The evidence contradicts the popular '21 days' rule. In the original habit-formation study, participants took a median of around 66 days for a new behaviour to become automatic, with individual times ranging from 18 to 254 days depending on the person and the habit. A later pooled review of the evidence found closely similar results — medians of roughly 59 to 66 days — and its authors state plainly that their findings refute the popular notion that habits form in approximately 21 days.

Worth knowingMissing a single day of practice did not meaningfully disrupt habit formation in the original study, which undercuts the urgency implied by a fixed deadline.

The idea that striking a brief high-power posture shifts your hormones traces to a single small laboratory study, which reported that holding an expansive pose for about a minute raised testosterone and lowered cortisol. That hormonal finding has not held up: several independent, larger, and more rigorously controlled attempts to reproduce it — including a close conceptual replication with blinded experimenters and a field study nested inside a real competition — found no significant change in testosterone or cortisol from power posing. The researchers behind the original study later co-authored a joint statement, alongside many other investigators who ran a coordinated set of preregistered replications, concluding that pose type has essentially no effect on any hormonal or behavioural measure. Standing in a power pose does not reliably change your hormones.

Worth knowingThis verdict addresses hormones only. The hormonal claim and the felt-power claim are different outcomes and must not be conflated: while testosterone and cortisol changes did not replicate, the effect of posing on self-reported feelings of power has been repeatedly confirmed, including in a dedicated Bayesian meta-analysis restricted specifically to that outcome. Power posing is not 'debunked' wholesale — only its hormonal mechanism is.

Controlled trials show a small, real reduction in time to fall asleep in older adults and magnesium-deficient individuals, but the RCT record is explicitly contradictory and evidence quality is rated low to very low across all meta-analyses. The popular claim implies broad, reliable benefit that the evidence does not support.

Worth knowingA large systematic review found that while observational data links magnesium status to better sleep quality, the RCT findings are contradictory — making a confident population-wide recommendation impossible on current evidence.

Grip strength is among the most replicated predictors of all-cause mortality in epidemiology, outperforming systolic blood pressure in large multi-country cohorts. Prospective studies spanning millions of participants consistently find that lower grip strength is associated with higher mortality from cardiovascular disease, respiratory disease, and cancer. The relationship holds as a predictive signal; grip strength reflects overall physiological reserve rather than acting as a direct cause of longer life.

Worth knowingPrediction is not causation. Grip strength functions as a proxy for overall muscular and physiological reserve. No randomised controlled trial has demonstrated that specifically training to improve grip strength extends lifespan.

Deliberate, structured practice is a genuine predictor of skill, but its power varies hugely by domain — explaining 26% of the variance in performance for games, 21% for music, 18% for sports, 4% for education, and less than 1% for professions. The specific '10 000 hour rule' popularised by Malcolm Gladwell is not, however, what the source science shows. It traces to a small study of 30 violin students and 12 pianists at one Berlin academy, which reported that skill tier corresponded to average accumulated practice time across the group — not a fixed per-person threshold for mastery. A pre-registered direct replication did not reproduce that core finding: by age 20, both the best and good violinists in the replication sample had already passed 10 000 hours of solo practice, with no reliable gap between the top two tiers. Ericsson himself later said there was no evidence for a 'magical number' of hours, and separately estimated that reaching elite international-level piano performance would take around 25,000 hours — roughly two and a half times the popularised figure.

Worth knowingThe original Ericsson study itself was small and correlational — 30 violinists and 12 pianists from a single Berlin conservatoire, sorted into skill tiers after the fact — and reported a group-average pattern, not evidence that any individual guaranteed mastery at 10 000 hours.

Melatonin produces a real but modest reduction in time to fall asleep (sleep onset latency) and acts on the circadian system, which suits sleep problems with a circadian component — jet lag, shift work, delayed sleep phase. But the American Academy of Sleep Medicine explicitly recommends against it for chronic primary insomnia, and its mechanism is chronobiotic (circadian timing) rather than sedative: it shifts sleep timing rather than inducing sleep. Calling it an effective treatment for insomnia without qualification overstates what the evidence supports.

Worth knowingThe sleep-onset benefit is statistically significant across multiple large meta-analyses but modest in magnitude — substantially smaller than the effects seen with approved hypnotic medications.

Spacing study sessions out over time produces better long-term retention than cramming them into one sitting — this is one of the best-replicated findings in cognitive psychology, drawn from a meta-analysis of 317 experiments and confirmed in a real-world field study on a large employee training dataset. But the advantage is conditional, not universal. First, how far apart sessions should be depends on how long you need to remember the material: a large factorial study found the optimal gap between sessions declined from about 20 to 40% of a 1-week test delay down to about 5 to 10% of a 1-year test delay — meaning spacing's advantage shrinks toward zero, and can favour massed practice instead, when the test is imminent. Second, the benefit is well-established for verbal and factual material but inconsistent for procedural skills: one study found spacing nearly doubled four-week retention of a maths procedure using 10 practice problems, while another equally well-powered study found no spacing benefit at all for a different maths procedure.

Worth knowingHow large the spacing effect is depends on the retention interval: the optimal gap between study sessions is not a fixed number of days but shrinks as a proportion of how long you need to remember the material — from about 20 to 40% of a 1-week delay down to 5 to 10% of a 1-year delay. This is also the mechanistic reason cramming can outperform spacing on an immediate test: massed repetition creates a close match between the study context and the test context that spaced study cannot offer.

