Binaural Beats How the illusion is built

Play a 200 Hz tone into your left ear and a 210 Hz tone into your right, and you hear a third sound: a slow pulse, ten times a second. Put a microphone between the two earcups and it records nothing of the kind. Illusion is the technical word for that, and it is not a put-down. Pitch is an illusion too.

The beat is assembled in the brainstem. Sound from each ear travels up its own pathway, and the two first meet at the superior olivary complex, the structure that works out where a sound is coming from by measuring arrival differences of tens of microseconds between the ears. Feed it two tones that never quite line up and it produces a rhythmic output at the difference between them.

Through a speaker none of that happens. Both ears hear both tones, the tones sum in the air before they reach you, and what arrives is an ordinary acoustic beat: a real amplitude wobble, physically present, no brainstem involved. Without headphones there is no binaural beat to hear.

The beat also survives conditions that would kill an acoustic one. Oster reports the Berlin audiologist E. Lehnhardt finding it still perceived when one of the two tones sits below the listener's threshold of hearing, and gives J. J. Groen's figure from Utrecht for how far below: one tone plainly audible, the other at minus 20 decibels, which Oster glossed as a hundredth the loudness of barely audible sound. The beat was still there. Oster's reading was that the brain detects and processes a signal too weak to reach consciousness. Send that same pair of tones into one ear and you hear the louder one, no beat.

Heinrich Wilhelm Dove described binaural beats in 1839, and for a long time nobody thought they needed explaining. As late as 1915 they were still being written off as a trivial special case of monaural beats, on the argument that each ear was simply picking up sound intended for the other.

What dislodged that was work on how the two ears are combined. Oster describes Robert Galambos's 1959 cat experiments as showing that loud clicks delivered to both ears generate nerve impulses meeting in the superior olivary nucleus: simultaneous signals reinforce each other there, while introducing a slight delay inhibits the result instead. The machinery a binaural beat depends on had been recorded directly.

Gerald Oster gathered the strands into a 1973 Scientific American article, “Auditory Beats in the Brain”, and that is the piece the modern industry traces back to. Oster was not selling relaxation. His interest was diagnostic: because the beat depends on machinery that compares the ears with microsecond precision, whether you hear it, and up to what pitch, is a cheap read on whether that machinery is working. He tested patients with a movement disorder and reported that a few could not hear binaural beats at all. One, a violinist, could not hear them on admission to hospital, and could hear beats up to about 650 Hz as his treatment progressed. His treatment produced the change; the beats only measured it. Oster's suggestion was that the range over which someone hears beats might expose physiological changes too subtle to catch another way.

The beats were the measuring instrument, pointed at the listener. Everything since has inverted that into something done to the listener.

200 and 210 Hz give you a 10 Hz beat. So do 400 and 410. So do 1,000 and 1,010. The beat frequency is the difference between the ears; the carrier is the pitch the whole thing rides on. The beat is the number that gets quoted. The carrier decides whether there is a beat to quote.

Oster's listeners heard beats most clearly around 440 Hz. Above that they became less distinct, above roughly 1,000 Hz he reported them vanishing altogether, and no subject of his reported a beat above 900 Hz. Below about 90 Hz the beat starts getting confused with the tones producing it. He was careful that this window is not a fixed property of hearing: the intensity of the tones and the type of earphones change the result, and he noted other investigators detecting beats from tones up to almost 1,500 Hz.

Schwarz and Taylor measured cortical response instead of asking people, and found a binaural-beat steady-state response at a low mean stimulus frequency of 400 Hz that became undetectable past 3 kHz. The two methods do not agree on where the ceiling sits. Their recording survived to 3 kHz, well above the pitch at which Oster's listeners heard nothing at all, which is the same gap that turns up at 40 Hz: a response you can record, no beat you can hear. Both agree it is a low-frequency effect. A 2 kHz carrier is not doing what a 400 Hz carrier does.

The 2025 Scientific Reports experiment by Melnichuk, Cooper and Hawk is unusual in varying the carrier on purpose: 340 Hz against 400 Hz, crossed with beta and gamma beats, two onset timings, and the presence or absence of background masking noise, across 80 participants, 64 of whom survived attrition and a performance exclusion to reach the analysis. Gamma beats on the low carrier beat a matched pure-tone control on percent hits (p = 0.002); the advantage of having white noise behind the tones (p = 0.033) was a separate contrast, not the same cell. Neither reduced the decline in performance over time, which is what sustained attention actually means. The EEG did show entrainment, though the carrier made no difference to it and the white noise cut it down.

There is a ceiling on the beat as well, and it moves with the carrier. Oster puts the fastest perceptible binaural beat at about 26 per second, from measurements by Licklider, Webster and Hedlum, and that maximum holds only over one band of pitches: tones higher or lower bring it down. Past about 30 Hz of separation there is no beat at all, and you hear two tones. Which is awkward for gamma presets, because at a 40 Hz beat there is nothing to hear. Schwarz and Taylor still recorded a 40 Hz cortical response.

