Sleep Architecture

Sleep architecture is what a night looks like when it is scored rather than merely counted. Polysomnography records electroencephalography, eye movements and chin muscle tone, and epochs of thirty seconds are assigned to wake, N1, N2, N3 or REM under standard scoring criteria. Those stages cycle roughly every ninety minutes, four to six times a night, and the composition of the cycles changes through the night: deep N3 dominates the first third, while REM periods lengthen toward morning. The derived metrics are total sleep time, sleep efficiency as the ratio of sleep to time in bed, sleep onset latency, wake after sleep onset, REM latency and the percentage of the night in each stage.

In a healthy adult, N1 accounts for only a few percent, N2 for roughly half, N3 for perhaps fifteen to twenty percent and REM for about a fifth to a quarter, with sleep efficiency above roughly eighty-five percent. Those proportions shift with age, most visibly as a progressive loss of N3. Drugs redistribute them: benzodiazepines and z-drugs increase N2 spindle activity while suppressing N3, and orexin receptor antagonists tend to increase REM time.

That separation is the reason architecture is reported separately from duration. A hypnotic can extend total sleep time while shifting the night toward lighter stages, so total sleep time alone cannot tell you whether sleep improved. Conversely a drug that consolidates sleep may leave total time unchanged and still be doing something worthwhile.

The practical trap is consumer wearables. Their summary screens report deep and REM minutes derived from movement and heart rate, and epoch-level agreement with polysomnography for individual stages is poor even where total sleep time is reasonable. Screenshots of those numbers are not stage data, whatever the label on the graph says.