A hospital presents two descriptions of the same hypothetical treatment choice. One describes how many people survive; the other describes how many die. The numbers may be mathematically equivalent, yet people can respond differently. A framing effect occurs when a change in how an option is described changes judgment or choice while the underlying information relevant to the decision remains the same. The effect is important because communication can shape preference, but the phrase covers several distinct phenomena and does not mean that wording is always more powerful than facts.
The classic risky-choice problem
Tversky and Kahneman’s 1981 paper showed how preferences can shift when a decision involving uncertain outcomes is stated in terms of gains or losses. Their well-known examples compare logically equivalent outcomes while varying whether people think about lives saved or lives lost. The result challenges a simple account in which preferences depend only on the final outcomes and their probabilities. It does not show that every real choice is irrational, nor that every gain frame makes people cautious in every setting.
The frame supplies a reference point. A gain description makes a certain outcome look like something securely preserved; a loss description can make the same outcome look like a loss to be avoided. Under uncertainty, people may evaluate deviations from that reference differently. Prospect theory provides one influential account of such patterns, but the framing effect is the observation that equivalent descriptions can yield different responses. Theory and observation should remain distinct when judging how much one experiment establishes.
Logical equivalence needs careful checking. If one description omits an important probability, changes the time period, or suggests a different population, the options are no longer simply two frames of the same choice. A genuine test holds the relevant quantities constant. In daily communication, wording often changes more than one thing at once: emphasis, trust, emotional tone, and perceived recommendations may all shift. A reader who notices a changed decision should ask what information actually stayed constant.
A health decision is also not a laboratory vignette. Real patients may know their values, have prior experience, and speak with clinicians. They may care about quality of life, side effects, uncertainty, and the credibility of the evidence as well as survival figures. The classic demonstration identifies a vulnerability in representation, not a prediction that any individual patient will choose one treatment. Ethical communication should present absolute numbers clearly and make room for questions rather than trying to exploit whichever frame produces a desired answer.
Different effects share the name
Levin, Schneider, and Gaeth argued that framing research includes at least three important types: risky-choice framing, attribute framing, and goal framing. Risky-choice studies alter descriptions of options under uncertainty. Attribute framing changes whether a characteristic is expressed positively or negatively, such as a percentage of lean versus fat content. Goal framing emphasizes the benefits of performing an action or the costs of not performing it. These variations need not have one mechanism or the same size of effect.
Attribute framing can influence the evaluation of a single object even when no gamble is involved. Calling a product “80 percent successful” may draw attention differently from “20 percent unsuccessful,” though the underlying rate is equivalent. The wording may change an impression, but the exact context matters. If success and failure refer to different definitions or time periods, the phrases are not equivalent. Good analysis checks the denominator and outcome before concluding that a psychological frame alone caused a difference.
Goal framing appears in public information and behavior-change messages. A message can say that screening may help detect a condition, or that not being screened may miss it. Rothman and Salovey developed a theory about how gain- and loss-framed messages could interact with perceived risk of health behaviors. The theory is informative for designing studies, but it cannot justify a general rule that all loss-framed health messages improve action. Effects depend on behavior, audience, baseline beliefs, and whether the message accurately describes benefits and harms.
Keeping these types separate also prevents confusing framing with deception. An honest account can present the same data in multiple ways. A deceptive account can omit the denominator or make a relative difference appear large while concealing a small absolute difference. Both may affect choice, but only the first is a clean framing contrast. The ethical goal is not to eliminate language, which is impossible; it is to make the representation transparent enough that a reader can reconstruct what the numbers mean.
How strong and general is the effect?
Kühberger’s meta-analysis of risky-decision framing examined how effect sizes varied with features of the task and study design. The synthesis supports the existence of framing effects while showing that details matter: the type of risky option, the wording, the problem domain, and whether people see one or both frames can change outcomes. A single classroom demonstration should not be treated as a universal estimate. The research question is not simply whether framing exists, but under what conditions it is large enough to matter.
Participants in many studies answer hypothetical questions. That is useful for experimental control, but it is a boundary when translating to costly action. People facing a real purchase, medical decision, or public vote may seek information, consult others, or reconsider. They may also be more vulnerable when time is short or numeracy is limited. Neither assumption should be made without evidence from the relevant setting. A controlled vignette isolates the wording effect; real decisions add context and constraints.
Within-person and between-person designs can also differ. If one person sees both the saved and lost versions, they may notice equivalence and compare them. If different groups each see only one version, the contrast reveals how initial presentation shapes response. Neither design is automatically superior for every purpose. The first tests recognition and correction; the second tests initial framing. Reports should say which was used because the interpretation of a changed preference depends on exposure.
De Martino and colleagues examined framing-related choices with functional brain imaging and reported activity associated with the bias in a controlled risky-choice task. Such evidence can help connect choice behavior to neural processes, but it does not locate a universal framing center or divide decisions into “emotional mistakes” and “rational successes.” Brain regions are involved in multiple functions, and a scanner task is narrower than a policy debate or clinical consultation.
Frames in communication and design
A dashboard might show a service as 95 percent available or 5 percent unavailable. Both can be useful, but each directs attention to a different question. A manager needs to know whether the service meets a target and when outages occurred, not merely which wording feels reassuring. Showing both rates, the time period, and the number of users affected can make the presentation more complete. The aim is to help people understand the system rather than engineer a preferred reaction through selective language.
