Computing the missing 99 percent

Science is running a brain-probing essay by Marcus Raichle, titled "The brain's dark energy." The basic idea is that most of the activity of the brain appears to have little to do with responding to environmental inputs:

Human functional neuroimaging, first with positron emission tomography (PET) and now largely with functional magnetic resonance imaging (fMRI), allows the brain's responses to controlled stimuli to be studied by measuring changes in brain circulation and metabolism (energy consumption). Surprisingly, these studies have revealed that the additional energy required for such brain responses is extremely small compared to the ongoing amount of energy that the brain normally and continuously expends (2). The brain apparently uses most of its energy for functions unaccounted for--dark energy, in astronomical terms. What do we know about this dark energy?
The adult human brain represents about 2% of the body weight, yet accounts for about 20% of the body's total energy consumption, 10 times that predicted by its weight alone. What fraction of this energy is directly related to brain function? Depending on the approach used, it is estimated that 60 to 80% of the energy budget of the brain supports communication among neurons and their supporting cells (2). The additional energy burden associated with momentary demands of the environment may be as little as 0.5 to 1.0% of the total energy budget (2). This cost-based analysis implies that intrinsic activity may be far more significant than evoked activity in terms of overall brain function.

OK, but the question is how that intrinsic activity is structured. Is it all about neurons held in balanced readiness to respond in finely nuanced ways to various inputs, or is it all about a vivid internal life of consciousness, or something else entirely?

They raise a very important point about the technical limits of fMRI to detect low-level activity:

A prominent feature of fMRI is that the unaveraged signal is quite noisy, prompting researchers to average their data to reduce this "noise" and increase the signals they seek. In doing this, it turns out that a considerable fraction of the variance in the blood oxygen level-dependent (BOLD) signal of fMRI in the frequency range below 0.1 Hz, which reflects fluctuating neural activity, is lost. This activity exhibits striking patterns of coherence within known networks of specific neurons in the human brain in the absence of observable behaviors (see the figure).

This implies that current technology cannot probe for the kind of sensitive networks that may enable most responses to stimuli. Of course, this also implies that many of the integral parts of cognitive function are also invisible -- averaged out of the analysis. Increasingly fine sampling in a context sensitive to time might make a difference -- instead of global oxygen consumption, a pattern of activity cascading into different regions might be picked up.

But at some point we also will reach the point where people are essentially heterogenous in their processing networks. Averaging across people for different treatments also tends to eliminate idiosyncratic things -- even if everyone has the same thing instantiated in a slightly different way. So there may be a limit on detection that is intricately connected to the heterogeneity of processing, which is itself a likely consequence of any learned (including cultural) elements of behavior.

References:

Raichle ME. 2006. The brain's dark energy. Science 314:1249-1250. DOI link