The underlying finding is real: people with poor social connection face a meaningfully higher risk of earlier death, a result replicated across three separate meta-analyses. But the specific 'as harmful as smoking 15 cigarettes a day' figure does not come from the meta-analysis usually cited for it — that paper describes the effect only in general terms, as comparable with quitting smoking, and reports a 50% relative-survival advantage for people with adequate social relationships, without ever naming a cigarette count. Researchers examining the claim's origins describe the '15 cigarettes/day' figure only as an oft-repeated claim whose derivation is unclear. The claim also collapses three distinct exposures into one: loneliness alone carries the smallest mortality risk (OR 1.14 to 1.26 across the two most recent meta-analyses), smaller than social isolation (OR 1.29 to 1.32) or living alone (OR 1.32) — yet the popular claim treats 'loneliness' as though it carried the combined weight of all three. When smoking is benchmarked on a comparable relative-risk scale, its own mortality risk in the moderate range (RR 2.02 for 10 to 20 cigarettes a day) is substantially larger than any of these odds ratios, undercutting the claimed equivalence even on its own quantitative terms.

Worth knowingThe mortality association between poor social connection and earlier death is genuine and consistently replicated — this is not a case of loneliness being harmless. The issue is specifically with the popular quantification, not the existence of a real effect.

People do have stable preferences for how they like information presented — nobody disputes that. What the popular claim actually asserts, though, is something stronger: that giving someone lessons tailored to their assessed 'learning style' produces better learning than a mismatched format. Tested properly, that specific mechanism does not show up. The foundational review of the field found no adequate evidence base for using learning-styles assessments in teaching, a direct test in a real course found no relationship between students' assessed style and their exam performance, and a classroom trial built specifically to detect a matching benefit found none. Even the most sympathetic meta-analysis designed to rehabilitate the idea could only confirm the required pattern in 26% of the outcome measures it examined, and its own authors concluded that was too small and inconsistent to justify style-matched teaching.

Worth knowingThe foundational critical review of the learning-styles literature concluded there is no adequate evidence base to justify incorporating learning-styles assessments into general educational practice, despite the idea's popularity in classrooms.

False as stated. A study of 132,515 adults aged 21-60 directly tested the idea that personality traits stop changing by age 30 and found that Conscientiousness and Agreeableness kept increasing throughout early and middle adulthood, well past 30. A separate meta-analysis found 4 of the 6 broad trait categories it studied showed significant mean-level change in middle and old age — change does not stop at any fixed adult cutoff. Personality can also be shifted deliberately: a meta-analysis of intervention studies found meaningful trait change (d = .37) after an average of 24 weeks, persisting beyond the intervention itself. Personality is not fixed by adulthood, and it keeps changing well beyond it.

Worth knowingThe claim survives because it mistakes a different, real phenomenon for trait fixity: people's personality relative to their peers becomes fairly consistent with age. Test-retest stability of trait rankings rises from .31 in childhood to .54 during the college years, to .64 at age 30, and plateaus around .74 between ages 50 and 70 — high, but not perfect. That is about relative ranking staying similar, not about trait levels staying the same.

Four decades of meta-analyses agree: mentally rehearsing a motor skill improves how well you later perform it, compared with no practice at all. The first major synthesis, pooling 60 studies, found mental practice improved performance with an average effect size of .48. A second, highly cited synthesis confirmed the effect was positive and significant, and found it was moderated by the type of task, the gap between practice and performance, and how long the mental practice lasted. A follow-up meta-analysis found an even larger average effect of .68, and identified that imagining the movement from the inside, as if you are the one performing it, works better than picturing yourself from the outside. A more recent, bias-corrected replication of the whole field, run in the shadow of psychology's reproducibility crisis, still found a small but significant positive effect (r = 0.131), smaller than the earlier estimates but the same direction and still real. And in an applied test with real stakes, novice surgeons who mentally rehearsed a laparoscopic procedure scored significantly higher on technical-skill ratings than those who did not, across every one of their practice sessions. So the effect holds up under scrutiny: it just shrinks once the field is corrected for publication bias and re-tested rigorously.

Worth knowingThis verdict covers rehearsing a specific motor skill, a sport movement, a surgical procedure, a musical piece, in your mind before performing it. It does not cover imagining yourself having already achieved a goal or outcome, which is a separate claim with different and largely opposite evidence, and the two should not be treated as the same thing.

Standardised ashwagandha extracts reliably lower serum cortisol in stressed adults, with reductions ranging from 11% to 32.63% across multiple independent RCTs and systematic reviews. The effect is specific to proprietary standardised formulations — not raw root powder or arbitrary products labelled ashwagandha. Critically, lower serum cortisol does not reliably translate to less felt stress: one meta-analysis confirmed cortisol fell significantly across trials but perceived stress scores did not improve.

Worth knowingThe cortisol-lowering effect is extract-specific: all consistent trial evidence uses standardised, withanolide-calibrated root extracts of the kind used in clinical trials. The evidence cannot be generalised to uncharacterised products sold as ashwagandha.

Beta-alanine raises muscle carnosine, which buffers hydrogen ions during intense exercise and delays fatigue-driving acidosis. The performance benefit is real but modest, and it is concentrated in sustained high-intensity efforts lasting roughly one to four minutes — efforts where muscle acid build-up is the limiter. Very short maximal sprints under a minute and extended aerobic efforts beyond roughly twenty-five minutes show no consistent benefit.

Worth knowingThe mechanism is well-established: beta-alanine supplementation elevates muscle carnosine, an intracellular proton buffer that slows the pH drop limiting high-intensity output — making it genuinely effective within its duration window.

Breathwork is broadly comparable to mindfulness meditation for reducing stress and anxiety, based on the limited direct comparative evidence available. The one rigorous head-to-head trial found that cyclic sighing matched mindfulness meditation for anxiety reduction and exceeded it for positive mood. However, this single trial covered one specific breathwork technique over a short duration, and the broader meta-analytic literature could not address the comparison for lack of qualifying head-to-head studies. Whether the equivalence extends to other breathwork types, longer durations, or clinical populations remains unknown.

