(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Transfer of motor learning is associated with patterns of activity in the default mode network [1]

['Ali Rezaei', 'Center For Neuroscience Studies', 'Queen S University', 'Kingston', 'Corson N. Areshenkoff', 'Department Of Psychology', 'Daniel J. Gale', 'Anouk J. De Brouwer', 'Joseph Y. Nashed', 'J. Randall Flanagan']

Date: 2025-08 

An often-desired feature of motor learning is that it generalizes to untrained scenarios. Yet, how this is supported by brain activity remains poorly understood. Here we show, using human functional MRI and a sensorimotor adaptation task involving the transfer of learning from the trained to untrained hand, that the transfer phase of adaptation re-instantiates a highly similar large-scale pattern of brain activity to that observed during initial adaptation. Notably, we find that these neural changes, rather than occurring at the level of sensorimotor regions, predominantly occur across distributed areas of higher-order transmodal cortex, specifically in regions of the default mode network (DMN). Moreover, we show that these learning-related neural changes relate to the structural properties of transmodal cortex (its myelin content and neurotransmitter receptor density), and that intersubject differences in DMN activity relate to both adaptation- and transfer-phase task performance. Together, these findings suggest that the transfer of learning across the hands is supported by the re-expression of the same activity patterns in the DMN as those that support initial learning. Collectively, these results offer a unique characterization of the whole-brain macroscale changes associated with sensorimotor learning and generalization, and establish a key role for higher-order brain areas, such as the DMN, in the transfer of learning to untrained scenarios.

By disentangling effector-specific versus effector-independent changes in manifold structure, we found that the connectivity of DMN areas, rather than sensorimotor regions, was predominantly modulated across both learning and transfer phases. Specifically, we found that the transfer phase of learning re-expressed a highly similar pattern of changes in DMN manifold structure as were observed during initial adaptation. These results suggest that the transfer of learning across the limbs is supported by the re-instantiation of the same activity patterns that support initial learning. Together, these findings offer new insights into the role of higher-order brain networks, such as the DMN, in real-world motor behavior, and contribute to a more comprehensive understanding of the neural substrates underlying the transfer of motor learning.

To explore this idea, we investigated changes in the manifold structure derived from FC patterns across cortical, subcortical, and cerebellar brain regions during sensorimotor adaptation and generalization. In our study, human participants learned to adapt their reaching movements during a visuomotor adaptation task performed with their right hand (learning phase), prior to performing the same task with their untrained left hand (transfer phase). To characterize learning- and transfer-related changes at the whole-brain level, using our manifold embedding approach, we examined the extent to which different brain regions shifted their connectivity profiles across different phases of sensorimotor adaptation and generalization.

Despite the recognized importance of explicit strategies in motor learning and generalization, the specific neural substrates supporting these processes are not fully understood. The default mode network (DMN)—a collection of interconnected brain areas within higher-order transmodal cortex—has been implicated in various explicit cognitive functions, including memory retrieval, planning, and decision-making [ 40 – 42 ]. While traditionally associated with internally focused tasks such as autobiographical memory and future thinking [ 40 , 43 – 45 ], emerging evidence suggests that the DMN may also play a role in tasks requiring the formation of strategies [ 46 – 48 ]. The DMN’s unique neuroanatomical and neurophysiological characteristics may enable it to support flexible strategy use: it occupies the top position within the brain’s cortical processing hierarchy, allowing it to integrate information across diverse cognitive and sensory systems [ 13 , 48 , 17 ]; its regions exhibit prolonged temporal integration windows, which may facilitate the maintenance and implementation of strategies over extended periods [ 17 , 49 – 51 ]; and its high receptor density and complex dendritic morphology suggest a heightened capacity for computational flexibility and learning [ 17 , 52 , 53 ]. Given that explicit cognitive strategies are crucial for the transfer of visuomotor adaptation, it is plausible that the DMN uniquely contributes to this process, perhaps reflected by specific changes in its manifold embedding.

