3 Don’t waste energy

Merely receiving a notification — without looking at or interacting with the phone — noticeably impairs performance on attention tasks. The degree of impairment is comparable to actually answering calls or sending messages. These findings come from laboratory studies involving university students. Two additional experiments also show that even when participants successfully refrain from looking at their phone, having it nearby still reduces their available mental resources; the drop is largest among people strongly dependent on their phones.

61 university students kept records for 30 days. The least regular fifth had a body clock (circadian rhythm) about 2.5 hours later than the most regular fifth. Specifically, the time melatonin began to rise was 00:08 versus 21:32. The stretch when they felt sleepiest was also about 1.8 hours later. The more regular the schedule, the better academic performance (correlation coefficient r = 0.37).

48 healthy adults aged 21 to 38 slept only 6 hours a night for 14 days straight. Their cognitive performance fell cumulatively, to about the level of 2 full nights with no sleep. Self-rated sleepiness rose only in the first few days, then barely rose further. Self-rated sleepiness did not differ between the 6-hour and 4-hour groups. This was a randomized controlled trial (RCT). Separately, short total sleep loss (acute sleep deprivation, no more than 48 hours) had an effect size g = -0.776 on simple attention lapses and g = -0.125 on reasoning accuracy. That part pooled 70 studies.

Taking a tablet with 400 mg of caffeine 6 hours before bed, total sleep time measured by instruments was 1.1 to 1.2 hours less than with a placebo tablet. Counting a 250 mL cup at 107 mg, 400 mg is about three or four cups of coffee. The placebo tablet looked the same and contained no caffeine. The sleep diaries participants wrote themselves reported no difference; they could not tell. Pooling several studies: caffeine cut total sleep time by 45 minutes on average. The share of time in bed that was actually sleep (sleep efficiency) fell by 7%. A 250 mL cup of coffee (107 mg) has to be finished 8.8 hours or more before bed if it is not to affect sleep.

In a two-week self-controlled trial involving 124 adults, those who limited email checks to three times daily experienced significantly lower daily stress than when they checked freely (effect size: Cohen’s d = 0.37). The overall volume of emails received and replied to did not differ between the two conditions. Without restrictions, participants checked emails 15.5 times per day on average.

On average, a 2.8-second interruption doubles the error rate for sequential tasks, while a 4.4-second interruption triples it. This finding comes from lab experiments. Another study observed 24 office workers (information professionals) in real-world settings. They switched topics every 11 minutes on average, and 57% of work topics were interrupted. Getting back to the interrupted task on the same day took 25 minutes and 26 seconds on average, with 2.26 other tasks added in between. A separate experiment with 48 participants showed that interrupted workers completed tasks faster, but reported higher stress, frustration, effort, and time pressure.

Switching between tasks clearly slows reaction time and raises the chance of mistakes. This conclusion comes from a review of multiple studies. One lab study involved university students. Those who frequently use several types of media at the same time are known as heavy media multitaskers. They’re more easily distracted by unrelated things and unrelated memories, and they also perform worse on task-switching tests.

In the lab the same people did both conditions in turn (crossover). One was reading an e-book on an iPad before bed, brightness at maximum, about 4 hours, for 5 nights in a row. The other was reading a paper book in the same dim room. Reading the glowing e-book delayed sleep onset by about 10 minutes versus paper (25.7 minutes versus 15.8 minutes). Evening melatonin was suppressed by 55%. The next day’s body clock (circadian rhythm) delayed by 1.5 hours or more; melatonin onset was 22:31 versus 21:01. The next morning they were sleepier and took longer to become fully awake.

In 2025 China’s residents averaged 6.97 hours of nighttime sleep, with average sleep onset at 0:10 (Chinese Sleep Research Society). People who used portable devices before bed were about 2.17 times as likely to get insufficient sleep (OR 2.17), about 1.46 times as likely to have poor sleep quality (OR 1.46), and about 2.72 times as likely to be sleepy by day (OR 2.72). Devices only left in the bedroom and not used gave similar numbers, about 1.79, 1.53, and 2.27 times (OR 1.79/1.53/2.27). That part pooled 20 same-time surveys of 125198 children and adolescents. Another same-time survey had 844 adults aged 18 to 94. Those who still used a phone after lights out fell asleep more slowly, spent a lower share of time in bed actually asleep, and were more fatigued by day.

