Effortful verb retrieval from semantic memory drives beta suppression in mesial frontal regions involved in action initiation

Abstract

The contribution of the motor cortex to the semantic retrieval of verbs remains a subject of debate in neuroscience. Here, we examined whether additional engagement of the cortical motor system was required when access to verbs semantics was hindered during a verb generation task. We asked participants to produce verbs related to presented noun cues that were either strongly associated with a single verb to prompt fast and effortless verb retrieval, or were weakly associated with multiple verbs and more difficult to respond to. Using power suppression of magnetoencephalography beta oscillations (15-30 Hz) as an index of cortical activation, we performed a whole-brain analysis in order to identify the cortical regions sensitive to the difficulty of verb semantic retrieval. Highly reliable suppression of beta oscillations occurred 250 ms after the noun cue presentation and was sustained until the onset of verbal response. This was localized to multiple cortical regions, mainly in the temporal and frontal lobes of the left hemisphere. Crucially, the only cortical regions where beta suppression was sensitive to the task difficulty, were the higher order motor areas on the medial and lateral surfaces of the frontal lobe. Stronger activation of the premotor cortex and supplementary motor area accompanied the effortful verb retrieval and preceded the preparation of verbal responses for more than 500 ms, thus, overlapping with the time window of verb retrieval from semantic memory. Our results suggest that reactivation of verb-related motor plans in higher order motor circuitry promotes the semantic retrieval of target verbs.

Keywords: beta oscillations; embodied cognition; magnetoencephalography (MEG); motor system; supplementary motor area (SMA); verb generation.

Figure 1

(a) Verb generation task design. Subjects were required to name a verb in response to a noun cue, which had either a single strong association (SA) verb or many weakly associated (WA) ones. Time point 0 is the onset of the vocal response. The other time points represent the group means of time intervals between noun cue presentation and verb production under WA and SA conditions. The blue rectangle represents the time window of interest used in the subsequent analysis. (b) Statistical Parametric Mapping space‐frequency statistical maps show scalp topography and frequency range of magnetoencephalography power suppression (p < .05, family wise error corrected). A—anterior, P—posterior, L—left, and R—right parts of the sensor array. The dashed lines correspond to 15–30 Hz frequency range, which showed significant suppression against the baseline. (c) Reconstructed cortical sources of beta suppression (15–30 Hz) for SA and WA trials pooled together. The color bar represents the strength of beta suppression in arbitrary units. The colored areas indicate the cortical regions with significant differences from baseline (p < .05, Bonferroni corrected) [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 2

Differences in beta suppression between strong (SA) and weak association (WA) trials. (a) Sensor‐based Statistical Parametric Mapping space–time images show three statistical clusters demonstrating significantly stronger response‐related beta suppression in WA compared to SA trials (p < .05, family wise error corrected). A—anterior, P—posterior, L—left, and R—right parts of the sensor array. (b) Time courses of baseline‐normalized beta (β) power changes under SA and WA conditions calculated from the sensors nearest to the maxima of the two most significant clusters. The grand average time courses are aligned to the onset of the noun cue (left panel) and to the onset of the vocal response (right panel). The shaded areas denote the SE of the mean. Note. Stronger beta suppression corresponds to lower (more negative) values of beta power change. The light orange rectangle denotes the time period where beta suppression was significantly stronger under WA than under SA condition. (c) The individual peak values of beta suppression at the same sensors under SA and WA conditions. The thick black line represents the group medians (Wilcoxon matched pairs test, p < .05). (d) The reconstruction of cortical sources corresponding with the three sensor‐space spatio‐temporal clusters. The images were thresholded at p < .05 (uncorrected) [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 3

Response‐locked analysis of spectral power in strong and weak association trials. (a) Topographical maps at the left represent grand averages of baseline‐normalized spectral power within beta range (15–30 Hz) for both conditions (a,b) and their contrasts (c). A sensor showing maximal statistical difference between conditions is marked in white. Time‐frequency plots from this sensor are shown at the right panels. For the contrast plot, non‐significant time‐frequency tiles are masked (also note the change in amplitude). Time point 0 marks response onset. (d) The grand average root mean square (RMS) from the orthogonal channels of the accelerometer placed on the throat for both conditions. Paired t test at each time point within −1 to +0.25 s time interval of RMS time courses showed no significant differences between conditions (p < .05, uncorrected) [Color figure can be viewed at http://wileyonlinelibrary.com]