What Is the Framing Effect?
The framing effect is a change in judgment, preference, or choice caused by the way substantively equivalent information or decision outcomes are described. A person may favor a relatively certain option when its results are expressed as gains, yet favor a risky option when the same results are expressed as losses. Another person may evaluate a treatment more favorably when told that 90 of 100 patients survive than when told that 10 of 100 die. What changes is the informational perspective, even when the numerical relationship remains constant. The foundational account by Tversky and Kahneman (1981) demonstrated that description can alter preferences under controlled experimental conditions.
Framing is a research construct, not a diagnosis, personality disorder, or universal defect of rationality. The term names an experimentally testable relation among a description, a specified decision context, and an observed response. A rigorous finding establishes which substantive features were held constant, which linguistic or presentation features changed, what outcome was measured, and what normative or experimental benchmark makes the difference meaningful. The wider field distinguishes risky-choice framing, attribute framing, and goal framing because these manipulations influence different psychological processes and outcomes Levin, Schneider, and Gaeth (1998).
The central question is how identical probabilities, quantities, or consequences can produce different evaluations when attention is drawn to what is gained, what is lost, what survives, what fails, or what action might prevent an unwanted result. That question belongs to the psychology of meaning, risk, numerical reasoning, decision theory, affect, communication, and choice architecture. For the larger scientific context, see the published Cognitive Biases, Judgment and Decision-Making research chapter.
The Classic Example: Two Ways to Describe the Same Outcomes
The disease problem and the mathematics
Imagine a public health decision affecting 600 people. In a gain-oriented version, Program A certainly saves 200 people. Program B offers a one-third probability of saving all 600 and a two-thirds probability of saving nobody. In a loss-oriented version, Program C certainly results in 400 deaths. Program D offers a one-third probability of no deaths and a two-thirds probability of 600 deaths. A and C specify the same final distribution; B and D also specify the same distribution. The gain and loss statements describe matching prospects from different linguistic viewpoints.
Expected survivors for A and C equal 200. Expected survivors for B and D also equal 200 because one-third times 600 plus two-thirds times zero equals 200. The corresponding expected deaths equal 400 in each option. The options do differ in the distribution of risk: one fixes an intermediate outcome, while the other makes an all-or-none outcome possible. Expected-value equality therefore does not compel every decision maker to be indifferent between the safe and risky option. The specific consistency question is whether the same decision maker changes preference between the equivalent gain and loss representations of the same two prospects, holding relevant information and preferences fixed.
In the 1981 experiments, the characteristic aggregate pattern was greater preference for the sure option in the gain description and greater preference for the gamble in the loss description. This was an important empirical challenge to description-invariant models of choice; the exact percentages in one historical experiment should not be mistaken for population constants Tversky and Kahneman (1981).
What the invariance principle requires
The normative principle of description invariance says that alternative formulations carrying the same decision-relevant facts should lead to the same preference ordering, provided the person's underlying evaluation criteria remain fixed. It is a consistency requirement across representations. It differs from the rule to maximize expected money or expected lives, and it does not prescribe risk neutrality. A person who consistently prefers the certain program in both versions can satisfy invariance; so can a person who consistently prefers the gamble. The experimentally important reversal arises when otherwise equivalent descriptions induce different preferences.
Researchers must test whether the descriptions truly communicate equivalent information to participants. A technically complementary statement can suggest a different causal story, social norm, degree of uncertainty, implied reference point, or omitted outcome. If readers interpret the statements differently, a measured preference difference may partly reflect these pragmatic inferences. Identifying those features is essential to separating a nominally equivalent experimental design from the actual decision problems participants represent DeKay and Dou (2024).
Three Main Types of Framing
Risky-choice framing
Risky-choice framing asks whether people select a certain outcome or a probabilistic one when the choices are represented as gains versus losses. The dependent variable is typically the selection of the risky prospect, a rating of its attractiveness, or willingness to accept risk. The manipulation often changes the verbal encoding of logically corresponding results. The disease problem is the standard example, although studies also use monetary gambles, public safety decisions, and other domains. Gain frames commonly favor certainty and loss frames can increase gambling, but effect size and even direction vary with descriptions, task designs, and populations Kühberger (1998).
