The Ghosts in the Speakers: Diana Deutsch and the Strange Science of Hearing Voices That Aren’t There

A long walk through the psychology of phantom sound — how two loudspeakers, a tired brain, and a hungry need for meaning can conspire to make you hear words, music, and maybe even something that feels like a visitor from another world.

There is a particular kind of silence that isn’t silent at all, the kind you notice at three in the morning when the refrigerator hums in the kitchen and your brain, bored and unsupervised, starts arranging that hum into syllables, and you lie there telling yourself it’s nothing, it’s just the compressor, it’s just the pipes settling, and still some older and less reasonable part of you sits up and listens, because it sounded — just for a second — like somebody said your name. This is not a ghost story. Or rather, it is a ghost story, but the haunted house is your own auditory cortex, and the ghost is you.

For more than four decades a psychologist named Diana Deutsch has been building a career out of exactly this kind of moment — the crack in the wall where perception and invention meet — and turning it into rigorous, peer-reviewed, laboratory-grade science. She is not a mystic. She is a Professor of Psychology at the University of California, San Diego, a researcher trained in the cold clean light of cognitive science, and what she found, working with tones and loudspeakers and the unglamorous machinery of stereo sound, was something that reads like folklore and behaves like physics: that under the right conditions, almost anyone’s brain will manufacture voices out of nothing, confidently, vividly, and without the slightest awareness that it is doing so.

This is her story, and it is also the story of every person who has ever wondered, lying awake, whether the voice they just heard came from outside them, from inside them, or from somewhere that doesn’t have a name in the psychology textbooks yet.

I. The Woman Who Heard Phantoms on Purpose

Diana Deutsch did not set out to study hallucinations. She built her reputation on musical illusions — the octave illusion, the tritone paradox, the scale illusion, the glissando illusion — a whole gallery of acoustic trick mirrors that she discovered largely by accident, the way a person exploring a cave finds a chamber nobody mapped before, simply because they kept walking when everyone else turned back. She was born Diana Sokol in London in 1938, the daughter of a sculptor of the expressionist school, and she has said that her strong interest in the relationships between art, science, and philosophy grew out of many conversations with her father in childhood. That detail matters, because everything Deutsch has spent her life proving is, at bottom, an argument that perception is not a passive recording device — it is a creative act, closer to sculpture than to photography, chiseled out of raw sound by expectation, memory, and desire.

Her most unsettling discovery, the one that concerns us here, is called the phantom words illusion. Using stereo loudspeakers, Deutsch presented listeners with repeating two-syllable words and phrases, alternating the syllables between speakers so that one syllable came from the right and the other from the left. Listeners, hearing this looping, spatially fractured sound, began to “hear” entire words and phrases that had never actually been presented — and these phantom words were often related to the listener’s own memories and expectations. One person hears “I love you,” another hears something ominous, a third hears their own name, and none of them are lying, and none of them are crazy, and all of them are simply doing what a healthy brain does when it is fed an ambiguous, looping, spatially confusing signal: it fills in the blanks with whatever is already sitting closest to the surface of the mind.

Deutsch has described the effect this way: after continuous exposure to repeating words and phrases arriving offset in time from different points in space, listeners are handed a palette of raw sound fragments, and from that palette they unconsciously assemble combinations that were never actually presented — genuine phantom words, generated by the brain in its attempt to extract meaning out of chaos. Say that sentence to yourself twice, slowly, because it is the whole thesis of this article compressed into one line: the brain does not wait for meaning to arrive, it goes out and builds meaning out of debris, the way a person starving in a desert will swear they see water, not because they are foolish but because the wanting is stronger than the seeing.

There is a companion phenomenon, the speech-to-song illusion, in which an ordinary spoken sentence, looped enough times, stops sounding like speech and starts sounding, unmistakably, like singing — the exact same acoustic waveform, no editing, no trickery, just repetition wearing down the listener’s perceptual categories until speech tips over into music. And there’s the cambiata illusion, in which two repeating patterns of tones, one delivered to each ear, produce yet another kind of auditory mirage. Deutsch’s entire body of work amounts to a catalogue of the places where the seams of human hearing show, where you can see — hear — the machinery underneath, still running, still guessing, even when there’s nothing real left to guess about.

II. What the Word “Phantom” Actually Means in a Clinic

Now, the magazine version of this story would stop there, with a neat little party trick — two speakers, a loop of syllables, a haunted kitchen at 3 a.m. But Deutsch herself pushed the idea further, into territory that is not a party trick at all, and that is where this gets serious, and where it earns the word “psychology” rather than “curiosity.”