34 healthy participants completed the Iowa Gambling Task after going 49.5 hours without sleep. This task is an experiment that measures risk decision making. When well-rested, they quickly learned to steer clear of high-risk options; after sleep loss, they opted for high-risk choices more often. Older participants showed a stronger response to this effect. This is one experiment, and it is the only study.

After a night of insufficient sleep, a 10-minute nap immediately reduces drowsiness, fatigue, and improves energy and mental performance; some of these effects persist beyond 155 minutes. A 5-minute nap yields virtually no benefit, and the advantages of a 20-minute nap appear only after 35 minutes. A 30-minute nap first causes sleep inertia, so performance is actually poorer right after waking, before any improvement occurs.

Students worked in a simulated open office. When the background noise rose from 39 dB to 51 dB (dB, LAeq), they recalled fewer words, rated themselves as more fatigued, and felt less inclined to keep working. Watching nature videos with sound during breaks restored their energy ratings better than listening only to river sounds or office noise.

Data from British munitions factories in World War I demonstrates that output is proportional to working hours up to 49 per week. After 49 hours, the rate of growth slows. Peak output is reached at roughly 63 hours. At 70 hours, output remains essentially the same as at 56 hours, meaning those extra 14 hours are wasted effort.

A review of multiple studies shows that repetitive rumination on distressing events worsens depression. This pattern, known as rumination, also strengthens negative thinking and weakens problem-solving abilities. It hampers real-world action and wears out people around you who are willing to listen. It is linked to anxiety, binge eating, alcohol misuse, and self-harm.

This study followed 2,978 twins aged 50 or older in Australia for an average of 20 years; 1,068 of them died over that time. For each standard deviation increase on the pessimism scale, the relative risk of death rose by about 13% (hazard ratio 1.134). The 95% confidence interval for this figure is 1.065–1.207, with a p-value of 8.85×10⁻⁵. Cardiovascular mortality rose by roughly 20% (hazard ratio 1.196; 95% CI 1.045–1.368; p=0.0093). No link was found between pessimism and cancer mortality. Similarly, higher optimism scores showed no meaningful association with mortality rates.

A survey of 120 widowed women aged 60–89 found that unpleasant interactions showed a stronger and more consistent link to psychological well-being than pleasant ones. By this result, the negative impact of a draining person outweighs the positive effects of a supportive one.

Pooling 306 sets of experiments for comparison (a meta-analysis): re-explaining what happened from another angle (cognitive reappraisal) had an effect on emotion of d = 0.36. Standing in the other person's place had d = 0.45. Shifting attention had d = 0.27. Holding down the outward show of emotion (not letting it show) worked, d = 0.32. Forcing the feeling down did not work (d = -0.04). Forcing yourself not to think about the thing also did not work (d = -0.12). These figures pool inner feeling, outward expression, and bodily response. The abstract does not give them separately. Three other experiments show that people who punished the offender kept turning the other person over for longer afterward, and people who did not punish moved on faster. People overestimate in advance how good revenge will feel.

Pooling 218 randomized controlled trials and 14,170 people (network meta-analysis). The comparison group was people doing other activities. Versus them, walking or jogging had an effect size on depression of g = -0.62 (95% confidence interval -0.80 to -0.45). Yoga -0.55, strength training -0.49. Higher intensity worked better. Pooling several light-therapy studies: seasonal depression effect size 0.84 (0.60 to 1.08), non-seasonal 0.53 (0.18 to 0.89). Another analysis pooled individual-patient data across studies. It looked at internet cognitive behavioral therapy (online self-help courses). Guided and unguided both beat controls. For moderate to severe (PHQ-9 above 9), guided was better. 12356 is the national mental-health assistance hotline, open no fewer than 18 hours a day.