An acoustic beat swings the sound from full loudness to silence and back. A binaural beat does nothing of the sort. Oster described it as only a slight modulation of a loud background, and put the depth of that modulation at about six decibels, which he glossed as roughly a tenth the loudness of a whisper. He had to prime his subjects to hear it at all: play monaural beats, switch suddenly to the binaural mode, and it still took two or three seconds for the beat to register.

Schwarz and Taylor found the same asymmetry at the cortex. The steady-state response to a binaural beat was smaller than the response to a real acoustic beat, and it vanished at carrier frequencies where the acoustic response survived. That study measures cortical signal. It is not an outcome anybody felt.

Noise ought to make it worse. It does the reverse. Oster reported that noise just loud enough to obscure the two tones makes the beats more distinct, and at the right level the tones themselves disappear while the noise seems to be modulated by the beat. Do the same with both tones in one ear and you hear only the noise. Oster attributes the explanation to Jeffress and colleagues at the University of Texas, who read it as chance reinforcement: the noise is more likely to reinforce when the two signals coincide and to cancel when they oppose. The 2025 trial below found its attention gain on the low carrier, and separately in the groups that had white noise running behind the tones — the same broadband layer Oster had described half a century earlier.

The stimulus is faint. Sensitivity to interaural timing differs from person to person, and sessions run ten to thirty minutes. Inconsistent findings are what you would predict from that.

The most thorough look at the entrainment question is the 2023 systematic review in PLOS ONE by Ingendoh, Posny and Heine, which gathered fourteen studies of binaural beat stimulation and brain oscillatory activity. Five supported entrainment. Eight contradicted it. One was mixed. The authors concluded that the question cannot be settled on the evidence available, which is a straight statement that the central mechanism claimed for this whole category is unproven.

A randomised controlled trial published in Anaesthesia in 2005 asked a different question. Padmanabhan, Hildreth and Laws played binaural beat audio to day-case surgery patients waiting for a general anaesthetic. Anxiety scores fell 26.3% in that group (95% CI 19–33%), against 11.1% for the same audio with the beats removed and 3.8% for no audio at all. One trial, one setting, self-reported scores, two decades old. What it measured was how people felt in a waiting room, which is a different question from the one the review asked. It is not a finding about treating anxiety.

The mechanism runs on a comparison between two ears, so anything asymmetric changes the input: an unequal hearing loss, a blocked ear, an old pair of earbuds where one side sits deeper than the other. A phone in a call mode that folds both channels to mono removes the effect completely and silently, with the audio still playing.

The low-frequency window follows from what the machinery is for. A sound whose wavelength is longer than your head bends around it and reaches both ears, so direction has to be worked out from the phase difference between them. Short wavelengths cast an acoustic shadow instead, and localisation switches to comparing loudness. Beat perception fades as the tones climb towards 1,000 Hz, which Oster read as a sign that the beat rides on the phase mechanism. Galambos's cats point the same way: a slight delay between the ears weakened the signal at the superior olivary nucleus, which is presumably why you turn your head towards a sound. You are nulling the difference out. In headphones, turning your head does nothing.

Age is less of an obstacle than you would guess. Hearing loss with age takes the high frequencies first, and binaural beats live below 1 kHz, so the window they need stays largely intact. Oster reported older listeners detecting beats and locating sounds almost as well as young ones.

Attention matters more than any of that. The beat is faint.

The general claim is unsettled, so the only useful question is whether this does anything for you. That is testable in an afternoon.

Prove the illusion first. Start a beat, then pull one earbud out. The pulse should disappear and leave a plain tone behind. If it keeps pulsing, you are hearing an acoustic beat: your source is mixing to mono, or the sound is coming from a speaker.

Conversation volume is enough. Turning it up does not deepen the modulation, which Oster estimated at about six decibels, and a loud carrier gets tiring to sit under.

Give it a fixed job rather than a mood. Twenty minutes of the same task, same time of day, two sessions with beats and two without, and one number written down afterwards: pages read, or how wound up you feel out of ten. Compare the four. It is crude, and it still beats a vibe.

If the answer turns out to be nothing, that is a real result, and it agrees with the majority of the studies in that 2023 review.

Arkhitec ships 41 binaural presets across all five bands, delta through gamma, with the carrier and the beat already set so there is nothing to dial in.

The binaural layer is one of seven in the mixer. Put a soundscape over the top, forest or thunder or Tibetan bowls, and the Smart Mix Engine keeps the levels balanced so the beat sits under the sound instead of fighting it. Save a combination and it comes back with one tap. Set the sleep timer and the session ends without you reaching for anything.

If you want to hear one before installing anything, the generator on this site is free.