This is a great paper. I like the experiment, I like the data, and I like almost all of the conclusions.
The Experiment
TMS is applied to regions in the left prefrontal and left superior temporal areas determined by peak activations in a previous fMRI study using a speech listening task. Here's where they stimulated:
During TMS the subject is asked to discriminate voiceless stop consonants in single syllables presented in noise (i.e., the task is relatively hard), or in a control condition, discriminate colors, or in another control, discriminate tones. All tasks are matched for difficulty. I'm not a fan of discrimination tasks for assessing speech networks, and it appears that to get an effect, the authors had to make the task more difficult by adding noise, but let's not worry about this here.
Performance is assessed in baseline and under TMS.
The Data
Here's what they found:
TMS to premotor cortex disrupted speech discrimination relative to baseline. Stimulation to STG did not disrupt speech discrimination. TMS anywhere did not disrupt color discrimination.
Two things to note about these effects. 1. The effect is small. Performance went from 78.9% correct in baseline to 70.6% correct during TMS -- a drop of 8 percentage points. Not a whopping effect, but reliable, and who knows what kind of disruption TMS actually achieves... 2. The STG stimulation site (see fig above) looks pretty dorsal. I suspect they might have seen a mild effect if they had targeted STS.
The tone control was run on only a subset of subjects: Baseline accuracy: 85.5%. premotor stimulation: 80.7% (not significant). STG stimulation: 79.1% (significant decrement). Looks like the premotor stimulation just missed significance (no p-value reported). So I suspect that if they had run a full set of subjects, premotor stimulation would have interfered with tone discrimination as well, which is an interesting result.
The Conclusions
1. Conclusion #1 is that TMS disrupts speech discrimination. Yes, I agree. So do frontal lesions, so this isn't exactly news, but it's nice to know that a blob of fMRI activation is actually doing something.
2. The role of the premotor site in speech perception may be that "premotor cortex generates forward models... that are compared within the superior temporal cortex with the results from initial acoustic-speech analysis .... Premotor cortex provides top-down information that facilitates speech perception in circumstances such as when the acoustic signal is degraded...." p. 1694. I LOVE it! Yes, exactly! Premotor cortex is NOT the seat of speech perception, but can facilitate the work of systems in the superior temporal lobe.
3. Regarding the lack of an effect of STG stimulation on speech perception, they suggest that this may result from bilateral organization of perceptual speech processes. Bingo! (But it also could be that they missed the critical spot.)
4. "... sensory regions are not sufficient alone for human perception." p. 1695. Depends on what you mean by human perception. If "human perception" refers to those 8 percentage points during discrimination of voiceless stop consonants presented in noise, then yes, I agree.
5. "... the perceptual representations of speech sounds, and perhaps sensory stimuli more general, are fundamentally sensory motor in nature." Damn. And they were on such a roll! :-) I disagree with the "representations" part. Change that word to "processing" and I'm OK with the statement -- well, better change "are fundamentally" to "can be in part," while we're at it. There's no evidence that the representation of speech sounds have a motor component. There's only evidence that the representations can be mildly affected by motor factors. For now, until we have some evidence, let's leave speech sound representations in the temporal lobe.
No comments:
Post a Comment