In medicine, absolute counts can clarify a framed percentage. “Ten of one hundred people experienced this effect” makes the denominator visible. If a clinician also explains what the effect means for a person’s circumstances, the patient can consider more than a favorable or unfavorable label. A balanced display does not guarantee neutral judgment, because any order, graph, or comparison creates some emphasis. It does make assumptions open to scrutiny and allows the person to ask which outcomes are missing.
Organizations use frames when describing risk. A project may be called a 70 percent chance of success or a 30 percent chance of failure. The two statements can be equivalent if success and failure exhaust the outcomes, but they may conceal severity, cost, and uncertainty. A prudent decision also asks what happens in the failure state, how the estimate was produced, and whether alternative projects have different risks. Rephrasing the same number is a useful check, not a substitute for better evidence.
Framing is not always an error to be corrected. People have legitimate reasons to focus on losses when a harm is severe or irreversible, and on gains when an opportunity is meaningful. The problem arises when an irrelevant presentation difference makes a choice unstable while the decision maker has not recognized the equivalence. Even then, the correction should respect values. A preference that survives transparent presentation of both descriptions may reflect a real priority rather than a bias.
Testing a framing claim
A well-designed study defines the information held constant, randomizes the wording, and measures a choice or judgment relevant to the hypothesis. It reports the size and uncertainty of the difference, not just whether one comparison crossed a significance threshold. Researchers should also check comprehension: if participants misunderstood the denominator, the effect may be partly informational. If the experiment changes social cues or advice alongside wording, it becomes harder to attribute the result to the frame itself.
For practical decisions, a simple audit can help. Translate percentages into frequencies using the same denominator. State both the gain and loss versions. Ask whether the options remain the same when written neutrally. If a preference changes, identify what changed in understanding or feeling. This is an Institute synthesis for decision hygiene, not a validated intervention that makes every person frame-resistant. A decision may still depend on legitimate risk attitudes and values after the arithmetic is clarified.
Another useful question is who chose the frame. A public agency, employer, advertiser, or clinician may have incentives that differ from the audience’s. Transparency about source and purpose lets readers judge the message in context. A communicator should not hide costs by choosing a favorable wording, and a critic should not assume malicious intent from every positive description. The substantive issue is whether the presentation enables informed choice with the relevant alternatives and uncertainties visible.
Digital interfaces create repeated framing opportunities. A default button, headline, progress meter, or notification can alter what feels normal or risky. Some designs may help users understand a choice; others may steer them toward a benefit for the platform. The classic laboratory effect does not quantify the impact of every interface choice. It offers a reason to test displays against user understanding and to inspect whether the same information is available in complementary forms.
There is also a difference between a frame and a reference class. A survival rate for people in one age group may be numerically correct but inappropriate for someone in another. Changing the comparison group changes the information, not merely the wording. A fair presentation should state who was included in the data and whether the figure applies to the decision at hand. Otherwise an apparently elegant gain-versus-loss comparison can distract from a more serious problem of external validity.
Timing can affect interpretation too. A 90 percent short-term success rate may coexist with substantial long-term complications. If the loss version uses a different interval, the two statements are not equivalent. A decision aid should align time horizons, denominators, and outcome definitions before testing the effect of positive or negative labels. This is especially important when several outcomes compete. Patients, workers, and citizens need the terms of comparison, not merely a persuasive headline.
Finally, recognizing a framing effect should not make people distrust every immediate preference. An initial response can highlight a value that matters, such as fear of a severe loss. The next step is to ask whether that value still guides the choice when the same evidence is represented another way. If it does, the preference may be stable and considered. If it does not, the difference identifies an assumption or reaction worth examining. The test supports reflection without requiring a mechanically frame-invariant human being.
What the evidence does not show
The evidence does not show that framing always dominates facts, that all framing effects share one mechanism, or that an individual will switch choices whenever wording changes. Brain-imaging contrasts do not prove that a single emotion circuit causes the bias. Hypothetical risky-choice studies do not directly establish clinical or political outcomes. Message-framing theories do not justify misleading communication. The strongest claims specify the task, equivalent descriptions, measured outcome, and population.
Framing effects reveal that information has a form as well as a quantity. A choice can look different when the same outcomes are described from different reference points. The practical response is to make equivalence inspectable, disclose denominators, and invite comparison. A person who can see both survival and mortality figures may still prefer one option for sound reasons. The goal is a preference informed by the actual tradeoff, rather than one produced unknowingly by the first phrasing encountered.
Research references
- Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453–458. View source
- Levin, I. P., Schneider, S. L., & Gaeth, G. J. (1998). All frames are not created equal: A typology and critical analysis of framing effects. Organizational Behavior and Human Decision Processes, 76(2), 149–188. View source
- Kühberger, A. (1998). The influence of framing on risky decisions: A meta-analysis. Organizational Behavior and Human Decision Processes, 75(1), 23–55. View source
- De Martino, B., Kumaran, D., Seymour, B., & Dolan, R. J. (2006). Frames, biases, and rational decision-making in the human brain. Science, 313(5787), 684–687. View source
- Rothman, A. J., & Salovey, P. (1997). Shaping perceptions to motivate healthy behavior: The role of message framing. Psychological Bulletin, 121(1), 3–19. View source