She downloaded the app in October 2013 because she wasn’t sleeping well.

This is a common enough reason — the sleep-tracking app market exists precisely because a significant portion of the population has a complicated relationship with the hours between lying down and waking up, and because the quantification of things that have traditionally resisted quantification has become, in the smartphone era, both possible and appealing. The app she chose was called Sleep as Android. It tracked sleep cycles through the phone’s accelerometer, generated graphs of light and deep sleep phases, and recorded audio throughout the night — clipping and flagging any sounds that deviated from ambient silence, so that in the morning a user could review what their sleeping hours had actually contained.

She set it up on the first night and found, in the morning, the expected things: some shifting, some ambient noise from outside, the ordinary acoustic texture of a night in an apartment. She reviewed the graphs. She noted the patterns. She found the whole enterprise, as she would later write, interesting in the way that data about yourself is always interesting — the specific pleasure of being shown what you cannot observe directly.

For the first month, nothing unusual appeared in the clips.

In November, she started hearing the clicking.

It appeared in the morning review — a sound in one of the flagged clips, brief, irregular, without a consistent cadence. Not a knock. Not a voice. Something closer to the sound of a small object being dropped, or a mechanism engaging, or a surface being tapped — specific enough to register as distinct from ambient noise, ambiguous enough to resist easy identification.

Her first assumption was the fan. She had a ceiling fan in the bedroom, and the pull-chain on a ceiling fan will, under certain conditions of air movement, tap against the housing in a rhythm that is irregular enough to be noticeable and mechanical enough to be mistaken for something intentional. She tested the theory — turned the fan on, watched the chain, listened for the sound. She could not reproduce what the app had recorded.

She tried other explanations. The building settling. A pipe in the wall. Something in the hallway. She was not alarmed, she would later say — not yet — only curious in the way that unexplained small things are curious, the mild intellectual itch of a pattern that doesn’t resolve.

Over the next several weeks the clicking appeared a few more times. Always at night. Always in the bedroom. Always in the clips the app flagged as anomalies rather than in the continuous background recording.

She filed it away as something she hadn’t figured out yet.

On the morning of December 30th, 2013, she opened the app and found a flagged clip from 2:04 AM.

She pressed play.

She heard the clicking — the same clicking, the sound she had been hearing for weeks, familiar enough now that her brain had categorized it and moved past it. Then she heard herself.

She heard herself sit up in bed. The shift of weight on the mattress, the change in the acoustic environment that comes when a person moves from horizontal to upright. Then her own voice, calm and clear — not the slurred approximation of someone half-asleep, not the disoriented mumble of a person disturbed by a noise, but a clear, coherent question directed at something in the room with her:

What are you doing?

A pause.

Then a voice — male, distinct, close enough to the phone’s microphone to be recorded clearly — said:

Nothing.

Then the clicking again.

Then the voice said something else. The exact words of this second statement have been transcribed various ways by various people who have listened to the recording, with no consensus on what was said. The audio quality at that point degrades. She has said she believes she heard a word or two but cannot be certain.

Then silence.

She listened to the clip twice. Then she put the phone down and checked every room in the apartment. Every door was locked from the inside. Every window was secured. There was no sign of forced entry — no damaged locks, no displaced furniture, nothing missing that she could identify. The apartment was exactly as she had left it when she went to sleep.

She was alone.

She posted the recording to Reddit.

This was her friend’s suggestion — not because Reddit was likely to solve the problem, but because the specific combination of thousands of people with varied expertise and the anonymity that encourages honest engagement sometimes produces useful analysis of things that conventional channels won’t touch. She uploaded the clip, explained the context, and waited.

The response was substantial and predictably varied. The majority of the comments fell into the categories that such posts always generate: elaborate theories, dismissals, questions about audio equipment, speculation about sleepwalking and somniloquy, suggestions that she had dreamed the whole thing, suggestions that she had fabricated the whole thing. The range of human response to an unexplained recording is wide and tends toward the poles.

