Alma was asleep by nine.
Maren had picked her up from David's at six one of her nights, a Wednesday, their custody arrangement parceled out in a spreadsheet that assigned parenthood with the clinical precision of a dosing schedule. They'd eaten mac and cheese from a box because Alma preferred it to anything Maren made from scratch, a preference that Maren accepted with the resignation of a woman who could restructure hippocampal architecture but could not compete with powdered cheddar.
Bath. Teeth. Story Charlotte's Web, chapter eleven, read in a voice that Maren modulated carefully because Alma was afraid of the part where Charlotte talks about death, and Maren knew that reading to a child was a form of translation, turning the unbearable into the survivable one sentence at a time.
"Mama?"
"Mm."
"Do spiders really die after they make the egg sac?"
"Some of them do, bug."
"That's sad."
"It is. But the babies carry her forward. All the things she knew, all the things she was they keep going."
Alma considered this with the seriousness of a child for whom abstraction is not abstract. Then she rolled over, pulled the blanket to her chin, and said: "I'm going to name all my spiders Charlotte."
Maren kissed her forehead. Turned off the light. Stood in the doorway for a count of ten her own ritual, unshared and then went to the kitchen table, opened her laptop, and plugged in the USB drive.
The drive contained a single folder labeled P0.
Inside the folder: fourteen files, each one a polysomnography recording in a format Maren recognized immediately the PSG-4 standard, a data architecture used by the open-source sleep research platform SleepStack. This was not the format Sollen used. Sollen's proprietary system, which Maren had helped design, recorded in PSG-7 with embedded metadata layers and AES-256 encryption. These files were raw. Unencrypted. And they were in a legacy format that SleepStack had deprecated three years ago.
Three years ago. Before the Phase III trial. Before Phase II. Before Maren had joined Sollen.
She opened the first file: P0-001.psg4.
The recording showed a standard overnight polysomnography session eight hours of sleep data, four channels (EEG, EOG, EMG, ECG), sampled at 256 Hz. The subject was unlabeled, identified only by the code P0-001. The session date was March 14, two years and eight months ago.
Maren scrolled through the data with the efficiency of someone who had read thousands of polysomnography recordings in her career. The sleep architecture was unusual but not alarming a slightly elongated slow-wave phase in the first cycle, irregular REM onset in the second. The kind of profile she'd associate with a subject under mild pharmacological influence.
She opened the second file. P0-002. Same format. Same approximate date March 15. Similar architecture.
She opened six more. The dates clustered in a two-week window, March 12 through March 26. All fourteen subjects P0-001 through P0-014 had been recorded during this period.
She checked the file metadata. The originating institution field was blank. The researcher ID was blank. The consent-form reference was blank.
This was a drug trial. Fourteen subjects. Two weeks. No institutional attribution. No consent documentation. And it predated Maren's involvement with Sollen by six months.
She went back to the first file and ran a spectral analysis on the EEG channel the same analysis she'd run on her Phase III subjects earlier that day, looking for artifacts. In the Phase III data, she'd found nothing.
In this file, she found something.
Beneath the surface layer of normal sleep data the delta waves, the sleep spindles, the K-complexes there was a second signal. Faint. Regular. A pulse at 6 Hz, embedded in the EEG trace like a watermark in paper, invisible unless you knew to look for it. It appeared during the deepest point of slow-wave sleep and persisted for exactly forty-two seconds before fading. It returned every ninety minutes, synchronized to the subject's REM cycle.
Maren stared at it.
She was, in this moment, intensely aware of two things: the sound of Alma's breathing coming through the wall slow, even, the breathing of a child who slept the way children were supposed to sleep, without assistance, without pharmacology, without fear and the fact that the 6 Hz signal in this file was not a recording artifact. It was too regular, too precisely timed, too cleanly embedded. It was artificial.
Someone had layered a secondary signal into the sleep data during the recording session. Which meant someone had been broadcasting a signal into the subject's brain while they slept.
She opened the remaining files. All fourteen showed the same 6 Hz embedded signal. Same timing. Same duration. Same precision.
She pulled back. Hands off the keyboard. The apartment was quiet around her the radiator, the street, Alma's breathing, the low hum of the refrigerator. She sat in the pool of laptop light and thought.
Phase 0.
