We Define First, Then We See
Walter Lippmann wrote in 1922 that we define first and then see, and he never said when that stops being true. The condition arrived sixty-four years later in a laboratory, and it is far more severe than the slogan admits.
Answer to the question
What is an ambiguous product experience?
It is evidence from your own trial of a product that honestly supports more than one conclusion, such as a moisturiser you cannot judge after a fortnight. Stephen Hoch and Young-Won Ha found in 1986 that advertising had dramatic effects on quality judgements under those conditions, and no detectable effect when the evidence was unambiguous, where judgement followed the physical evidence alone.
Key takeaways
- In Hoch and Ha's first 1986 experiment, where the evidence from the trial was unambiguous, quality judgements depended on the objective physical evidence alone and were unaffected by advertising.
- Change the same trial to an ambiguous product experience and advertising had dramatic effects on perceived quality, with inspection protocols showing people searching for confirmation of the claim.
- A 2025 meta-analysis led by Tom Schiebler, pooling 150 records covering 58,597 people, found expectations and satisfaction moving together to a moderate degree, with no evidence for a contrast effect.
- Leonard Lee's 2006 three-condition beer study found that disclosing the balsamic vinegar cut preference only when the disclosure came before tasting; told afterwards, preference did not move in any way the study could detect.
You buy the moisturiser, use it for 14 days, and cannot tell whether your skin has changed. That is an ambiguous product experience, and what you conclude from it was mostly settled before you opened the jar.
Walter Lippmann put the inversion into print in 1922. For the most part, he wrote, we do not first see and then define, we define first and then see. He was describing what he called a stereotype, meaning a pattern already in the head that imposes character on raw sensation before it reaches the intelligence, and the sentence sits at Part III, Chapter VI, section 1 of Public Opinion.
Read what the sentence actually claims
Two things fall away whenever that line is quoted. The first is the hedge at the front of it, for the most part, which concedes that sometimes we do see first. The second is more serious. He never says when. There is no condition in the book under which the inversion fails, no test that would tell a reader whether they are in the ordinary case or the exception, and no measurement of how often either happens.
He also meant something far less hostile by the word than a modern reader does. "The abandonment of all stereotypes for a wholly innocent approach to experience would impoverish human life," he wrote. "What matters is the character of the stereotypes, and the gullibility with which we employ them." In the chapter that follows he makes the stronger claim, that the pattern is a defence of social position rather than only an economy of effort, which is one reason a correction can land like an insult.
What the book does not contain is evidence. Across twenty-eight chapters the only controlled demonstration is a staged brawl at Göttingen, which Lippmann received at third hand through F. van Langenhove, and which measured how badly witnesses reported a sudden event rather than how a prior belief shaped what they saw. He supplies the formulation and the history of the idea. The experiments have to come from somewhere else.
Put the claim in a laboratory
Somebody eventually ran it. In 1986 Stephen Hoch and Young-Won Ha published two experiments in the Journal of Consumer Research, treating an advertisement as a tentative hypothesis that a shopper then tests by using the product. The design turned on a single variable, which was whether the trial produced clear evidence about quality or an ambiguous product experience.
> When consumers have access to unambiguous evidence, judgments of product quality are dependent only on the objective physical evidence and unaffected by advertising.
That is the authors' own wording, and it is flatter and harder than any summary of it. Give those same people an ambiguous product experience and advertising had "dramatic effects on perceptions of quality", with inspection protocols showing that people hunted for the very thing the advertisement had told them to expect.
Judgements of product quality were dependent only on the objective physical evidence and unaffected by advertising.
Their second experiment asked where in the sequence the advertising was acting. It worked, they found, "by affecting the encoding of the physical evidence", with some retrieval of advertisement-consistent evidence as well, although less of that.
The quotations above are from the published abstract. We cannot tell you how many people took part, or which two product categories they handled, because the paper is closed, the abstract states neither, and no open version exists in any index we checked. So no sample size appears here.
Name the condition he never stated
Three years later the same pair returned to the moderator and gave it a structure, with the author order reversed, so the credit for that one belongs to Young-Won Ha. Their 1989 paper sets out "structural characteristics that render evidence about product quality open to either one or multiple interpretations", which matters because it makes ambiguity something a researcher can specify in advance rather than discover afterwards.
