = 74.5%) of its variance. For dprimeLOW, none of the variables remained in the model. 4. Discussion In the present study, we demonstrated that the ability of older adults to acquire syntactic knowledge was positively correlated with FA of white matter microstructure underlying left and right BA 44/45 and tracts originating in left BA 44/45, whereas negative correlations were found with FC strength between left and right BA 44/45. Notably, performance was best predicted by a comprehensive model that incorporated white matter microstructure underlying left BA 44/ 45 and FC strength between left and right BA 44/45. 4.1. Association of white matter underlying left and right BA 44/45 with grammar learning in older adults In this study, we found that the grammar learning ability of older adults relied on FA of white matter microstructure underlying left BA 44/45 (including Broca's area) and pathways arising from this area, similar to previous results from young adults (Flöel et al., 2009). In addition, white matter microstructure of right-hemispheric homologue areas as measured by FA was associated with AGL performance in our older participants. Our study provides further support that right hemispheric homologues might play an important role in mediating language learning. A bilateral correlation of white matter microstructure with behavior in older adults was also found by Obler et al. (2010) who showed that higher FA values in left- as well as righthemisphere perisylvian and mid-frontal white matter structures were associated with better naming ability in healthy older participants.

Likewise, Stamatakis et al. (2011) confirmed an association of FA values in bilateral language-related brain areas and performance on the tip-oftongue-task in older adults. Note however that in our study, tracts originating from right BA 44/45 did not show a relationship with grammar learning, indicating a preserved dominant role for left BA 44/45 and its connections with posterior language areas for acquiring syntactic knowledge. This interpretation is supported by the strong trend for correlation between performance and FA values in the left but not right IPL. Also, in Stamatakis et al. (2011), statistical peaks of clusters correlating with performance were located in the left hemisphere, likewise indicative of left-hemisphere dominance of white matter microstructure for language processes. In young subjects, Flöel et al. (2009) observed a specific correlation of left-hemispheric white matter microstructure and performance on rule-based grammar acquisition that depends on implicit learning (Reber, 1967). For older adults, the present study revealed correlations of left-hemispheric white matter microstructure particularly for chunkbased learning, known to rely on both implicit and explicit elements (Lieberman et al., 2004; Meulemans and Van der Linden, 1997). Thus, it seems that in the elderly, not only implicit but also explicit learning ability in the verbal domain relies on intact white matter microstructure in the left hemisphere.

4.2. Negative correlation of inter-hemispheric functional coupling with grammar learning in older adults We found a negative correlation between performance of older adults and the inter-hemispheric functional coupling, indicating that more synchronous spontaneous fluctuations between left and right BA 44/45 during task-independent resting-state fMRI are linked to lower grammar learning ability. This association was evident in both explicit and implicit elements of the task, reflecting a relation of functional coupling with both superficial and rule-based learning abilities. As the acquisition of syntactic knowledge is assumed to be a highly lateralized function that depends on left-hemispheric perisylvian language networks (Wingfield and Grossman, 2006; Xiang et al., 2010), a higher temporal correlation of parts of this network with righthemispheric areas might determine decreased behavioral performance. Indeed, higher functional connectivity within specialized perisylvian language networks has been shown to be associated with superior syntax processing in older adults (Peelle et al., 2010) and superior reading abilities in young to middle-aged adults (Koyama et al., 2011). However, temporal correlations of spontaneous fluctuations to brain areas outside these specialized regions, such as inter- rather than intra-hemispheric coupling, were associated with poorer performance (Koyama et al., 2011). Koyama et al. (2011) hypothesized that functional specialization may reflect more efficient neural processing. We suggest that greater functional correlation between bilateral prefrontal areas in our study might be explained by a lack of inhibition between functionally connected prefrontal brain regions, mediated by bilateral deterioration of white matter structures (Fling et al., 2011; Netz et al., 1995). This would be in line with a recent study that assessed resting-state FC in older adults and found a negative correlation between FC and verbal learning performance; i.e., higher FC of bilateral subcortical areas (basal ganglia and thalamus) was correlated with lower learning ability (Ystad et al., 2010). However, it needs to be noted that these authors assessed subcortical structures that are distinct from cortical tissue and only one recent study using magnetoencephalography (MEG) reported that enhanced resting-state FC of cortical areas may be associated with reduced cognitive performance (e.g., Schlee et al., 2012). This may be particularly important in the context of highly lateralized functions like syntax learning which have been shown to rely on left-hemispheric specialized networks. Here, reduced interhemispheric coupling, in combination with more pronounced intra-