Attribute framing
Attribute framing changes the evaluation of a single option by describing one characteristic positively or negatively. An item described as 80% successful can be judged differently from the same item described as 20% unsuccessful. This is not a choice between a risky and a safe prospect; it is a judgment about an attribute. Researchers can ask for quality ratings, attractiveness, confidence, or purchase intentions. The relevant benchmark is the identical underlying attribute distribution, assuming the two statements communicate the same denominator and time horizon. The effect also depends on whether the attribute really has complementary categories and whether the omitted remainder can be inferred correctly Levin, Schneider, and Gaeth (1998).
Goal framing
Goal framing alters the persuasive emphasis surrounding a behavior. A prevention message might highlight the benefits of completing an action, while another highlights the harms of failing to complete it. Outcomes include attitudes, intentions, and observed behavior, which must be analyzed separately. Unlike pure attribute framing, the two messages can activate different goals, concerns, emotional responses, and perceived responsibilities. Therefore a stronger immediate attitude does not establish a stronger real-world behavioral effect. Reviews of health communication show why a single broad slogan about loss messages or gain messages being superior misrepresents the evidence Akl et al. (Cochrane review).
Other uses of the word frame
Framing in political communication and media studies can mean selective emphasis, issue definition, causal attribution, or narrative context. Those forms can change the actual information a reader receives rather than merely restating identical outcomes. Defaults, anchors, order effects, relative-risk ratios, and visual salience may interact with framing, but they should be identified as additional manipulations. The scientifically precise question is always what was held equal across conditions, what changed, and whether the outcome is a judgment, preference, intention, or behavior Guenther, Gaertner, and Zeitz (2021).
How Researchers Explain Framing Effects
Prospect theory and reference-dependent value
Prospect theory models evaluation relative to a reference point rather than treating every outcome as a context-free final wealth or welfare level. Its classic value function is typically concave for gains, convex for losses, and steeper on the loss side in many studied settings. This structure can help account for risk aversion over moderate gains and risk seeking over some losses. Probability weighting adds another component: psychological decision weights need not equal stated numerical probabilities. Yet a formal value function is only part of an explanatory theory; a framing experiment must also specify how words determine the psychological reference point and how outcomes are encoded Kahneman and Tversky (1979).
There is an important theoretical distinction between an experimental framing effect and a generalized claim that losses always loom larger than gains. Loss aversion concerns valuation across gains and losses relative to a reference; risky-choice framing concerns a change in choice across equivalent descriptions. They can be connected without being interchangeable. The recent evidence on complete and incomplete descriptions reveals explanatory limits when a model leaves the interpretation of omitted information or linguistic negation unspecified DeKay (2026).
Gist, verbatim information, and fuzzy-trace theory
Fuzzy-trace theory distinguishes precise verbatim representations from the bottom-line meaning, or gist, that people extract from a problem. In the disease example, people may encode the certain gain as some lives saved, while the risky gain includes a possibility that nobody survives; in the loss description, they may focus on a chance that nobody dies. These qualitative contrasts can influence choice even when numerical expectations are equal. The theory treats meaning extraction as a fundamental cognitive function, and it generates predictions about when deleting zero outcomes, changing comparisons, or emphasizing categorical distinctions alters choices Reyna and Brainerd (1991).
Gist-based explanation is compatible with the observation that people can understand the numbers yet choose differently, but it also requires direct tests. Researchers measure probability judgments, interpretations of sentences, memory for exact numbers versus meanings, and patterns of choice under specially designed wording variants. Model comparisons in 2021 and 2026 demonstrate that theoretical accounts disagree about the representations responsible for particular framing variants Reyna et al. (2021) and DeKay (2026).
Language pragmatics, omitted information, and implied meaning
Descriptions can be mathematically compatible yet pragmatically different. The statement that 200 people will be saved may suggest a lower bound or leave open the fate of the other 400; the statement that 400 will die can suggest a different lower bound. An incomplete description can direct attention to one component and induce the reader to supply a missing component. Researchers therefore distinguish complete and incomplete descriptions, positive and negative terms, matched and mismatched options, and true changes in reference point. This is a major reason identical-looking framing studies sometimes yield different results DeKay and Dou (2024).
