Gamma Waves The Fastest, Faintest Band

Gamma waves are the fastest activity an EEG picks up, 30 to 100 Hz, and the faintest. Amplitude falls as frequency rises across the EEG spectrum, so gamma reaches the scalp as a small signal sitting on top of everything slower. Delta, at the far end, is tall enough to read off a raw trace from across the room. Gamma is not.

Gamma is also by far the widest band. Delta, theta, alpha and beta together cover roughly thirty hertz. Gamma covers seventy on its own, and people still write it as though it named one thing.

Gamma is not a state. Delta describes a stage of sleep you occupy for a stretch of the night. Gamma arrives in bursts of tens to hundreds of milliseconds, riding on the slower rhythms without replacing them. A burst is over before you could notice it had started.

Your visual system takes a scene apart before you ever see it. Colour, motion, orientation and edges are handled by different populations of neurons, in different places, on slightly different timings. Nothing reassembles them at one location, because there is no screen and nobody watching it. Yet you do not experience red, leftward motion and car-shaped as three separate facts. You see one red car moving left.

How the brain marks features as belonging to the same object is the binding problem, one of the older open questions in the field.

The best-known proposal is that timing does the tagging: neurons representing the same object fire together, and the synchrony is what marks them as one thing. The frequency that synchrony turned up at is gamma. Gray, König, Engel and Singer reported in Nature in 1989 that oscillatory responses in cat visual cortex, in the 40 to 60 Hz range, synchronised across spatially separate cortical columns with on average no phase difference between them. The synchronisation depended on how far apart the cells were and on their orientation preferences, and it was influenced by the global properties of the stimulus. Timing that shifts with the layout of what is out there is the behaviour a binding code would need.

That is why gamma, and no other band, became the frequency attached to consciousness research. If a unified percept is assembled by synchrony, gamma is the carrier. Thirty-seven years later the hypothesis is still a hypothesis.

Two facts collide. Gamma is the smallest signal in the EEG, and the frequency range it occupies is crowded with electrical activity that is not brain activity at all.

Skeletal muscle produces broadband noise that runs straight through the gamma range and well past it. Your electrodes sit on a head wrapped in muscle. Forehead, temples, jaw, neck. Holding your head up generates EMG. A slight clench generates more. Before you can ask whether audio raised somebody's gamma, you have to be sure you measured their cortex and not their jaw.

In 2007 Whitham and colleagues put a figure on it. Two volunteers were recorded doing cognitive tasks in two states, normally and then under complete neuromuscular blockade with the dominant arm left unparalysed. Power above 20 Hz differed between the two states by ten- to two-hundred-fold. Two subjects is a very small sample, and the reason it is two is that the manipulation is paralysing a conscious volunteer. There is no gentler way to get the comparison. Their conclusion in Clinical Neurophysiology was that most of the scalp EEG recording above 20 Hz is of EMG origin, and that previous studies measuring gamma EEG needed to be re-evaluated.

Then there are your eyes. Even holding a steady fixation, they twitch once or twice a second, and each twitch throws an electrical spike from the extraocular muscles into the gamma band. In 2008 Yuval-Greenberg and colleagues showed in Neuron that the classic induced gamma-band response, the broadband bump seen 200 to 300 milliseconds after a stimulus, is time locked to the onset of these miniature saccades and reflects the saccadic spike potential. The tell had been sitting in plain sight for years. The bump arrived at a latency that wandered from trial to trial, because it was tracking the eyes. The paper drew a published reply in the same journal, so this is a live argument and not a settled verdict.

Trustworthy gamma work therefore leans on intracranial electrodes, MEG, or scalp EEG with the artefact control to match: eye tracking to catch the saccades, separate EMG channels, careful filtering. It is also the reason to hold gamma entrainment claims to a higher standard than claims about slower bands.

Mains electricity adds its own line. Fifty hertz in Australia and most of the world, 60 Hz in North America, plus harmonics, all inside gamma. The standard fix is a notch filter, which takes real signal out along with the hum.

Forty hertz is the figure that stuck. It sits low in the gamma range and it is roughly where the cat-cortex synchrony clustered, so it inherited the significance of the binding work.

A separate line of laboratory research then went much further, delivering 40 Hz stimulation to brain tissue and measuring what changed. Almost all of the attention-getting results there used flickering light. Animals, not people. Tissue-level readouts, not anything a person notices from the inside.

Consumer 40 Hz audio has borrowed the number from that work and none of the conditions. Different modality, different species, different intensity, different measurement. Anyone selling you the second by pointing at the first is skipping four separate leaps.

