TMS: "Virtual lesion" or "virtual excitation"?

Yesterday's introduction to paired-pulse transcranial magnetic stimulation elicited an insightful comment from reader "Kix":

As you mention, TMS can be used in order to disrupt or to read out some areas of the brain. I don't see why these functions should be mutually exclusive. For instance, delivering single pulse TMS over the primary motor cortex during movement preparation allows one to guess the direction of the future movement from the size of the motor evoked potential. However, the fact of reading out the prepared response might also disrupt it (Michelet et al. 2010).

I think we agree here. If TMS is disrupting the function of an area and also "reading out" its functioning, then we don't know whether the "read out" is a reflection of the area's intrinsic functioning or the dysfunction induced by TMS. This is a serious problem: how can we pin the tail on the donkey?

For example, Kix cites a beautiful study by Michelet et al in which the authors infer that response inhibition (in a flanker task) occurs via a process of biased competition or "response replacement." I fully believe this is actually the way response selection occurs; but those who don't could easily claim that TMS disrupted a targeted inhibition of the incorrect response that normally occurs prior to the excitation of the correct response. If TMS is both disrupting and inducing normal motor function, this just further complicates any conclusions we might draw about how motor control works in an intact, behaving organism.

Behavioral outcomes from TMS can only be measured through either the primary motor cortex or the primary visual cortex (where TMS induces perception of phosphenes). So, assessing the state (rather than function) of some areas is always done through TMS on the area of interest plus on the motor cortex in order to elicit MEPs. Therefore, those paradigms strongly rely on the state of the primary motor cortex itself. Excitability (or state) of the primary motor cortex is influenced by many factors such as reward (Kapogiannis et al. 2008, EJN). Clearly, the fact that MEPs are the single possible outcome of all this new paradigms is very limiting.

I agree this is all very problematic.

But being restricted to MEP's as an outcome measure for TMS is somewhat irrelevant to the interpretational issue I'm emphasizing. For example, the intensity of TMS to PMv (aka rIFG) might be calibrated so as to produce a certain increase in a measure correlated with PMv/rIFG functioning: stop signal reaction time (SSRT). We could then estimate the probability that a subject will fail to inhibit at a particular stop signal delay given a test pulse to PMv/rIFG, and test whether a conditioning pulse to another area (e.g., TPJ) causes a deviation from the expected probability of inhibition. One could do this - that's not the problem; the problem is that the result wouldn't have an unambiguous interpretation.

To demonstrate this, suppose the test PMv/rIFG pulse reduces SSRT by 50ms but this is countermanded by a conditioning TPJ pulse ... what are we to conclude? That the TPJ normally inhibits the function of rIFG, and this is being induced via a conditioning stimulus? Or that the TPJ normally supports the function of rIFG and this is being disrupted? Or actually that rIFG and TPJ only typically interact in other contexts, while the conditioning pulse both disrupted that segregation of function and induced an out-of-context interaction?

Finally, I'd like to point out that TMS requires a lot of training and is much more complicated that it sounds. The perfect coil placement is required in order to obtain consistent responses to the stimulation. In addition, stimulating both the premotor and motor cortices with two different coils requires their placement over a very small area on the top of the head of the subjects. Let me tell you that it is quite a achievement to be able to modulate M1 and PMv at the same time.

I didn't mean to imply this work was easy - in fact I've been incredibly impressed with the rigorous methods and extreme technical expertise in the TMS literature. Bravo!

My complaint is about the inherent uncertainty of the conclusions we can draw from this data without knowing whether the TMS is inducing or disrupting function (or worse, some bizarre combination of both!)