Developing Intelligence

A continuing challenge in cognitive neuroscience is determining which neural structures are actually responsible for certain thoughts and behaviors. For example, fMRI and other neuroimaging techniques cannot tell us if a certain region of visual cortex is necessary for perceiving motion, or if it is merely coactivated whenever motion is perceived. Such distinctions are both particularly important and particularly difficult to achieve in domains thought to be uniquely human: we cannot simply lesion human brains and observe the consequences as we can with animals trained to perform lower-level cognitive tasks.

Human brain damage caused by stroke, neurological disease, and other of “nature’s experiments” are of some help in this endeavor but are also associated with their own problems (e.g., the brain can drastically reorganize and compensate for damage; widespread rather than isolated regions of the brain are often affected, making localization difficult; the adoption of varying strategies by human subjects can confound the results).

In 1985, a new tool – transcranial magnetic stimulation, or TMS – has surfaced which allows us to temporarily disrupt neural activity in a relatively small region of cortex (approximately 1 square centimeter). The only problem is that no one knows exactly how TMS performs this “disruption.” For example, TMS might deactivate or even overexcite a particular region of the brain. Interpreting the cognitive effects of TMS requires a better understanding of its mechanism of action.

TMS’s effects have recently been clarified in a Science article by Allen, Pasley, Duong & Freeman. The authors administered short trains of magnetic stimulation (i.e., 1 to 4 seconds) to an anesthetized cat’s visual cortex, while measuring neural, vascular, and metabolic characteristics of the stimulated tissue after various lengths of time and various lengths and frequencies of TMS. Here’s the basic time-line of neural events after TMS:


Immediate effects.
– Spotaneous spike rate increases by 200% (similar increases occur among local field potentials oscillating at frequencies above 40 Hz). This increase scales with the frequency at which TMS is applied.
– Evoked spike rates decrease by 50% (similar reductions occur among local field potentials oscillating below 40 Hz).
– Tissue oxygen increases, at a level that scales with the frequency with which TMS is applied.

10-15 seconds.
Peak of tissue oxygen increase; now begins to decrease below normal levels for next 2 minutes; this reduction is greater for faster frequencies of TMS

0-30 seconds.
– The coherence of spontaneous neuronal firing (i.e., phase locking) is maximally decreased across a wide range of frequencies.
– The coherence of evoked neuronal firing is decreased at low frequencies (delta: 1-4 Hz) but increased at high frequencies (gamma & high gamma: 30-300 Hz).

1 minute.
Spontaneous spike rate decreases to normal levels.

1-1.5 minutes.
– There is a significant increase in the amount of spontaneous neural activity occuring in the gamma range (30-90 Hz); other frequency bands return to normal levels in terms of spontaneous activity.
– Low frequency delta-wave evoked spiking remains at below baseline levels, but other frequency bands return to baseline levels.

~5 minutes.
Evoked spike rate returns to normal levels; the minimum spike rate attained scales with TMS frequency.

These results are an important step in understanding exactly how TMS affects the brain, but there are many remaining issues. Most obviously, there are not great models for understanding how changes in some frequency bands but not others may affect neuronal computation, much less the cognitive processes which arise from it. Likewise, although there are interesting theories about the role of spontaneous firing, these theories are not developed enough to allow for a clean-cut interpretation of how TMS might affect cognitive processing.

There are also more methodological issues. For example, the the effects of TMS may depend on the site of cortical stimulation (neighboring regions of cortex can respond differently to identical stimulation). TMS may also affect underlying white matter. Finally, this work shows that fairly minor differences in stimulation frequency and duration can alter the effects of TMS; a wide variety of techniques are often already used in humans, including long durations of repetitive TMS (sometimes up to 15 minutes in length) and a number of different stimulation frequencies (more than twice the frequencies used here.) So it’s not clear that the effects observed here change uniformly throughout that parameter space – in my opinion, it’s even unlikely.

Given such highly heterogeneous effects of TMS on even a single cortical area in a single species, nature’s experiments may currently be a more straightforward tool for uncovering causal brain-behavior relationships in higher-level cognitive neuroscience.