Sleep has been proposed to be a physiological adaptation to conserve energy, but little research has examined this proposed function of sleep in humans. We quantified effects of sleep, sleep deprivation and recovery sleep on whole-body total daily energy expenditure (EE) and on EE during the habitual day and nighttime. We also determined effects of sleep stage during baseline and recovery sleep on EE. Seven healthy participants aged 22 ± 5 years (mean ± s.d.) maintained ∼8 h per night sleep schedules for 1 week before the study and consumed a weight-maintenance diet for 3 days prior to and during the laboratory protocol. Following a habituation night, subjects lived in a whole-room indirect calorimeter for 3 days. The first 24 h served as baseline – 16 h wakefulness, 8 h scheduled sleep – and this was followed by 40 h sleep deprivation and 8 h scheduled recovery sleep. Findings show that, compared to baseline, 24 h EE was significantly increased by ∼7% during the first 24 h of sleep deprivation and was significantly decreased by ∼5% during recovery, which included hours awake 25–40 and 8 h recovery sleep. During the night time, EE was significantly increased by ∼32% on the sleep deprivation night and significantly decreased by ∼4% during recovery sleep compared to baseline. Small differences in EE were observed among sleep stages, but wakefulness during the sleep episode was associated with increased energy expenditure. These findings provide support for the hypothesis that sleep conserves energy and that sleep deprivation increases total daily EE in humans.

Caffeine’s wakefulness-promoting and sleep-disrupting effects are well established, yet whether caffeine affects human circadian timing is unknown. Here we show that evening caffeine consumption delays the human circadian melatonin rhythm in vivo, and chronic application of caffeine lengthens the circadian period of molecular oscillations in vitro primarily via an adenosine receptor/cyclic AMP-dependent mechanism. In a double-blind, placebo controlled, ~49-day long within-subject study, we found the equivalent amount of caffeine as that in a double espresso 3 hours before habitual bedtime induced a phase delay of the circadian melatonin rhythm in humans by ~40 minutes. This magnitude of delay was nearly half of the magnitude of the phase-delaying response induced by exposure to 3-hours of evening bright-light (~3000 lux; ~7 Watts/m2) that began at habitual bedtime. Furthermore, using human osteosarcoma U2OS cells expressing clock gene luciferase reporters, we found a dose-dependent lengthening of circadian period by caffeine. By pharmacological dissection and siRNA knockdown we established that perturbation of adenosine receptor signaling, but not ryanodine receptor or phosphodiesterase activity, is sufficient to account for caffeine’s effects on cellular timekeeping. We also used a cyclic AMP biosensor to show that caffeine increased cyclic AMP levels, indicating that caffeine can influence a core component of the cellular circadian clock. Taken together, our findings demonstrate that caffeine influences human circadian timing and gives new insight into how the world’s most widely consumed psychoactive drug impacts upon human physiology.

Multisynaptic neural and endocrine pathways from the suprachiasmatic nucleus of the hypothalamus have been hypothesized to communicate circadian and photic information to the adrenal glands. In humans, light exposure has been reported to have no effect, increase, or decrease cortisol levels. These inconsistent findings in humans may be related to differences among studies including the intensity (∼500 to 5500 lux), duration (15 min to 4 h), and circadian phase of light exposure. The authors assessed the influence of exposure to bright light on cortisol levels in humans during the rising and descending phases of the circadian rhythm of cortisol, that is, when cortisol levels are high. Twenty healthy men and women were studied using a within-subject research design. Subjects were studied in an environment free of time cues for 9 to 10 days. Subjects received a 6.7-h exposure of bright light (∼10,000 lux; equivalent to ambient light intensity just after sunrise or just before sunset) or dim light (∼3 lux; equivalent to candlelight) during the biological night and morning. Bright light exposure significantly reduced plasma cortisol levels at both circadian phases studied, whereas dim light exposure had little effect on cortisol levels. The finding of an acute suppressive effect of bright light exposure on cortisol levels supports the existence of a mechanism by which photic information can acutely influence the human adrenal glands.

To date, no detailed examination of the pattern of change in reaction time performance for different sensory modalities has been conducted across the circadian cycle during sleep deprivation. Therefore, we compared sustained auditory and visual attention performance during 40h of sleep deprivation assessing multiple metrics of auditory and visual psychomotor vigilance tasks (PVT). Forty healthy participants (14 women) aged 30.8±8.6 years were studied. Subjects were scheduled for an ~8h sleep schedule at home prior to three to six laboratory baseline days with 8h sleep schedule followed by 40h sleep deprivation. Visual and auditory PVTs were ten minutes in duration and were administered every 2h during 40h of sleep deprivation. Data were analyzed with mixed model ANOVA. Sleep deprivation and circadian phase increased response speed, lapses, anticipations, standard deviation of response times and time on task decrements for visual and auditory PVTs. In general, auditory vigilance was faster and less variable than visual vigilance with larger differences between auditory and visual PVT during sleep deprivation versus baseline. Failures to respond to stimuli within 10 seconds were 4 times more likely to occur to visual versus auditory stimuli. Our findings highlight that lapses during sleep deprivation are more than just long responses due to eye closure or visual distraction. Furthermore, our findings imply that the general pattern of change in attention during sleep deprivation (e.g., circadian variation, response slowing, lapsing and anticipations, time on task decrements and state instability) is similar among sensory-motor behavioral response modalities.

Information from light and melatonin appear to be combined by the human circadian clock. The ability to combine circadian time cues has important implications for understanding fundamental physiological principles of the human circadian timing system. Knowledge of such principles is important for designing effective countermeasures for phase-shifting the human circadian clock to adapt to jet lag, shift work, and for designing effective treatments for circadian sleep-wakefulness disorders.

Settings such as lending and policing can be modeled by a centralized agent allocating a scarce resource (e.g. loans or police officers) amongst several groups, in order to maximize some objective (e.g. loans given that are repaid, or criminals that are apprehended). Often in such problems fairness is also a concern. One natural notion of fairness, based on general principles of equality of opportunity, asks that conditional on an individual being a candidate for the resource in question, the probability of actually receiving it is approximately independent of the individual's group. For example, in lending this would mean that equally creditworthy individuals in different racial groups have roughly equal chances of receiving a loan. In policing it would mean that two individuals committing the same crime in different districts would have roughly equal chances of being arrested.In this paper, we formalize this general notion of fairness for allocation problems and investigate its algorithmic consequences. Our main technical results include an efficient learning algorithm that converges to an optimal fair allocation even when the allocator does not know the frequency of candidates (i.e. creditworthy individuals or criminals) in each group. This algorithm operates in a censored feedback model in which only the number of candidates who received the resource in a given allocation can be observed, rather than the true number of candidates in each group. This models the fact that we do not learn the creditworthiness of individuals we do not give loans to and do not learn about crimes committed if the police presence in a district is low.As an application of our framework and algorithm, we consider the predictive policing problem, in which the resource being allocated to each group is the number of police officers assigned to each district. The learning algorithm is trained on arrest data gathered from its own deployments on previous days, resulting in a potential feedback loop that our algorithm provably overcomes. In this case, the fairness constraint asks that the probability that an individual who has committed a crime is arrested should be independent of the district in which they live. We empirically investigate the performance of our learning algorithm on the Philadelphia Crime Incidents dataset.

No abstract

No abstract