D. Antonenko et al. / NeuroImage 62 (2012) 1667–1674

hemispheric coupling (Koyama et al., 2011) might be beneficial for performance. 4.3. Comprehensive model: prediction of grammar learning performance by white matter underlying left BA 44/45 and inter-hemispheric functional coupling in older adults In a regression model that incorporates both structural and functional connectivity data, higher white matter microstructure underlying left BA 44/45 and lower FC between left and right BA 44/45 were the best predictors for grammar learning performance in older adults. These findings support the hypothesis that grammar learning ability of older adults is reduced if left-hemisphere white matter underlying BA 44/45 is lower and inter-hemispheric functional coupling between left and right BA 44/45 is higher. Our results complement previous studies that showed age-related performance differences in healthy aging to be best predicted by regional structural and inter-regional functional variations in task-related fMRI (Andrews-Hanna et al., 2007; Chen et al., 2009; Madden et al., 2004; Ystad et al., 2010). Thus, combining structural and functional assessments might allow insights into age-related cognitive changes surpassing the explanatory values of either method alone. 4.4. Limitations First, our cross-sectional approach is able to test and describe associations between inter-individual differences in brain imaging parameters and behavior, but does not allow inferences about causality of relationships. Statistical models which explained a considerable amount of variance in dependent variables can however provide indicators of directionality of associations. Second, it has to be kept in mind that our results did not emerge from a whole brain analysis using fully corrected statistics, but were based on ROI approaches, without correcting the level of significance for the number of ROIs. However, we have increased confidence in the present findings for a number of reasons: we had a strong a priori hypothesis of the involvement of BA 44/ 45 area in this task, given previous findings from our own (Flöel et al., 2009) and other groups (Forkstam et al., 2006; Petersson et al., 2004) using the same task (AGL). In addition, we found specific effects not only for the hypothesis-driven pre-defined ROIs (as compared to control ROIs), but also for the regression model using combined functional and structural imaging data. Moreover, similar associations between white matter microstructure and grammar learning ability were obtained from extraction of regional FA values, from TBSS, and from tract-specific FA values. Third, only elderly individuals were included in the present study, so a direct comparison with young individuals was not possible. However, previous results from our own group on young subjects, using the identical task, provide information on structural correlates of AGL performance in the young. Results from other groups (Koyama et al., 2011) that provide information on functional coupling suggest that young subjects would show less interhemispheric and more intra-hemispheric coupling compared to the elderly, a hypothesis to be explored in future studies. Fourth, we also did not find correlations of inter-hemispheric FC with inter-hemispheric FA as indicated by the microstructure of the anterior corpus callosum. This lack of a direct association may be most likely explained by the fact that functional communication between the left and right IFG is not solely mediated by a direct connection through the corpus callosum, but also by complex polysynaptic connections through bilateral thalamic, basal ganglia, and pre-supplementary connections (e.g., Crosson et al., 2003). In order to test the possibility that there is a lack of interhemispheric inhibition from left to right IFG, a task-related paradigm in combination with a hypothesis driven functional connectivity analysis would have been required (e.g., Dynamic Causal Modelling, Stephan et al., 2010). This needs to be scrutinized in future studies.

1673

4.5. Conclusions and outlook In summary, our results suggest that the variability in the acquisition of syntactic knowledge in older adults relies on intact white matter microstructure in language-related areas as well as their interhemispheric functional coupling, offering further insight into interindividual differences of brain-behavior relationships in the elderly. Given that relationships between brain imaging parameters and behavioral performances likely differ between age-groups and cognitive domains, future studies should apply multivariate cognitive assessments to investigate neural correlates of behavioral variability in young and elderly individuals. Additionally, future studies should try to modulate microstructural features of white matter (e.g., through intense training approaches, Scholz et al. (2009)), or inter-hemispheric functional coupling (e.g., through noninvasive electrical brain stimulation, Meinzer et al. (2012a)), to draw conclusions about the causality of relationships and to possibly identify strategies to revert agerelated functional and structural changes.

