EEG Theta Dynamics within Frontal and Parietal Cortices for Error Processing during Reaching Movements in a Prism Adaptation Study Altering Visuo-Motor Predictive Planning

Abstract

Modulation of frontal midline theta (fmθ) is observed during error commission, but little is known about the role of theta oscillations in correcting motor behaviours. We investigate EEG activity of healthy partipants executing a reaching task under variable degrees of prism-induced visuo-motor distortion and visual occlusion of the initial arm trajectory. This task introduces directional errors of different magnitudes. The discrepancy between predicted and actual movement directions (i.e. the error), at the time when visual feedback (hand appearance) became available, elicits a signal that triggers on-line movement correction. Analysis were performed on 25 EEG channels. For each participant, the median value of the angular error of all reaching trials was used to partition the EEG epochs into high- and low-error conditions. We computed event-related spectral perturbations (ERSP) time-locked either to visual feedback or to the onset of movement correction. ERSP time-locked to the onset of visual feedback showed that fmθ increased in the high- but not in the low-error condition with an approximate time lag of 200 ms. Moreover, when single epochs were sorted by the degree of motor error, fmθ started to increase when a certain level of error was exceeded and, then, scaled with error magnitude. When ERSP were time-locked to the onset of movement correction, the fmθ increase anticipated this event with an approximate time lead of 50 ms. During successive trials, an error reduction was observed which was associated with indices of adaptations (i.e., aftereffects) suggesting the need to explore if theta oscillations may facilitate learning. To our knowledge this is the first study where the EEG signal recorded during reaching movements was time-locked to the onset of the error visual feedback. This allowed us to conclude that theta oscillations putatively generated by anterior cingulate cortex activation are implicated in error processing in semi-naturalistic motor behaviours.

Fig 1. Experimental set up and protocol.

A. Experimental set up. For sake of simplicity only the target at 0° has been reported. B. Experimental protocol. Upper panel: series of reaching (black dots) and pointing (aftereffect, AF; red dots) trials performed by a representative participant during the entire experimental session. The y axis represents the angular error associated with each movement. Reaching and pointing trials were grouped into blocks delimited by dashed vertical lines. Each block contains 20 reaching and 12 pointing trials, i.e. 32 movements. Note that, for sake of simplicity, only the reaching trials are numbered on the x axis. Each block was executed under different lens conditions as indicated by labels above each block (Sham and Prisms conditions). Lower panel: schematic representation of the trial structure. In particualar, the x axis represents the temporal sequence of events (in seconds, s) within a trial, and the y axis the sequence of trials under different lens conditions (labels on the left represent the number of reaching trials; labels on the right indicate the different lens conditions under which the movements were executed). C. Trajectory of a typical reaching movement executed under visual distortion induced by 10° prism lenses (at the -20° target). The dashed black line represents the hand movement trajectory; the dotted blue line represents the projection of mask limits on the movement plane, as empirically determined by fitting a polynomial curve across the points (indicated by asterisks) where the participant starts to see the tip of his right finger. The onset of curvature was identified as the point on the trajectory path where the distance between the real hand trajectory and the ideal trajectory (dashed red line) reached its maximum (continuous red line). D. Profile of tangential velocity of the same movement illustrated in C. The red arrow indicate a clear indentation on the velocity profile due to a corrective sub-movement. In both bottom panels the coloured dots indicate the position (C) of the right finger tip as well as its tangential velocity (D) at the selected time points, as reported in panel D. The target position in C is indicated by the light blue dot.

Fig 2. Evolution of the angular error of reaching (A) and pointing (B) movements in subsequent trials performed during baseline and prism blocks.

