Behavioral Science Dictionary

Attribute framing

Choice, Risk & Value

Calling something '90% lean' beats '10% fat,' though they're identical.

What it means

A framing effect in which an attribute or object is evaluated more favorably when described in positive terms than when the logically equivalent negative description is used. No gamble is involved, unlike risky-choice framing; the manipulation simply relabels the same characteristic. Of the three framing types it is the most consistent in direction — positive labels reliably beat logically equivalent negative ones — though it is also the smallest in magnitude (meta-analysed at about d = 0.26, against roughly d = 0.44 for risky-choice and goal framing), and it shrinks or reverses once the audience has direct experience of what is being described. The usual account is associative encoding: '90% lean' calls to mind leanness and health, while '10% fat' primes fat — so attitudes shift even when respondents can perform the trivial arithmetic linking the two. It matters across marketing, health communication, and policy: the chosen valence predictably moves perceived quality, risk, and acceptability — a tool for honest clarity and a lever for manipulation.

Why it happens

The interest of the standard account lies less in explaining the effect than in predicting where it stops. Levin, Schneider and Gaeth read the label as a retrieval cue: the frame appears to do its work at encoding, supplying the associations the judgment then rests on. That is an interpretation rather than a demonstrated sequence — the framing data show the shift, not the mechanism producing it — but it carries a prediction. A cue matters only while it is the best evidence available. Once something more diagnostic arrives — a taste of the product, a vaccine the reader has already had twice — the associations the label supplied are no longer the best thing the judgment has to go on, and the gap between frames narrows. The effect should therefore run strongest with an unfamiliar object and weakest with a familiar one, which is roughly what the evidence below shows.

What the evidence shows

Levin, Schneider and Gaeth's typology rests on a claim about consistency: of the three framing types, attribute framing is the one whose direction almost always holds, with positive labels beating negative ones, while goal framing's direction remains contested. Consistent is not the same as large. Piñón and Gambara's meta-analysis of 51 studies and roughly 13,500 participants puts attribute framing at d = 0.26 — the smallest of the three types, with risky-choice and goal framing both nearer d = 0.44. A dependable nudge, not a dramatic one. A preregistered replication by Zhu and Feldman with 860 participants reproduced both the framing effect and the claim that numerate people are somewhat less swayed by it — but the numeracy-by-framing interaction was tiny (partial eta-squared = .01, 90% CI [.00, .02]). Numeracy buys a slight buffer, not protection.

Where it breaks down

Levin and Gaeth's original study contains its own limit. The gap between '75% lean' and '25% fat' narrowed once people actually tasted the beef: a frame is decisive only while it is the best evidence in the room. Culture moves it too. Cheon and colleagues ran an attribute-framing task on Korean and North American samples — how many positive reviews a product needed before participants would accept it, with the reviews framed positively or negatively — and found the effect several times larger among Koreans, who required 34% and 67% more reviews under negative framing across two studies, against 9% and 7% for Americans; they attribute the difference to a prevention rather than a promotion regulatory focus. Two implications follow. Effect sizes lifted from Western student samples travel badly, and the frame matters least exactly where you might most want it to work — with an audience that already has direct experience of the thing being described.

Using it in practice

Barnes and Colagiuri gave 1,222 UK adults genuine manufacturer Patient Information Leaflets for COVID-19 boosters, with the side-effect rates inside them framed either positively or negatively — '40% will get a sore arm' against '60% will not get a sore arm' — across three vaccine-familiarity conditions: the same vaccine as prior doses, a familiar one, and an unfamiliar one. Positive framing raised intention only for the vaccine people were least familiar with; for the familiar ones it backfired, cutting intention among the least enthusiastic readers — the authors warn that positively reframing a familiar vaccine should be treated with caution. The lever works where the audience has nothing else to go on — also where relabelling shades most easily into manipulation. Two habits keep the use honest. Pick the frame that matches the decision the reader faces, not the one that flatters the product, then read the opposite frame: if it tells a story you would be embarrassed to put in front of the same person, the flattering label is doing work the evidence does not support. Where the stakes are high, give both.

Examples

Ground beef labeled '75% lean' is rated tastier and less greasy than the same product labeled '25% fat.'

A surgeon who says 'nine in ten patients are alive five years on' wins more consent than one who says 'one in ten are dead within five years' — the same figure, differently coloured.

A courier advertising '98% of parcels arrive on time' wins business from one admitting '2% arrive late,' though the two firms have identical records.

A recruiter working from the scorecard alone rates a candidate who passed eight of ten technical modules above one described as having failed two of ten; a recruiter who has already interviewed the same candidate barely moves between the two descriptions.

A mortgage advertised on the share of applicants it approves draws more enquiries than the identical product disclosing the share it rejects.

First described in Levin & Gaeth (1988); typology by Levin, Schneider & Gaeth (1998).

Key references

  1. Zhu, M., & Feldman, G. (2023). Revisiting the links between numeracy and decision making: Replication Registered Report of Peters et al. (2006) with an extension examining confidence. Collabra: Psychology, 9(1), 77608. doi.org/10.1525/collabra.77608
  2. Barnes, K., & Colagiuri, B. (2022). Positive attribute framing increases COVID-19 booster vaccine intention for unfamiliar vaccines. Vaccines, 10(6), 962. doi.org/10.3390/vaccines10060962
  3. Cheon, J. E., Nam, Y., Kim, K. J., Lee, H. I., Park, H. G., & Kim, Y.-H. (2021). Cultural variability in the attribute framing effect. Frontiers in Psychology, 12, 754265. doi.org/10.3389/fpsyg.2021.754265
  4. Piñón, A., & Gambara, H. (2005). A meta-analytic review of framing effect: Risky, attribute and goal framing. Psicothema, 17, 325-331. www.psicothema.com/pdf/3107.pdf
  5. 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. doi.org/10.1006/obhd.1998.2804
  6. Levin, I. P., & Gaeth, G. J. (1988). How consumers are affected by the framing of attribute information before and after consuming the product. Journal of Consumer Research, 15(3), 374-378. doi.org/10.1086/209174

Where this comes up

Related terms

← All 1001 terms