Disclosure statements All authors declare that they have no conflicts of interest.

Acknowledgments The authors thank Lucia Kerti and Henrike Rupp for help with data collection, and Dr. Nadine Külzow and Angela Winkler for helpful comments on previous versions of the manuscript. This work was supported by grants from the Deutsche Forschungsgemeinschaft (to AF: Fl-379-8/1; and by DFG-Exc-257), the Bundesministerium für Bildung und Forschung (to AF: FKZ0315673A; to AF and MM 01EO0801), and the Else-Kröner Fresenius Stiftung (to AF: 2009-141).

References Andrews-Hanna, J.R., Snyder, A.Z., Vincent, J.L., Lustig, C., Head, D., Raichle, M.E., Buckner, R.L., 2007. Disruption of large-scale brain systems in advanced aging. Neuron 56, 924–935. Anwander, A., Tittgemeyer, M., von Cramon, D.Y., Friederici, A.D., Knosche, T.R., 2007. Connectivity-based parcellation of Broca's area. Cereb. Cortex 17, 816–825. Behrens, T.E.J., Berg, H.J., Jbabdi, S., Rushworth, M.F.S., Woolrich, M.W., 2007. Probabilistic diffusion tractography with multiple fibre orientations: what can we gain? NeuroImage 34, 144–155. Bickel, C., Pantel, J., Eysenbach, K., Schroder, J., 2000. Syntactic comprehension deficits in Alzheimer's disease. Brain Lang. 71, 432–448. Biswal, B.B., Mennes, M., Zuo, X.-N., Gohel, S., Kelly, C., Smith, S.M., Beckmann, C.F., Adelstein, J.S., Buckner, R.L., Colcombe, S., Dogonowski, A.-M., Ernst, M., Fair, D., Hampson, M., Hoptman, M.J., Hyde, J.S., Kiviniemi, V.J., Kötter, R., Li, S.-J., Lin, C.-P., Lowe, M.J., Mackay, C., Madden, D.J., Madsen, K.H., Margulies, D.S., Mayberg, H.S., McMahon, K., Monk, C.S., Mostofsky, S.H., Nagel, B.J., Pekar, J.J., Peltier, S.J., Petersen, S.E., Riedl, V., Rombouts, S.a.R.B., Rypma, B., Schlaggar, B.L., Schmidt, S., Seidler, R.D., Siegle, G.J., Sorg, C., Teng, G.-J., Veijola, J., Villringer, A., Walter, M., Wang, L., Weng, X.-C., Whitfield-Gabrieli, S., Williamson, P., Windischberger, C., Zang, Y.-F., Zhang, H.-Y., Castellanos, F.X., Milham, M.P., 2010. Toward discovery science of human brain function. Proc. Natl. Acad. Sci. U. S. A. 107, 4734–4739. Cabeza, R., 2002. Hemispheric asymmetry reduction in older adults: the HAROLD model. Psychol. Aging 17, 85–100. Carter, A.R., Astafiev, S.V., Lang, C.E., Connor, L.T., Rengachary, J., Strube, M.J., Pope, D.L., Shulman, G.L., Corbetta, M., 2010. Resting interhemispheric functional magnetic resonance imaging connectivity predicts performance after stroke. Ann. Neurol. 67, 365–375. Catani, M., Jones, D.K., ffytche, D.H., 2005. Perisylvian language networks of the human brain. Ann. Neurol. 57, 8–16. Chen, N.-k., Chou, Y.-h., Song, A.W., Madden, D.J., 2009. Measurement of spontaneuos signal fluctuations in fMRI: adult age differences in intrinsic functional connectivity. Brain Struct. Funct. 213. Crosson, B., Benefield, H., Cato, M.A., Sadek, J.R., Moore, A.B., Wierenga, C.E., Gopinath, K., Soltysik, D., Bauer, R.M., Auerbach, E.J., Gokcay, D., Leonard, C.M., Briggs, R.W., 2003. Left and right basal ganglia and frontal activity during language generation: contributions to lexical, semantic, and phonological processes. J. Int. Neuropsychol. Soc. 9, 1061–1077. Davis, S.W., Dennis, N.A., Daselaar, S.M., Fleck, M.S., Cabeza, R., 2008. Que PASA? The posterior–anterior shift in aging. Cereb. Cortex 18, 1201–1209.