Worth knowingThe comparative claim rests on a single randomised controlled trial; at the meta-analytic level, no qualifying head-to-head trials between breathwork and meditation existed at the time of the most recent breathwork meta-analysis, which could only establish that breathwork outperforms non-breathwork control conditions.

Cold water immersion reliably reduces delayed-onset muscle soreness and biochemical markers of muscle damage, making you feel recovered faster after hard training. But when applied habitually after resistance training it suppresses the molecular signals that drive muscle growth, trading long-term adaptation for short-term comfort.

Worth knowingThe soreness-relief benefit is well-supported across a large body of RCTs: cold water immersion was most effective for biochemical markers and neuromuscular recovery, and best for alleviating muscle soreness.

Collagen supplementation produces a small-to-moderate, well-supported reduction in joint pain and functional impairment in osteoarthritis, backed by moderate-to-high certainty evidence from a large trial sequential meta-analysis. The skin half of the claim does not hold up under scrutiny: pooled positive effects in major meta-analyses disappear entirely when restricted to non-industry-funded or high-quality trials, leaving the popular skin claim without independent support.

Worth knowingFor joint health in osteoarthritis, the evidence is relatively robust: a large trial sequential meta-analysis confirmed consistent pain relief and function improvement at moderate-to-high certainty by GRADE — making joints the stronger, more defensible half of the compound claim.

The honest answer turns on a distinction the popular claim erases: who emerges as a leader versus who leads effectively once in the role. A large meta-analysis across many studies confirms extraversion is the single most consistent Big Five personality correlate of leadership, but that same analysis found extraversion tracks who gets seen and chosen as a leader more strongly than it tracks who actually performs well as one -- the emergence effect is not the same as the effectiveness effect. A field study of store managers, paired with a laboratory experiment, found the relationship between extraverted leadership and objective group performance actually reverses depending on the people being led: extraverted leadership lifts measurable output such as profit only when followers are passive, but when followers are proactive -- voicing ideas, taking initiative -- extraverted leadership is linked to lower objective performance, because extraverted leaders are perceived as less receptive to that input. A large modern update of the original meta-analysis, spanning many countries, corroborates that the extraversion-effectiveness link is not a fixed law: it is stronger in collectivist cultures and appears to work through leader behaviour rather than personality alone. So 'extroverts make better leaders' holds for who tends to be recognised and selected as a leader, but the case for extroverts leading more effectively once in charge is conditional -- on follower proactivity and on cultural context -- not a blanket truth.

Worth knowingThe meta-analysis pools results from many independent samples across business, government or military, and student settings, and separates leader emergence from leader effectiveness as distinct outcomes -- extraversion's link to effectiveness is real but consistently weaker than its link to emergence.

The evidence is genuinely split, and the honest picture is small-to-null on average, with any real signal concentrated in specific groups rather than in students generally. The largest, best-designed randomised trial — a nationally representative U.S. study (Yeager and colleagues, National Study of Learning Mindsets) — found a real intervention effect, but it improved grades specifically among lower-achieving students, and it only held in schools where peer norms aligned with the intervention's message; it was not a general effect across all students. An earlier longitudinal study of 373 7th-graders found a similar pattern: believing intelligence is malleable predicted rising grades, and a brief classroom intervention reversed a grade decline seen in the control group. Set against this, the field's most rigorous systematic review (Macnamara and Burgoyne, 63 intervention studies, N=97,672) found only a small overall effect (d=0.05, 95% CI 0.02-0.09) that became statistically nonsignificant once publication bias was corrected for, and the authors traced the apparent benefit substantially to weak study design and to researchers with a financial incentive in the outcome publishing larger effects. A separate pair of meta-analyses (Sisk and colleagues) likewise found weak overall effects, with any real benefit confined to students who are low-income or academically at risk. A construct-validation study testing the theory's core premises directly found little support for them, concluding that mind-set claims 'appear to be overstated.'

Worth knowingThis is not evidence that mindset beliefs are irrelevant to learning — the effect is real but narrow: it shows up in lower-achieving students and depends on classroom or school peer-norm context, not as a general lever that raises achievement for students at large.

The evidence is genuinely split, not simply for or against. The original jam-display experiment found a dramatic effect in one setting: only 3% of shoppers bought when shown an extensive display, versus 30% who bought from a limited display. But the largest attempt to average across the whole choice-overload literature — 63 experimental conditions from 50 studies, pooling 5,036 participants — found the effect shrinks to virtually zero once pooled, with considerable unexplained variance between individual studies. A separate, larger re-analysis of that same body of research argues the pooled-null result is misleading because it averages over very different conditions, and that assortment size does reliably matter once factors such as how complex the choice set is and how settled the decision-maker's preferences already are get modelled — a conclusion its authors present as directly contradicting the pooled-null finding. The two research teams have not reconciled: whether 'more options' reliably backfires depends on unresolved, actively disputed questions about which conditions matter, not on one settled number.

Worth knowingNo peer-reviewed, Crossref-verifiable direct replication of the original jam-display study could be located. Reports that replication attempts failed trace only to secondary sources (blog posts, a magazine feature) rather than a citable primary study, so this verdict does not assert a failed replication.

Making a specific if-then plan measurably increases follow-through, but how much depends heavily on who is measuring it and what kind of behaviour is being changed. The two largest meta-analyses supporting the effect (94 independent tests, d = .65; a later synthesis of 642 tests) come from the theory's own originators. Independent researchers replicate a similarly strong effect for one-off, discrete actions — a clinical/mental-health meta-analysis outside that research group found an even larger effect (d + = 0.99) after excluding one outlier. But independent researchers studying repeated, habitual behaviours like exercise find the opposite: a pooled analysis of randomised physical-activity trials found no significant effect (0.15) without added reinforcement, and a large field experiment on gym attendance found a tightly estimated null result. The claim holds well for one-off, novel actions; it does not reliably hold for sustained behaviour change on its own.

