What the Research Says About Sleep

The Most Effective Sleep Intervention Isn't a Pill, a Herb, or a Cold Plunge — It's a Set of Boring Rules That Cut Time to Fall Asleep by 16 Minutes

Here’s the number that should stop you from buying another sleep supplement: a meta-analysis of 11 randomized controlled trials found that stimulus control — a behavioral protocol that essentially tells you to stop using your bed for anything except sleep — reduced sleep onset latency (SOL) by an average of 16 minutes compared to doing nothing. That’s a larger effect than most over-the-counter sleep aids, herbal extracts, or expensive gadgets can claim. And unlike those alternatives, the evidence for stimulus control comes from a systematic review with a pooled sample of 363 adults, not a single underpowered pilot study. If you want to sleep better, the most reliable lever you can pull is boring, free, and uncomfortable: get out of bed when you can’t sleep.

What the research actually shows

The sleep intervention with the strongest evidence base isn’t a drug or a device — it’s cognitive behavioral therapy for insomnia (CBT-I), and its most potent component appears to be stimulus control. The meta-analysis on stimulus control found that, compared to waitlist or no-treatment controls, the intervention produced a standardized mean difference of -0.83 for sleep onset latency — a large effect by conventional benchmarks. In real terms, that’s the 16-minute reduction mentioned above. Total sleep time also increased, but the effect was smaller and less consistent across studies.

Digital delivery of CBT-I works almost as well as face-to-face therapy. A separate meta-analysis of 37 RCTs with 13,227 participants found that eHealth-delivered CBT-I (apps, websites, phone sessions) reduced insomnia severity on the Insomnia Severity Index by a mean difference of -4.61 points compared to inactive controls (waitlist, sleep education, usual care). That’s a clinically meaningful change — the threshold for "minimally important difference" on the ISI is typically around 3 points. When compared directly against in-person CBT-I, the digital versions performed roughly equivalently, with no statistically significant difference in insomnia severity outcomes.

What about the popular alternatives? Cold-water immersion — ice baths, cold plunges, cold showers — shows modest effects on mood and fatigue, but the evidence is low-to-moderate quality, and optimal protocols (temperature, duration, frequency) remain unclear. The herbal medicine Suanzaoren (jujube seed) improved self-reported sleep quality on the PSQI in some trials, but the evidence base has serious methodological limitations, and a meaningful subset of participants had cancer-related sleep disturbances, limiting generalizability to otherwise healthy people. Most drug treatments for sleep in specific clinical populations — including commonly prescribed antidepressants for fibromyalgia patients — showed little to no benefit in a meta-analysis of 40 RCTs totaling over 10,000 participants.

The nuance most people miss

The stimulus control meta-analysis has a dirty secret: 10 of the 11 included trials were published before 1998. The total sample across all studies was only 363 adults — smaller than a single decent-sized modern RCT. And none of the studies formally assessed psychiatric or medical comorbidity using validated structured interviews. So while the effect is real, the evidence base is older and smaller than you’d expect for something that works this well.

More importantly, stimulus control performed no better than other active insomnia treatments when compared head-to-head. Relaxation training, paradoxical intention (trying to stay awake), and imagery training all produced similar improvements. This suggests that the specific mechanism — breaking the association between bed and wakefulness — may matter less than the general effect of doing something structured and consistent.

The digital CBT-I meta-analysis also has limitations worth noting. All outcomes were self-reported; no objective measures like polysomnography or actigraphy were included. And 54% of studies required a formal clinical diagnosis of insomnia, meaning the results may not fully apply to people with mild or occasional sleep problems. The average participant was 46 years old and 71% female — not a perfect match for every reader.

For the cold-water immersion data, the effect on sleep specifically was not the primary focus of the meta-analysis. The review covered mood, fatigue, and quality of life broadly, and the evidence quality was rated low-to-moderate. If you’re considering ice baths for sleep, you’re extrapolating from indirect evidence.

Practical implications

• Try stimulus control before anything else. The protocol is simple: only go to bed when sleepy; use the bed only for sleep (no phones, no reading, no worrying); if you can’t fall asleep within roughly 10 minutes, get out of bed and return only when sleepy again; set a fixed wake time every morning; don’t nap during the day. This is uncomfortable for the first 1–2 weeks, but the meta-analysis shows it works within that timeframe for most people.

• If you can’t do in-person therapy, use a digital CBT-I program. The meta-analysis of 37 RCTs found that app- or web-based CBT-I reduced insomnia severity by 4.61 points on the ISI — comparable to face-to-face therapy. Look for programs that include the core components: stimulus control, sleep restriction, cognitive restructuring, and relaxation training. Avoid apps that only offer sleep tracking or white noise.

• Skip the supplements and cold plunges unless you’re curious. The evidence for Suanzaoren is methodologically weak, and cold-water immersion’s effects on sleep specifically are not well-established. If you want to experiment with these, treat them as secondary interventions, not primary strategies. The one exception: if you have fibromyalgia, CBT-I is the only intervention with moderate-quality evidence for improving sleep quality in that population.

Design your own experiment

What to test: Stimulus control protocol (the full version, not a partial implementation). This is the intervention with the largest effect size and the strongest evidence base.

How long to run it: Minimum 2 weeks, ideally 4 weeks. The meta-analysis included trials ranging from 2 to 8 weeks, and the effects were detectable within the first 2 weeks. If you can’t tolerate the protocol for 2 weeks, it’s probably not for you.

What to measure: Sleep onset latency (SOL) in minutes, measured via a daily sleep diary. Record the time you get into bed and the estimated time you fall asleep. Also record total sleep time (TST) and wake after sleep onset (WASO) if you want a fuller picture. The primary outcome is SOL — that’s where the largest effect was observed.

What confound to watch for: Sleep restriction is often bundled with stimulus control in clinical protocols, but the meta-analysis specifically examined stimulus control as a standalone intervention. If you also start restricting your time in bed, you’re testing a combined intervention, and you won’t know which component is driving the effect. Stick to the stimulus control rules alone for the first 2 weeks, then add sleep restriction if you want to experiment further.

What a positive result looks like: A reduction in SOL of at least 10–15 minutes from your baseline average over the first week. The meta-analysis found a mean reduction of 16 minutes, but individual responses vary. If your SOL drops by less than 10 minutes after 2 weeks, the protocol may not be effective for you, or you may need to check your adherence — the most common reason for failure is not actually getting out of bed when you can’t sleep.