1674

D. Antonenko et al. / NeuroImage 62 (2012) 1667–1674

Davis, S.W., Kragel, J.E., Madden, D.J., Cabeza, R., 2011. The architecture of crosshemispheric communication in the aging brain: linking behavior to functional and structural connectivity. Cereb. Cortex 22, 232–242. de Vries, M.H., Monaghan, P., Knecht, S., Zwitserlood, P., 2008. Syntactic structure and artificial grammar learning: the learnability of embedded hierarchical structures. Cognition 107, 763–774. de Vries, M.H., Barth, A.C.R., Maiworm, S., Knecht, S., Zwitserlood, P., Flöel, A., 2009. Electrical stimulation of Broca's area enhances implicit learning of an artificial grammar. J. Cogn. Neurosci. 2427–2436. Fling, B.W., Peltier, S.J., Bo, J., Welsh, R.C., Seidler, R.D., 2011. Age differences in interhemispheric interactions: callosal structure, physiological function, and behavior. Front. Neurosci. 5, 38. Flöel, A., de Vries, M.H., Scholz, J., Breitenstein, C., Johansen-Berg, H., 2009. White matter integrity in the vicinity of Broca's area predicts grammar learning success. NeuroImage 47, 1974–1981. Flöel, A., Meinzer, M., Kirstein, R., Nijhof, S., Deppe, M., Knecht, S., Breitenstein, C., 2011. Short-term anomia training and electrical brain stimulation. Stroke 42, 2065–2067. Forkstam, C., Hagoort, P., Fernandez, G., Ingvar, M., Petersson, K.M., 2006. Neural correlates of artificial syntactic structure classification. NeuroImage 32, 956–967. Frey, S., Campbell, J.S., Pike, G.B., Petrides, M., 2008. Dissociating the human language pathways with high angular resolution diffusion fiber tractography. J. Neurosci. 28, 11435–11444. Friederici, A.D., 2009. Pathways to language: fiber tracts in the human brain. Trends Cogn. Sci. 13, 175–181. Friederici, A.D., 2011. The brain basis of language processing: from structure to function. Physiol. Rev. 91, 1357–1392. Goh, J.O., 2011. Functional dedifferentiation and altered connectivity in older adults: neural accounts of cognitive aging. Aging Dis. 2, 30–48. Helmstaedter, C., Lendt, M., Lux, S., 2001. VLMT: Verbal Learning and Memory Test. Beltz Test GmbH, Göttingen, Germany. Henry, J.D., Crawford, J.R., Phillips, L.H., 2004. Verbal fluency performance in dementia of the Alzheimer's type: a meta-analysis. Neuropsychologia 42, 1212–1222. Jenkinson, M., Smith, S., 2001. A global optimisation method for robust affine registration of brain images. Med. Image Anal. 5, 143–156. Kaestle, C.F., Campbell, A., Finn, J.D., Johnson, S.T., Mikulecky, L.J., 2001. Adult Literacy and Education in America: Four Studies Based on the National Adult Literacy Survey. U.S. Department of Education, National Center for Education Statistics, Washington, DC, pp. 2001–2534. Koyama, M.S., Di Martino, A., Zuo, X.N., Kelly, C., Mennes, M., Jutagir, D.R., Castellanos, F.X., Milham, M.P., 2011. Resting-state functional connectivity indexes reading competence in children and adults. J. Neurosci. 