Worth knowingThe two headline meta-analyses defining this effect (94 independent tests, d = .65, and a later update spanning 642 independent tests) were conducted by the theory's own originators (Gollwitzer and Sheeran), so on their own they cannot settle whether the effect replicates independently of the lab that built it.

The core mechanism behind this claim is well established: the brain does not run two demanding tasks in true parallel, it switches between them, and each switch carries a measurable cost in speed and accuracy. That cost, however, is not universal — it shows up specifically when the tasks compete for the same underlying cognitive resource. When two tasks draw on separate resources instead (for example one perceptual-motor, one cognitive), modelling of that exact task pairing shows they can eventually be interleaved with little or no measurable interference — but only after many trials of practice; the same pairing still showed real interference before that training took hold. So the escape hatch from switch costs is narrow: it requires both resource-disjoint tasks and a well-practised pairing, which does not describe the everyday case of switching between demanding, overlapping tasks, and 'multitasking always costs you' still overstates what the switch-cost research shows for that narrow, trained exception. The stronger popular version of this claim — that habitual heavy media multitasking causes a lasting decline in general cognitive ability — rests on much shakier ground: pre-registered replication attempts of the founding study mostly failed, and a meta-analysis correcting for small-study bias found the association was no longer significant. Finally, every citation behind this verdict measures laboratory reaction time, accuracy, or distractor filtering — none directly measures real-world workplace output, so extending these findings to a claim about 'productivity' specifically involves a genuine gap beyond the direct evidence.

Worth knowingSwitch costs are a robust, replicated laboratory finding: alternating between two tasks produces slower, more error-prone performance than repeating the same task, and this cost is only partly reduced by advance preparation or cueing.

The picture is genuinely two-sided, and both sides are well evidenced. In controlled experiments, expert performers on a proceduralised sensorimotor skill -- golf putting -- reliably choked when a cash incentive raised the stakes, while performers on a task that stays in working memory rather than becoming automatic did not choke under the same pressure; choking was eliminated, and performance even improved under pressure, when performers had trained under conditions that raised self-consciousness during practice. An earlier, foundational experiment found the same pattern from a different angle: golfers who learned putting implicitly, without building up a large store of explicit step-by-step rules, were less likely to break down under evaluative and financial pressure than golfers who learned the same skill explicitly. A review of the wider behavioural and neuroimaging literature situates this within a broader pattern -- that high incentives can produce performance decrements rather than the performance boost simple economic models predict. But a field study of actual PGA, Senior PGA and LPGA Tour final rounds, where the pressure and stakes are as real as competitive golf gets, found no support for the choking hypothesis at all: players leading going into the final round won more often than not rather than folding. So choking is a real, mechanistically demonstrated phenomenon for skills that have become automatic, and it is trainable away -- but it is not a universal law that applies uniformly whenever stakes rise, and the clearest direct test among elite professionals under real competitive pressure did not find it.

Worth knowingThe core choking-under-pressure experiments used undergraduate and collegiate golfers in controlled laboratory settings with cash incentives standing in for real stakes -- strong for isolating the explicit-monitoring mechanism, but a different context from professional, career-stakes competition.

The popular claim usually means one specific thing: repeating a fixed positive self-statement about yourself, such as ‘I am a lovable person’. Tested directly, that practice backfired for people with low self-esteem -- the exact population it is marketed to help -- leaving them feeling worse, not better, while people who already had high self-esteem gained only a small benefit (the same study's own meta-analytic comparison found the low-self-esteem effect, d = 0.72, was even larger than the high-self-esteem benefit, d = 0.66). That backfire finding is not settled: a later two-study replication attempt found no self-esteem-moderated effect at all, and in its second study found no self-esteem benefit from repeating positive self-statements or from writing about personal values either -- so neither harm nor benefit for this specific practice is currently well established. A different, often-confused intervention that researchers actually call 'self-affirmation' -- writing about your own core personal values, not reciting a scripted positive trait -- has a real applied evidence base in education and health, and one large adolescent trial found it improved self-esteem specifically. But that is a different exercise from reciting affirmations, and defenders of the popular practice regularly borrow its evidence without acknowledging the swap.

Worth knowingThe strongest single experimental test of the popular practice (repeating a fixed positive self-statement) found it made low-self-esteem participants feel worse, not better -- the reverse of the marketed benefit -- with an effect size (d = 0.72) larger than the modest gain seen in high-self-esteem participants (d = 0.66).

Frequent sauna use — several sessions per week sustained over years — is consistently linked to lower risk of fatal cardiovascular events and incident hypertension in long-running prospective cohort data (large groups of people tracked over many years, with researchers recording who develops disease), with a clear dose-response pattern that persists after adjustment for conventional risk factors. Causation is not established: the supporting studies all draw from the same overlapping Finnish cohort, healthy-user confounding (the possibility that sauna users are already healthier in ways that statistics cannot fully account for) cannot be fully excluded, and short-term randomized trials of passive heating have not confirmed cardiometabolic benefit.

Worth knowingAll three major supporting studies draw from the same Finnish research cohort (the Kuopio Ischaemic Heart Disease Risk Factor Study), so the dose-response association reflects one population studied repeatedly — not independent replication across diverse groups.

This claim bundles two different things, and the evidence treats them very differently. Actual sleep — falling asleep overnight and waking up — has solid, replicated support: in a controlled sleep-lab study, people who slept between learning a hidden rule and being retested were more than twice as likely to consciously discover that rule as people who stayed awake for the same stretch, and a separate study found sleep deprivation measurably pushed people toward riskier, worse choices on a gambling task relative to their own rested baseline. The other half of the claim — that a brief spell of quiet, unconscious 'mulling it over' without actually sleeping produces demonstrably better complex decisions than consciously thinking it through — traces to an influential study on consumer choice, but has since failed its largest, best-powered, pre-registered direct replication and an earlier independent meta-analysis, neither of which found a reliable unconscious-thought advantage. In short: if 'sleep on it' means literal sleep, the advice holds up; if it means merely stepping away to let your unconscious work on it, the evidence does not support a reliable benefit.

