Distinct patterns of brain activity characterise lexical activation and competition in spoken word production

Abstract

According to a prominent theory of language production, concepts activate multiple associated words in memory, which enter into competition for selection. However, only a few electrophysiological studies have identified brain responses reflecting competition. Here, we report a magnetoencephalography study in which the activation of competing words was manipulated by presenting pictures (e.g., dog) with distractor words. The distractor and picture name were semantically related (cat), unrelated (pin), or identical (dog). Related distractors are stronger competitors to the picture name because they receive additional activation from the picture relative to other distractors. Picture naming times were longer with related than unrelated and identical distractors. Phase-locked and non-phase-locked activity were distinct but temporally related. Phase-locked activity in left temporal cortex, peaking at 400 ms, was larger on unrelated than related and identical trials, suggesting differential activation of alternative words by the picture-word stimuli. Non-phase-locked activity between roughly 350-650 ms (4-10 Hz) in left superior frontal gyrus was larger on related than unrelated and identical trials, suggesting differential resolution of the competition among the alternatives, as reflected in the naming times. These findings characterise distinct patterns of activity associated with lexical activation and competition, supporting the theory that words are selected by competition.

Figure 1. Induced brain responses.

A. The panels in the right column show the stimulus-locked time-frequency representations of relative power change for Stroop-like (related vs. identity, upper right) and semantic (related vs. unrelated, lower right) effects, averaged over the sensors highlighted in white in the corresponding topographic maps to the left. Dashed lines indicate the clusters. B. The middle panel shows the estimated sources in the whole-brain analysis for the Stroop-like (upper) and semantic (lower) effects. The left and right panels show the time-frequency representation of the activity in the estimated sources. Dashed rectangles enclose the spectrotemporal cluster of interest (4–8 Hz, 350–650 ms). In this cluster, relative power increase was observed for the Stroop-like effect in the left superior frontal gyrus (upper right panel) and in the left postcentral gyrus (upper left panel). Relative power increase was observed for the semantic effect in the left superior frontal gyrus (lower right panel), but not in the left postcentral gyrus (lower left panel).

Figure 2. Induced brain responses time-locked to the onset of the naming responses.

The right-hand panels show the response-locked time-frequency representations of relative power change for Stroop-like (related vs. identical, upper right) and semantic (related vs. unrelated, lower right) effects, averaged over the sensors highlighted in white in the corresponding topographic maps. Dashed lines indicate the clusters.

Figure 3. Evoked brain responses.

A. Event-related fields (combined planar gradient) for the distractor types, averaged over the left temporal sensors highlighted in the layout in the middle. The grey area indicates the window tested for statistical significance. The Stroop-like effect (related vs. identity) was characterised by an amplitude increase in left temporal channels, as shown in the scalp topography to the right (upper topography), between 375–430 ms. The semantic effect (related vs. unrelated) was characterised by amplitude decrease in left temporal channels, as shown to the right (lower topography), between 375–400 ms. The scalp topographies show the difference between conditions averaged in the time window of the corresponding significant temporal cluster (shown below each topographical map) with the sensors participating in the cluster highlighted in white. B. Estimated sources of the Stroop-like (upper) and semantic (lower) effects in the whole-brain analysis in the time window of the corresponding significant temporal cluster (shown to the left of each source map). The difference t –value maps were thresholded at ±2.16 (13 degrees of freedom, alpha = .05). C. Activity from the left temporal cortex (averaged over the estimated sources in B) for the distractor types.

Figure 4. Observed results and WEAVER++ simulations.

A. Differences in picture-naming times as empirically observed and from the simulations for the related condition (black bar) and identity condition (white bar) relative to the unrelated condition. B. Differences in signal amplitude of left temporal cortex activity for the related condition (black bar) and identity condition (white bar) relative to the unrelated condition and corresponding priming effects in the simulations. RT =  response time; unr =  unrelated; rel =  related; iden =  identity.