Emotion, attention, and motivational relevance
Terms such as saved, died, gained, and lost differ in emotional connotation. They can orient attention toward attractive outcomes, unwanted consequences, or anticipated regret. The emotional response may interact with task urgency, personal experience, and perceived control. These pathways are plausible mechanisms and may be observed in specific designs; a difference in emotion is not automatically the cause of every framing effect. Strong mechanistic evidence requires measuring or manipulating the proposed mediator, showing temporal ordering, and testing alternative explanations. Numeracy research also demonstrates that people can extract affective meaning from numbers while using their numerical structure Peters et al. (2006).
Processing style and dual-process accounts
Some descriptions of framing invoke a fast, intuitive process opposed by a slow, reflective process. Dual-process theories provide useful hypotheses about cognitive control, attention, and effort, but the intuitive-versus-deliberative distinction alone cannot predict the magnitude of a particular gain-loss reversal. Fuzzy-trace theory, for example, assigns sophisticated meaning-based processing a central role; additional deliberation can preserve a gist representation rather than simply replace it with arithmetic. Explanations improve when they identify the exact representation and decision rule instead of treating the labels System 1 and System 2 as a completed causal account Reyna and Brainerd (2011).
The Research Evidence: From Classic Experiments to 2026
The original experiments and early medical research
Tversky and Kahneman's 1981 experiments established that preferences could reverse when equivalent alternatives were presented from different perspectives. The paper used monetary and life-outcome problems to place descriptive models of human choice in direct tension with the assumption of invariant preferences. It remains foundational because it specified experimentally manipulable descriptions and measurable decisions, not because every individual responds in one predetermined way Tversky and Kahneman (1981).
McNeil and colleagues studied treatment descriptions among 238 patients, 491 graduate students, and 424 physicians asked to choose hypothetical lung-cancer treatments. The appeal of surgery relative to radiation was affected by survival versus mortality language and by other features of the presentation, including whether treatment names were supplied and whether outcomes appeared as life expectancy or cumulative probabilities. These contrasts also warn against attributing every observed difference to one isolated word: information format and contextual cues were studied alongside outcome valence McNeil et al. (1982).
The 1998 synthesis: heterogeneous but replicable patterns
A major meta-analysis synthesized 136 empirical papers, almost 30,000 participants, and 230 effect sizes. Kühberger reported an overall small-to-moderate risky-choice framing effect alongside consequential differences between experimental designs. Reference-point manipulations, outcome salience, response mode, and problem structure were especially relevant. This synthesis supports treating framing as a family of conditional experimental effects. Its average should not be exported uncritically to a particular medical consent form, negotiation, financial decision, or interface Kühberger (1998).
Levin, Schneider, and Gaeth's typology further showed why risky-choice, attribute, and goal-framing studies cannot be treated as measures of one homogeneous mechanism. A study of product ratings, a choice between lotteries, and a campaign to increase screening behavior make different theoretical and practical demands. A synthesis combining their effect sizes without distinction can obscure more than it clarifies Levin, Schneider, and Gaeth (1998).
The 2024 comparison of wording variants
DeKay and Dou directly tested how the details of option descriptions shape framing effects. Their 2024 investigation used online and student samples of 906 and 521 participants and examined 81 risky-choice framing variants based on combinations of nine gain-frame and nine loss-frame descriptions. Whether positive or negative outcomes were explicitly mentioned explained substantial variation: some designs amplified the usual effect, others weakened or reversed it. Nonetheless, a residual framing difference remained in a balanced design. These results challenge explanations based solely on a fixed gain-versus-loss value curve or solely on the mismatch of descriptions DeKay and Dou (2024).
The 2025 meta-analysis and interpretation of its odds ratio
A 2025 random-effects meta-analysis combined 40 studies with 17,416 participants and found a pooled association between framing and risky decisions (odds ratio 2.467 under the study's coding). Domain and age moderated its results; effects were stronger for life-and-death problems than for some monetary or study-related problems, and greater in adolescent than adult samples in the included comparisons. Culture was not detected as a moderator in that analysis. An odds ratio describes a contrast in odds under a specified outcome definition, not a probability ratio, and it does not mean that exactly 2.467 times as many people everywhere will take a risk. Interpretability depends on the coded direction, base rate, task designs, and between-study variation Ding, Li, and Qiao (2025).
The 2026 replication with complete descriptions
DeKay's 2026 replication and extension used 1,697 participants in a sample well matched to U.S. Census characteristics, including 477 participants making choices between completely described options. It reproduced the 2024 patterns closely and found that framing persisted even when gain and loss descriptions were matched and complete. This matters because merely adding the complementary outcome does not reliably erase every risky-choice framing difference. The author compared theory specifications and found that a rescaled explicated valence account accommodated the replicated patterns better than the other specifications considered. This is evidence about the tested formulations and models, rather than a final settlement of all framing mechanisms DeKay (2026).