One well-studied form of motor generalization in humans examines how visuomotor adaptation of reaching movements transfers across the hands. Such adaptation is known to involve at least two parallel processes: An automatic, implicit process that adapts gradually and a strategic, explicit process that adapts rapidly [ 27 – 30 ]. While the implicit process is thought to involve the recalibration of internal models within sensorimotor cortex and the cerebellum, the explicit process is associated with the conscious application of strategies and is hypothesized to engage higher-order cognitive regions [ 29 – 32 ]. Critically, behavioral studies have indicated that the successful transfer of visuomotor adaptation from the trained to the untrained hand is predominantly mediated by explicit strategy use rather than implicit recalibration [ 28 , 33 – 36 ]. This is noteworthy as recent computational work indicates that context—the specific set of environmental and internal cues associated with a given experience, which is often highly explicit in nature—plays a pivotal role in how the brain manages and generalizes learning [ 37 – 39 ].

Understanding these processes requires examining not just isolated brain regions, but how large-scale neural systems dynamically coordinate their activity. Traditional fMRI approaches often focus on localized activation changes or connectivity between specific pairs of regions. However, motor learning and generalization engage widely distributed circuits spanning cortical, subcortical, and cerebellar areas [ 1 , 7 – 10 ]. Investigating changes within such distributed systems benefits from methods that capture the holistic organization of brain-wide functional organization. To achieve this, we employed manifold learning, a powerful dimensionality reduction technique [ 11 – 13 ]. This approach, which builds upon key observations from both neuronal [ 11 , 14 – 16 ] and fMRI [ 17 – 21 ] studies, embeds the unique functional connectivity (FC) profile of each brain region (its pattern of covariance with all other regions) into a common low-dimensional space. The resulting ‘manifold’ provides a representation of the brain’s overarching functional geometry, revealing the principal axes along which functional network organization varies [ 13 , 21 ]. By analyzing how regions are positioned within this manifold across different task phases, we can track large-scale network reconfigurations—identifying shifts in functional integration and segregation—that underpin learning and its transfer [ 22 – 26 ].

The central nervous system’s ability to adapt our movements to new and changing environments is fundamental to daily life. This adaptability not only involves mastering movements within specific trained contexts but also extends to transferring learned skills to novel scenarios, a phenomenon known as generalization [ 1 , 2 ]. Understanding how our motor memories generalize to new situations, such as using one’s non-dominant hand for a well-practiced task, provides crucial insights into the neural bases that support behavior and has significant implications for patient treatment and rehabilitation [ 3 – 6 ]. Yet, the specific neural processes that underpin motor generalization remain unresolved.

