Night and reward
Two ways to strengthen what you practise by day. At night, the day's codes are replayed softly in deep sleep, and only there: below waking level, stopping at the first sign of waking, with a nightly limit. By day, a success (a ball through the rim, a right answer, a clean move) is answered within a fraction of a second by a reward code, which can also return at night.
Tonight's plan
Choose the codes you practised today. Mark any code linked to a painful memory: it is never played during sleep. If a nightmare is detected, the device plays a separate safety code or wakes you gently.
Finding deep sleep needs a sleep headband or another device that reads sleep stages; a phone on the mattress cannot. The plan is the device's instructions, in the RSOS night format.
A night, simulated
A simulated sleeper with a device that sometimes misreads the stage. Compare closed-loop cueing with cueing at random times, on the same night.
Did it help memory
In the evening, learn a list (words of a language, say), test yourself, and split the items into two halves: one half gets codes tonight, the other none. Test both halves in the morning. Over several nights the difference in what you kept is the effect of the night cues.
Night diary
Each morning, note whether cueing ran, how many bad dreams you remember, and how distressing they were. Cueing nights and nights without are compared.
Reward codes
Each learning area has its own family of codes, all starting with the area's prefix, so the body knows the area from the first pulses. A success plays the area's reward code at once: tap the button the moment you succeed, or let the phone's motion sensor recognise a good move.
Is a cue noticed sooner with light
Ten short trials. Keep your eyes on this box; at a random moment a cue comes, as vibration alone or as vibration with a flash of light. Tap as soon as you notice it. (Vibration works on Android phones; elsewhere you will hear a tone instead.)
Across many simulated people
Closed-loop cueing put nearly every cue in deep sleep and rarely woke the sleeper; cueing at random times woke them many times a night and, in the memory model, turned a benefit into a loss. A memory benefit was found in most groups of twenty people, rarely in a single person over a week. Reward codes and night replay are small effects that need large training studies.
Full results: rsso/results/night.md. To run it with real people and devices: the protocol kit.