My 2 a.m. habit

For decades I’ve woken up in the middle of the night. Wide awake — not groggy, not drifting. I’d get up and walk around the house. Overseas, working, it was the same thing, except I’d walk the compound instead. Almost every time, I ended up in the kitchen.

I snacked for a reason. I’d read that eating fat and protein before bed helps you stay asleep, so I’d go for cottage cheese or a slow-release casein shake. I’d also read that waking up in the night is caused by low blood sugar, so other nights I’d eat a bowl of cereal — chase the sugar dump, ride it into what I thought of as a sugar coma, and use that to force myself back down. I did exactly that last night. And like almost every time, I woke up this morning feeling like garbage.

For years I chalked the whole pattern up to my time in the Army — irregular sleep schedules, sleeping in combat zones with generators running and explosive wake-up calls, the kind of conditioning that doesn’t just switch off when you retire. That’s a real factor. But it didn’t explain why the pattern never left, decades later, in my own bed, with nothing going off outside.

So I went looking for the why. Here’s what the research actually says.

The historian who found it

Roger Ekirch, a professor of history at Virginia Tech, wasn’t looking for a sleep pattern. He was researching nighttime life in preindustrial Europe when he came across a 1699 English court deposition from a nine-year-old girl named Jane Rowth. Testifying about the night her mother was murdered, Rowth described waking with her mother around 11 p.m. — after their “first sleep” — before her mother left with two men who arrived at the window. Her mother never returned; the case was never solved.

The phrase caught Ekirch’s attention because the court didn’t ask Rowth to explain it. Everyone in the room already understood what “first sleep” meant.

That sent him digging. His 2005 book, At Day’s Close: Night in Times Past, cited more than 500 references to a two-phase sleep pattern drawn from diaries, court records, medical texts, and literature. He has since said he’s found over 2,000 references across a dozen languages, some going back to ancient Greece.

A sample of what he found:

Chaucer, The Canterbury Tales, “The Squire’s Tale” (late 1300s): the character Canacee “slept her first sleep, and then awoke.” The line is used as an ordinary plot detail, not an oddity requiring explanation.
Laurent Joubert, 16th-century French physician: in a medical manual, Joubert advised couples that the best time for intercourse was not at the end of a long day but “after the first sleep,” when they “have more enjoyment” and “do it better.”
Classical antecedents: Ekirch and other researchers trace the concept back further still — the Romans had a term for it, primus somnus (“first sleep”), and Thucydides used a Greek equivalent in the 5th century BCE. This suggests the pattern (or at least the vocabulary for it) predates medieval Europe by well over a thousand years.

Ekirch’s own explanation for why the pattern disappeared: the spread of street lighting and, later, electric light pushed bedtimes later, compressed the two segments into one, and made a full night’s uninterrupted sleep both possible and socially expected.

It’s not just historical — it shows up in the lab

In the early 1990s, psychiatrist Thomas Wehr, then at the National Institute of Mental Health, ran a controlled experiment on this question. He placed healthy adult volunteers in a setting with 14 hours of darkness a day (roughly a natural winter night, with no artificial light) for a month, and let them sleep as much as they wanted.

By the fourth week, a consistent pattern emerged: subjects slept in two blocks of about three to five hours each, separated by one to three hours of quiet wakefulness, for roughly eight hours of total sleep. Wehr published the results in the Journal of Sleep Research in 1992 (“In short photoperiods, human sleep is biphasic,” J Sleep Res, 1:103–107). It’s one of the more frequently cited studies in the sleep literature and is generally treated as independent, biological corroboration of what Ekirch found in the historical record — extended darkness, on its own, nudges human sleep toward two segments rather than one.

The part that doesn’t get repeated as often: this is disputed

Ekirch’s segmented-sleep theory is widely cited, but it isn’t uncontested, and a fact-based account should say so plainly.

In 2023, researcher Niall Boyce published a peer-reviewed reconsideration in the journal Medical History (“Have we lost sleep? A reconsideration of segmented sleep in early modern England,” Medical History, 67(2), 91–108). Boyce went back through a number of the primary sources Ekirch cited and argued that several of them are more ambiguous than Ekirch’s reading suggests — that phrases translated or interpreted as “first sleep” and “second sleep” don’t always clearly refer to two literal periods of sleep separated by wakefulness. Boyce didn’t claim segmented sleep never existed; his argument was narrower — that the evidence for it being the dominant, near-universal pattern in early modern England is thinner than commonly presented.

Ekirch responded directly in a 2024 article in the same journal (“Reflections on ‘Have we lost sleep?'”), defending his original readings and arguing Boyce’s critique rests on a narrow slice of his research. Boyce published a further reply the same year.

Bottom line: the existence of “first sleep” and “second sleep” as terms in the historical record is not in dispute — that’s textual fact. How universal the underlying sleep pattern actually was, and how confidently we can read every citation as literal, is an active, unresolved academic argument between specialists. Anyone citing this history as settled fact is skipping over that.

Where “insomnia” as a diagnosis actually starts

Separate from the history, there’s a defined clinical threshold. The DSM-5 (and its current text revision, DSM-5-TR) sets out specific criteria for Insomnia Disorder. In summary, the diagnosis requires:

Difficulty falling asleep, staying asleep, or early waking with inability to return to sleep
Occurring at least 3 nights per week
Present for at least 3 months
Causing clinically significant distress or impairment in daily functioning
Occurring despite adequate opportunity and circumstances for sleep
Not better explained by another sleep, medical, or mental disorder, or by substance use

A single, predictable nighttime awakening that resolves on its own and doesn’t come with daytime impairment does not, by this definition, meet the bar for a diagnosis. That’s not an opinion — it’s the literal wording of the criteria as published by the American Psychiatric Association.

What this leaves you with

Three separate, verifiable things:

Historical texts across thirteen centuries and several languages use “first sleep” and “second sleep” as ordinary, unremarked-upon terms — strong evidence the pattern existed and was familiar, at least to some populations, in some eras.
A controlled 1992 lab study found that removing artificial light pushes human sleep toward that same two-segment pattern within about a month.
The clinical definition of insomnia has specific frequency, duration, and impairment thresholds — a single nightly awakening, on its own, doesn’t automatically clear them.

What the record does not establish: that segmented sleep was universal, that everyone should expect or aim for it, or any claim about industry motives behind how insomnia gets diagnosed today — that last one wasn’t in the sources and isn’t something I could verify, so it’s left out here rather than asserted.

Where that leaves me

I still don’t know whether my own pattern is “first sleep” in Ekirch’s sense, a habit built by decades of irregular military schedules, or some combination of both — the research doesn’t let me claim more certainty than that. What it does tell me is that waking once a night, on its own, isn’t automatically a disorder by any clinical definition on the books. Whether the fat, the protein, or the sugar bowl are actually doing anything useful for me, versus just being a ritual I’ve built around the waking, is a separate question — one worth asking a doctor rather than guessing at.

Primary sources for further reading: A. Roger Ekirch, At Day’s Close: Night in Times Past (W.W. Norton, 2005); Ekirch, “Sleep We Have Lost: Pre-Industrial Slumber in the British Isles,” American Historical Review, 2001; Thomas Wehr, “In short photoperiods, human sleep is biphasic,” Journal of Sleep Research, 1992; Niall Boyce, “Have we lost sleep? A reconsideration of segmented sleep in early modern England,” Medical History, 67(2), 2023; American Psychiatric Association, DSM-5-TR, Insomnia Disorder criteria.