Framing is one of the most replicated effects in behavioural science and one of the most oversold in practice. The same facts, presented differently, do change what people choose, but there is not one framing lever, there are three that behave differently, and the effect that looks huge in a lab shrinks hard the moment a competing message is in the room.
That framing changes choices, logically equivalent descriptions producing different preferences, is one of the most replicated findings in the field, and the mechanism is treated in L4-14 (Tversky and Kahneman, 1981); communication research frames the same idea as selection and salience, making some aspects of reality more prominent than others (Entman, 1993). The applied work begins where that leaves off, with two questions the phenomenon on its own does not answer: which kind of frame you are actually using, and how large the effect will be once it leaves the lab.
The first advance is the recognition that framing is not one lever but several. Levin, Schneider and Gaeta sorted framing effects into three types that behave differently, and confusing them is a common practical error (Levin, Schneider and Gaeta, 1998). Risky-choice frames are the Asian Disease kind: the same uncertain prospect described in terms of gains or of losses, where the effect runs through prospect theory and depends heavily on the person's reference point. Attribute frames describe one feature of a single thing in a positive or a negative light (like the same cut of beef labelled "90% lean" or "10% fat") and the positive label reliably makes people judge the thing more favourably. Of the three types this is the most dependable. Goal frames push a specific action by stressing either what you gain by doing it or what you lose by not: a health message that says "screen early and you catch problems in time" versus "skip the screening and you could miss something". These are the weakest and least predictable of the three, and the easiest for a competing message to cancel out. Knowing which type you are using tells you how reliable the effect will be and what it depends on.
The honest and applied heart of the matter is that framing is robust as a phenomenon but wildly context-dependent in magnitude. The large, clean effects come from low-information, low-stakes settings where a single frame is presented in isolation, exactly the conditions of the classic experiments and almost never the conditions of real communication. Druckman's work showed that framing effects are sharply attenuated by competing frames, by the credibility of the source, and by people's prior attitudes: when an opposing frame is also present, or the framer is not trusted, or the audience already has a view, the neat one-frame effect largely dissolves (Druckman, 2001). The mature review of the field reaches the same conclusion, that framing effects are real but conditional, and that the interesting question is not whether framing works but when and for whom it survives contestation (Chong and Druckman, 2007).
The practical consequence is blunt: the controlled single-frame effect is an upper bound, not a forecast. A marketing or messaging experiment that tests one frame against no frame, without the competing messages and real stakes of the field, systematically overestimates what the frame will do in the wild. And the three frame types differ in exposure to this: attribute framing, working on evaluation of a single object, is the most reliable; risky-choice framing depends on the reference point and can even reverse if that point moves; goal framing is the most fragile and the most easily neutralised by a counter-frame.
Effect magnitudes vary by domain, population, and stakes, and shrink under the competition and deliberation that characterise real decisions, so lab estimates travel poorly to the field. Some framing effects reverse when the reference point or the audience changes. And the whole area is culturally contingent: which frames resonate, and which competing frames are already in the air, differ by society and issue, so a frame validated in one market is a hypothesis in another.
Can we predict which frame wins when several compete in a real information environment, rather than measuring one frame in isolation? How do frames interact, cancel, or compound when audiences meet them repeatedly over time? And how much of the measured framing literature would survive being re-run under realistic competition and stakes?
The usable core: framing genuinely changes choices, but it is three different tools with different reliabilities, and every estimate you have seen is probably an overestimate of the field, so match the frame type to the job and test it against the frames it will actually meet. What follows is the toolkit, type by type, then the discipline that keeps you honest about size.
The ethics matter because framing works on logically identical facts, which places it squarely in manipulation territory whenever the framing hides something the audience would want to know. The honest line is the one that runs through this series: a frame that simply makes a true and relevant feature salient is fair game, while one that survives only because the audience does not notice the equivalent, less flattering description ("10% fat" behind "90% lean" is harmless; a financial or health risk reframed to vanish is not) fails the test of whether it would survive being seen clearly. And because framing invites counter-framing, the durable position is usually the honest one.