Stephen Hoch separated two senses of the word in a retrospective essay in 2002, and the distinction is the one trade coverage of this literature never makes.
| Sense of ambiguity | What the shopper faces | What it does to learning |
|---|---|---|
| Vagueness | The evidence is unclear and they know it is | Little. The experience gets discounted as noise |
| Openness to interpretation | The evidence honestly fits more than one reading | A great deal. The prior belief supplies the reading |
Vagueness, in his account, "is not a problem when it comes to learning from experience as long as the consumer recognizes the lack of clarity". It is the second meaning "that makes experience seductive, openness to multiple interpretations". And his wider claim about markets is what makes this the normal case rather than a curiosity: "Quality parity is the norm in many categories." Where two shampoos genuinely work about as well, every trial of either is an ambiguous product experience by construction.
Where two shampoos genuinely work about as well, every trial of either is an ambiguous product experience by construction.
The framework around all of this arrived in 1989, when Hoch and John Deighton described learning from product use as four stages, hypothesising, exposure, encoding and integration, moderated by familiarity with the category, motivation to learn, and the ambiguity of the information environment. It is an argued framework rather than a study, with no sample and nothing tested, and it should be cited as a way of thinking. Deighton had reached the territory first, in 1984, with a sole-authored paper on advertising and evidence whose abstract we could not retrieve, so we can credit his priority and say nothing about his findings.
Watch the effect stop working
The boundaries of the effect are sharper than the slogan survives, and not every murky trial is an ambiguous product experience in the sense that matters. Leonard Lee led a study in 2006 in which pub patrons judged ordinary beer against "MIT brew", the same beer with a few drops of balsamic vinegar in it, either blind, or told beforehand, or told immediately after tasting but before giving a verdict. Disclosure cut preference only when it came before the first sip. Arriving after the experience, it moved nothing the study could detect.
That is the distinction this whole argument needs. Expectation reshaping a sensation as it happens and expectation overriding a verdict already formed are two different claims, and this design separates them.
Expertise can turn the effect around completely. Elizabeth Cowley and Eunika Janus had people taste grapefruit juice and then showed them advertisements describing the brand as orange juice. Less familiar drinkers were somewhat more likely to misremember having tasted a blend. More familiar drinkers became significantly less likely to misremember it, apparently using the misleading advertisement as a cue to pin down what they had actually drunk.
Memory is where the effect does travel. Kathryn Braun showed in 1999 that advertising presented after a trial can be mistaken for the person's own recollection, which she described as an instant source-forgetting problem, so that language and imagery from the advertisement become confused with experiential memory and people come to believe their past use of the product had been as the advertisement suggested.
Even the mechanism has a defender. Joshua Klayman argued in 1987, with Young-Won Ha as co-author, that much of what gets filed as confirmation bias is better described as a positive test strategy, which is the habit of testing cases you expect to have the property you are interested in. Under realistic conditions it is a good heuristic, and its appropriateness, they insist, "must be understood in terms of the interaction between the strategy and the task at hand".
Weigh the limits, then take our side
The only large synthesis available points the same way and is worth reading carefully. Tom Schiebler led a meta-analysis published in 2025 that pooled 150 records covering 58,597 people, and found expectations and satisfaction moving together to a moderate degree, with no evidence at all for the backlash that would occur if a disappointed expectation soured the verdict. William Boulding and colleagues had modelled that pattern in 1999, proposing that prior belief bends the reading of the evidence and then bends the updating that follows. What the meta-analysis supplies is a pooled correlation rather than a test of their mechanism.
Two caveats arrive with that figure and they pull in opposite directions. The first cuts against it: the association was strongest in cross-sectional studies, which is to say in the weakest designs, and weaker in longitudinal and experimental ones. The second cuts the other way, because the authors report an unusual downward publication bias, which in their reading means the true correlation is higher than the one they report rather than lower.