Worth knowingThe two literatures are easy to conflate in popular advice, and 'sleep on it' is often defended by citing unconscious-thought research — the weaker of the two bodies of evidence — rather than the sleep research, which is the stronger one.

This is one of the better-supported classic decision biases. In controlled negotiation experiments, whichever side named the first number obtained a better final outcome, and that first offer strongly predicted the eventual settlement price, in both face-to-face and email negotiations. Anchoring more generally is also one of the few classic psychology effects that held up in a large, preregistered replication project spanning many independent samples worldwide, with very large effects — unlike several other classic findings tested in that same project, which failed to replicate. But the negotiation advantage is not unconditional: the same research that established it also found the first-offer advantage was eliminated, not just reduced, when the other side deliberately focused on their own walk-away alternatives, the other side's likely reservation price, or their own target figure instead of anchoring on the number just named. And it is not a novice-only effect — professional real-estate agents were anchored by a manipulated listing price just as much as inexperienced participants, so expertise alone does not confer immunity.

Worth knowingThe first-offer advantage was eliminated, not merely reduced, when the other party used one of three specific counter-strategies: focusing on their own alternatives, the other side's likely reservation price, or their own target outcome — so a first offer is a strong default, not a guaranteed win.

Zone two training reliably builds an aerobic base — low-intensity endurance exercise increases skeletal muscle mitochondrial content and fat-oxidation capacity (the cellular engines that burn fat and sustain aerobic effort), and elite endurance athletes structure the bulk of their training at low intensity. But the superlative 'best way' is unsupported: the only direct comparative trial found that a polarized model pairing low-intensity volume with high-intensity sessions outperformed the high-volume low-intensity approach on every key performance measure, and a recent review found no controlled evidence that zone two beats higher intensities for the mitochondrial adaptations the claim implies it excels at.

Worth knowingThe mechanistic and observational case for zone two as a core base-building method is solid: regular endurance exercise drives increases in skeletal muscle mitochondrial content and respiratory capacity, and elite endurance athletes consistently structure the bulk of their training at low intensity.

High-blocking amber lenses worn before bed may benefit people with existing insomnia symptoms, but the broad consumer claim is not supported by the overall evidence base. Meta-analytic syntheses find no statistically significant effect on objective sleep outcomes across healthy and mixed adult populations.

Worth knowingThe mechanistic basis is real: evening blue-light exposure suppresses melatonin and delays sleep onset, giving glasses a plausible theoretical rationale.

Afternoon caffeine can reduce sleep time and quality for typical consumers — a standard coffee (107 mg) is best consumed at least 8.8 h before bedtime. But 'ruins' overstates: at 100 mg, evidence shows no significant sleep disruption at the pre-bed timing tested, and individual metabolism varies 5-6 fold, making a flat time rule unreliable.

Worth knowingThe 8.8 h cutoff applies specifically to a 107 mg coffee; higher-dose products require an even earlier cutoff.

Intermittent fasting is a genuine fat-loss tool — it outperforms no diet — but when calories are equated, head-to-head trials and pooled meta-analyses consistently find it produces no more fat loss than continuous calorie restriction. The popular claim mistakes 'IF works' for 'IF works better.'

Worth knowingMultiple randomised trials — including alternate-day fasting and sixteen-eight time-restricted eating protocols — find no significant between-group difference in weight or fat loss compared with calorie-matched continuous restriction.

The specific figure of twenty-three minutes does not appear in either of the two peer-reviewed papers it is almost universally credited to. A field study of information workers measured an average of 11 minutes 4 seconds spent in a working sphere before switching to another task or being interrupted, and a separate average of 25 minutes 26 seconds for same-day resumption of an interrupted task -- though that resumption figure was reached only after roughly two other working spheres intervened, so it is not a single interruption's recovery time. A separate lab study found the opposite of the popular framing: people who were interrupted completed the same task in less time than an uninterrupted baseline of 22.77 minutes, at the cost of more self-reported stress. An independently logged field study at a technology company found roughly 10 minutes to handle an alert plus a further 10 to 15 minutes to return to focused work; for an immediate response to an email alert specifically, the resumption phase averaged 16 minutes 33 seconds. Real interruption costs are well documented across all three studies and roughly cluster in a ten-to-twenty-five-minute range; the specific number popularly quoted is not.

Worth knowingThe nearest a source comes to the popular number -- the field study's 25 minutes 26 seconds same-day resumption figure -- is a same-day-resumption average reached only after roughly two intervening working spheres, not a clean single-interruption recovery time, so even the closest real number differs in what it actually measures.

The original Israeli parole-board study found favourable rulings fell from around 65% at the start of a session to nearly zero by the end, then jumped back to around 65% after each food break — the 'hungry judge' pattern behind this claim. But this is a genuinely contested single-dataset finding, not a settled effect. A reanalysis of the same case records found that scheduling is not random: unrepresented prisoners are systematically heard last within each session, right before a break, and are less likely to be granted parole regardless of any hunger effect. The original authors dispute this, reporting that the meal-break pattern survives once legal representation is added as a control. Separately, a later simulation study argues that even a purely rational, non-fatigued judge — one who simply takes longer to write up a grant than a denial, with only limited foresight about session length — could produce an order effect of a similar size through statistical artifact alone. A swing this large, from a majority of favourable rulings to almost none, is an unusually big effect for any single psychological mechanism to carry on its own.