In her later work, Deutsch used her findings on musical and auditory illusions to illustrate how perceptual experiences of sound recruit both bottom-up and top-down processes to construct what we hear. The tritone illusion, involving two Shepard tones separated by half an octave, can be perceived as rising or falling depending on the listener — and Deutsch and her colleagues found that this perception varies according to a listener’s exposure to different spoken languages, implicating language itself as a top-down influence on what the ear reports as fact. And she drew an explicit line from there to clinical hallucination: research on both phantom words and auditory hallucinations implicates this same role of language-related, top-down influence on sound perception — listeners are inclined to interpret ambiguous acoustic signals as semantically coherent, often reporting intelligible words and phrases emerging from sequences of white noise or a single repeating word.

This is not a metaphor. It is a genuine bridge, built with careful hands, between a laboratory illusion that happens to healthy college students wearing headphones and the auditory verbal hallucinations — clinicians call them AVHs — reported by patients living with schizophrenia, with severe depression, with post-traumatic stress, with dementia, with Parkinson’s disease, or simply with profound hearing loss. Deutsch’s own “high-low” word illusion, a dichotic listening task in which different information is fed to each ear simultaneously, has actually been used as a research probe inside psychiatric hospitals. A 2016 study out of Zhejiang University in China used the Deutsch high-low word illusion, paired with event-related potential recordings, to compare sixteen patients with first-episode paranoid schizophrenia experiencing auditory hallucinations against sixteen matched healthy volunteers, finding measurable differences in the timing of their brain’s electrical responses to exactly this kind of illusion. In other words: a stereo trick invented to study why we mishear song lyrics turned out to be sensitive enough to register, on an EEG readout, the neural fingerprint of schizophrenia. That is not a small thing. That is a psychologist’s parlor game turning, quietly, into a diagnostic instrument.

III. Two Speakers, One Skull, and the Case for Stereo-Induced Hallucination

Here is where the idea gets genuinely unsettling, and I want to sit with it rather than rush past it, because it is the beating heart of what you came here to read: stereo sound itself — not madness, not demons, not a chemical imbalance, but ordinary two-channel stereo audio — can be a manufacturing condition for hallucinated voices.

Think about what stereo actually does to a nervous system built over millions of years for a world with exactly one source of sound coming from exactly one direction at a time. Your auditory system evolved to triangulate — a twig snapping to your left, a predator’s breath to your right — and it is extraordinarily good at using the tiny time delay and volume difference between your two ears to compute, almost instantly, where a sound “really” comes from. Stereo playback exploits that machinery and, in a sense, lies to it: it delivers two different, carefully engineered signals into your two ears simultaneously, asking your brain to fuse them into a single coherent auditory scene that never existed anywhere in physical space. Most of the time your brain performs this fusion so smoothly you never notice the sleight of hand — that is the entire magic trick behind headphones, behind surround sound, behind the illusion that a singer is “standing” between your speakers even though no singer is anywhere in your living room.

But push that system — loop the material, strip away visual confirmation, fragment the signal between left and right the way Deutsch did in her phantom words experiments — and the fusion starts to misfire in interesting ways. The brain, still obligated to produce a single coherent story out of two competing streams, starts inventing. It doesn’t announce that it’s inventing. It hands you the finished product — a word, a name, a phrase — with the same confident, unflagged certainty it uses for real speech, because as far as your perceptual system is concerned, there is no separate “hallucination channel” and “real channel.” There is only the channel, and whatever arrives in it gets treated as true until something else — reason, context, a second listen — comes along to argue otherwise.

This is worth saying plainly, because it has practical consequences for the world we actually live in, a world soaked in headphones and earbuds and binaural audio and ambient soundscapes piped into every gym and dentist’s office and meditation app: long, repetitive, looping stereo audio, especially at low volume, especially heard while tired, especially heard through headphones that isolate you from competing real-world sound, is close to a laboratory recipe for exactly the kind of phantom perception Deutsch spent her career cataloguing. You do not need to be unwell to experience it. You need only be human, and tired, and listening.