Results

To study the neural processes that support sensorimotor adaptation and generalization, we had human participants (N = 38) perform a classic visuomotor rotation (VMR) task [54] in the MRI scanner. In this task, participants were required to move a cursor, representing their finger position, to hit a target that could appear in one of eight locations (along an invisible ring), using an MRI-compatible touchpad (Fig 1A). Participants began the study by performing two separate ‘baseline’ blocks (64 trials each; 6 s per trial)—the first with their non-dominant left hand (LH Baseline) and the second with their dominant right hand (RH baseline)—in which the motion of the cursor directly matched their finger movements. Following these baseline blocks, subjects performed a ‘learning’ block with their right hand (160 trials; RH Learning) in which the relationship between finger movement on the touchpad and movement of the cursor was rotated clockwise by 45°. Consequently, subjects had to learn to adapt their hand movements in a 45° counterclockwise direction to accurately hit the target. Following this RH learning block, participants performed 16 “report” trials, which allowed us to obtain a direct index of subjects’ explicit strategy in counteracting the VMR [28]. In these report trials, subjects were asked to indicate their aim direction using a dial on a joystick, controlled by their left hand, prior to executing a target-directed movement via their right hand. Finally, following these report trials, subjects completed the testing session by performing the exact same learning block (80 trials) but using their left hand (LH Transfer), allowing us to examine the intermanual transfer (i.e., generalization) of sensorimotor adaptation to the untrained hand.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Task structure and overview of fMRI analysis. (A) MRI setup (top) and task structure (bottom). (B) Average participant learning curves (median across 8-trial bins). Orange and green traces denote periods during the task in which the left hand (LH) and right hand (RH) were used, respectively. Banding around each trace denotes ±1 standard error of the mean (SEM). Inset bar graphs at right compare the initial error (first 16 trials) for RH learning and LH transfer trials. ** denotes p < 0.01 (C) Neural analysis approach. Right, for each participant, we extracted time series data from the Schaefer 400 cortical parcellation, Tian 32-region subcortical parcellation, and Nettekoven 32-region cerebellar parcellation for six equal-length task epochs (indicated by the coloured boxes) and then computed functional connectivity (FC) matrices for each epoch. We then estimated connectivity manifolds for each task epoch using PCA with centered and thresholded connectivity matrices (see Materials and methods). To visualize how the dominant pattern of FC changed across task phases before alignment to a common space, the surface brain maps depict the first Principal Component (PC1) loadings calculated from a representative subject’s covariance matrices for each specific task epoch (e.g., LH Baseline, RH Baseline, etc.). The color bar indicates the PC1 loading value for each brain region, reflecting the strength and direction of that region’s contribution to the epoch’s primary connectivity pattern. These maps serve an illustrative purpose and will differ slightly from the first PC of the common template Baseline manifold shown in Fig 2, which provides the reference space for all subsequent alignment and analyses. Left, construction of this template Baseline manifold. All manifolds were aligned using Procrustes transform to a common template manifold created from a group-averaged FC matrix based on the mean across the LH and RH Baseline epochs. This allowed us to assess learning-related changes in manifold structure from this Baseline architecture. (D and E) Subject-level clustering is abolished through a Riemannian centering approach. UMAP visualization of the similarity of FC matrices, both before centering (D) and after (E) centering. In these plots, each point represents a single FC matrix, color-coded either to subject identity (left panels) or task epoch (right panels), with its location in the two-dimensional space based on the similarity between matrices. Note that the uncentered connectivity matrices in D show a high degree of subject-level clustering, thus obscuring any differences in task structure. By contrast, the Riemannian manifold centering approach (in E) abolishes this subject-level clustering. The data and code needed to generate this figure can be found in https://zenodo.org/records/15648991. https://doi.org/10.1371/journal.pbio.3003268.g001

As shown in Fig 1B, we found that subjects, on average, were able to reduce their visuomotor errors during each of the learning and transfer periods. Moreover, consistent with prior work on the generalization of sensorimotor adaptation across the hands [28,34,55], we found that subjects on average exhibited lower initial errors during the transfer than learning phase (see right inset in Fig 1B). This result has been taken to indicate that the neural representation of learning is to some extent shared across the two hands (i.e., is effector-independent [27,56,57]).

Next, in order to explore the changes in brain activity that underpin sensorimotor adaptation and generalization, we examined task-related FC within six distinct, equal-length 32-trial epochs (i.e., 96 imaging volume windows) over the time course of the task. We defined the LH and RH Baseline epochs as the first four 8-trial bins (32 trials) of each functional scan, respectively, and defined ‘Early’ and ‘Late’ learning epochs as the first and last 32 trials of each of the RH Learning and LH Transfer functional scans, respectively. For each subject and each of these six epochs, we extracted region-wise mean blood oxygenation level-dependent (BOLD) time series data from the Schaefer 400 cortical parcellation [58], Tian 32-region subcortical atlas [59], and Nettekoven 32-region cerebellar atlas [60]. From this data, we then estimated covariance (FC) matrices for each epoch (LH Baseline, RH Baseline, RH Learning Early, RH Learning Late, LH Transfer Early, and LH Transfer Late; see bottom Fig 1C; [for a similar approach, see [22,24]).

