EEG stands for electroencephalography. Break that word down and you get electro – encephalo – graphy, which roughly translates as “electricity – head – writing”. EEG is one of the simplest ways (keeping in mind we’re talking brain science here) of measuring brain activity, in that the technology was basically invented 90 years ago and isn’t a lot more complicated than plugging a microphone into an audio recorder. You simply attach electrodes to the head—and these can be made out of simple metal such as tin, connected to a copper wire, although there are now much more sophisticated systems—and attach these to an electrical amplifier that boosts the really tiny voltage from your head to levels that we can easily record and on a computer. In this post, I’m going to focus on what EEG measures, and how it’s used in research.
What are we measuring?
EEG is the electrical activity produced by tens or hundreds of thousands (or more) of neurons in the brain. We believe that most of what we measure comes from a type of neuron called pyramidal cells, which occur primarily in the cerebral cortex (the folded outer surface of the brain) and are oriented perpendicular to its surface. Neurons use electricity to communicate with each other, and each neuron acts as a type of “computer” to decide when to fire. This decision is based on the sum of all its inputs within a short time frame (there are a lot of inputs by the way—a typical neuron gets inputs from 10,000 other neurons!). The outer “skin” (cell membrane) of each neuron has an electrical charge—also called potential—and the inputs from other neurons change the neuron’s potential. If they change it enough, this triggers the neuron to fire, and send an electrical signal down its axon to in turn send an electrical signal to other neurons.
Where this generates EEG signals is not so much the actual action potentials, but the overall inputs to the pyramidal neurons—the changes in their membrane potentials that occur in between action potentials. These are called post-synaptic potentials because they are electrical potentials on the receiving (post) side of a synapse—the junction between two neurons. While action potentials happen very quickly and then are over, post-synaptic potentials change more slowly. So when inputs to a brain area are relatively synchronous—meaning lots of similar inputs come in at the same time, in waves—the post-synaptic potentials of lots of neurons in the receiving brain area change in synchrony, and this common activity over lots of neurons sums to create a large enough signal that we can measure it from electrodes on the scalp.
When I say “large enough”, however, it’s still not very large. EEG is measured in microVolts, or about one millionth the amount of voltage in a tiny watch battery. The amount of electricity in the brain is actually a lot more than this, but EEG is a significantly smaller signal than if we could place electrodes directly on the surface of the brain (every cognitive neuroscientist’s dream!), especially because the skull is a terrible conductor of electricity. The tiny size of EEG signals is actually a big problem—possibly the biggest problem—with EEG in research: the size of the signals we can record from electrodes on your head are very small compared to a lot of other signals that also get picked up by those electrodes. This includes the muscles in your face and neck, eye movements, blinks, and even electromagnetic noise produced by other electrical devices around the person whose EEG we’re recording. The electrodes pick up all these signals along with EEG from the brain, and these “artifacts” are often 10–100x the size of the brain signals. Formally, we say that EEG is a “low signal-to-noise ratio (SNR) technique”, which basically means there’s a lot of noise relative to the amount of brain signal we pick up. This means that it can be hard to get good EEG data—optimally applying EEG electrodes is a lab skill that requires training and careful attention to detail.
What is EEG used For?
EEG can be used both for research and for clinical diagnostic purposes. Clinically, it’s most commonly used to diagnose epilepsy, although it can be used for other neurological conditions as well. There are quite a few startup companies right now, for instance, who are looking at using EEG to diagnose concussion and to determine when someone has recovered from a concussion. In most cases, though, clinical EEG is done in a hospital and might take anywhere from a few hours to several days. The reason it can take so long is both that setting up to get really good diagnostic recordings takes some time, but then monitoring brain activity can take time too. In many cases of epilepsy, the abnormal brain activity occurs only occasionally and so long periods of recording are necessary to detect those abnormal bursts of activity—and see them repeatedly so as to distinguish them from random events. Doctors may keep someone in the hospital for days, waiting for them to have a full seizure or else just to get lots of examples of their brain activity.
In cognitive neuroscience research, which is what I’m focusing on here, we use EEG as a way of trying to figure out how brain activity relates to things like language, perception, attention, and memory, and more generally to explain how human minds work. Typically we present carefully-controlled stimuli—such as pictures, movies, sounds, words, etc.–to people and record the brain activity triggered by these. This can range from really boring experiments, to somewhat less boring ones like having people engage in a conversation or watch a musical performance, to normal day-to-day activities like riding a bicycle or driving. But using EEG to try to understand complex real-world tasks is really complicated, so typically we try to control the experience of the participant quite closely, so that we can make sense of the data.
Because EEG is, quite frankly, a pretty messy measurement of brain function, it’s limited in the kinds of things it can tell us. It’s actually kind of amazing that people have figured out much at all. Indeed, when Hans Berger, the inventor of EEG, initially presented his results, he was widely ridiculed and discredited because what he presented looked nothing like what brain activity looked like when recorded directly from the exposed brains of other animals. It was only when other researchers, Adrian and Matthews, did more systematic experiments and showed how EEG could relate to direct recordings that people came to believe in EEG.
Even today, the information we can get from EEG is useful, but it’s also quite limited. EEG is not very good at telling us where in the brain signals are coming from. This is because our brain is full of water and electrolytes, so it’s an excellent conductor of electricity. This means that all the electrical signals inside the head that EEG measures, conduct out to the scalp in all directions. So anything we measure with EEG is all of the combined activity that is happening at once. There are ways of doing fancy signal processing that try to separate the signals from different brain areas, and determine where each one is coming from, but this is usually pretty unreliable. What EEG is used for a lot in cognitive neuroscience is for getting markers of specific sensory and cognitive processes. For instance, the appearance of any visual stimulus will elicit a characteristic set of sensory responses. The size of these responses are sensitive to simple features like the brightness of the stimulus, and some also differ in size depending on whether the person is paying attention. So we can use EEG to understand the process of attention, what people are paying attention to, and things like that. Often it’s enough to know this, without caring where in the brain attention is “happening”.
In cognitive neuroscience research, there are two broad ways in which we use EEG to try to understand how the brain and mind work. One is looking in what we call the time domain and the other is looking at the frequency domain. There’s also a combination of the two called time-frequency analysis. It’s important to realize though, that the way we actually record EEG is the same for both time- and frequency domain research—although the way you design your experiment might be different. The difference is in how you view and analyze the data. Time-domain EEG is looking at EEG as we would intuitively think to look at it: as activity from each electrode on the scalp, over time. In frequency-domain EEG, we apply a mathematical operation called the fast Fourier transform (FFT) that allows us to measure the oscillations in the data—basically counting the number of peaks and troughs that occur in the data in a given amount of time, like one second. It turns out that different cognitive processes, mental states, and brain areas operate on different frequencies, which makes this a useful way of looking at things.