IV. Naming the Thing: A Short, Honest Glossary

Psychology is a discipline obsessed with precision partly because imprecision, here, can cost someone their reputation, their freedom, or their peace of mind. So before going further, here are the terms doing real work in this territory, defined plainly:

• Auditory Verbal Hallucination (AVH): the clinical term for hearing speech — words, sentences, commanding voices — with no external speaker present, perceived by the listener as coming from the ordinary “outer” world, as real as any voice from another person.
• Pseudohallucination: a term with a long and contested history in psychiatry, referring to voice-hearing experiences located in the “inner,” subjective space — closer to a vivid inner monologue than to an external event — as opposed to true hallucinations, which are experienced as arising from “outer,” objective space. This inner/outer distinction has long served as one of the main criteria psychiatry uses to try to separate genuine hallucinations from pseudohallucinations.
• The Phantom Words Illusion: Deutsch’s term for the specific stereo-induced mishearing described above — illusory words and phrases assembled by the brain out of looping, spatially fragmented syllables.
• Earworm (involuntary musical imagery): the maddeningly familiar experience of a song fragment looping uncontrollably in the mind — a milder, near-universal cousin of true musical hallucination.
• Musical Hallucination: a more severe, sustained perception of music, often orchestral, often with no controllable off-switch, frequently associated with hearing loss, seizure disorders, or certain medications.

That last category belonged, for years, to the neurologist Oliver Sacks, whose case studies read less like medical charts and more like small, aching short stories. Sacks described one deaf patient whose auditory cortex, deprived of its usual input, began to generate spontaneous activity of its own, taking the form of musical hallucinations, mostly memories from earlier in her life — because, he wrote, the brain needs to stay incessantly active, and if it is not getting its usual stimulation, it will create its own stimulation in the form of hallucinations. Read that twice as well. It is the same law operating on a different instrument: a brain denied real input does not go quiet. It goes looking, and it finds — or invents — something to listen to.

V. Not Everyone Who Hears a Voice Is Diagnosable — and That Fact Alone Should Humble Us

Here the story takes its strangest, most humane turn, and it belongs to an anthropologist rather than a psychologist: Tanya Marie Luhrmann, of Stanford University, who spent years sitting not in laboratories but in churches, temples, and psychiatric wards across three continents, listening — literally — to how people talk about the voices they hear.

In a landmark comparative study of people diagnosed with serious psychotic disorder in California, in Chennai, India, and in Accra, Ghana, Luhrmann and her colleagues found real, consistent differences in how their voices were experienced. Voice-hearers in the California sample were more likely to describe their voices as intrusive, unreal thoughts; those in the South Indian sample more often described their voices as providing useful guidance; and those in the West African sample were more likely to describe their voices as morally good and causally powerful. Luhrmann has summarized the American pattern bluntly: many Americans think of their minds as a private fortress, so that hearing an audible voice feels like proof the fortress has been broken — a verdict she says people find terribly upsetting, in a way that the different social contexts of Chennai and Accra do not necessarily produce.

This is not an argument that hallucinations are “just cultural” or that biology doesn’t matter — it plainly does, and the EEG differences Deutsch’s high-low illusion revealed in schizophrenia patients prove there is real, physical machinery underneath all of this. But it is an argument, a well-evidenced one, that the meaning a person assigns to an unbidden voice — friend, foe, ancestor, angel, demon, symptom — is not handed down by the voice itself. It is supplied afterward, by the listener, out of whatever cultural and personal materials happen to be lying around. Luhrmann’s later research on religious communities found that groups teaching people to “discern,” or accurately identify, the legitimate voice of a spirit or invisible being generally rely on two underlying methods: training a person’s attention deliberately toward inner experience, and repeated practice. Put a different label on that process, strip away the religious vocabulary, and you are looking at something close to what a cognitive psychologist would call attentional training and top-down expectancy — the very same top-down forces Deutsch found sculpting the tritone paradox.

None of this is an insult to faith, and it should not be read as one. It is, instead, an invitation toward humility on all sides: the believer who hears a divine voice and the patient who hears a hostile one may be running very similar neural software, interpreted through radically different cultural operating systems.

VI. So — How Do You Tell the Difference? A Practical, Unflinching Guide

This is the question underneath all the others, the one nobody quite wants to ask out loud: if the brain is this good at manufacturing convincing voices out of nothing, how does a person — any person, not just a patient in a study — tell the difference between a real external voice, their own inner monologue wearing a disguise, a stereo-induced phantom-words illusion, and something they might be tempted to call demonic, prophetic, or otherwise supernatural?