What replication does and does not establish
Replication strengthens confidence that a pattern arises under specified procedures; transport to other populations and real-life decisions remains a separate question. A repeated questionnaire effect does not automatically establish durable influence on a person's financial welfare, medical treatment, or civic behavior. Conversely, a weak effect in one setting does not refute the existence of the effect in a distinct type of framing. Strong evidence synthesis separates stimulus wording, denominators, task incentives, within-person versus between-person designs, participant demographics, and the outcome measure. It also looks for publication bias, preregistration, open materials, and model predictions that distinguish competing accounts.
How Framing Is Measured
Randomized wording experiments
The common experimental design randomly assigns participants to versions of a choice problem and compares responses. Researchers may compute the difference in the percentage choosing the risky option, an odds ratio, or a model-estimated coefficient. In a between-subjects design, each participant sees only one frame, reducing direct recognition of the manipulation; in a within-subjects design, the same participant can reveal preference reversals but may also notice equivalence, remember prior answers, or respond to repetition. Preregistration should identify the primary outcome and analysis before the data are examined.
Construct validity and denominator checking
Experiments must specify whether outcomes are identical in every materially relevant respect. A survival percentage must refer to the same population and period as the corresponding mortality percentage. A 90% treatment success rate and a 10% failure rate are complementary only if success and failure are exhaustive categories defined the same way. Relative-risk reductions and absolute-risk reductions are generally different statistics; presenting them as frames of the same fact requires reconstructing the same underlying event probabilities. Good methodology distinguishes altered language from altered information content.
Size, heterogeneity, and individual differences
A well-designed report gives outcome percentages for both frames, an uncertainty interval for their contrast, and enough procedural detail to reproduce it. Investigators should distinguish statistical uncertainty from variation among tasks. Age, numeracy, personal involvement, expertise, cultural context, prior beliefs, and emotional salience may modify framing in particular datasets. Peters and colleagues linked numeracy to reduced susceptibility in several judgment tasks, but numerical skill does not ensure frame-invariant behavior in every decision. The 2025 meta-analysis detected specific age and domain moderators but did not establish a universal cultural null effect Peters et al. (2006); Ding, Li, and Qiao (2025).
Response measures and real-world endpoints
Judged attractiveness, risk perception, stated intentions, hypothetical choice, actual behavior, and long-term consequences are different dependent variables. Effects that appear in rating scales can be smaller, larger, or qualitatively different in incentivized choices. Researchers should report the behavioral measure actually observed. In health contexts, whether a message changed a stated preference is distinct from whether treatment uptake, adherence, comprehension, decisional conflict, or health outcomes changed.
Framing in Medicine and Public Health
Survival, mortality, and treatment preferences
Health choices make framing consequential because people must integrate benefits, harms, probabilities, and personally important outcomes. Expressing an outcome as a survival rate or a mortality rate can alter the felt attractiveness of treatment, even when both quantities describe the same event distribution. Patients and clinicians can be affected; expertise does not eliminate the need for well-designed communication. The study by McNeil and colleagues established a historical demonstration, while later research has separated outcome perception from the harder endpoint of real treatment behavior McNeil et al. (1982).
Why the Cochrane review matters
The Cochrane review of health information framing included 35 studies with 16,342 participants and distinguished attribute from goal framing and understanding from behavior. It found little consistent evidence that either type of positive-versus-negative framing reliably changes health consumers' behavior. Some intermediate measures varied; for instance, selected screening messages changed perceived effectiveness, but pooled behavioral effects were small and uncertain. The conclusion supports balanced patient information and undermines universal claims that negative messages always outperform positive ones Akl et al. (Cochrane review).
A separate meta-analytic review of gain- and loss-framed health messages reported a small advantage for gain frames in some prevention behaviors (r = 0.075), without corresponding general advantages for attitudes, intentions, or detection behaviors; this corrected estimate incorporates the published 2013 erratum. This result is domain-specific and cannot be applied as a rule to every vaccine message, screening choice, or individual treatment decision Gallagher and Updegraff (2012; corrected 2013).
