Not vibes.
Real ground.

The classic lab version — subtle death reminders shifting behaviour — failed a large multi-lab replication in 2022. What the field still argues for is the different thing this card asks: open, deliberate reflection. That case rests on a 2012 review whose own authors call it tentative, not on a replicated experiment.

The self-versus-behaviour split is well supported, and Tangney credits Helen Block Lewis (1971) for it herself. What has moved on is the idea that shame is simply the harmful one — researchers now ask when each emotion helps and when it hurts.

Festinger’s 1954 theory covers the comparing — of opinions and abilities. The social-media half is a separate, much newer literature, and it is correlational: it shows the link, not which way it runs.

Bandura’s evidence is that doing the thing is the strongest source of self-belief. That each broken promise measurably erodes something called ’self-trust’ is our extrapolation — that isn’t a measured construct, and the causal direction between confidence and performance is still disputed.

The energetic cost is best evidenced in emotional-labour research: across 95 studies and 494 correlations, “surface acting” — performing an emotion you don’t feel — tracks with impaired well-being at ρ = .39 to .48 (Hülsheger & Schewe, 2011, Journal of Occupational Health Psychology 16(3), 361–389, doi:10.1037/a0022876). That is workplace research under organisational display rules, both sides self-reported by the same person, and the authors say it does not allow causal inference. Authenticity research itself is self-report and correlational, and researchers still disagree on how to measure it.

Sartre and Camus are philosophy, not evidence. The psychology adds a twist worth knowing: searching for meaning predicts more distress, while having found it predicts less.

The strongest finding here is the confidence illusion: searching for an explanation inflates your rating of your own understanding (Fisher 2015, independently replicated in 2023), and using a search engine inflates confidence in your own memory (Ward 2021). Weaker: offloading reduces later recall of what you offloaded — a small lab effect. The 2017 ’phone on the desk drains your thinking’ finding failed to replicate. We claim nothing about judgment; nobody has shown that.

The self-serving bias is large in Western samples and much smaller in East Asian ones. Its reversal is a well-documented feature of depression — but that link is correlational, not proof that one causes the other.

One small 2009 study (49 people), never directly replicated. Later work points the other way — telling people can just as easily increase commitment. It seems to depend who you tell.

That avoiding a feeling tracks with feeling worse is well replicated. That the avoidance is the cause, rather than a companion, is less settled — and the standard questionnaire has been criticised for partly measuring distress itself.

Experimentally raising someone’s intolerance of uncertainty raises their worry — real causal evidence, which is rare here. The same intolerance also shows up in depression and OCD, so it isn’t worry’s private mechanism.

A clinical model rather than a lab finding — and in the same model some anger is primary and healthy, not a cover for anything.

A widely accepted model, built to replace stage theories and the idea that grief must be worked through. The back-and-forth itself is the part still being tested.

A taught DBT skill that Linehan adapted from Zen practice — not a laboratory finding, and no trial has isolated it from the rest of DBT. What lab work does support is the neighbouring move: accepting a feeling beats suppressing it for what happens afterwards.

Measured as a belief, not a behaviour, and the research is correlational. It is also markedly more common in some cultures than others — Joshanloo’s own cross-cultural work is about exactly that difference.

Body-based stress signals are genuinely measurable, and noticing them is a trainable skill. The popular idea that trauma is ’stored’ in muscle or tissue is not established — we are not making that claim.

The strongest evidence on this page: people randomly assigned to ruminate got more depressed, and those assigned to distraction got less. That is a causal design, not a correlation.

Compassion-focused therapy does reduce self-criticism in controlled trials, but the effects are small and most trials compare it to a waiting list rather than to another therapy.

The hunger-and-thirst comparison is the researchers’ own wording, not ours. The evolutionary account is a theory; the health consequences are well replicated.

The link between holding a feeling in and worse outcomes is correlational and modest in size, and it is found in Western samples far more than Eastern ones. We are not claiming feelings build up under pressure — that idea is folk psychology, not a finding.