Worth knowingThe dispute is not resolved: the original authors' reply reports the meal-break pattern survives after controlling for legal representation, while the reanalysis authors maintain the pattern is better explained by non-random case scheduling than by hunger or fatigue.

Heart rate variability (HRV)-guided training — adjusting daily session intensity based on readiness — has genuine support across multiple RCTs and a meta-analysis: it produces similar or superior fitness adaptations compared to fixed periodization plans. The kernel of truth ends there. Every studied protocol responds to a depressed HRV reading by prescribing a low-intensity session, not complete rest; furthermore, single-day readings are unreliable noise — protocols require multi-day rolling averages to detect genuine readiness shifts. The specific consumer rule 'low HRV = skip your workout' maps onto no tested protocol and exceeds what the evidence actually endorses.

Worth knowingNo published RCT protocol instructs athletes to skip their session on low-HRV days — every studied protocol prescribes a low-intensity session as the readiness-informed response to a depressed reading, not complete rest.

Decades of observational studies produced a consistent J-shaped association between moderate drinking and cardiovascular risk — the real foundation of the popular belief. Two more rigorous methodological approaches undermine that apparent benefit: correcting for the misclassification of former drinkers as abstainers eliminates the mortality advantage, and Mendelian randomization studies find that genetic variants predicting lower alcohol intake are associated with better cardiovascular outcomes, not worse. The best available causal evidence indicates that lower alcohol consumption benefits cardiovascular health: it does not merely fail to support a net heart benefit — it points the opposite way, toward harm.

Worth knowingThe observational J-curve literature is large and internally consistent: dozens of prospective cohort studies found that light-to-moderate drinkers had lower cardiovascular mortality and coronary heart disease incidence than abstainers. This is the genuine observational kernel the popular claim rests on.

There is a genuine kernel of truth here, but 'drives' claims more direct causal force than the evidence supports. In the foundational field study, psychological safety predicted team learning behaviour, and it was learning behaviour — not psychological safety on its own — that predicted team performance: once both were entered into the same statistical model, psychological safety's own direct effect on performance became non-significant (B = .25, p = .42), while learning behaviour remained a significant predictor (B = .60, p < .05). An independent replication two decades later, in South Korean sales teams rather than a US manufacturer, found the same pattern: no significant direct effect of psychological safety on team effectiveness (β = 0.037), with the relationship holding only through a full double-mediation pathway via learning behaviour and team efficacy. A further meta-analysis shows the link is also task-contingent, stronger in complex, creative, sensemaking-heavy work and possibly absent where tasks do not require learning. So the popular claim's real mechanism — psychological safety enabling the learning behaviours that in turn improve performance, mainly in complex or creative work — is well supported; the popular claim's implied direct, universal 'drive' is not.

Worth knowingThe performance link is mediated, not direct: psychological safety predicts team learning behaviour, and it is learning behaviour that predicts performance, not psychological safety acting on its own. Once both are entered into the same model, psychological safety's direct effect on performance is non-significant while learning behaviour's effect remains significant.

There is a real 'undermining effect', but it is narrower than the popular claim suggests. A large meta-analysis found that engagement-contingent, completion-contingent, and performance-contingent rewards significantly reduced people's later free-choice engagement and interest in an activity (d = -0.40, -0.36, and -0.28, respectively). But the same meta-analysis found the opposite for a different kind of reward: positive verbal feedback increased both later free-choice engagement (d = 0.33) and self-reported interest (d = 0.31). A rival meta-analysis concluded that, overall, reward does not decrease intrinsic motivation, and traced the only reliable negative effect to expected, tangible rewards given to people simply for doing a task. So 'rewarding people destroys their motivation' holds for a specific, narrow kind of reward — tangible, expected, and tied to doing or finishing the task — not for reward in general, and not for praise.

Worth knowingOther reviews in this literature converge on the same narrower conclusion from the opposite rhetorical direction: detrimental reward effects are real, but occur only under highly restricted, easily avoidable conditions, and reward can even be used to boost generalised creativity.

The original study, published in the Journal of the Association for Consumer Research, found that people scored worse on tests of working memory and reasoning when their own smartphone sat nearby on the desk rather than in another room, even though it was silenced and never touched. But the claim has not held up well since it was published. A pre-registered study using the exact same tasks and the exact same phone-location conditions found no difference in performance at all. A separate study testing short-term and prospective memory instead found no overall effect of phone presence either. A large meta-analysis pooling many studies since then did find a negative effect overall, but one that varies substantially depending on which cognitive skill is being tested, rather than a single clean 'brain drain'. So the mere-presence effect is real in the sense that it was published with a significant result, but the most direct replication attempts have failed to reproduce it, and the wider evidence since is mixed rather than settled.

Worth knowingPhone notifications are a different, better-established distraction mechanism from mere presence: pings measurably disrupt attention even without the phone being touched, but that is a separate claim from the passive 'sitting nearby' effect this verdict addresses.

Standing at a desk burns a fraction of a calorie more per minute than sitting — a real but trivially small difference that falls far below what 'significantly' implies to a lay reader. No randomised trial has shown meaningful caloric benefit or body-composition change from standing desk use alone.

Worth knowingThe best available meta-analysis confirms a real energy expenditure advantage for standing, but the effect is so small it is unlikely to produce detectable weight change in practice, especially given evidence that workers may compensate with increased sedentary time outside work.

The original study found that students who took notes on laptops performed worse on conceptual questions than students who took notes longhand, and pinned this on laptop note-takers transcribing lectures word for word instead of processing and reframing the material in their own words. But the specific performance advantage has not held up well since. A large preregistered replication reproduced the behavioural pattern behind the claim, laptop users writing more and copying more verbatim, but did not find that longhand users actually scored better on the later test, and an accompanying meta-analysis of several similar studies echoed the same null result. A separate preregistered replication, which added an e-writer condition and even a no-notes condition, found no consistent difference between any of the note-taking methods. So the mechanism behind the claim, that verbatim transcription is a shallower way to process material, does appear to be real and reproducible, but the specific benefit, that handwriting beats typing on a later test, has not held up under direct replication.

