Comparing cortical activations for silent and overt speech using event-related fMRI

Abstract

At present, functional magnetic resonance imaging (fMRI) of cortical language functions favors "silent" task paradigms with no overt speaking, due to severe motion artifacts in MR images induced by vocalization. To the extent that the neural substrate of silent speaking might differ from that of overt speaking, this is a problem for understanding spoken language. The present study combined event related fMRI methodology with a set of techniques for motion reduction, detection, and correction to further investigate overt speech and compare it to silent speech. The purpose of the study was two-fold. We aimed to test a multiple-step image processing protocol involving discrimination and separation of motion-induced signals from activation-induced signals and we aimed to use this multi-step image processing protocol to compare the similarity of activation of cortical pathways potentially relevant to language production during silent and overt speech, focusing on Broca area and primary motor cortex as test cases. If the problem of motion artifact can be handled effectively, fMRI can add greatly to the tools available to investigate human language.

Copyright 2001 Wiley-Liss, Inc.

Figures

Figure 1

Topographically illustrates the three regions of interest, the Broca area (BA), the “mouth, lips, and tongue” region of the primary motor cortex (MLT‐PMC), and the “inferior vocalization” region of the primary motor cortex (IV‐PMC). CS, central sulcus; Pre‐CS, precentral sulcus; SF, Sylvian fissure.

Figure 2

The time course of signal changes in a voxel of the primary motor cortex of a subject during 12 trials of speaking a letter overtly is plotted in panel a (solid line). For comparison, the time course of signal changes obtained in a voxel near the mouth during the same task in the same subject is also plotted (dashed line). Note that the BOLD signal changes occur 4–6 sec after the signal changes induced by articulatory motion. Also, the magnitude of the BOLD signal changes at peak response is approximately 1.5%, whereas the motion‐induced signal changes vary from 30–70%. Within the first 6 sec of each trial, there was no spike‐like signal change in the time course of the BOLD responses. In some of trials, e.g., Trials 5, 7, 8, and 9, there was a noticeable shoulder‐like signal increase (see arrows). The average time courses of the BOLD responses and articulatory motion averaging over the 12 repeated trials (a) are plotted in (b). Again, the peak of the BOLD response occurs 6 sec after the peak of signal changes induced by articulatory motion. Error bars = SEM.

Figure 3

The average time courses of signal intensity changes over subjects in the left and right of the “mouth, lips, and tongue” region of the primary motor cortex obtained from the ROI analysis during speaking a letter name silently and overtly (a), and during generating an animal name silently and overtly (b). The ROI masks were delineated from the most activated clusters in the overt conditions. L, left; R, right.

Figure 4

The average time courses of signal intensity changes over subjects in the left and right of the “inferior vocalization” region of the primary motor cortex obtained from the ROI analysis during speaking a letter name silently and overtly (a), and during generating an animal name silently and overtly (b). Note that the magnitudes of the BOLD responses were approximately 50% less than ones observed in the “mouth, lips, and tongue” region of the primary motor cortex during the same tasks (Fig. 3). The ROI masks were delineated from the most activated clusters in the overt conditions. L, left; R, right.

Figure 5

The average time courses of the signal intensity changes over subjects in Broca area and its right opposite homologue obtained from the ROI analysis during speaking a letter name silently and overtly (a), and during generating an animal name silently and overtly (b). Overtly speaking a letter name increased activity in Broca area but decreased activity in the right homologous region, compared to silently speaking a letter name (a). Conversely, overtly generating an animal name decreased activity in both Broca area and its right homologue (b). Note that the magnitude of the BOLD signal changes was small and the line‐width was narrow during overt animal‐name generation, compared to during silent animal‐name generation (b). The dip within the first 2 sec of the trial was artificial and possibly caused by articulatory motion. The ROI masks were generated from the most activated clusters in the silent conditions. L, left; R, right.

Figure 6

(a) Typical activation patterns obtained during silently speaking a letter name (two left panels), and overtly speaking a letter name (two right panels). Broca area was activated in both silent and overt condition, but with a large extent in the overt condition (right, top panel). The middle‐inferior potion of the primary motor cortex (PMC) was activated bilaterally during overtly speaking a letter name, but not during silently speaking a letter name. (b) Typical activation patterns observed during silently generating an animal name (two left panels) and overtly generating an animal name (two right panels). Note that Broca area was not activated during overt animal‐name generation (right, top panel), but activated to a large extent during silent animal‐name generation (left, top panel). The “mouth, lips, and tongue” region of the primary motor cortex was activated bilaterally during overt speech, but not during silent speech. In addition, overt speech activated the left most inferior portion of the primary motor cortex, and the activated cluster was disconnected from one in the middle portion of the primary motor cortex (right, top panel). MLT‐PMC, the “mouth, lips, and tongue” region of the primary motor cortex; IV‐PMC, the “inferior vocalization” region of the primary motor cortex; CS, central sulcus; BA, Broca area; RH, right homologue of Broca area.