Prior work [61,62], including our own [22,24,32,63], has shown that the majority of variance in patterns of whole-brain FC can be attributed to static, subject-level differences, and that these trait-like, subject-level features can mask any task-related (state-dependent) differences in brain activity. To mitigate this issue, using established procedures, we centered all subjects’ connectivity matrices using the Riemannian manifold approach [22,32,63,64] (see Materials and methods). To demonstrate the impact of this centering method and its importance for being able to extract any learning-related effects within our data, we projected all subjects’ FC matrices, both before and after our centering procedure, using the unsupervised nonlinear dimensionality-reduction technique uniform manifold approximation (UMAP; [65]). In this UMAP space, nearby points represent more similar patterns of whole-brain covariance compared to distant points. As shown in Fig 1D, prior to the centering procedure, nearly all the FC matrices cluster according to subject identity. This observation aligns with previous work using the Midnight Scan Club dataset [61] showing that subject-level structure can impede the detection of any task-related effects. However, after use of our centering procedure, this subject-level clustering is completely eliminated (Fig 1E), thus allowing us to explore modulations in connectivity patterns associated with the different task epochs.

Next, to investigate changes in whole-brain connectivity during the learning task, we used the centered matrices to estimate whole-brain (cortical, subcortical, and cerebellar) connectivity manifolds for each participants’ FC matrices from the six epochs [see also 22,24]. Following from prior work [12,66,67], each matrix was initially transformed into an affinity matrix by calculating the pairwise cosine similarity between regions after row-wise thresholding (see Materials and methods). Next, we applied dimensional reduction to the affinity matrices using principal component analysis (PCA) to obtain a set of principal components (PCs) that provide a low-dimensional representation (i.e., manifold) of cortical-subcortical-cerebellar connectivity structure for each epoch. The right panel in Fig 1C shows the spatial pattern of the first PC (PC1) loadings derived from the centered FC matrix of a representative subject for each task epoch. Each brain region’s color indicates the loading value (see Fig 1C caption), illustrating shifts in the primary axis of connectivity organization across the task phases. Next, to allow for direct comparison across epochs, we used Procrustes transformation to align each manifold onto a template Baseline manifold, constructed from the mean of all LH Baseline and RH Baseline connectivity matrices by first averaging within, and then across, all subjects (Fig 1C, left). This approach served two purposes: First, the common Baseline manifold served as a reference for manifold alignment [12], enabling direct comparisons among all participants within a single, unified task-based neural space. Second, it facilitated the identification of any learning- and transfer-related deviations from this baseline manifold architecture [see also 22,24].

Whole-brain manifold structure during Baseline trials The top three PCs that comprise the template Baseline manifold (Fig 2A) describe the overarching functional organization of whole-brain activity during Baseline trials with the left and right hand. PC1 mainly distinguishes cortical sensorimotor regions (positive loadings in red) from visual cortex and the cerebellum (negative loadings in blue; see also Fig 2C). Meanwhile, PC2 mainly separates visual, somatomotor and cerebellar regions (in red) from higher-order transmodal, DMN regions (in blue; see also Fig 2C). Finally, PC3 primarily distinguishes visual and posterior cingulate regions (in red) from the cerebellum and subcortex (in blue; see also Fig 2C). Collectively, these top three PCs explain the majority of the total variance (i.e., 50.90%) in the Baseline FC data (Fig 2B), and we retained these top three PCs for our further analyses. [However, note that the inclusion of PCs 4 and 5 in our analysis, which explained 9.68% and 5.87% of the variance, respectively, did not significantly impact any of our findings; see Fig A in S1 Text]. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Main connectivity axes extracted from Baseline trials. (A) Region loadings across cortex (top), subcortex (middle) and cerebellum (bottom) for the top three PCs. (B) Variance explained for the first 10 PCs. Black trace shows the cumulative variance explained across PCs, whereas the blue bars denote the variance explained for each individual PC. (C) The Baseline (template) manifold in low-dimensional space, with regions colored according to the Yeo and colleagues, 17-network assignment [58,70]. (D) Illustration of how eccentricity is computed. A single region’s eccentricity along the manifold is calculated as the Euclidean distance (dashed line) from manifold centroid (black square). The eccentricity of three example brain regions is highlighted. (E) Regional eccentricity during Baseline trials. Each brain region’s eccentricity is color-coded in the low-dimensional manifold space (left) and on the cortical, subcortical, and cerebellar surfaces (right). Black square denotes the center of the manifold (manifold centroid). L = left; R = right. https://doi.org/10.1371/journal.pbio.3003268.g002 We then aimed to characterize the embedding of each cortical, subcortical, and cerebellar brain region in this Baseline-derived manifold space, such that we could then assess subsequent shifts in the positioning of these regions during the learning and transfer phases of the task. To this end, following from previous methods [22–24], we computed the Euclidean distance of each region from the manifold origin (coordinates [0,0,0]; see Fig 2D). This measure, which we term ‘Manifold Eccentricity’, provides us with a multivariate index of each brain region’s relative positioning in the three-dimensional manifold space (Fig 2E). From this whole-brain manifold perspective, regions with higher eccentricity (e.g., located at the extremes of the manifold) can be interpreted as generally having greater functional segregation from other networks in the brain, whereas regions with lower eccentricity (e.g., situated near the manifold’s center) can be interpreted as generally having higher functional integration (lower segregation) with other networks [23,68,69]. Congruent with this interpretation, we found that our eccentricity measure was significantly correlated with several classic graph theoretical measures of integration and segregation (Fig B in S1 Text). Thus, changes in manifold eccentricity during sensorimotor adaptation and generalization provide insight into the functional segregation versus integration of different brain regions over time.