Worth knowingThe clearest replicated finding is behavioural rather than a performance benefit: laptop note-takers reliably write more and copy more verbatim, but this has not been shown to reliably translate into worse quiz performance.

Dunbar's own foundational analysis, extrapolating from a primate neocortex-to-group-size regression, predicts a human group size of 147.8 -- the figure popularly rounded to 150 -- and Dunbar cross-checked it against hunter-gatherer groupings, farming-community splits and army units that cluster near the same range. But even Dunbar's own confidence interval around that estimate was wide, running from roughly a hundred to well over two hundred, and a modern re-analysis using updated primate datasets and different statistical methods found the underlying regression cannot reliably produce one number at all, concluding that a cognitive limit on human group size cannot be derived this way. Independent evidence does support the idea of tiered relationship layers: one large analysis of mobile-phone call patterns found strong evidence for a layered social structure broadly consistent with Dunbar's tiers -- inner five, middle fifteen, outer 150 -- though with large variability in the middle layers, and a large online survey found real person-to-person variation in how people allocate relational energy across those layers, with extraversion unrelated to the pattern. Taken together, 150 is a genuine, historically corroborated central estimate from Dunbar's own methodology, not a fabricated number, but treating it as a precise, hard ceiling overstates the evidence: the original analysis carried a very wide margin of error, and a modern statistical re-analysis could not derive a reliable single number at all.

Worth knowingThe commonly cited number traces to Dunbar's original Journal of Human Evolution paper, which is paywalled and could not be independently verified from any available source; the 147.8 figure and its confidence interval used here instead come from Dunbar's own companion paper, which reports the identical regression and result with fully verified text.

Walking more steps each day cuts mortality risk — a consistent finding across large cohorts and meta-analyses. But the ten-thousand-step target has no physiological basis; it traces to a decades-old Japanese pedometer marketing campaign, not to physiology. The mortality-benefit curve levels off for adults over sixty at roughly 6,000–8,000 steps per day, and at approximately 7,500 steps per day in older women specifically. The claim that you 'need' ten thousand steps overstates the evidence: the majority of the survival benefit accrues well below that figure.

Worth knowingThe mortality-benefit plateau for older adults (aged ≥60 years) falls at approximately 6,000–8,000 steps per day according to a meta-analysis of fifteen international cohorts; the ten-thousand-step target lies above this range without delivering proportionally greater survival benefit in that age group.

There is no peer-reviewed evidence for a general claim that 93% of communication is non-verbal. The figure traces to two 1967 experiments in which 37 female psychology majors judged a speaker's feelings from a single ambiguous spoken word ('maybe') paired with varying vocal tones and facial photographs — a design built to make the verbal channel almost irrelevant, not a test of communication as a whole, and no single study measured all three channels together. Mehrabian himself has called extending his 7% figure to all verbal communication absurd, and a widely used nonverbal-communication handbook states plainly that the popular 93% estimate rests on faulty analysis. Yet a content analysis of 79 public websites citing the figure found 63 of them (80%) using it as a general statement about communication overall — the opposite of what the original narrow research actually measured.

Worth knowingThe two foundational 1967 studies (Mehrabian & Ferris; Mehrabian & Wiener) could not be quoted directly for this verdict — Crossref carries no abstract for either record and no accessible full text was located — so the scope facts here are sourced via a later peer-reviewed paper that itself quotes both originals, and Mehrabian's own correspondence, verbatim; the 1967 papers were not read directly.

Serum creatinine rises with creatine supplementation — because creatine is metabolised to creatinine — but this is a metabolic artefact, not organ damage. GFR, the true measure of kidney filtration, is unchanged across meta-analyses of randomised trials and a gold-standard radioisotope clearance RCT. The popular claim conflates a benign biomarker shift with kidney harm.

Worth knowingThese studies predominantly recruited healthy adults; individuals with pre-existing renal disease were largely excluded. The refutation applies to healthy populations; caution in that sub-group cannot be assumed away from this evidence base.

The strongest direct evidence runs against this claim. Two intervention field studies used wearable sociometric badges plus email and messaging logs to measure face-to-face interaction before and after a switch to open-plan offices at large corporate headquarters: contact fell sharply post-redesign -- a drop of roughly 70% -- with electronic messaging rising to compensate, the opposite of the collaboration boost the redesign was meant to produce. A systematic review pooling a large body of the existing comparative literature on open-plan versus enclosed offices corroborates this at scale, finding open-plan associated with more negative outcomes across health, satisfaction, productivity and social-relationship measures. One narrower single-firm survey found a different pattern for one specific slice of the picture -- high office density and low privacy were positively linked to expressive personal relations among coworkers -- but the same sample showed worse satisfaction, engagement and well-being overall, so it does not amount to independent, generalisable support for the claim as stated. No sound peer-reviewed evidence surfaced showing open-plan conversions increase collaboration; the industry claims that circulate to that effect are anecdotal and were excluded from this evidence base.

Worth knowingThe two headline sociometric-badge field studies were natural experiments drawn from a single broad corporate-redesign context, not randomised trials, and used small samples -- at most about a hundred employees per study -- so they demonstrate strong within-company change but limited generalisability across industries or office types.

