Skip to content Skip to navigation

Ask a Neuroscientist: Does the brain have an energy budget?

by Ada Yee

Graduate students take questions from the public and answer them on the blog Neuwrite West as part of their Ask the Expert series.


“My question concerns the relative aggregated firing rate for three classes of neurons within the human brain. The context of the question is my interest in the relative energy budget that has evolved in brain functionality in these areas, with the possible extension to artificial stand-alone energy-limited systems. (I’m a Silicon Valley hardware guy.) The three neuron classes I’m defining for the purpose of this question are 1) Internal body monitoring and management; 2) Background monitoring of external sensory inputs, including “priority interrupt” generation, and 3) active attention-related processing correlating to FMRI scans. Has there been any research in this area? I respect biology for usually having found the optimum solution for basic issues like this. Thanks!”  — Ron Cline


Hi Ron,

You ask a fascinating question about how the brain uses energy, and identify some of the brain’s most critical functions. While there is no readily available, quick answer to this challenging question, I will try to address it from a few different angles. I’ll start by talking briefly about how the three functions you name (internal monitoring, external monitoring, and attention) are actually organized in the brain. Then, I will introduce three possible ways of assessing how the brain distributes its resources: first, by direct measurement of oxygen flow and glucose uptake; second, by examining what processes the brain sacrifices when energy gets low; and third, by calculation from known simple properties of neurons.


You’ve defined three “types” of neurons for each of three processes: internal monitoring, external monitoring, and attention. Let me point out that most neuroscientists don’t classify neurons according to these three functions. Instead, each of these functions involves multiple areas of the brain, and every area contains a heterogeneous mixture of neuron types. Some of these neurons have very high basal firing rates and some low.

One possibility is to think of the brain as having several networks of multiple processors serving these functions, but all processors have the same repertoire of resistor types; it’s not so much the resistors themselves that make these processors serve a specific function, but how they are uniquely arranged.

Finally, as you have pointed out, neurons code information by firing spikes. Spiking rate changes drastically as a neuron performs its function: for example, a neuron involved in monitoring hunger may go quiet when the animal has just eaten, while a neuron involved in vision might let out a burst of spikes when it encounters an object moving a certain direction.


Could we determine the brain’s energy budget from functional magnetic imaging (fMRI) experiments, in which measurements of blood oxygen level (a.k.a. BOLD signal) are measured during performance of different tasks?

Whether BOLD signal truly represents spiking activity is still being tested. Moreover, as we’ll talk about later, spiking isn’t the only activity taking up a neuron’s energy. On the other hand, if we’re simply interested in how the brain uses energy, then maybe BOLD signal is our messenger—since we’re measuring oxygen delivery, which is required for metabolism, then by definition we’re watching where the brain is working hardest and needing replenishment most.

However, for reasons I won’t explain here, BOLD is an indirect and relative measure of oxygen consumption. It is most useful when comparing the same region to itself across short time periods, and inappropriate for comparing different brain regions, because blood vessels don’t reach all areas equally. So, comparing the energy consumed by an area involved one process over another is not straightforward (also: this).

An older method called positron emission tomography (PET) can track brain metabolism by measuring the uptake of “tagged” glucose throughout a patient’s brain. While most studies using this technique focus on aging or Alzheimer’s patients, buried in one of these studies are measurements of energy use in normal patients (also: this).

This result, which should be taken with caution because it was found from only a small number of patients under certain conditions, suggests that certain brain areas (namely, the frontal, cingulate, and occipital cortices, which are roughly involved in attention, emotion, vision, and the basal ganglia, which is important for motor planning) consume at least 20% more glucose than areas such as the hippocampus (involved in memory).


Another approach is to consider how the brain prioritizes its work during times of extreme stress or starvation. When energy is limited, what functions does the brain sacrifice in order to save energy?

One creative approach has been to study ultra-marathoners. These runners, who run nearly 3,000 miles over 64 days without rest, withstand extreme stress but remain healthy. Therefore, any changes during the marathon could be thought to reflect a healthy challenge, but not damage, to the brain.

MRIs before, during and many months after the race showed that grey matter temporarily decreases in many regions of cortex, including regions implicated in some of the sensory intake processes, attention and external monitoring functions you’ve named. Also decreased are parts of the basal ganglia.  On the other hand, it’s hard to imagine that centers involved in internal monitoring have shut down in these runners, since essential functions such as breathing and sweating (temperature regulation) must occur during intense physical activity.

Thus, we might conclude that under duress, the brain prioritizes internal functions over sensory intake and attention. However, this study examined only a dozen individuals and needs to be repeated to ensure that the findings hold up.


Some scientists have drafted estimates of the brain’s energy expenditure based on the basic properties of spikes. A single spike in a neuron is generated by the passive flow of positive charges (in the form of sodium ions) into the neuron. The neuron balances this inflow by actively pumping ions back out. The “energy cost” of a single spike is estimated by measuring how much energy the neuron consumes with this pumping action.

Further calculations estimate the energy needed for signaling between two neurons. This involves a whole separate set of ion flows, pumps, and packaging of signaling molecules needed to communicate with other cells. It may surprise you to learn that for most neurons, this inter-neuronal communication--called “synaptic transmission”--actually takes up more energy than spiking. Both theoretical estimates and recent data collected from lab experiments have shown this to be true (see also: this).

In fact, there are many tasks within neurons that each require a different share of energy, including not only spiking and talking to other neurons, but also maintaining stores and generating new signaling molecules. As we discussed, there is a variety of neurons types within the brain. As might be guessed, bigger neurons have higher overall consumption than small ones, but also, different neurons have different energy distributions: my 6’ older brother might eat a lot more than me, but I spend more of my time chatting on the phone, while he spends more energy at the gym building biceps.

Now we can see how deriving the brain’s energy costs is a nuanced exercise. Nevertheless, putting everything together, scientists have constructed some general “energy budgets.” The latest editions of these suggest that the cortex, whose functions include external monitoring of sensory information as well as top-down attention processes, takes up some 44% of the brain’s total energy. Within the cortex, some 55% of energy is consumed by cell-to-cell communication, 21% on spikes, and 24% on maintaining ion balancing and molecules involved in transmission (see: here).


As you can see, you’ve posed a tough question, but one that plenty of current neuroscience research is tackling from many angles. We’ve gone through a few ways to think about how the brain distributes its energy, and I hope this helps you consider the question in even more depth. If you’re interested, I encourage you to take a look at this paper (note: paywall) which includes and interactive spreadsheet that allows you to make some estimates of your own!