That was what the P0 designation meant. Phase 0 a preclinical human study, the kind of thing that existed in a regulatory grey zone. Small sample, limited scope, sometimes run before formal IND filing, sometimes run to generate data that would never appear in an FDA submission. Pharmaceutical companies did them. Not always. Not officially. Not with the kind of rigorous oversight that later phases required.
Sollen had run a Phase 0 trial two years and eight months ago. Fourteen subjects. An unknown drug or perhaps, Maren thought with a coldness that started in her sternum and radiated outward, a known drug. Her drug. An early version of Somniplex-R, tested before she was hired, tested before she was told the compound existed.
And during that trial, someone had broadcast a secondary signal into the subjects' brains while they slept.
She opened file P0-014. The last subject.
The session date was March 26. A Saturday.
Maren checked her own calendar. She kept meticulous digital records every appointment, every meeting, every weekend obligation in a color-coded system that David used to mock gently and now, post-divorce, mocked not at all because it was part of the custody documentation.
March 26. A Saturday. Nineteen months ago.
She searched her calendar. The entry for that Saturday read: Alma aquarium, 10 a.m.
She remembered this. She remembered it clearly. Alma in her yellow boots, pointing at the beluga tank. The blue underwater light on her daughter's face. The whale turning, slow and ancient, and Alma pressing her nose to the glass and whispering hello as though the whale might hear. Maren could feel the temperature of the aquarium cool, slightly damp, smelling of salt water and hot dogs from the concession stand. She could hear the ambient sound children's voices rising and falling like a tide. She could see the whale.
She remembered this the way she remembered her own name: completely, without doubt.
She picked up her phone and opened the photo gallery. She searched by date: March 26.
There were no photos.
She scrolled backward, forward. She had photos from March 24 a picture of Alma eating a granola bar in the kitchen, no context, the kind of snapshot parents take because the light is good and the child is still. She had photos from March 28 a screenshot of a journal article, because Maren used her camera roll as a filing system. But March 26 was empty.
She put the phone down.
She picked it up again.
She opened the text chain with David and searched for "aquarium." One result: a thread from fourteen months ago in which she'd mentioned taking Alma to the aquarium as a suggestion for a birthday outing. David had replied: she loved it last time. Maren had written: I know, she talked about the beluga for weeks.
This exchange confirmed the memory. Except that the exchange took place five months after March 26, and the last time David referenced could have been any time. There was no message from March 26. No text saying heading to the aquarium now. No photo sent to David of Alma with the whale. Nothing that pinned the memory to that specific date.
Maren sat very still.
She was a scientist. She dealt in evidence. And the evidence in front of her said: on March 26, the day P0-014 was recorded, she had a detailed, vivid, sensory-rich memory of an aquarium visit with her daughter and no external evidence that it had occurred.
This could mean several things. She might have the date wrong. People misremember dates constantly; it was one of the most documented failures of autobiographical memory. She might have gone to the aquarium on a different Saturday and her brain had simply mis-filed the receipt. That was the rational explanation. That was the explanation she would offer a subject in her own trial who reported temporal confusion about a personal memory.
But.
There was a file on this USB drive labeled P0-014. It was recorded on March 26. It contained a polysomnography session with a 6 Hz artificial signal embedded in the slow-wave phase. And it was nineteen months old exactly the period during which Maren had experienced no anomalies, no strange memories, no gaps or distortions. The period during which she had been, by her own assessment, fine.
She closed the laptop.
She went to Alma's room and stood in the doorway and watched her daughter sleep. The nightlight cast a faint orange glow across the bed a plastic owl, chosen because Alma believed owls were "the wisest animal and also the most dressed-up." Alma's breathing was slow and regular. Her hand was curled around a stuffed rabbit named Clover who had been through the washing machine eleven times and showed it.
Maren leaned against the doorframe and listened to her daughter breathe and felt something she could not quantify a fear that had no data set, no spectral analysis, no incidence report. A fear that lived in her body, not her mind. The specific, located terror of a woman who has just been shown a hairline crack in the foundation of her own consciousness and cannot tell if the crack is new or if it has been there all along and she simply hadn't been given the right instrument to see it.
She went back to the kitchen. She opened the laptop. She opened file P0-014 again and stared at the 6 Hz signal, pulsing in the slow-wave trough like a second heartbeat artificial, precise, patient and she thought: Whose brain is this?
She did not sleep that night.
Not because she couldn't. Because she was afraid to.