31, 8617–8624. Kürten, J., de Vries, M.H., Kowal, K., Zwitserlood, P., Flöel, A., 2012. Age affects chunkbased, but not rule-based learning in artificial grammar acquisition. Neurobiol. Aging 33, 1311–1317. Li, S.-J., Lindenberger, U., 1999. Cross-level unification: a computational exploration of the link between deterioration of neurotransmitter systems dedifferentation of cognitive abilities in old age. In: O'Connor, D. (Ed.), Dementia and Normal Aging. Cambridge University Press, Cambridge, pp. 331–365. Lieberman, M.D., Chang, G.Y., Chiao, J., Bookheimer, S.Y., Knowlton, B.J., 2004. An eventrelated fMRI study of artificial grammar learning in a balanced chunk strength design. J. Cogn. Neurosci. 16, 427–438. Madden, D.J., Whiting, W.L., Huettel, S.a., White, L.E., MacFall, J.R., Provenzale, J.M., 2004. Diffusion tensor imaging of adult age differences in cerebral white matter: relation to response time. NeuroImage 21, 1174–1181. Madden, D.J., Bennett, I.J., Song, A.W., 2009. Cerebral white matter integrity and cognitive aging: contributions from diffusion tensor imaging. Neuropsychol. Rev. 19, 415–435. Meinzer, M., Wilser, L., Flaisch, T., Eulitz, C., Rockstroh, B., Conway, T., Crosson, B., 2009. Neural signatures of semantic and phonemic fluency in young and old adults. J. Cogn. Neurosci. 21, 2007–2018. Meinzer, M., Harnish, S., Conway, T., Crosson, B., 2011. Recent developments in functional and structural imaging of aphasia recovery after stroke. Aphasiology 25, 271–290. Meinzer, M., Antonenko, D., Lindenberg, R., Hetzer, S., Ulm, L., Avirame, K., Flaisch, T., Flöel, A., 2012a. Electrical brain stimulation improves cognitive performance by modulating functional connectivity and task-specific activation. J. Neurosci. 32, 1859–1866. Meinzer, M., Seeds, L., Flaisch, T., Harnish, S., Cohen, M.L., McGregor, K., Conway, T., Benjamin, M., Crosson, B., 2012b. Impact of changed positive and negative taskrelated brain activity on word-retrieval in aging. Neurobiol. Aging 33, 656–669. Meulemans, T., Van der Linden, M., 1997. Does the artificial grammar learning paradigm involve the acquisition of complex information? Psychol. Belg. 37, 69–88. Murphy, K.J., Rich, J.B., Troyer, A.K., 2006. Verbal fluency patterns in amnestic mild cognitive impairment are characteristic of Alzheimer's type dementia. J. Int. Neuropsychol. Soc. 12, 570–574. Netz, J., Ziemann, U., Homberg, V., 1995. Hemispheric asymmetry of transcallosal inhibition in man. Exp. Brain Res. 104, 527–533. Nichols, T.E., Holmes, A.P., 2002. Nonparametric permutation tests for functional neuroimaging: a primer with examples. Hum. Brain Mapp. 15, 1–25.

