Food-to-sleep observation set / Source cutoff 28 September 2026

22 days, 22 following nights.

Dinner and later intake, movement modality, reported bedtime and next-morning Garmin outcomes are linked by date. This is a descriptive n=1 record, not an intervention result.

Linked observations

Detailed sleep outcomes

Valid reported bedtimes

Current reading

Twenty-one observations have a Garmin Sleep Score (mean 67.7, median 71, range 45–83). The 21 nights with detailed records averaged 379 minutes (6:19); 17 of 21 were shorter than seven hours. Sleep duration remains the clearest recurring constraint in the active series. These are descriptive device readings, not a diagnosis or personal sleep-need estimate.

The timing question remains unresolved. Among the 17 observations with both an unambiguous final-calorie time and a reported bedtime, the food-to-bed interval had essentially no linear association with Garmin Sleep Score (r = 0.01) or total sleep time (r = 0.02). Reported bedtime is not verified sleep onset, and meal composition, activity and other context vary, so this does not rule out a timing effect under more controlled comparisons.

The September 26 San Jacinto exposure produced the most disrupted multi-metric night in the current detailed series: score 45, total sleep 5:05, awake time 3:53, HRV 24 ms, average overnight heart rate 68 bpm and Body Battery change +9. The following recovery night rebounded to score 71, sleep 6:25, awake time 16 minutes, HRV 37 ms, overnight heart rate 61 bpm and Body Battery +55. This is a recovery-trajectory candidate, not a causal hike effect: the acute night was confounded by an unusual elevated-leg sleeping position, and two other long hikes in this series did not reproduce the crash.

Food observations: hypotheses, not rankings

The meal record is still heterogeneous, so these are descriptive component-level checks, not food effects. Crude means below are unadjusted for activity, bedtime, prior sleep or co-occurring foods. The purpose is to identify what deserves a matched repeat—and what currently looks like a false lead.

Exposure	Current record	Descriptive pattern	Status / next test
Grapes	11 scored nights	Scores 47–83; crude mean 69.6 versus 67.7 overall. High and low outcomes both occur across different activity contexts.	Working hypothesis: at the observed portions, grapes are not a dominant driver of sleep outcome. Continue within-modality comparisons.
Orange / citrus	6 exposures; 5 scored nights	Scores 54–79; crude mean 67.8, essentially the same as the full scored series.	Unresolved / no directional signal. Keep recording final-fruit exposures and compare similar rest or low-load days.
Quinoa-containing meals	12 exposures; 11 scored nights	Scores 47–83; crude mean 68.6. This common meal base spans nearly the full outcome range.	Background candidate: useful as a recurring dietary backdrop rather than a current explanatory variable.
Mooncake	3 exposures; 2 scored nights	The two scored late-mooncake nights were 75 and 83; the newest half-mooncake exposure lacks a Sleep Score.	Counterexample only: the record does not support “late mooncake always worsens sleep,” but it does not support a benefit either.
Pork / char siu	7 exposures; 6 scored nights	Crude scored-night mean is 62.2 versus 69.9 on non-pork nights. But two low pork nights also carried major competing exposures: late popcorn after a hike and the San Jacinto/high-fragmentation night.	False lead for now: removing those two confounded nights raises the pork-night mean to 70.3. Do not treat pork as a current adverse signal; collect matched repeats.
Alcohol	1 natural exposure	The sake night scored 46 with HRV 25 ms, RHR 60 bpm and Body Battery +34, followed by a rebound night.	Single perturbation: strongest food-related one-night disturbance so far, but not a replicated result. Observe natural recurrence only; do not manufacture an exposure.

Current food-layer conclusion: no food has earned a “sleep-promoting” or “sleep-disrupting” label. The most useful early result is narrower: several recurring foods span both good and poor nights, while an apparent pork signal fails a simple sensitivity check. That shifts the next phase toward matched comparisons rather than rankings.