Global changes in manifold structure during learning and transfer We next sought to establish changes in manifold architecture across the different phases of the task. To this end, we first applied multivariate analytical methods to investigate, at the global whole-brain level, potential differences in manifold eccentricity across epochs. Fig 3A visualizes the mean (across subject) whole-brain eccentricity for each of the six task epochs using two-dimensional UMAP [71]. In this two-dimensional projection, nearby points represent epochs with more similar whole-brain eccentricity patterns. This visualization suggests that multiple types of task-related information might be encoded in the global manifold structure. For instance, UMAP dimension 1 suggests that relevant task epoch-related information is being encoded in the whole-brain eccentricity data, with LH and RH Baseline trials (black dot and red square) represented differently from the LH Transfer epochs (cyan and purple dots) and the RH Learning epochs (green and blue squares; read Dimension 1 from left to right). By contrast, UMAP dimension 2 also suggests that, independent of the hand being used, the early learning and early transfer epochs of the task (cyan and green dots) are represented differently from both the Baseline and Late Learning epochs (read Dimension 2 from top to bottom). Together, this suggests a multiplexing of both effector and effector-independent information in the manifold eccentricity data. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 3. Multiplexing of task-based information in global manifold architecture. (A) UMAP visualization of the similarity in neural states across task epochs (based on the data in B). In this plot, each point represents the across-subject mean whole-brain eccentricity for different task epochs (see legend at bottom), with nearby points representing more similar patterns of whole-brain eccentricity. Note that the two UMAP dimensions appear to capture different types of information about the task structure (see text). (B) Average across-subject representational similarity matrix (RSM) for the patterns of whole-brain eccentricity associated with the six task epochs (the value in each cell represents the Pearson spatial correlation). Note that the RSM is symmetric about the diagonal (black squares, which represent self-correlations). (C) Idealized model RSMs representing different hypotheses about task-related structure. (D) An idealized model RSM representing a control hypothesis based on the passage of time. In these model RSMs (C and D), values of 1 indicate predicted similarity between epochs according to the model, while 0 indicates predicted dissimilarity. These models are compared to the empirical data RSM shown in B. (E) Model comparisons. Each bar indicates the across-subject mean of cosine similarity between the RSM of the whole-brain eccentricity data (in B) and each of the models (in C and D; higher values mean higher similarity). Inference was performed via bootstrap resampling of subjects (1,000 bootstrap samples). Error bars indicate the standard error of the mean. The significance of each model for a one-sided comparison against zero is marked by white dew drops on the horizontal axis, and against the lower-bound estimate of the noise ceiling (gray icicles; FDR-corrected for 4 models). Models’ Pairwise differences are summarized by stars. ** denotes q < 0.05. The data and code needed to generate this figure can be found in https://zenodo.org/records/15648991. https://doi.org/10.1371/journal.pbio.3003268.g003 To complement these UMAP visualizations and quantitatively test specific hypotheses about the information represented, we performed Representational Similarity Analysis (RSA; [72,73]) on the eccentricity data. Using established methods [74] we constructed, for each subject and epoch, a representational similarity matrix (RSM) by calculating the Pearson correlation between the vectors of regional eccentricity values (464 regions) for every pair of the six task epochs (Fig 3B shows the group-averaged RSM). This RSM captures the pairwise similarity between the neural states associated with each epoch based on their whole-brain eccentricity patterns. We next tested how well this empirical RSM structure (Fig 3B) was explained by three different a priori theoretical models representing distinct hypotheses about task structure (Fig 3C and 3D): The first model predicted the encoding of hand information only (‘Hand’ model, Fig 3C, left), the second predicted the encoding of the task epoch only (Baseline, Early or Late; ‘Epoch’ model, Fig 3C, middle), and the third predicted the encoding of the Baseline versus Learning periods only (‘Learning model’; Fig 3C, right). These binary model RSMs represent theoretical similarity structures (where 1 denotes predicted similarity, 0 predicted dissimilarity) corresponding to each hypothesis. We compared each participant’s empirical RSM to these theoretical model RSMs using cosine similarity, with higher values indicating a better fit between the model and the eccentricity data. The group-averaged results are shown in Fig 3E. Statistical inference via bootstrapping revealed that both the ‘Hand’ and the ‘Learning’ model performed better than the ‘Epoch’ model, with the ‘Learning’ model performing best (though not statistically better than the Hand model; see Fig 3E). Note that one potential explanation of these above results is that they could simply reflect the passage of time; i.e., the learning/transfer epochs occur later in time than the baseline epochs, and thus any learning/transfer-related structure in the neural data could simply reflect subjects’ waning attention or fatigue during MRI testing. To explore this possibility, we tested a fourth model that merely encoded the passage of time (‘Time’ model in Fig 3D), in which time-adjacent epochs were coded as being the most similar. Note that this RSM model had the lowest performance of all four models, and had significantly less explanatory power than both the ‘Learning’ and ‘Hand’ models (Fig 3E). Taken together, these RSAs further support the notion that both effector and learning/transfer-related information are present in the whole-brain manifold data. This prompted us to next consider what specific sets of brain regions are responsible for these different representational structures in the data.