Ethical presentation and informed choice
An ethical decision aid gives the relevant absolute probabilities, common denominators, a consistent time horizon, and both benefit and harm descriptions when useful. It communicates evidence quality and the range of uncertainty without selectively highlighting only the preferred outcome. Balanced disclosure supports the reader's own values, including legitimate differences in risk tolerance. In high-stakes medical contexts, an accurate bilingual or accessible presentation, numeracy support, and opportunities to ask questions can matter more than optimizing a slogan for immediate persuasion.
Framing in Consumer Choices, Organizations, and Public Communication
Product information is often presented as a positive quality attribute or its negative complement. A company can highlight customers who renewed, while a competitor highlights those who left. The comparability of these statements depends on population definition and observation period. Framing can affect perceived value and attention, but claims about purchases require tests of actual purchasing behavior and attention to price, reputation, and availability. Such results belong to consumer and economic psychology when the central intent concerns marketplace valuation, while the general mechanism remains within judgment and decision-making.
Organizations also present projects as gains to be secured or losses to be avoided. Executives and teams may respond differently to structurally equivalent forecasts when one report emphasizes opportunity and another emphasizes downside. Yet organizational decisions introduce incentives, authority, accountability, group norms, and unequal information. A change in final decisions cannot be assigned to framing without measuring those competing influences. In legal settings, context effects on forensic interpretation and institutional judgments have their own specialized literatures; see Forensic Evidence and Cognitive Bias: How Context Can Contaminate Judgment and Judge Decision-Making: Psychology, Expertise, and Bias.
Risk communication in public policy must confront competing outcomes and different social distributions of benefits and losses. Two descriptions that are equivalent for an aggregate total can mask distributional differences that matter morally or politically. A framing audit must therefore examine whose outcomes count, which comparisons are salient, and whether the options really contain the same values and consequences. Psychological experiments establish mechanisms under controlled conditions; ethical and policy evaluation additionally requires explicit values and institutional context.
Framing in Human–AI Decision Support
Digital systems select words, rankings, charts, and summaries before presenting evidence to users. An AI-generated explanation may describe a prediction as a chance of success or a chance of failure, place a possible gain beside an avoidable loss, or omit the complementary probability. These design choices can alter users' subjective interpretation even when the underlying model output is unchanged. The framing effect therefore supplies testable hypotheses for human–AI interaction and interface evaluation, rather than a ready-made estimate of the impact of every AI system.
A responsible audit preserves numerical equivalence, displays denominators and uncertainty, checks whether one option receives more complete descriptions than another, and compares outcomes across experimentally randomized interface variants. It evaluates calibrated understanding and consequential choices as well as satisfaction ratings. Authority cues, automation bias, and confidence displays are additional mechanisms that should be tested separately. For the specialized issue of algorithmic authority, see AI as Authority: Trust, Expertise, Automation Bias, and Human Decision-Making.
The wider Artificial Era creates a setting in which human and artificial systems increasingly participate in decision environments. Psychological findings about human susceptibility to wording remain empirical claims that require direct testing; broader philosophical characterizations of the Artificial Era belong to the author's published theoretical work and should be attributed and assessed on their own terms. The present article concerns the measured psychology of human judgment, including its interaction with technological presentations.
Can Framing Effects Be Reduced?
Translate both frames into a common representation
A practical first step is to write both complementary statements: 90 of 100 survive and 10 of 100 die; 200 of 600 saved and 400 of 600 not saved. Then compare actual distributions of outcomes rather than their first verbal labels. Explicit denominators, time horizons, and consistent units help expose statements that only appear equivalent. This procedure follows the normative requirement of description invariance and the scientific finding that omitted or selectively emphasized outcome components matter, but an improvement in transparent presentation should not be treated as proof that every individual becomes immune to framing DeKay and Dou (2024).
Separate the probability distribution from the value judgment
For risky choices, write each option's outcomes and probabilities before assessing preferences. A person can explicitly value security, freedom from catastrophic outcomes, or opportunities for an unusually good result. Their values may legitimately favor the sure prospect or the gamble. The objective is to make those preferences consistent across equivalent descriptions, rather than to force expected-value maximization. For health decisions, use absolute numbers and discuss harms as well as benefits; the Cochrane review supports balanced communication as a prudent practice while recognizing limits in behavioral evidence Akl et al. (Cochrane review).