### What this cannot show
Twelve of the fourteen papers behind this piece are closed, which is why no sample size appears above except the meta-analytic one. Nobody has replicated Hoch and Ha; we looked, and found extensions, a formal model and a boundary condition instead, which is not the same thing. The claim that advertising moves a verdict only during an ambiguous product experience has apparently never been meta-analysed on its own.
Ralph Allison and Kenneth Uhl's 1964 beer-tasting study, run by and for a brewer, is the ancestor of every blind tasting since, and because its abstract gives neither sample nor result we are naming it and quoting nothing from it.
Where we come down is with the condition rather than the slogan. Lippmann's sentence is right in the half of the world he hedged towards, which is larger than it sounds, because quality parity, in his words the norm in many categories, makes an ambiguous product experience the ordinary case in all of them. It is wrong wherever the evidence is clear, where advertising did nothing the experiment could detect, and it is unreliable where the person is expert or the claim arrives too late. This is a live effect that was published without its boundary condition, which is a different failure from an effect that died.
Keep the two claims apart, too. That the dearer wine tastes better, and that your coffee tastes different in a white cup, are findings about expectation reaching into a sensation. This article is about something else, which is what you decide your own trial proved, and the same instinct runs on when you defend what you just bought. So the question to carry out of this is which of the judgements you currently hold about the things you own were made from evidence clear enough to have contradicted the claim that sold them, and whether, standing in the shop next time, you would be able to tell the difference.
Common questions
What is an ambiguous product experience?
Evidence from your own trial of a product that honestly supports more than one conclusion, such as a moisturiser you cannot clearly judge after two weeks. Stephen Hoch and Young-Won Ha found in 1986 that advertising shifted quality judgements dramatically under those conditions, and had no detectable effect when the evidence was unambiguous and judgement followed the physical evidence alone.
Does advertising change what you taste?
Sometimes, and the timing decides it. Leonard Lee's 2006 study found that telling pub drinkers their beer contained balsamic vinegar cut their preference only when they were told before tasting. Told straight after tasting, before giving a verdict, their preference did not shift in any way the study could detect, which separates perception from retrospective judgement.
Can expertise protect you from misleading claims?
It can reverse the effect. Elizabeth Cowley and Eunika Janus gave people grapefruit juice, then advertisements calling the brand orange juice. Less familiar drinkers were somewhat more likely to misremember a blend, while more familiar drinkers became significantly less likely to, using the misleading advertisement to fix what they had really drunk. Familiarity was already a named moderator.
Sources
Hoch, S. J., & Ha, Y.-W. (1986). 'Consumer Learning: Advertising and the Ambiguity of Product Experience.' Journal of Consumer Research, 13(2), 221-233. doi:10.1086/209062. [Abstract read in full; author order confirmed at Crossref and OpenAlex, Hoch first; Hoch at University of Chicago Graduate School of Business, Center for Decision Research]: two experiments using product categories giving either ambiguous or unambiguous evidence about quality. With unambiguous evidence, 'judgments of product quality are dependent only on the objective physical evidence and unaffected by advertising'; with ambiguous evidence, 'advertising had dramatic effects on perceptions of quality' and 'induced consumers to engage in confirmatory hypothesis testing and search'. Experiment 2: advertising worked 'by affecting the encoding of the physical evidence', with retrieval of ad-consistent evidence 'to a lesser degree'. NOT VERIFIED: sample sizes, the two product categories, any effect size. Crossref deposits first page 221 only; the range 221-233 comes from RePEc.
Ha, Y.-W., & Hoch, S. J. (1989). 'Ambiguity, Processing Strategy, and Advertising-Evidence Interactions.' Journal of Consumer Research, 16(3), 354-360. doi:10.1086/209221. [Abstract read in full; author order confirmed at Crossref and OpenAlex and it REVERSES, Ha first]: two moderators of advertising-evidence interaction, ambiguity of the evidence and consumer information processing strategies, with 'a theoretical account of ambiguity, identifying structural characteristics that render evidence about product quality open to either one or multiple interpretations'. Experiment 1: ambiguity of the decision environment 'played a key role'. Experiment 2: processing strategies influenced advertising's effect on interpretation. NOT VERIFIED: sample sizes, the processing-strategy manipulation. Pages 354-360 from RePEc; Crossref carries 354 only.
