Remembered for proving that "personality does not exist", he actually proved something more useful: behaviour comes from the person and the situation together, and who you are predicts your averages, not your next single act.
Walter Mischel's 1968 book Personality and Assessment is one of the most misremembered works in psychology. The story people tell is that Mischel proved personality is a myth. That's just wrong. What he actually showed is narrower and more interesting: if you measure a trait and then try to predict a single action from it, the correlation is usually around 0.30. Modest, quite weak. His argument was that behaviour is produced by the person and the situation together, and that trait scores predict broad tendencies rather than isolated acts.
The work that followed filled in the rest. Epstein (1979) demonstrated the principle of aggregation: any single action is noisy, but average a person's behaviour over many occasions and the stability climbs to high levels. You cannot predict whether someone will be talkative at 3pm on Tuesday, but you can predict how talkative they are in general, and predict it well. Funder and Ozer (1983) then turned the tables on the situationists by recomputing the most famous "the situation overwhelms the person" experiments, the obedience and bystander studies, in the same statistical terms used for traits. Those celebrated situational forces averaged a correlation of about 0.40, not particularly larger than personality traits. Kenrick and Funder (1988) went through the deflationary explanations one by one ("traits are just in the eye of the beholder, just shared stereotypes, just labels...") and found the evidence rejected them. By the 1990s the field had settled on interactionism, and Mischel himself supplied the cleanest version of it. His cognitive-affective model (Mischel & Shoda, 1995) reframed personality as stable if-then signatures: if at work then driven, if at home then disorganised, etc... The behaviour swings by situation, but the pattern of swings is the consistent thing. Fleeson (2001) confirmed it with weeks of real-life data: people show nearly the full range of every trait within a single fortnight, yet each person's average sits almost perfectly still. And traits, far from being useless, predict mortality, divorce, and career success about as well as wealth or IQ do (Roberts et al., 2007).
The remaining argument is about the truce itself. Everyone now says behaviour is "person times *'situation," but that slogan can hide the hard question: when do traits dominate, and when does the situation swamp them? The rough answer is that traits show through in weak, ambiguous, unscripted moments and get flattened in "strong situations" that pull everyone the same way. The effect sizes on both sides are real but moderate, roughly 0.2 to 0.4, and people still argue about how much better measurement could push them.
The 0.30 figure that started it all is about single acts, and quoting it as proof that "personality barely matters" is the original misreading repeated. Aggregation changes the entire picture. And a trait label tells you about someone's tendencies, not what they will do in one specific room on one specific day.
The live question is the interaction itself: when do dispositions win, when do situations win, and can we model the two together well enough to predict a specific act rather than a long-run average?
The practical mistake this layer guards against is treating a person as a fixed type whose behaviour you can read off a label. People are consistent in their averages and their if-then patterns, and genuinely variable moment to moment. Strategy that ignores either half goes wrong.
Segmentation by broad trait labels predicts tendencies, not specific purchases at specific moments. "This is our high-Openness segment" will not tell you what they buy on a given Tuesday. If-then descriptions (this category, this occasion, this mindset) are far more actionable than personality typologies. And because the situation carries real weight, designing the moment of choice, the shelf, the default, the context, often beats trying to select the perfect personality to target.
The dream of reading a voter's psychology off their trait profile and steering them oversells what traits can do at the level of a single action. Personality predicts broad leanings weakly, and the situation (who asks, when, in what frame) carries as much of the load. Treat act-level psychographic micro-targeting as a weak lever, not a magic one.
Behaviour is person times situation, and the situation is the half you can actually redesign. If you want to change what people do, shaping the situation (defaults, friction, prompts) is usually more reliable than sorting people by disposition and hoping the type predicts the act.
Stop asking "what type of person is this?" and start asking "this person, in which situation?" Read tendencies and averages off traits. Read the specific moment off the situation. And when you want consistency, look for the if-then signature rather than a single global label.
.png)