Consider alternative descriptions and counterfactuals
When a decision feels compelling, reconstruct it from the opposite verbal perspective: what would be lost, gained, prevented, or preserved? Ask whether a change of wording alone changes the answer and, if so, which assumption or value accounts for the difference. In formal work, teams can specify in advance how alternative descriptions will be generated and test whether decisions are stable across them. These are transparent decision-quality practices. Evidence of persistent effects in fully described experiments shows that they should be evaluated empirically, not promoted as universally successful debiasing therapies DeKay (2026).
Improve numerical communication without overpromising training transfer
Numeracy and interpretation of quantities are relevant to framing, yet the effectiveness of a particular debiasing lesson depends on its content, target task, retention period, and transfer outcome. A score improvement immediately after a training task is weaker evidence than improved decisions months later in a new setting. Researchers and designers should assess comprehension, calibration, actual behavior, and possible adverse effects of information overload. The evidence on numeracy and framing supports careful testing of quantitative aids, not a claim that intelligence, reflection, or expertise alone eliminates framing Peters et al. (2006).
Scientific Limitations and Open Questions
Framing experiments depend on the details of their language, participant interpretation, incentive structure, social setting, and dependent variables. The broad body of evidence demonstrates that some descriptions systematically influence choice, while the size and direction of effects vary enough to demand precise classification. A meta-analytic average is not a transferable causal coefficient for a specific webpage or consultation. Between-subject differences cannot always establish within-person preference reversals, and hypothetical choices are not a direct measure of actual personal outcomes.
A second limitation concerns the meaning of equivalence. Two descriptions may be formally equivalent to a researcher yet not convey identical information to a participant because of implied context or unmentioned possibilities. The growing experimental literature treats this problem as an object of investigation rather than simply dismissing it as participant error. Recent work with matched and complete wording supports the continued relevance of framing after important linguistic confounds are addressed DeKay and Dou (2024); DeKay (2026).
A third limitation concerns generalizability. The finding of no detected cultural moderator in one 2025 meta-analysis cannot establish equivalence across all cultures, translations, institutions, or decision domains. Age effects depend on which life stages and tasks were studied; expertise effects depend on what information professionals use. Evidence should be labeled established for well-replicated laboratory patterns, limited or context-dependent for specific practical transfers, and contested where rival cognitive models produce different predictions. The most productive research specifies observable contrasts that could discriminate mechanisms.
Frequently Asked Questions
Is the framing effect a cognitive bias?
It is commonly classified among cognitive biases when a systematically different response violates a well-specified representation-invariance criterion. The classification depends on the experimental benchmark and on whether the descriptions really convey the same decision-relevant information. A framing response by itself is not a mental-health symptom or diagnosis.
Does positive framing always produce safer decisions?
No. The familiar sure-gain and risky-loss pattern occurs in many studied risky-choice problems, but alternative descriptions can weaken, eliminate, or reverse the effect. Health-message experiments also show different results depending on whether the outcome is perception, stated intention, or actual behavior DeKay and Dou (2024); Akl et al. (Cochrane review).
Is framing the same as loss aversion?
No. Loss aversion refers to a particular pattern of valuation relative to a reference point. Framing describes a response difference between presentations of a decision. Reference-dependent valuation may contribute to some framing effects, but linguistic interpretation and gist-based processing provide additional, experimentally examined explanations.
How does framing differ from anchoring?
Anchoring concerns the influence of an initial numerical or informational reference on subsequent judgments. Framing changes how an outcome or choice is described or represented. A task can contain both: an initial benchmark can act as an anchor while gain-loss language changes the emotional or categorical meaning of options. Their causal contributions require separate controls.
Are experts or people with high numeracy immune?
No single qualification guarantees invariance. Some studies associate numerical skill with lower susceptibility in specific tasks; clinicians and other trained professionals have also shown framing-dependent responses. The relevant outcome is performance under a given decision design, not a fixed judgment about someone's intelligence or competence Peters et al. (2006); McNeil et al. (1982).
Can a complete explanation of both outcomes eliminate the effect?
Showing all components helps remove some linguistic asymmetries, but controlled replication in 2026 found framing effects even with matched, complete descriptions. Complete information improves transparency without guaranteeing a particular psychological outcome DeKay (2026).
What is the best way to recognize framing in everyday decisions?