Hoch, S. J. (2002). 'Product Experience Is Seductive.' Journal of Consumer Research, 29(3), 448-454. doi:10.1086/344422. [FULL TEXT READ at faculty.wharton.upenn.edu; sole author; Wharton School, University of Pennsylvania]: an essay, no data, no sample, no hypothesis test. 'Product experience seduces consumers into believing that they learn more than is actually so.' Distinguishes two ambiguities: vagueness, which 'is not a problem... as long as the consumer recognizes the lack of clarity', and 'openness to multiple interpretations', which is the one that bites. 'Quality parity is the norm in many categories.' Cites Boulding, Kalra and Staelin for the claim that prior beliefs 'also influence interpretation of new information, something not consistent with a Bayesian updater'. Hoch does not cite Hoch and Ha (1986) in this paper's reference list, which was read in full; he cites Klayman and Ha (1987) instead.
Deighton, J. (1984). 'The Interaction of Advertising and Evidence.' Journal of Consumer Research, 11(3), 763-770. doi:10.1086/209012. [Published record verified at OpenAlex and the Oxford Academic landing text; sole author; received 1 October 1983, published 1 December 1984. ABSTRACT NOT AVAILABLE and not read]: record, title, sole authorship, journal, volume, issue and pages verified; design, sample and findings NOT VERIFIED. Note this paper is two years EARLIER than Hoch and Ha, not later. Beware the DOI 10.1086/209011, which belongs to a different article.
Hoch, S. J., & Deighton, J. (1989). 'Managing What Consumers Learn from Experience.' Journal of Marketing, 53(2), 1-20. doi:10.1177/002224298905300201. [Abstract read in full; author order confirmed, Hoch first; both at University of Chicago]: a framework, not a study, with no sample and no test. 'what consumers learn from the experience of using products is not a simple matter of discovering objective truth.' Four-stage process (hypothesising, exposure, encoding, integration) with three moderating factors: familiarity with the domain, motivation to learn, and ambiguity of the information environment.
Allison, R. I., & Uhl, K. P. (1964). 'Influence of Beer Brand Identification on Taste Perception.' Journal of Marketing Research, 1(3), 36-39. doi:10.1177/002224376400100305. [Abstract read in full; author order confirmed at Crossref and OpenAlex, Allison first; Uhl at University of Iowa]: run by and for Carling Brewing Company, 'a relatively inexpensive experiment to help dichotomize contributing influences as being either product or marketing oriented'; 'groups of beer drinkers that tasted (drank) and rated beer from nude bottles and from labeled bottles'. NOT VERIFIED: sample size, number of brands, and the result. The abstract states none of them; the paper is closed at SAGE with no open-access location.
Braun, K. A. (1999). 'Postexperience Advertising Effects on Consumer Memory.' Journal of Consumer Research, 25(4), 319-334. doi:10.1086/209542. [Abstract read in full; SOLE author, Kathryn A. Braun; Cornell University]: post-experience advertising conceptualised 'as an instant source-forgetting problem where the language and imagery from the recently presented advertising become confused with consumers' own experiential memories', so that 'through a reconstructive memory process... Consumers may come to believe that their past product experience had been as suggested by the advertising.' NOT VERIFIED: number of experiments, sample sizes, any figure. Pages are 319-334. She later publishes as Braun-LaTour.
Klayman, J., & Ha, Y.-W. (1987). 'Confirmation, disconfirmation, and information in hypothesis testing.' Psychological Review, 94(2), 211-228. doi:10.1037/0033-295X.94.2.211. [Abstract read in full; author order confirmed, Klayman first, University of Chicago; the second author is the same Young-Won Ha as the 1986 and 1989 consumer papers]: a theoretical and analytical paper with no sample. 'many phenomena labeled confirmation bias are better understood in terms of a general positive test strategy', which 'can be a very good heuristic for determining the truth or falsity of a hypothesis under realistic conditions. It can, however, lead to systematic errors or inefficiencies.' Their own scope: 'must be understood in terms of the interaction between the strategy and the task at hand.'
