Binding during sequence learning does not alter cortical representations of individual actions

Abstract

37As a movement sequence is learned, serially ordered actions get bound together into sets in order 38 to reduce computational complexity during planning and execution. Here we examined how the 39 binding of serial actions alters the cortical representations of individual movements. Across five 40 weeks of practice, healthy human subjects learned either a complex 32-item sequence of finger 41 movements (Trained group, N=9) or randomly ordered actions (Control group, N=9). After five 42 weeks of training, responses during sequence production in the Trained group were correlated, 43 consistent with being bound together under a common command. These behavioral changes, 44 however, did not coincide with plasticity in the multivariate representations of individual finger 45 movements, assessed using fMRI, at any level of the cortical motor hierarchy. This suggests that 46

Consolidation alters motor sequence-specific distributed representations

“…Importantly, no significant changes in representational magnitude were found in the contralateral primary somatosensory cortex after consolidation. This is in line with the fact that M1 representational geometry has been shown to be strongly shaped by ecological finger co-activations (Ejaz et al, 2015), and to be resistant to extensive training of a sequence built on a new co-activation structure (Beukema et al, 2018). This negative result complements that of Figure 2, supporting that primary representation uncovered there might not reflect sequential features but the interaction of pre-existing somatotopic organization and single finger position in each sequences.…”

Finger representation in the cortex of the congenitally blind

“…This may explain why the anophthalmic individuals’ S1 hand representation is not different from that of sighted controls. As a measure strongly correlated to everyday hand use (Ejaz et al 2015), the canonical structure of hand representation has been found to be remarkably stable, despite hand loss (Kikkert et al 2016, Wesselink et al 2019), extensive behavioural training (Beukema et al 2019), or expert musicianship (Ogawa et al 2019). While average dissimilarity between fingers may vary, indicating altered signal strength, the representational structure is stable.…”

Neural representations underlying mental imagery as unveiled by representation similarity analysis

“…As a first step, separately for each run, we suppressed correlated noise across voxels by applying the multivariate noise normalization. We then computed the dissimilarity (d i,j ) between the activation patterns (u) of a pair of stimuli (i,j) using a leave-one-out cross validation scheme as follows (Beukema et al 2019 ): where M represents the independent partitions (cross-validation folds), and T the number of time points. Distances were computed in each pair of runs ( l , m ) and then averaged across each possible combination of runs.…”

“…Cross-validation also allows negative crossnobis distances, if the pattern of activity of each stimulus is not consistent and, consequently, a given area is not reliably able to encode differences between them. The meaningful zero point allows testing cross-validated Mahalanobis distances against zero to assess whether an area significantly discriminates between a pair of stimuli (i.e., the average distance will be significantly higher than zero) or not (Beukema et al 2019 ; Diedrichsen et al 2016 ; Diedrichsen and Kriegeskorte 2017 ; Walther et al 2016 ). Furthermore, it is also possible to compare distances between two or more pair of stimuli by means of two-sample or paired t -tests, to assess whether a distance is higher than others (Diedrichsen et al 2016 ; Walther et al 2016 ; Yokoi et al 2018 ).…”

“…For each question, we first performed one-sample t -tests to check whether average distances were higher than zero in the areas of HVC and in the HC, i.e., to test whether these regions were able to discriminate pairs of exemplars selected according to each experimental subquestion. When an area significantly encoded the dissimilarity in more than one category or domain, as a second step, we checked whether in that area there was a better discrimination in one category or domain over the others, by performing paired or two-sample t -tests (i.e., two categories) or ANOVA (more than two categories), according to our experimental hypotheses (for similar procedures, see Beukema et al 2019 ; Yokoi et al 2018 ).…”