Write the options with absolute numbers, common denominators, explicit probabilities, and the same time horizon. Re-express gains as corresponding losses and ask whether any preference change reflects new information, a real value difference, or only a changed description. In high-stakes decisions, use independent, balanced evidence and allow time to examine uncertainties.
Conclusion
The framing effect establishes that the linguistic and conceptual representation of a choice can become part of the psychological decision process. Controlled research has documented preference shifts across equivalent gain and loss descriptions, the influence of positive and negative attributes, and conditional effects of goal-oriented persuasion. Decades of evidence, including the 2024 experiments, 2025 meta-analysis, and 2026 replication, support the phenomenon while sharpening the distinctions among reference points, gist, pragmatics, valence, and actual information content. Better decisions begin with transparent representations, explicit probabilities, and stable criteria for comparing the same outcomes. Their effectiveness remains a matter for measurement in the settings where choices are actually made.
Related Articles
Decision-Making Under Risk and Uncertainty: Models and Examples — Expected utility, subjective probability and descriptive models for comparing choices with known and unknown outcomes.
Cluster Cognitive Biases, Judgment and Decision-Making — the central scientific chapter on judgment, risk, heuristics, and decision theories.
Forensic Evidence and Cognitive Bias: How Context Can Contaminate Judgment — application of contextual bias to forensic interpretation.
Judge Decision-Making: Psychology, Expertise, and Bias — specialized evidence on institutional judgment.
AI as Authority: Trust, Expertise, Automation Bias, and Human Decision-Making — technological authority and human evaluation of recommendations.
References
Akl, E. A., Oxman, A. D., Herrin, J., Vist, G. E., Terrenato, I., Sperati, F., Costiniuk, C., Blank, D., and Schünemann, H. (2011; Cochrane presentation 2021). Framing of health information messages. Cochrane Database of Systematic Reviews, CD006777. DOI.
DeKay, M. L. (2026). Risky-choice framing effects persist when option descriptions are matched and complete: A replication and extension of DeKay and Dou (2024). Psychonomic Bulletin and Review, 33, Article 139. DOI.
DeKay, M. L., and Dou, S. (2024). Risky-choice framing effects result partly from mismatched option descriptions in gains and losses. Psychological Science, 35(8), 918–932. DOI.
Ding, X., Li, M., and Qiao, J. (2025). Meta-analysis examining the relationship between framing effect and risky decisions. Journal of Behavioral and Experimental Economics, 116, 102351. DOI.
Gallagher, K. M., and Updegraff, J. A. (2012). Health message framing effects on attitudes, intentions, and behavior: A meta-analytic review. Annals of Behavioral Medicine, 43, 101–116; corrected in 46(1), 127 (2013). Original DOI. Erratum.
Guenther, L., Gaertner, M., and Zeitz, J. (2021). Framing as a concept for health communication: A systematic review. Health Communication, 36(7), 891–899. DOI.
Kahneman, D., and Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291. DOI.
Kühberger, A. (1998). The influence of framing on risky decisions: A meta-analysis. Organizational Behavior and Human Decision Processes, 75(1), 23–55. DOI.
Levin, I. P., Schneider, S. L., and 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. DOI.
McNeil, B. J., Pauker, S. G., Sox, H. C., Jr., and Tversky, A. (1982). On the elicitation of preferences for alternative therapies. New England Journal of Medicine, 306, 1259–1262. DOI.
Peters, E., Västfjäll, D., Slovic, P., Mertz, C. K., Mazzocco, K., and Dickert, S. (2006). Numeracy and decision making. Psychological Science, 17(5), 407–413. DOI.
Reyna, V. F., and Brainerd, C. J. (1991). Fuzzy-trace theory and framing effects in choice: Gist extraction, truncation, and conversion. Journal of Behavioral Decision Making, 4(4), 249–262. DOI.
Reyna, V. F., and Brainerd, C. J. (2011). Dual processes in decision making and developmental neuroscience: A fuzzy-trace model. Developmental Review, 31(2–3), 180–206. DOI.
Reyna, V. F., Brainerd, C. J., Chen, Z., and Bookbinder, S. H. (2021). Explaining risky choices with judgments: Framing, the zero effect, and the contextual relativity of gist. Journal of Experimental Psychology: Learning, Memory, and Cognition, 47(7), 1037–1053. DOI.
Tversky, A., and Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453–458. DOI.