Boulding, W., Kalra, A., & Staelin, R. (1999). 'The Quality Double Whammy.' Marketing Science, 18(4), 463-484. doi:10.1287/mksc.18.4.463. [Abstract read in full; author order confirmed, Boulding first, Duke University]: a normative Bayesian updating model extended so that 'a person's perception of a complex service encounter is a blend of the objective dimensions of the service encounter and the person's prior overall belief about the quality level of the firm', weighted by customer experience level and encounter complexity. 'these prior beliefs also influence their perceptions of the data themselves, which in turn affect their new (updated) overall opinion.' NOT VERIFIED: sample sizes. Scope as the authors state it is service encounters and complexity, not products in general.
Cowley, E., & Janus, E. (2004). 'Not Necessarily Better, but Certainly Different: A Limit to the Advertising Misinformation Effect on Memory.' Journal of Consumer Research, 31(1), 229-235. doi:10.1086/383438. [Abstract read in full; author order confirmed at OpenAlex, Cowley first, The University of Sydney; the second author is Eunika Janus, not Janiszewski]: participants tasted grapefruit juice then saw advertisements claiming the brand was orange juice. 'Less familiar consumers... were somewhat more likely to remember that they tasted an orange-grapefruit juice blend. More familiar consumers were significantly less likely to remember tasting an orange-grapefruit juice blend after seeing the misleading advertisements. The results suggest that the more familiar consumers use the misleading advertisements to help them remember exactly what product was consumed.' NOT VERIFIED: sample sizes, effect sizes.
Lee, L., Frederick, S., & Ariely, D. (2006). 'Try It, You'll Like It: The Influence of Expectation, Consumption, and Revelation on Preferences for Beer.' Psychological Science, 17(12), 1054-1058. doi:10.1111/j.1467-9280.2006.01829.x. [Abstract read in full; author order confirmed, Leonard Lee first, Columbia University; Frederick and Ariely at MIT]: pub patrons evaluated regular beer and 'MIT brew' (regular beer plus a few drops of balsamic vinegar) blind, told before tasting, or told immediately after tasting but before indicating preference. 'Disclosure of the secret ingredient significantly reduced preference only when the disclosure preceded tasting, suggesting that disclosure affected preferences by influencing the experience itself, rather than by acting as an independent negative input or by modifying retrospective interpretation of the experience.' NOT VERIFIED: sample size, preference figures.
Schiebler, T., Lee, N. J., & Brodbeck, F. C. (2025). 'Expectancy-disconfirmation and consumer satisfaction: A meta-analysis.' Journal of the Academy of Marketing Science, 54(1), 91-112. doi:10.1007/s11747-024-01078-x. [Abstract read in full; OPEN ACCESS, full PDF at link.springer.com; author order confirmed, Schiebler first, LMU Munich; Nick J. Lee at Warwick]: meta-analytical path analysis of 150 records, N = 58,597. 'an overall positive relationship between expectations and consumer satisfaction (r = .29 [0.24, 0.34]) and no evidence supporting contrast effects.' The positive correlation was stronger for predictive than normative expectations, for services than goods, and for cross-sectional than longitudinal and experimental studies. 'we found an unexpected downward publication bias, which suggests that the true correlation between disconfirmation and consumer satisfaction is higher than the (already high) estimate we found.' The only quantitative synthesis in the file.
Lippmann, W. (1922). Public Opinion. New York: Harcourt, Brace. [FULL TEXT READ, Project Gutenberg ebook 6456, 634,181 characters, logged at Books/Public Opinion/Public Opinion.md]: Part III, Chapter VI, section 1 for 'For the most part we do not first see, and then define, we define first and then see'; Part III, Chapter VII for the stereotype as a defence of social position and for 'The abandonment of all stereotypes for a wholly innocent approach to experience would impoverish human life'; Part V, Chapter XV, section 4 for 'manufacture of consent'. THIS EDITION HAS NO PAGINATION: cite part, chapter and section only. The book contains no psychological finding of its own; the Göttingen demonstration reaches it at third hand via F. van Langenhove and measures eyewitness accuracy.