There is no single test that settles this with total certainty — psychiatry itself, as the research on pseudohallucinations shows, has never managed to draw one clean unambiguous line. Clinical observation has repeatedly found that patients’ actual reports of voice-hearing do not easily fit into the neat categories theorists once proposed, and even a patient stabilized on antipsychotic medication may go on hearing voices that sound highly external and perceptually vivid while no longer believing those voices are real. That finding alone should temper anyone’s confidence in a tidy checklist. Still, clinicians, researchers, and contemplatives across very different traditions have converged, independently, on a handful of genuinely useful distinguishing questions. Here they are, offered not as dogma but as a lantern for a dark room.

Location. Does the voice seem to arrive from a specific point in outer space — behind you, to your left, through the wall — the way a real voice would, obeying the physics of distance and direction? Or does it seem to arise from inside your head, indistinguishable in location from your own thoughts? Voices heard in that “inner” space have consistently been described in the clinical literature as more distressing, less controllable, and less likely to remit than voices perceived as coming from the outside. — a counterintuitive finding, since one might expect an “external” voice to feel scarier, and yet the opposite tends to be reported.

Control and dialogue. Real conversation, with a real person or in genuine prayer or contemplative practice, has a certain elastic, responsive quality — you can ask a question you don’t know the answer to and receive something that surprises you, something with texture and unpredictability, the hallmark of an independent mind. A stereo-induced phantom word, by contrast, is a closed loop: it is built from the raw material you already possess — your own vocabulary, your own memories, your own fears — assembled instantly and without genuine novelty, the way Deutsch’s listeners each heard different phantom phrases shaped precisely by their own personal associations, not by any content that wasn’t already latent in their minds.

Repetition and fatigue as triggers. Ask honestly: did the experience arrive after long exposure to looping, ambiguous, or monotonous sound — a fan, an engine drone, a repeated song, hours of headphone use, sleep deprivation, a long stretch of sensory isolation? This is precisely the fingerprint of the phantom words mechanism and of Sacks’s musical hallucinations, both of which are reliably summoned by exactly these conditions. A voice that shows up reliably after fatigue and repetitive sound, and fades with rest, silence, and real human company, is behaving like a manufactured perception rather than an encounter.

Content and character over time. Genuinely significant experiences — whatever their ultimate source, spiritual or otherwise — tend, across most serious contemplative traditions, to leave a person more oriented toward clarity, compassion, and steadiness. Experiences that escalate toward isolation, command a person toward harm to themselves or others, or systematically cut a person off from the trusted people in their life warrant urgent, serious clinical attention — regardless of whether the person interprets them as psychiatric, spiritual, or something else entirely. This is not a theological claim; it is simply the most consistent practical marker used across both religious and clinical traditions to separate an experience worth exploring from one that requires real medical help right away.

When in doubt, get a second observer and a second set of ears — literally. If a voice seems to be coming from a real external source, the simplest test in the world is also the most powerful: change rooms, change speakers, ask another person if they hear it too. Deutsch’s whole body of work exists because she built situations where the “external” source was demonstrably, verifiably producing nothing but two ordinary, separable audio tracks — the test that dissolved the illusion, every time, was simply listening to each channel alone. Real voices survive that test. Phantom ones do not.

VII. Coda: The Brain as an Unreliable, Devoted Narrator

There is an old, unglamorous truth sitting underneath all of this, and it is the one worth carrying out of this article and into whatever noisy, looping, headphone-saturated life you happen to be living: your brain is not a passive microphone recording the world exactly as it is. It is a narrator, and like every narrator worth reading, it has opinions, it has memory, it has hunger, and sometimes, late at night, tired and underfed on real information, it starts telling you a story before it has actually finished listening. Diana Deutsch spent a career proving this with nothing more exotic than two ordinary loudspeakers and a loop of syllables, and in doing so she built a small, sturdy bridge between a college dorm-room party trick and the deepest, oldest human question there is — how do we know, in the end, whose voice we are actually hearing?

The honest answer is: carefully, patiently, with more humility than certainty, and with real trust in the people around you who can tell you what they hear too. The refrigerator hums. The syllables loop. And somewhere in the dark, your own mind, faithful and overeager, reaches out to finish the sentence for you.

___________________________________

This article discusses auditory hallucinations for informational and educational purposes. If you or someone you know is experiencing distressing voices, sudden changes in perception, or thoughts of self-harm, please reach out to a licensed mental health professional or a crisis line in your area — these experiences are common, often treatable, and nothing to face alone.