Disclosure plus a response is the researched route to being known. How concealment reads to other people is our own line, not part of that model.

Bowen’s idea, from clinical family-therapy theory rather than experiment. His own word is differentiation, not boundaries. The self-report measure built from it tracks with lower anxiety and better adjustment — a correlation, not proof of cause.

The claim is well evidenced, but under a different name than most people expect: unmitigated communion, the research on over-involvement with others to the exclusion of the self.

Research separates high standards from fear of falling short. It is the second one that reliably tracks with harm; the first, on its own, often doesn’t. And you cannot randomise people into perfectionism, so this is correlational throughout.

Meta-analysis confirms the rebound after you stop suppressing a thought. The idea that suppression makes it worse while you are doing it only held when the mind was already loaded.

In the lab the cost lands immediately and on the other person — partners of people suppressing showed raised blood pressure and the conversation went worse. It is not universal: the social cost is much smaller in cultures where restraint is the norm.

The research explains why help feels threatening to accept — it can imply inferiority. Calling refusal a wall is our reading, not theirs. People also badly underestimate how willing others are to help.

Sociotropy is not one thing. It splits into liking people and fearing their disapproval, and only the fear half tracks with depression — measured across two samples of about a thousand outpatients. Wanting closeness is not the problem.

Guilt as an interpersonal repair signal is Baumeister’s well-supported thesis. That an unmade apology is that signal with nowhere to go is our image, not a finding either researcher reports.

People who reach out underestimate how much it lands — reliably, though the effect is modest. That finding has been independently replicated in a pre-registered study, which is rare. How silence is read from both ends is our inference.

The reasoning is sound — spent is spent. How reliably the bias can be measured in any one person is less clear: recent work found the classic 1985 scenarios hang together poorly. The original studies were about money and projects, not relationships.

Measured in the lab while people deliberately dwell on an old hurt: heart rate, blood pressure and sweat all rise. That is an immediate reaction, not proof that grudges make you ill. Forgiveness therapy itself outperforms not just waiting lists but other active treatments.

The behaviour holds up. The explanation did not: in 2016 the original authors showed passivity is the default response to prolonged adversity, and what is actually learned is control. Citing researchers who corrected themselves is the point.

Patterns do pass down families. The epigenetic mechanism has not been demonstrated in humans, is widely overstated, and this card does not rest on it — the headline studies are small and hard to separate from ordinary social transmission.

Staying scanned for threat after the danger has passed is hypervigilance, a listed PTSD symptom described long before polyvagal theory existed. We used to credit polyvagal theory here; a 2023 paper argues all five of its founding premises are refutable, and this card never needed the mechanism.

The researchers who named self-forgiveness also named its counterfeit — letting yourself off by denying responsibility or blaming the other person. The benefits show up only when the responsibility is kept.

From one 2004 lab study using a clicking game with virtual doors, and no direct replication has been published. What it showed is that people will pay real money just to stop a door closing.

This is philosophy, not a psychological finding. The story is *Mengzi* 2A.2 (trans. James Legge, public domain); we describe it rather than quote it, because the memorable English wording differs by translator. Psychology has no construct for forcing a process that has its own timescale; what it does have is narrower — spaced learning beats cramming, and grief counselling for ordinary bereavement shows no lasting benefit. We previously credited the marshmallow test here. That was wrong: it studies a person choosing to wait, not a process that cannot be hurried.

Across 92 studies and 6 participant datasets — 564,497 people in 25 countries — roughly one in three avoid medical information they could have, and over 40% do for Alzheimer’s and Huntington’s. The equivalent finding in finance is contested.

Jung’s model comes from clinical work, myth and symbolism, not from experiments. There is no measure of ’the shadow’ and no trial behind it — this is the oldest and least tested idea on the page, and we would rather say so.

The rest cards carry no psychology because they ask nothing. Giving a card that says “stop” a scientific justification would be the exact thing this deck refuses.