With our introduction to paired pulse transcranial magnetic stimulation (ppTMS) out of the way, we now turn to a 2010 PNAS paper by Neubert, Mars, Buch, Olivier & Rushworth in which conditioning TMS is applied to the right inferior frontal gyrus (rIFG) as well as the pre-supplementary motor area (pre-SMA) between 3 and 18ms prior to TMS over the contralateral primary motor cortex (M1). Note that all of these areas are known to interact in exceedingly complex ways based on the task subjects are performing, the intensity of the magnetic stimulation, the relative intensities of the paired pulses, and the subjects’ proximity to the initiation of movement at the time of conditioning TMS (all described yesterday).
Skip the blockquotes unless you want to know about the detailed methods & results (they are actually quite interesting). The “cognitive-level” take-home is below the block quote.
Neubert et al asked subject to press a button with either the right or left index finger depending on whether a central stimulus matched the color of stimulus presented to the left or right of it. Responses were always “congruent”: if red was on the left and green was on the right, and the center stimulus was red, subjects would respond with their left index finger. Every trial began with a fixation stimulus (for 1s), followed by the flankers (for 450-600ms); the center stimulus was then presented until subjects’ responded. While the flankers tended to change from trial to trial (always red and green, but variously assigned to left or right of the central stimulus), the central stimulus tended to stay the same across multiple trials. Only infrequently would it change color (every 3 to 7 trials).
This is clever – it means that subjects will likely process the identity of the flankers and prepare a button press by assuming that the central stimulus will be the same color as it was previously. Most of the time, this assumption is correct (the current trial ends up being what we’ll call a “Stay” trial) but sometimes the assumption is wrong, and subjects must overcome their assumption to respond according to the other color (i.e., a “Switch” trial), potentially leading to “action reprogramming.”
To investigate the neural substrates of such “action reprogramming,” a conditioning TMS pulse (the “CS”) was applied to either rIFG or pre-SMA with a variable stimulus onset-asynchrony (SOA; 75, 125 or 175ms) after the onset of the flankers. 8ms later, the test pulse was applied to left M1 (the “TS”), and motor evoked potentials were measured from the right first dorsal interosseus (a muscle responsible for moving the index finger). The strength of the pulses was 110% and 100% of the resting motor threshold for the CS and TS respectively. The idea here is that action programming and reprogramming is required on Stay and Switch trials, respectively, and so by stimulating rIFG, we should be able to “boost” those processes. In other words, TMS to rIFG at the right time might make that programming/reprogramming all the more visible in motor-evoked potentials.
But for a moment, let’s put these theoretical predictions aside. Based purely on previous research and without invoking complex concepts like action reprogramming, what do we expect to happen in this task, given yesterday’s crash course in ppTMS? Well, first, subjects were substantially slower on Switch than Stay trials (416ms vs. 302ms respectively), meaning that differences in motor preparation might be important here, as they are in all other ppTMS studies. Second, because the authors used a suprathreshold CS, we know to expect MEP facilitation as a result of pre-SMA stimulation (absent any previous evidence that pre-SMA stimulation is dependent on motor preparation). Based on previous information about the result of motor preparation on suprathreshold CS’s to rIFG, we know to expect MEP suppression as a result of rIFG stimulation if subjects are farther from the initiation of movement (as in resting state tasks discussed yesterday and perhaps “Switch trials” in this study), but enhanced MEPs resulting from rIFG stimulation if subjects are closer to the initiation of movement (e.g., Stay trials in this study).
And what we can predict is precisely the pattern the authors observed, except it was only observed at certain SOAs: the pre-SMA CS had the expected effect only at the 125ms SOA, whereas the rIFG CS had the expected effect only at the 175ms SOA. Moreover, the authors observed a reduction in the MEP of the right hand regardless of whether that hand was being switched away from, or switched to. Thus, this effect is unlikely to be reflective of some selectively controlled “reprogramming” mechanism – i.e., to inhibit the MEP of the not-to-be-used hand.
They next manipulated the interval between CS and TS to determine how that parameter, well known to affect ppTMS at other sites, might influence the results observed at the 175ms SOA for rIFG CS, and at the 125ms SOA for pre-SMA CS. The results demonstrated a robust facilitation of MEP with pre-SMA stimulation at intervals from 6-12ms, and suppression of MEP with rIFG stimulation at intervals of 6 and 12ms, but not the intermediate 3, 9 or 12ms intervals. Nonetheless, for both rIFG and pre-SMA, the effects at earlier intervals were correlated with each other (3ms & 6ms), as were the effects at later intervals (9-18ms), but these two “clusters” were not correlated with one another – suggesting that at least two distinct mechanisms are being engaged at these differing intervals, regardless of site.