Change how you deal with institutions: from “pleading, getting angry, retelling grievances,” to “taking documents, noting deadlines, appealing through channels.” What you save is time spent in repeat talk and rumination. This is a one-time mindset shift, not daily time saved. What you can check against are the deadlines themselves, all written in the files. For public-order cases (治安案件), counting from the day accepted, the handling period is no more than 30 days; for major or complex cases, with approval from the public-security organ one level up, it may extend 30 days. When a case with a complainant is not filed, the notice of non-filing must be delivered within 3 days. The complainant may apply for reconsideration within 7 days; the public-security organ decides within 30 days. If still unsatisfied, apply for review one level up within another 7 days.

Pooled results of many studies (meta-analysis): 54 independent samples, 94 effect sizes, N = 36,583, using a three-level random-effects model. The average correlation between upward social comparison online and psychological maladjustment was r = 0.330 (95% CI 0.289–0.370). Split by specific outcome: socially evaluative negative emotions such as fear of being judged 0.438, anxiety 0.382, depression 0.306, lower well-being 0.268, lower self-esteem 0.263. Age, cultural background and year of data collection did not change the strength of this link. Results from long-term follow-up studies were no stronger than those from single-time-point surveys. There was also a randomized controlled trial of 2,743 US Facebook users. Of these, 61% were willing to accept deactivation for less than 102 US dollars. They were randomly assigned to a four-week deactivation group or a control group. For the average participant, deactivation freed up 60 minutes a day. The subjective well-being index rose by 0.09 standard deviations. By component: life satisfaction 0.12, anxiety 0.10, depression 0.09, happiness 0.08. After removing what multiple comparisons can produce by chance, these results still counted. The authors gave a reference point: psychological interventions that make people more positive. Such interventions improve subjective well-being by about +0.34 standard deviations and depression by about -0.23 standard deviations. So this effect is about 25% to 40% of that of a usual psychological intervention. The cost over the same period: the news knowledge index fell by 0.19 standard deviations, and political polarization fell by 0.16 standard deviations. After the experiment ended, the deactivation group still used Facebook 12 minutes a day less than the control group (0.31 standard deviations, a drop of 23%). There was also a randomized controlled trial of 143 undergraduates, randomly split into two groups. One group limited each of Facebook, Instagram and Snapchat to 10 minutes a day; the other used them as usual. After three weeks, loneliness and depression fell clearly in the limited group. In both groups, anxiety and fear of missing out were lower than at the start.

Experiment 1: Participants wore a T-shirt with Barry Manilow’s face and walked into a room with other participants. Wearers estimated on average that 46% of bystanders could correctly name who was on the shirt. Bystanders’ actual correct rate was 23%; the estimate was exactly twice the actual. Subtracting actual from estimate room by room, the average overestimate was 23 percentage points (95% CI 9–38). After dropping the two rooms with only two or three bystanders, it was 27 percentage points (11–43). Experiment 2: Switch to a self-chosen T-shirt the wearer saw as positive. Average estimate 48%, bystanders’ actual correct rate 8%, a sixfold gap. In a footnote the authors note wearers’ estimates were almost the same across both experiments (48% and 46%); the gap was bystander accuracy (8% and 23%). They suspect the main reason is that Experiment 2’s bystander questionnaire was harder, not that a decent shirt is easier to overestimate. Experiment 3: Undergraduates in groups of 3 to 7 discussed an assigned topic for 20 minutes, then wrote a policy proposal together. Right after, each person separately estimated how others in the group would rank everyone; lower rank numbers are better. On “how many verbal mistakes,” self-estimated rank averaged 2.30, others’ actual rank 2.96, difference −0.66 (p < .01). So people think their mistakes look more obvious to others. “Possibly offensive remarks” was the same (2.35 versus 2.90, p < .01); “remarks others would pick on” differed less (2.61 versus 2.82, p < .05). Another series had three settings: strangers paired for talk, new college roommates, and an adult workshop. People systematically underestimated how much the other liked them and how much the other enjoyed the talk. The authors call this the “liking gap.” It shows up in talks of different lengths and lasts months among new roommates.