The claim that the hot hand is a fallacy rests on an archival analysis of shot records that found essentially no advantage after a made shot: hit rate after a hit was, if anything, lower than after a miss (weighted mean: 51% versus weighted mean: 54%), and this was read for decades as proof that fans' and players' belief in shooting streaks is a cognitive illusion. Two independent, peer-reviewed statistical papers later showed that the method behind that finding is itself biased: conditioning on a streak of hits systematically undercounts hits in finite sequences, mechanically pulling the measured hit rate down. Reapplying a bias-corrected method to that same original archival data reverses the result, finding real streak shooting with large effect sizes, and the paper states directly that the hot hand is not a myth and the belief in it is not a cognitive illusion. Separately, shot-tracking data that models shot difficulty (defender distance, shot distance, game situation) finds a modest real hot-hand effect, in the range of 1.2 to 2.4 percentage points, once difficulty is held constant — though this comes from a working paper, not a peer-reviewed journal article. The picture is not fully settled: the newest peer-reviewed reanalysis, using a modern shot-difficulty model, finds a reverse hot-hand pattern (worse performance after streaks of difficult makes) rather than a clean confirmation of the folk hot-hand belief. Taken together, the specific statistical case for calling the hot hand a fallacy has been undercut by rigorous, peer-reviewed correction of the original method, even though the underlying phenomenon is still being actively remeasured.

Worth knowingTwo independent, peer-reviewed papers — one in Econometrica, one predating it in The American Statistician — prove, as a mathematical result rather than a matter of interpretation, that the classic method for detecting hot-hand streaks in finite shot sequences is a biased estimator that mechanically understates real streakiness.

The evidence points the opposite way from the popular advice. Longitudinal studies found that people who spent more time positively fantasising about a desired future put in less effort and did worse at reaching it, weeks to years later — while people who judged the outcome as realistically likely (a different mental act from fantasising about it) did better. Follow-up experiments identified a likely reason: imagining the successful outcome measurably lowers the physiological and behavioural energy needed to pursue it. A separate research programme that distinguished simulating the process of reaching a goal from simulating its successful completion found the same pattern: process simulation produced progress toward goals, but envisioning successful completion of the goal did not. The one future-oriented technique with meta-analytic support for goal attainment is structurally different from pure success-visualisation — it works only when the positive fantasy is deliberately paired with a real obstacle and a plan for overcoming it.

Worth knowingThis verdict rests on two independent research programmes: Oettingen's lab (a four-study longitudinal cohort finding that positive fantasies predicted lower effort and worse attainment, and a four-experiment causal follow-up identifying reduced energy as the mechanism) and Taylor and colleagues' programme distinguishing 'process' from 'outcome' simulation, which found that mental simulation of the process for reaching a goal produced progress toward it, while envisioning successful completion of the goal did not.

The idea that willpower is a single, finite tank that empties with use rests on small early laboratory experiments and a first meta-analysis of that literature — a later bias-correction re-analysis of that same meta-analysis described it as having concluded the depletion effect was 'robust and medium in magnitude (d = 0.62)'. That headline result has not held up under scrutiny. When the very same dataset was re-analysed using methods built to detect and correct for publication bias, the depletion effect became statistically indistinguishable from zero. Two large, independent, pre-registered multi-laboratory replication projects then tested the effect directly, each using a different protocol and far larger combined samples than any single study in the original literature. The first found a small pooled effect with a 95% confidence interval spanning zero (d = 0.04, 95% CI [-0.07, 0.15]). The second used a Bayesian analysis and found the data favoured no effect over even a modest true effect. A popular fallback explanation — that depletion only appears in people who believe willpower is limited — has also failed a pre-registered direct replication of that specific claim, and a proposed biological mechanism, that exerting self-control burns through blood glucose, has been tested directly and not supported. On the strongest currently available evidence, a literal, finite, depletable willpower resource is not what the data show.

Worth knowingThe claim is usually defended by pointing to the original small lab studies and the first meta-analysis of that literature, which a later bias-correction re-analysis described as having concluded willpower depletion was 'robust and medium in magnitude (d = 0.62)' — but that meta-analysis has since been shown to be an artefact of publication bias once bias-correction methods were applied to its own dataset, and it has not survived two much larger, pre-registered, multi-laboratory replication attempts that used different methodologies and both converged on a null result.

No scientific evidence supports the rule. A formal literature search found no proof that every person must 'drink at least eight glasses of water a day'; individual water needs vary substantially based on body size, activity, age, and climate.

Worth knowingHydration itself matters — the refuted element is the specific universal figure, not the importance of adequate fluid intake.

This claim has not actually been tested. No peer-reviewed study measures discipline, willpower, or self-control as an outcome of taking cold showers. The two most-cited cold-shower studies measure something else entirely: one is a randomised trial that looked at sickness absence and quality of life, not discipline, and the other is an untested hypothesis paper about mood and depression, not willpower. The closest indirect evidence, a meta-analysis on training self-control through repeated effortful acts in general, found only a small-to-medium effect that shrank further once publication bias was corrected, and its own authors said the mechanism driving any such effect is poorly understood. So the honest answer is not that cold showers fail to build discipline, but that nobody has actually measured whether they do, which means the claim cannot be confirmed, refuted, or even graded as nuanced or overstated.

Worth knowingThe best-known cold-shower trial did find a real health benefit, a reduction in self-reported sickness absence, but health outcomes are not the same as discipline or willpower, and the trial did not test either.

We decline to rule on this claim as stated. The only controlled evidence measures breathing proxies — apnoea events and snoring — in diagnosed mild-OSA habitual mouth-breathers, not sleep quality in general healthy sleepers, which is the claim's actual subject. Sleep quality was not a primary endpoint in any trial and no controlled study exists in healthy sleepers, so on current evidence the claim can be neither supported nor refuted.

Worth knowingAll positive evidence is confined to mild obstructive sleep apnoea patients who are habitual mouth-breathers; no controlled study has demonstrated benefit in healthy sleepers without a sleep-breathing disorder.