Meta-analysis of regional changes in manifold embedding To probe the putative behavioral-cognitive processes supported by the distinct patterns of brain activity identified above, we performed a meta-analysis on the regional eccentricity maps in Figs 4A and 5A using the Neurosynth database [76] and NiMARE library [75]. This meta-analytic decoding approach allows us to interpret the functional significance of our neural findings by linking the observed patterns of brain activity to cognitive processes identified in the broader neuroimaging literature. By doing so, we can gain insights into the potential behavioral and cognitive functions associated with the regions showing significant changes in eccentricity, thereby enhancing our understanding of the underlying processes supporting motor learning and generalization. The results of this meta-analytic decoding analysis are depicted in the form of word clouds in Figs 4D and 5D. In these visualizations, keywords/descriptors associated with whole-brain patterns of brain activity from other fMRI studies that closely resemble our own brain maps receive a higher weighting (and thus a larger font size), with the polarity (blueness or redness of the text) describing the direction of this resemblance. For instance, from the fMRI literature we find that our main effect of Hand F-stat brain map (in Fig 4A) is positively correlated with functional terms like ‘action’, ‘hand’ and ‘motor imagery’, and negatively correlated with terms like ‘social’, ‘mental states’ and ‘autobiographical’ (Fig 4D). This result validates our findings by confirming that regions associated with hand movements and motor control are indeed implicated when participants use different effectors. By contrast, for our main effect of Task Epoch F-stat map (in Fig 5A), we find that this brain map is positively correlated with terms like ‘social’ and ‘mental states’, as well as several other higher-order cognitive processes (e.g., ‘theory of mind’); whereas it is negatively correlated with terms like ‘movements’, ‘motor imagery’ and ‘action’ (Fig 5D). This suggests that the regions involved in learning and generalization are linked to higher-order cognitive functions rather than traditional motor functions. At first glance, this outcome may seem counterintuitive given that the Task Epoch F-stat map describes neural changes associated with motor learning and generalization. However, it is generally consistent with contemporary work positing that explicit cognitive processes, which underpin many abstract higher-level functions (e.g., theory of mind), also play a critical role in motor learning and generalization [30,34,79]. Thus, these analyses not only corroborate our findings by associating motor-related brain maps with motor functions but also reveal that the regions involved in learning and generalization are linked to higher-order cognitive processes.