Nucifora, P.G., Verma, R., Melhem, E.R., Gur, R.E., Gur, R.C., 2005. Leftward asymmetry in relative fiber density of the arcuate fasciculus. Neuroreport 16, 791–794. Obler, L.K., Rykhlevskaia, E., Schnyer, D., Clark-Cotton, M.R., Spiro, A., Hyun, J., Kim, D.-S., Goral, M., Albert, M.L., 2010. Bilateral brain regions associated with naming in older adults. Brain Lang. 113, 113–123. Oldfield, R., 1971. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9, 97–113. Park, D.C., Reuter-Lorenz, P.A., 2009. The adaptive brain: aging and neurocognitive scaffolding. Ann. Rev. Psychol. 60, 173–196. Peelle, J.E., Troiani, V., Wingfield, A., Grossman, M., 2010. Neural processing during older adults' comprehension of spoken sentences: age differences in resource allocation and connectivity. Cereb. Cortex 20, 773–782. Perruchet, P., Pacteau, C., 1990. Synthetic grammar learning: implicit rule abstraction of explicit fragmentary knowledge. J. Exp. Psychol. Gen. 199, 264–275. Petersson, K., Forkstam, C., Ingvar, M., 2004. Artificial syntactic violations activate Broca's region. Cogn. Sci. 28, 383–407. Poletiek, F.H., 2002. Implicit learning of a recursive rule in an artificial grammar. Acta Psychol. (Amst) 111, 323–335. Powell, H.W., Parker, G.J., Alexander, D.C., Symms, M.R., Boulby, P.A., WheelerKingshott, C.A., Barker, G.J., Noppeney, U., Koepp, M.J., Duncan, J.S., 2006. Hemispheric asymmetries in language-related pathways: a combined functional MRI and tractography study. NeuroImage 32, 388–399. Rajah, M.N., D'Esposito, M., 2005. Region-specific changes in prefrontal function with age: a review of PET and fMRI studies on working and episodic memory. Brain 128, 1964–1983. Reber, A.S., 1967. Implicit learning of artificial grammars. J. Verbal Learn. Verbal Behav. 6, 855–863. Schäfer, K., Blankenburg, F., Kupers, R., Gruner, J.M., Law, I., Lauritzen, M., Larsson, H.B., 2012. Negative BOLD signal changes in ipsilateral primary somatosensory cortex are associated with perfusion decreases and behavioral evidence for functional inhibition. NeuroImage 59, 3119–3127. Schlee, W., Leirer, V., Kolassa, I.T., Weisz, N., Elbert, T., 2012. Age-related changes in neural functional connectivity and its behavioral relevance. BMC Neurosci. 13, 16. Scholz, J., Klein, M.C., Behrens, T.E., Johansen-Berg, H., 2009. Training induces changes in white-matter architecture. Nat. Neurosci. 12, 1370–1371. Smith, S.M., Nichols, T.E., 2009. Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference. NeuroImage 44, 83–98. Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens, T.E.J., JohansenBerg, H., Bannister, P.R., De Luca, M., Drobnjak, I., Flitney, D.E., Niazy, R.K., Saunders, J., Vickers, J., Zhang, Y., De Stefano, N., Brady, J.M., Matthews, P.M., 2004. Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage 23 (Suppl. 1), S208–S219. Smith, S.M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T.E., Mackay, C.E., Watkins, K.E., Ciccarelli, O., Cader, M.Z., Matthews, P.M., Behrens, T.E.J., 2006. Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. NeuroImage 31, 1487–1505. Stamatakis, E.A., Shafto, M.a., Williams, G., Tam, P., Tyler, L.K., 2011. White matter changes and word finding failures with increasing age. PLoS One 6, e14496. Stephan, K.E., Penny, W.D., Moran, R.J., den Ouden, H.E., Daunizeau, J., Friston, K.J., 2010. Ten simple rules for dynamic causal modeling. NeuroImage 49, 3099–3109. Taler, V., Phillips, N.A., 2008. Language performance in Alzheimer's disease and mild cognitive impairment: a comparative review. J. Clin. Exp. Neuropsychol. 30, 501–556. Thiel, A., Schumacher, B., Wienhard, K., Gairing, S., Kracht, L.W., Wagner, R., Haupt, W.F., Heiss, W.D., 2006. Direct demonstration of transcallosal disinhibition in language networks. J. Cereb. Blood Flow Metab. 26, 1122–1127. Tyler, L.K., Shafto, M.A., Randall, B., Wright, P., Marslen-Wilson, W.D., Stamatakis, E.A., 2010. Preserving syntactic processing across the adult life span: the modulation of the frontotemporal language system in the context of age-related atrophy. Cereb. Cortex 20, 352–364. Uddén, J., Folia, V., Forkstam, C., Ingvar, M., Fernandez, G., Overeem, S., van Elswijk, G., Hagoort, P., Petersson, K.M., 2008. The inferior frontal cortex in artificial syntax processing: an rTMS study. Brain Res. 1224, 69–78. Wechsler, D., 1987. The Wechsler Memory Scale — Revised. Psychological Corporation, San Antonio. Wierenga, C.E., Benjamin, M., Gopinath, K., Perlstein, W.M., Leonard, C.M., Rothi, L.J., Conway, T., Cato, M.A., Briggs, R., Crosson, B., 2008. Age-related changes in word retrieval: role of bilateral frontal and subcortical networks. Neurobiol. Aging 29, 436–451. Wingfield, A., Grossman, M., 2006. Language and the aging brain: patterns of neural compensation revealed by functional brain imaging. J. Neurophysiol. 96, 2830–2839. Xiang, H.-D., Fonteijn, H.M., Norris, D.G., Hagoort, P., 2010. Topographical functional connectivity pattern in the perisylvian language networks. Cereb. Cortex 20, 549–560. Ystad, M., Eichele, T., Lundervold, A.J., Lundervold, A., 2010. Subcortical functional connectivity and verbal episodic memory in healthy elderly — a resting state fMRI study. NeuroImage 52, 379–388.