But among the thousands of comments was a message from someone who identified themselves as a speech and language specialist, who had reviewed the audio and wanted to discuss it further. She sent them a clean copy of the recording. She also sent recordings of her own voice and her son’s voice, as they requested, for comparison.

The specialist’s conclusion: the voice in the recording was physiologically inconsistent with either her or her son. The resonance, the formant frequencies, the specific acoustic signature of the voice — these were distinct enough from both comparison samples that the specialist said it would be impossible for either of them to have produced it.

Someone else’s voice.

In her locked bedroom.

At 2:04 AM.

Answering a question she asked while she was asleep.

The clicking continued after that night.

She kept the app running — partly from the habit of three months, partly because she understood now that the recordings were documentation of something, though she could not have said precisely what. In the weeks that followed, the app captured the clicking sound multiple more times. Always the bedroom. Always the night hours. Always the same irregular, unidentifiable quality.

No more voices.

She found some cold comfort in this, she said later — the absence of the voice felt, irrationally, like a reduction in whatever had been present. The clicking alone was unsettling. The clicking with a voice attached to it was something else, something that occupied a different category of her mind entirely.

She moved in the early spring of 2014.

She has not used a sleep-tracking app since.

There are rational explanations that the rational mind reaches for in cases like this, and they are worth naming honestly.

Somniloquy — sleep talking — is common, affecting a significant percentage of the population, and can produce coherent-sounding speech that the speaker does not remember upon waking. The question she asked — what are you doing? — could have been produced by her own sleeping brain responding to the clicking sound, constructing a narrative around it, generating a verbal response to a stimulus that her waking mind would have handled differently.

The male voice is harder to account for through this explanation. The speech specialist’s assessment was not a formal forensic analysis — it was the opinion of one professional, working from recordings of uncertain provenance, without the controlled conditions that rigorous acoustic analysis requires. The possibility that the assessment was wrong, or that the comparison recordings were inadequate, or that the vocal qualities attributed to a male voice were produced by some combination of recording artifacts and sleep-altered speech, cannot be entirely excluded.

The clicking has no consensus explanation. Ceiling fans, pipes, animals in the walls, settling buildings — all of these produce sounds. Whether any of them produce this specific sound, in this specific pattern, is unverifiable from a recording made on a smartphone three years ago.

These explanations exist.

They are available.

She is aware of them.

She moved anyway.

The thing that stays, after everything else is examined and filed and partially explained, is the question she asked.

What are you doing?

Not a cry. Not a startled sound. A question — directed, coherent, the kind of question you ask when you have noticed something and want to understand it. The kind of question that presupposes a presence capable of answering.

Some part of her, sleeping, in the dark, in a locked apartment, registered something in the room and asked it to account for itself.

And it answered.

And she went back to sleep.

And in the morning she woke up and made coffee and opened her phone and found out that any of it had happened.

There are things the sleeping mind perceives that the waking mind does not have access to. This is not controversial — the neuroscience of sleep is full of evidence that consciousness during sleep processes information from the environment in ways that don’t transfer to waking memory. The brain, during sleep, is not off. It is differently on.

What it was on to, in that bedroom, at 2:04 AM on December 30th, 2013 — that is the question that does not have an answer she has been able to locate.

She asked the room what it was doing.

The room, or something in it, said nothing.

It may have been telling the truth.

Or it may have been the most precise lie available.

Either way, she was already going back to sleep.

And whatever was there stayed until morning.

Or didn’t.

She has no way of knowing.

That is the part she thinks about.

👇👇👇

Full story · 13 min read · Dante Darkside

“Some part of her sleeping brain registered something in the room and asked it to account for itself. And it answered. I need to lie down.” — Reader, New York NY “The speech specialist said it was physiologically impossible for her or her son to have made that voice. In a locked apartment. At 2 AM.” — Reader, London “She kept recording after that night. The clicking continued. No more voices. She found cold comfort in the absence of the voice. I understand why.” — Reader, Chicago IL “The question that stays: what else did she sleep through, in the months before that night? The app had been running since October.” — Reader, Sydney “She moved in the spring. She has not used a sleep tracking app since. The correct decisions, in that order.” — Reader, Toronto