Pooling several studies (meta-analysis): 45 studies, N = 11747, samples of undergraduates, medical students, community adults, and psychiatric patients, random-effects model. Below are weighted average correlations with suicidal thoughts. Perfectionistic concerns r = .28 (95% CI .24–.32; 42 studies, 7936 people). Perfectionistic concerns is a bundle including socially prescribed perfectionism, concern over mistakes, doubts about actions, and gaps between reality and standards. Socially prescribed perfectionism alone is also r = .28 (.25–.32, 30 studies, 3640 people). Perfectionistic striving r = .10 (.07–.13, 31 studies, 4588 people), a bundle of self-oriented perfectionism and personal standards. Parental criticism r = .20 (.11–.29). Then links to suicide attempts. Perfectionistic concerns r = .12 (.07–.17, 15 studies, 5275 people). Socially prescribed perfectionism r = .19 (.08–.29, 8 studies, 689 people). Then the follow-up part. After subtracting suicidal thoughts present at the start, socially prescribed perfectionism still moved in the same direction as later follow-up suicidal thoughts, modestly. Self-oriented and other-oriented perfectionism showed no link.

Pooled results of 154 studies (meta-analysis), 184 samples, 10,189 people. Methods that calm the body, such as deep breathing, mindfulness and meditation, reduce anger and aggression. Effect size g = -0.63, 95% CI -0.82 to -0.43. This result was stable across people of different sex, race, age and culture. Methods that raise bodily arousal, such as punching a bag, jogging and cycling, were ineffective overall. Effect size g = -0.02, 95% CI -0.13 to 0.09. The authors conclude that venting anger and going for a run are not effective ways to reduce anger.

One analysis pooled 146 randomized studies (meta-analysis). People wrote or spoke about something important to them. The average effect was positive and significant, but small, effect size r = .075. Another pool took only randomized controlled trials; 30 trials together showed no significant effect on physical or mental health. Sixteen randomized trials in people with cancer and survivors, pooled, were non-significant on psychological, physical, and quality-of-life outcomes. On the psychological side g = 0.04, 95% CI -0.06 to 0.14.

A Cochrane systematic review pooled 80 randomized controlled trials and 5820 women with primary dysmenorrhea. Primary dysmenorrhea means period pain with no other disease found. Versus placebo, nonsteroidal anti-inflammatory drugs relieved pain better (OR 4.37, 95% CI 3.76–5.09). Converted, 18% of people on placebo got moderate or better relief, versus 45% to 53% on these drugs. The tradeoff is more side effects (OR 1.29, 1.11–1.51). 10% of people on placebo had side effects, versus 11% to 14% on these drugs. The extras were mainly stomach upset (OR 1.58) and nervous-system reactions (OR 2.74). Versus acetaminophen, these drugs also relieved pain better (OR 1.89, 1.05–3.43), though only 3 trials. Which of these drugs is best showed no difference. The UK NICE guideline (updated 2024) lists situations that should raise suspicion of endometriosis. They are long-term pelvic pain, period pain that affects daily activity and quality of life, and deep pain during or after sex. Also pain on bowel movements during periods, blood in urine or painful urination during periods, and infertility with these symptoms. Girls aged 17 and under are included. The guideline advises a trial of acetaminophen or one of these drugs first, for example 3 months. If pain relief is not enough, refer for further checks. A normal pelvic exam and ultrasound do not rule the disease out. One study surveyed 1418 women who had laparoscopy at 16 hospitals in 10 countries. From first symptoms to surgical diagnosis of endometriosis averaged 6.7 years, mostly lost at the primary-care stage. Women with a diagnosis lost an average of 10.8 hours of work a week, mostly presenteeism with lower efficiency. A Dutch online survey asked 32748 women aged 15 to 45. 13.8% had taken leave for menstruation, averaging 1.3 days a year. 80.7% were at work during periods with lower efficiency, equal to 8.9 days of lost work a year.