A. Reaching error (light blue dots: mean deg ± SD) at the time when visual feedback appears. Each dot represents the mean value of all participants (n = 12). Each participant performed subsequent reaching movements organized in blocks of 20 movements for each of the following conditions: baseline (0°), 5°, 10°, and 15° prisms exposure. Blocks are indicated by the labels on the x axis of the bottom graph (B). Note the stability of the performance during baseline block, the dependence of the error magnitude on the degree of prism deviation (15° > 10° > 5°) at the early prism exposure, and the progressive error reduction with the repetition of reaching movements. B. Pointing error (aftereffect, AF, light blue squares: mean deg ± SD) measured as angular deviation from midline. Each square represent the mean of all participants (n = 12) obtained after having averaged four consecutive pointing trials for each participant. The four pointing trials were inserted, within each block, after the 4^th, the 12^th and the 20^th reaching movement, obtaining three groups of pointing trials per block (see labels on the x axis). Note the progressive growth of the angular deviation from midline with prisms exposure, which could be considered a mark of adaptation.

Fig 3. Spatio-temporal dynamics of theta oscillations at Fz channel site time-locked to hand visual feedback appearance in the low- and high-error conditions.

Time-frequency representation of the EEG spectral power time-locked to visual feedback appearance (Time = 0 s, black vertical line) in the high- (a) and low-error (b) conditions. Colour bar indicate power variations in dB. In (c) the results of the statistical comparison between the two conditions, (a) and (b), are reported. Significant time-frequency bins (p-values < 0.05) were coloured in dark red, while others were masked, i.e. replaced, with green. The red ellipse highlights significant values within the theta frequency band. In a’ and b’ significant spectral power changes as compared to baseline (Time<0 s) are represented for high- (a’) and low-error (b’) conditions, respectively. Note that the frequency window is restricted in order to highlight changes within the theta frequency band. Non-significant time-frequency power bins are masked with green. In d, the time-course of IAF-based theta power at Fz site (averaged over the whole participants population) in the high- (blue) and low-error (red) conditions is represented. The green box shows the time window where the two curves differed maximally.

Fig 4. Scaling of IAF-based theta power by the magnitude of motor error.

Event-related spectral perturbation (ERSP) in the theta band at Fz channel time-locked to visual feedback appearance (Time = 0 s). Single trials were sorted by the magnitude of error and partitioned in 20 quantiles of increasing errors. In the upper right plot, significant spectral power changes (p< 0.05) as compared to baseline (Time<0 s) are represented. Non-significant time-frequency points (p>0.05) are masked with light blue. x axis: times (s); y axis: error progression; z axis: theta power (dB). Colour bar indicate power variations in dB.

Fig 5. Temporal evolution of theta power scalp maps.

Maps were computed at selected times after visual feedback appearance (Time = 0 s) in the high- and low-error conditions (upper and middle rows, respectively). The corresponding statistical comparison between high- and low-error conditions is shown in the lower row: red dots indicate channel locations where the comparison reached statistical significance (paired t test with Bonferroni correction, p≤0.0125). Colour bar indicate power variations in dB.

Fig 6. Spatio-temporal dynamics of theta oscillations at Fz channel site time-locked to the onset of movement correction in the high-error condition.

A. Time-frequency representation of EEG spectral power time-locked to the onset of movement correction (Time = 0 s, black vertical line) in the high-error condition. B. The same time-frequency representation as in A, but focused on the theta range. Unlike A, only the spectral changes which appeared to be statistically significant as compared to baseline (i.e., Time < 0) were plotted, while non significant time frequency points were masked, i.e. replaced by green. Colour bar indicate power variations in dB.

Fig 7. Evolution of theta power in the high-error condition at Fz channel site with respect to the onset of movement correction.

A. Evolution of theta power (in dB) from –200 ms to 450 ms, with respect to the beginning of the corrective movement (Time = 0 s, dashed black vertical line). Data were obtained after slicing this time-range into 13 bins of 50 ms each. Bins to the left and to the right of the dashed line precede and follow Time 0, respectively. Error bars represent the standard error of mean. B. Colour matrix of the p-values obtained from the Generalized linear mixed model regression statistics. Columns and rows numbers of the matrix correspond to the time bins of panel A. Each pixel of the matrix reports the corresponding colour-coded p-value obtained from each bin-to-bin comparison (13 x 13). Six non overlapping ranges of p-values were colour coded as indicated on the right. Dashed white lines mark the time bins which precede and follow, respectively, the beginning of the corrective movement (Time = 0 s).