Attribute framing, the most reliable, for shifting evaluation of one thing. Describe a single attribute in its favourable form ("90% lean", "95% success rate", "most of our energy is now renewable") and evaluation of that object shifts predictably in your favour (Levin, Schneider and Gaeta, 1998). When to use: product and option descriptions, anywhere you are presenting one thing's characteristics. Reliability: the highest of the three, because it works on simple evaluation. Caveat and ethics: it is the same object either way, so keep the framed attribute true and relevant, and remember sophisticated audiences mentally flip "90% lean" to "10% fat", which limits both the effect and the risk of deception.
Risky-choice framing, powerful but reference-point-dependent, for decisions under uncertainty. Frame an uncertain prospect as securing a gain or as risking a loss (Tversky and Kahneman, 1981). Because losses loom larger than equivalent gains, a loss frame tends to push people toward the risk-related option and a gain frame toward the safe one, but this runs entirely through the audience's reference point. When to use: genuine choices under risk (insurance, protective behaviours, investment). Reliability: real but conditional. Caveat: move the reference point and the effect can weaken or reverse, so you must know what the audience treats as the neutral baseline before you can predict which way the frame pushes.
Goal framing, the weakest and most contested, for motivating a specific action. Emphasise the benefit of acting or the cost of not acting (the gain-framed and loss-framed versions of the same appeal), most familiar from health messages. When to use: prompting a specific behaviour, especially detection or protective actions. Reliability: the lowest and most context-dependent of the three, and the most easily neutralised by a competing frame. Caveat: do not build a campaign on a goal-framing effect measured in isolation, because it is exactly the type most likely to evaporate under real-world competition.
Whichever type you use, the effect you will get is almost certainly smaller than the one you read about. Two habits protect you. Test in competitive context: never run a single frame against a blank, always test it against the alternative and opposing frames it will actually face, and with realistic stakes, because a single-cell framing test inflates the effect systematically (Chong and Druckman, 2007). And discount the published effect the way the nudge reckoning taught (L5-13): treat the lab number as a ceiling and plan for a fraction of it. A frame that still moves behaviour when a credible opponent is framing the other way is worth having; one that only works in silence is not a strategy.
Lean on attribute framing, the reliable member of the family, for how you describe products and options, and keep the framed attribute honest, because the "90% lean" move only stays legitimate (and only keeps working) while the flipped version would not embarrass you. Treat risky-choice and goal framing as weaker, more conditional tools, useful but not to be leaned on, and above all test every frame competitively: your real setting has competitors framing the other way, so a frame that wins in a one-cell test may do nothing on the shelf. Discount the effect sizes in the framing literature hard when you forecast, and remember that the durable brand position is the honest frame, since a flattering frame exposed as a half-truth costs more trust than it ever bought.
This is framing's most-studied and most-contested arena, and Druckman's findings are the essential corrective (Druckman, 2001; Chong and Druckman, 2007): the single-frame effects that look decisive in a lab shrink under real political competition, where an opposing frame, a distrusted messenger, and hardened prior attitudes are always present. The implication is that framing is real but not a magic wand, that source credibility often matters more than clever wording, and that a frame only earns its keep if it holds up when the other side is framing hard in the opposite direction. Note that this decision-and-message sense of framing is distinct from the collective-action framing (injustice, identity, agency) that mobilises movements, treated in L5-07. The honest campaign frames truthfully and expects a modest, contested effect, not the lab's headline.
Public-health and civic messaging leans heavily on goal framing (act versus fail to act) and on gain-versus-loss appeals, which is precisely the most fragile corner of the toolkit, so calibrate expectations accordingly and test in the real, noisy information environment rather than a clean survey. Where an attribute frame is available and honest (presenting a genuine statistic in its clearer form), it is the more reliable choice. Two disciplines matter most: test against the competing frames and misinformation the message will actually meet, since a public frame never travels alone; and hold the honesty line with particular care, because a government caught framing to obscure rather than clarify pays in the one currency public messaging cannot afford to lose, which is trust. The global caveat is sharp, since which frames resonate and which counter-frames dominate differ profoundly across societies, so a public-messaging frame validated in one system is only a starting hypothesis in another.
Three habits. Match the frame type to the job: attribute framing to shift evaluation of one thing (reliable), risky-choice framing for genuine decisions under uncertainty (reference-point-dependent), goal framing to prompt an action (weak and contestable). Test competitively, against the opposing frames and real stakes the message will face, never as a single frame against a blank. And keep it honest, making true features salient rather than hiding the equivalent unflattering description, because framing invites counter-framing and the frame that survives contact is almost always the truthful one.
.png)