Tubbs et al. — The Mind After Midnight: Nocturnal Wakefulness, Behavioral Dysregulation, and Psychopathology (2022)

Andrew S. Tubbs, Fabian-Xosé Fernandez, Michael A. Grandner, Michael L. Perlis and Elizabeth B. Klerman, Frontiers in Network Physiology vol. 1, published 3 March 2022, DOI 10.3389/fnetp.2021.830338. A mini review, which is the first thing to say about it: the paper assembles other people's results into a hypothesis and reports no new data of its own. Fetched and read in full after the ScienceAlert piece arrived by Telegram.

Four of the five authors were at the university-of-arizona (the Sleep and Health Research Program in the psychiatry department, and the McKnight Brain Institute); Perlis was at Penn, Klerman at Massachusetts General Hospital and Harvard Medical School.

The claim in one sentence

The authors' own: nocturnal-wakefulness-associated changes in attentional bias, affective regulation, reward processing and executive function "set the stage for dysregulated behaviors and psychiatric disorders." Not that tired people make worse decisions — that being awake at the wrong circadian hour is itself a risk state, separable in principle from sleep debt.

The four pieces they stack

Affect has a daily shape. Positive affect peaks in the day and bottoms out 1am–4am. Negative affect holds flat through the day and then rises sharply at night. So the two curves do not merely converge at night; they cross.

Reward processing moves with circadian phase. Striatal reward activation is greatest in the afternoon. Sleep deprivation produces greater ventral striatum activation alongside reduced anterior insula activity — more anticipation, less of the signal that would price the risk. The authors' summary of the resulting state is an "aberrant reward system" with elevated dopamine anticipation and a blunted response to actually receiving the reward.

Inhibition falls as the circadian night proceeds. Behavioral inhibition is maximal during the circadian day and declines into the night; working memory, complex attention, problem solving and response inhibition all deteriorate. Sleep inertia makes abrupt nighttime awakenings worse than continuous wakefulness in this respect.

The molecular floor drops out. Under the synaptic homeostasis hypothesis, continuous wakefulness saturates synapses and degrades signal transmission. dopamine peaks in the latter half of the dark period, and sleep deprivation reduces available D2 (inhibitory) receptors in the ventral striatum — the same receptor lewis-borrelli-reward-2021 tracks. An adrenergic stress surge further weakens prefrontal activity.

Put together: an overactive amygdala assigning too much emotional salience, a reward system that wants more and gets less from it, and a prefrontal cortex too degraded to veto either — with attention narrowed onto the negative and rumination running unchecked.

The behavioural evidence they cite

The methodological point worth keeping

Raw counts by hour are the wrong instrument, and the authors say so: "unadjusted counts are not appropriate to evaluate time-of-day differences in behavioral risk," because the size of the awake population is itself the biggest daily rhythm in the data. Adjusting for it is what turns a nighttime deficit in raw suicide counts into a threefold excess in risk. Every figure in the table above that this corpus can lean on is one where that adjustment was done.

The opioid line is the cleanest case: overdose risk rises 4.7-fold at night while heroin use falls, so the exposure and the outcome move in opposite directions and the raw count would have hidden it.

What the authors say is missing

They are unusually direct about it, which is the main reason this page tiers T1 on a paper that measured nothing:

• "All aspects of this hypothesis require empirical validation."
• No study has examined the interaction of sleep deprivation and circadian timing on reward processing. This is the load-bearing mechanism and nobody has tested it. (Re-checked 2026-09-28: still open. Kakoschke, Dickinson and Drummond (2025, doi:10.1080/21642850.2025.2520838) found steeper delay discounting at 4 am than at 4 pm, but they ran sleep restriction and time of day as two separate studies, so the interaction is still untested.)
• Most executive-function studies assessed subjects during the day, not during the night of the sleep deprivation — so the deficits being attributed to the circadian night were mostly measured at the wrong hour.
• The effects of prolonged wakefulness, abrupt nighttime awakening, sleep loss and circadian phase are not separated in the existing literature, and the hypothesis needs them separated.
• Data should be collected during the biological night, "when possible, using protocols that do not cause sleep loss" — the design constraint that makes the question hard.

How this spoke should hold it

A hypothesis paper whose authors list the experiments that would test it, and state that none of them has been done. That is a different object from a finding, and the corpus already has a vocabulary for the difference — see replication-crisis for the case where a literature was tested. The honest summary is that the behavioural epidemiology is real and adjusted, the mechanism is assembled from studies done at other hours for other reasons, and the paper says so.

The follow-up matters for exactly that reason: mind-after-midnight-nimh-grant records the five-year NIMH-funded study built to run the missing experiment.

Related

mind-after-midnight-hypothesis · mind-after-midnight-sciencealert · mind-after-midnight-nimh-grant · circadian-rhythm · reward-system · dopamine · lewis-borrelli-reward-2021 · synthesis