Because the authors also collected diffusion tensor imaging from the same subjects, they could correlate the effects of rIFG & pre-SMA stimulation with white matter integrity (assessed via dtMRI). MEP suppression following rIFG stimulation at 6ms was related to a small bit of white matter between rIFG and M1, whereas the effect at 12ms was related to a small bit of subcortical white matter (albeit both effects were significant only with a very lenient statistical threshold: one-tailed p<.001, with a 10-voxel extent threshold). Additional analyses revealed, as one might expect, that the little subcortical area was significantly more interconnected with white matter in other nearby subcortical areas than was the white matter between M1 and rIFG. A final experiment using repetitive TMS (1Hz for 15 minutes) to disrupt pre-SMA functioning (as compared to a control condition involving parietal TMS) revealed that functional BOLD connectivity between rIFG and M1 was dependent on intact pre-SMA function. In particular, only without rTMS to pre-SMA was functional connectivity between rIFG and M1 greater on "Switch" than "Stay" trials.
What does this extremely rigorous and careful study really tell us?
Consistent with previous work using ppTMS, rIFG stimulation has effects on the motor-evoked potentials (MEPs) elicited by TMS to primary motor cortex (M1). It’s not clear that there’s anything special about rIFG in this regard: the same effects have been observed with conditioning stimulation to the dorsal premotor area (PMd; as discussed yesterday).
Also consistent with previous work, we can see that motor preparation makes a difference. In particular, subjects were likely closer to response execution in “Stay” than “Switch” trials, and thus an expected pattern of facilitation/suppression (respectively) on MEPs was observed. Unfortunately, we don’t know whether the apparently distinct mechanisms operating at intervals of 3-6ms vs. 8-12ms in this paradigm reflect different combinations of short & long-interval intracortical inhibition & facilitation or interhemispheric inhibition & facilitation, or only some of those mechanisms.
We did learn that the suppression of MEP’s resulting from rIFG stimulation is relatively non-selective – that is, it was observed both if the right hand that was being switched away from or switched to. This suggests that it is something universal to “Switch” trials that is yielding the observed effect, and not something specific to the demand to “reprogram” a particular response. The obvious candidate here is that there are simply different reaction times on the two trial types. This is perhaps confirmed by alternative analyses showing no differential ppTMS effect across trial types that might be thought to differentially require reprogramming (“motor stay” vs. “motor switch”) but didn’t show differential reaction times.
We also learned that the asynchrony between the onset of the flankers and the onset of the conditioned stimulus matters. This is fairly interesting as it has been explored only in a few prior ppTMS studies. In particular, rIFG and pre-SMA stimulation is basically without effect unless delivered sufficiently long after visual stimuli are processed, suggesting a more high-level role for the rIFG and pre-SMA than purely perceptual processing.
But the major interpretational problem here is related to all the caveats with ppTMS that I mentioned yesterday and have continued to emphasize today. We just have no real reason to believe that the effects of a conditioning pulse to rIFG is inducing it’s normal function as opposed to simply disrupting it – the latter is a far more traditional assumption in TMS studies outside the motor cortex. In fact, a quite coherent story can be told if the conditioning pulse to rIFG is thought to impair normal rIFG function (I’ll leave that as an exercise to the reader).
I’ll end with this: it’s clear from decades of research that induction of function is the wrong way to interpret the effects of neurostimulation, at least with regard to cortical areas outside the motor and somatosensory strips. Consider, for example, the speech and motor arrest induced by macrostimulation of either left or right IFG in humans. Superficially this may seem to support an inhibitory role for the rIFG, but this kind of inferential logic also dictates that the fusiform gyrus should be interpreted to have an inhibitory role, given that stimulation of that region also led to motor arrest in the same study. By similarly fallacious logic, the medial temporal lobe should also be interpreted to have an inhibitory role in memory formation, because quite naturally stimulation of the medial temporal lobe inhibits memory formation. Using this logic we should also consider that the rIFG may normally yield complex hallucinations, because stimulation of this area led one patient to perceive his doctor’s face duplicated in distorted form across his visual field.
Clearly these kinds of inferences are unwarranted, which is perhaps why the authors of all these studies interpret their effects as reflecting the disruption, as opposed to induction, of normal function. And it’s really a classic issue in neuropsychology – known as the fallacy of inferring function from dysfunction. From Susan Greenfield’s book “Brain Story”:
“The fallacy of this conclusion is obvious if you imagine the same logic applied to a radio–if you took a valve out of a radio and it started to howl, that would not mean that the function of the valve was to inhibit howling.”