Evidence register

Question	Status	Current evidence	Next informative comparison
Sleep duration	Repeated descriptive pattern	17 of 21 detailed nights under seven hours; mean 6:19.	Capture sleep opportunity and wake time consistently on comparable days.
Food-to-bed timing	Unresolved	17 paired intervals; r=0.01 with Sleep Score and r=0.02 with total sleep.	Accumulate nights with similar activity and meal context but naturally different intervals.
Recurring food components	Mixed / unresolved	Grapes, citrus and quinoa-containing meals span broad score ranges; no repeated component shows a stable directional effect. A crude pork signal fails sensitivity to two confounded low nights.	Use matched activity/timing contexts and repeated component exposures before assigning direction.
High-load recovery	Single observation, confounded	San Jacinto night showed a broad acute disturbance followed by next-night rebound; other long hikes were materially different.	Repeat a high-load hike with normal sleeping position and complete timing data.

Observation ledger

The dates below are the sleep dates. A dash means the value was not recorded, not zero. Observation 1 has a score but no detailed sleep record or bedtime; observation 20 has detailed sleep metrics but no reported bedtime or Sleep Score. Download the 22-row CSV snapshot for food descriptions, documentation quality, activity dose, notes and source values. The master workbook retains additional Garmin fields.

Sleep date	Day activity	Last calories	Bedtime	Score	Sleep h:mm	HRV ms	RHR bpm
09-05	Unrecorded	18:45	—	69	—	—	—
09-06	Hike	19:00	23:00	72	6:06	34	58
09-07	Rest	19:15	23:30	70	7:05	42	54
09-08	Rest	19:10	22:20	81	6:59	44	56
09-09	Rucking	—	22:38	46	5:55	25	60
09-10	Run	18:30	22:38	77	6:07	40	54
09-11	Run	18:30	23:22	54	5:16	46	51
09-12	Hike	18:50	00:10	63	6:04	44	53
09-15	Rest	—	23:48	68	7:50	56	55
09-16	Ruck	18:50	22:51	74	7:45	35	60
09-17	Walk	21:00	22:38	75	6:03	39	57
09-18	Run	19:00	23:25	51	6:02	35	55
09-19	Hike	—	23:08	82	6:46	42	55
09-20	Rest	19:00	22:24	51	6:20	38	55
09-21	Hike	21:00	23:21	47	5:07	39	56
09-22	Walk	19:00	22:41	83	6:40	47	54
09-23	Ruck	18:40	22:53	80	6:39	40	54
09-24	Run	20:30	22:55	83	7:09	40	52
09-25	Rest	18:30	23:15	79	5:40	47	51
09-26	Rest	19:00	—	—	5:36	39	56
09-27	Hike	18:45	21:30	45	5:05	24	58
09-28	Recovery	19:15	22:30	71	6:25	37	55

Questions the next observations should test

Duration and opportunity: when bedtime and wake opportunity are comparable, what repeatedly distinguishes nights above and below seven recorded hours?
Food components: on matched rest or low-load days, do repeated grapes, citrus, protein type or dessert exposures show a stable effect that survives counterexamples?
Timing: does a food-to-bed interval signal emerge within comparable activity and meal contexts rather than across the pooled series?
Movement dose: why did the San Jacinto response differ from other 12–13 mile hikes, and does duration, elevation gain, heat, recovery state or another recorded factor repeat?
Recovery trajectory: does a high-load stimulus repeatedly produce an acute autonomic cost followed by a next-night rebound when sleep conditions are normal?

Sources and limits

Source: Sleepgenic_Food_to_Sleep_Master_Log_2026-09-28.xlsx, Master Log, observations 1–22. Date of this release update: 28 September 2026. Historical comparison: separate Sleep Data.xlsx baseline workbook; it is not added to these 22 observations. Three final caloric cutoffs and two bedtimes are missing; one of those bedtime-missing observations also lacks a Sleep Score. Photo-derived food descriptions can be uncertain and activity dose is incomplete on some days. Bedtime is reported time in bed, not verified sleep onset. Garmin stages are estimates. Correlations are exploratory Pearson correlations across available paired observations, with no missing values imputed.

Read the research method · Read the superseded ten-night interim report.