Evidence that the manifold patterns during early learning are re-expressed during transfer Accompanying the brain map results in Figs 4E and 5E are scatterplots (rightmost panels) that denote the eccentricity changes of each significant brain region across epochs for each pairwise contrast. In the case of the brain regions showing a main effect of Hand (shown in Fig 4A), these eccentricity plots clearly delineate the strong contralateral effects described above (and shown in Fig 4E); i.e., the eccentricity of left hemisphere regions is higher, on average, for RH trials, and vice versa for the right hemisphere regions on LH trials. However, in the case of the brain regions showing a main effect of Task Epoch (shown in Fig 5A), there are two important observations to be made. Firstly, unlike the contralateral hand effects (in Fig 4E), the eccentricity changes in both hemispheres are highly similar (compare left and right hemisphere eccentricity plots in Fig 5E). This indicates that the connectivity changes occurring across the baseline, early, and late learning/transfer epochs are largely bilateral in nature, and mainly localized to higher-order, non-motor regions of cortex. Secondly, there is a particularly large increase in the manifold eccentricity of these regions during the first early learning period, which is followed by a second (more moderate) increase in eccentricity during the early transfer period (see Fig 5E). This second increase in eccentricity, along with the resulting main effect, suggests that the manifold structure observed during initial RH learning may be selectively ‘re-expressed’ during initial LH transfer. To explore this idea directly, moving beyond simple magnitude comparisons, we used an RSA approach to test whether the spatial pattern of regional eccentricity was similar between the two epochs. Specifically, we tested if the spatial correlation between the pattern of regional eccentricity during the RH Learning Early and the LH Transfer Early epochs is significantly higher than the correlation between the RH Learning Early epoch and the other learning epochs that bookend the LH Transfer Early epoch (i.e., RH Learning Late and LH Transfer Late epochs). This RSA analysis employs a logic similar to that used to measure cortical reinstatement in episodic memory research, where neural activity patterns during encoding and retrieval are directly compared [80–89]. When we performed this analysis, we found that the correlation between the RH Learning Early epoch and each of the three subsequent epochs significantly differed (rmANOVA, F(2,74) = 3.14, p = 0.049). Follow-up paired-samples t tests revealed that the spatial correlation between the RH Learning Early epoch and the LH Transfer Early epoch was indeed significantly higher than compared to the RH Learning Early epochs’ correlation with each of RH Learning Late (one-tailed t test, t(37) =1.87, p = 0.035) and LH late learning (one-tailed t test, t(37) = 2.58, p = 0.007; Fig 6B) epochs. Together, these findings provide compelling evidence that the manifold structure established during RH Learning Early is most strongly re-expressed during the early phase of transfer to the untrained left hand. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 6. Re-expression of early learning manifold structure across epochs. (A) Mean eccentricity values for the RH Learning Early epoch for brain regions exhibiting a significant main effect of Task Epoch (i.e., areas from Fig 5A). (B) Pattern similarity (Pearson r) between RH Learning Early and the three subsequent task epochs—RH Learning Late, LH Transfer Early, and LH Transfer Late—computed across the regions shown in A. The black line shows the across-subject mean, and individual points represent single subjects. (C) Mean eccentricity values for the RH Learning Early epoch for brain regions exhibiting a significant main effect of Hand (i.e., areas from Fig 4A). (D) Pattern similarity between RH Learning Early and the same three comparison epochs as in (A), but computed across the Hand-significant regions shown in C. * denotes p < 0.05 for one-tailed paired-samples t tests. The data and code needed to generate this figure can be found in https://zenodo.org/records/15648991. https://doi.org/10.1371/journal.pbio.3003268.g006 Note that one alternative explanation of these above results could be that these ‘re-expression’ effects are not specific to higher-order cortical areas per se, but rather reflect a brain-wide phenomenon. To explore this possibility, we performed the exact same RSA on brain regions exhibiting a main effect of Hand (in Fig 4A). We reasoned that these brain areas, primarily localized to sensorimotor cortex and which predominantly showed contralateral effects (Fig 4E), should demonstrate hand-specific patterns distinct from the re-expression effect observed in the higher-order regions, such as the DMN. Specifically, we would predict that the correlation between the pattern of regional eccentricity during the RH Learning Early and RH Learning Late epochs should be significantly higher than the correlation between the former condition and the two LH transfer epochs (early and late). When we performed this RSA on the brain regions exhibiting a main effect of Hand, we again found that the correlation between the RH Learning Early epoch and each of the three subsequent epochs significantly differed (rmANOVA, F(2,74) = 5.08, p = 0.008). Consistent with our predictions, follow-up paired-samples t tests revealed that the highest correlation with the RH Learning Early epoch was for the RH Learning Late epoch, and that the correlations were subsequently lower for both the LH early (one-tailed t test, t(37) = −1.84, p = 0.037) and late (one-tailed t test, t(37) = −3.10, p = 0.002) transfer epochs (see Fig 6D). This contrasting hand-specific pattern in sensorimotor regions serves as a valuable control, as it further highlights the selective nature of the re-expression effects observed in the higher-order cortical areas, such as the DMN (Fig 6B).