Forty hertz also means two different things depending on how it reaches you. An isochronic tone at 40 Hz genuinely pulses. The amplitude of the sound rises and falls forty times a second, and you hear it as a rough flutter. A 40 Hz binaural beat is not in the sound at all. Each ear gets a steady tone, and the wobble is constructed from the phase difference between them by a brainstem mechanism that works on the fine timing of the waveform. That mechanism relies on the auditory nerve following the waveform cycle by cycle, which it manages well for low tones and progressively worse as the pitch climbs. Which is why binaural beats are demonstrated on low carriers, and why the pitch you put the beat on matters as much as the rate you set.

A pulse in the air and a pulse assembled behind your eardrums are two different mechanisms. Which one suits you is a question about your speakers and your ears.

Everything above is textbook physiology. What follows is contested, and the two should not be read at the same confidence.

The broad picture for binaural beats is unsettled. Ingendoh, Posny and Heine reviewed fourteen studies of what binaural beats do to EEG rhythms, for PLOS ONE in 2023, and refused to call the question closed. The balance across those fourteen sat against entrainment, and none of them was about gamma in particular.

For gamma specifically the most directly relevant trial is Melnichuk, Cooper and Hawk, in Scientific Reports in 2025. Eighty undergraduates were randomised across sixteen conditions, varying the beat band (beta or gamma), the carrier tone (340 or 400 Hz), whether the audio began before the task or with it, and whether white noise ran behind it. Twelve did not return for the second session and four scored two standard deviations below the mean, leaving sixty-four analysed, an average of four people per cell. Against a pure-tone control, gamma beats on the 340 Hz carrier came out ahead on percent hits in a continuous performance task, p = 0.002, averaged across audio onset and noise. That figure rests on the four cells sharing that beat and carrier, roughly sixteen people. On the 400 Hz carrier the same comparison found nothing and ran slightly the other way. Hits were also higher in the groups that had white noise behind the tones, p = 0.033, and not higher in the groups without it. What the beats did not do was flatten the vigilance decrement. Hits declined across blocks either way, with no interaction between condition and that decline.

The same paper has an awkward result in it. White noise attenuated the EEG entrainment, and attenuated it more for gamma beats than for beta. So noise went with better scores and weaker entrainment at the same time. Those two did not move together, which leaves open the possibility that whatever helped on the task was not entrainment.

Nobody has shown that gamma audio makes you think better. One trial found an edge on one task with one carrier, at four people per cell, and no effect on the decline in vigilance it was built to track.

Treat it as a decision about what to have in your ears while you work.

The arrangement with any evidence behind it is a gamma beat on a low carrier, with broadband noise or rain running behind it. Both halves of that come out of the same 2025 trial. The effect showed up on the 340 Hz carrier and not the 400 Hz one, and it showed up in the groups that had noise behind the tones and not in the groups without it. Four people a cell, so hold it loosely.

The carrier is the one setting that has moved a result in more than one study, and the older of the two is the more specific about it. Schwarz and Taylor, in Clinical Neurophysiology in 2005, did evoke a steady-state cortical response to a 40 Hz binaural beat, but only on a low carrier tone, a mean stimulus frequency of 400 Hz, and it became undetectable once the carrier went past 3 kHz. The monaural beat produced a bigger response than the binaural one did. That is an electrical response in auditory cortex, not evidence that anyone did anything better. Set the two side by side and the carrier is where both results turned: 2005 held at 400 Hz and vanished above 3 kHz, 2025 appeared at 340 Hz and not at 400. Four hundred hertz sits on both sides of that, which is a fair reminder that a cortical response and a task score are not the same measurement.

So keep the carrier in the low hundreds. Phase-following is the mechanism a binaural beat is built from, and a high carrier works against it. Arkhitec's gamma presets run on a 200 Hz carrier for binaural and monaural and 300 to 350 Hz for isochronic, so there is nothing to set.

Keep the beat quieter than the bed. You are not trying to hear it. On the isochronic and monaural presets, where 40 Hz is a genuine flutter in the audio, loud enough to notice is louder than it needs to be and it wears badly over an hour. A 40 Hz binaural preset has no audible pulse to judge by at all, so set that one by the carrier tone.

Give it one task and thirty to forty-five minutes. Judge it on what you produced, not on how the session felt. Feeling is where placebo lives, and any real gamma effect is small enough that your impression of it is not evidence.

Not late in the evening. Whatever gamma audio does, it will not help you sleep.

If you cannot tell whether the beat or the noise is doing the work, run the noise on its own for a few sessions.

Arkhitec's gamma presets run from 30 to 50 Hz, with the beat rate and carrier tone already paired: seven rates in the isochronic set, six binaural, four monaural. The band itself goes to 100 Hz. Nothing above 50 is offered, and there is no evidence asking for it. Binaural needs headphones. The other two run from any speaker.

The 7-layer mixer is what makes the arrangement above quick to build. Lay a gamma layer down, put broadband noise or rain over it, and set the levels independently so the beat sits underneath instead of in front.