The National Health Commission’s 2020 clinical norms, following the international classification, list the diagnostic points for insomnia disorder. One is trouble falling asleep, easy waking after sleep, or early waking. Two is daytime fatigue, sleepiness, and impaired function. Three is these at least 3 times a week for at least 3 months. Under 3 months is called short-term insomnia disorder. The norms set cognitive behavioral therapy for insomnia as first-line. It includes sleep restriction, stimulus control, cognitive therapy, relaxation training, and sleep hygiene — 5 parts — usually one cycle of 6 to 8 weeks. The norms say short-term effects match medication, and long-term outlook is better than medication. They also state sleeping pills generally should not continue past 4 weeks; if continued, reassess monthly. The American College of Physicians 2016 guideline strongly recommends: for adults with chronic insomnia, start with this therapy. If it alone is not enough, then discuss short-term medication with a doctor. One analysis pooled 20 randomized controlled trials and 1162 people (meta-analysis). After treatment, time to fall asleep shortened by 19.03 minutes (95% CI 14.12–23.93). Time awake after falling asleep fell by 26.00 minutes (15.48–36.52). The share of time in bed actually asleep (sleep efficiency) rose by 9.91% (8.09%–11.73%). Total sleep time rose only 7.61 minutes, not statistically significant (−0.51 to 15.74). No study reported adverse outcomes. Another analysis pooled 30 randomized controlled trials on how long effects last. Versus no treatment, improvement in insomnia severity shrank over time. At 3 months g = 0.64, at 6 months 0.40, at 12 months 0.25. Effects weaken gradually, but the authors still see clinical meaning within a year. For online delivery, one analysis pooled 11 randomized controlled trials and 1460 people. All sleep measures improved, effect sizes g from 0.21 to 1.09. The authors say this matches in-person effects; longer courses and more follow-up give larger effects. Another online randomized trial of 1711 people gave the control group only sleep-hygiene materials. Those on digital therapy had large gains in sleep-related quality of life and small gains in daytime physical and mental state. Then sleeping pills. One analysis pooled 24 randomized controlled trials and 2417 people with insomnia aged 60 and over. Each trial used at least 5 nights in a row. Versus placebo, people on pills slept 25.2 minutes more total and woke 0.63 fewer times at night. For every 13 who took pills, 1 had better sleep quality. For every 6 who took pills, 1 had a side effect (NNH 6, 4.7 to 7.1). Memory and thinking side effects were 4.78 times placebo (OR 4.78, 1.47–15.47). Daytime fatigue was 3.82 times (OR 3.82, 1.88–7.80).

The American Academy of Sleep Medicine 2017 diagnostic guideline lists who is more likely to have moderate to severe disease. Most days very sleepy by day, plus at least two of the next three: one, frequent loud snoring; two, someone has seen you stop breathing, gasp, or choke awake in sleep; three, already diagnosed hypertension. The guideline requires sleep monitoring for diagnosis; questionnaires alone are not enough. Five or more breathing pauses or shallow spells per hour plus symptoms count as the disease. Fifteen or more count whether or not there are symptoms. The American Academy of Sleep Medicine 2019 systematic review pooled 38 randomized controlled trials. Versus no treatment, people on CPAP cut self-rated sleepiness by 2.4 points (95% CI 1.9–2.8). That score is the Epworth Sleepiness Scale, max 24, higher means sleepier. Counting only people who were not sleepy to begin with, 5 trials cut 1.0 point; the authors saw no clinical meaning. The same review pooled 10 before-after treatment studies. After CPAP, crash rate was 0.3 times the pre-treatment rate (RR 0.3, 0.2–0.4), about seven tenths less. A 2010 analysis pooled 9 similar studies, result RR 0.278 (0.22–0.35). It also said daytime sleepiness dropped clearly after one night of use, and simulated driving improved within 2 to 7 days. A 2009 analysis by the same group said untreated patients’ crash risk was roughly 1.21 to 4.89 times others’. Then heart and lifespan. The SAVE trial enrolled 2717 people aged 45 to 75 with moderate to severe disease and coronary disease or prior stroke. After randomization they were followed 3.7 years on average. Cardiovascular events: 17.0% on CPAP, 15.4% without. No difference between groups (HR 1.10, 0.91–1.32). Users averaged only 3.3 hours a night. Their snoring, daytime sleepiness, quality of life, and mood all improved. A 2017 analysis pooled 10 randomized trials and 7266 people. Overall mortality also did not differ (RR 1.13, 0.99–1.29). A 2019 estimate said about 176000000 (1.76亿) people aged 30 to 69 in China have the disease, 23.6%. Of them about 65520000 have moderate to severe disease, 8.8%, the largest count worldwide.