Changes in regional manifold embedding relate to underlying brain structure A unique structural property of DMN brain areas is that, compared to other regions (e.g., sensory-motor cortex), they tend to exhibit lower myelin content. As a consequence, they also show greater complexity in their patterns of dendritic branching and arborization [90–92]. These structural characteristics have been hypothesized to facilitate the integrative capacity of brain regions and their potential for both learning [93,94] and contextual-processing [95,96]. Similarly, DMN areas have been shown to exhibit a higher density of neurotransmitter receptors (e.g., serotonin, dopamine, etc.) as compared to several other cortical areas [52,53], which is hypothesized to increase their capacity for cognitive flexibility. Given these neuroanatomical observations, we naturally wondered whether our Hand and Task Epoch effects might relate to differences in the distribution of both myelin content and receptor density across cortex. To examine this, we built null models through spin permutation testing procedures [97,98] that account for the spatial autocorrelation in our two F-stat brain maps (Hand and Task Epoch) and tested for their spatial correlation with (1) the Human Connectome Project 1200 (HPC) myelin map (T1w/T2w ratio) [98] and (2) the first principal gradient (PC1) of receptor density from Hansen and colleagues [52], which captures total neurotransmitter receptor density per brain area. This permutation ‘spin-testing’ approach revealed that brain areas selectively modulated by the effector (Hand) were associated with higher myelin content (r = 0.24, P spin = 0.013) and lower receptor density (r = −0.14, P spin = 0.048), whereas brain areas selectively modulated by the learning/transfer phases of the task tended were associated with lower myelin content (r = −0.27, P spin = 0.039) and higher receptor density (r = 0.20, P spin = 0.024; see Fig 7). These results imply that differences in the localization of our two main effects may result, in part, from the underlying structural features (i.e., myelination and receptor density) of those different brain areas. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 7. Effector-related and learning/transfer-related main-effects relate to underlying brain structure. Spatial correlations between our two main-effect F-stat maps (shown at top) with (A) cortical myelin concentration (T1w/T2w ratio) and (B) the first principal component of receptor density (B, from [52]). Spatial null model testing was performed using the Neuromaps toolbox [98,97]. L = left; R = right. https://doi.org/10.1371/journal.pbio.3003268.g007

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3003268

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/