{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-09-20T04:28:12.530Z"},"content":[{"type":"documentation","id":"7316f710-5763-47d4-9a61-80beea563af3","slug":"memory-conformity-group-research","title":"Memory Conformity: Why Participants Who Talk to Each Other Start Remembering the Same Thing (2026)","url":"https://www.koji.so/docs/memory-conformity-group-research","summary":"Memory conformity is when people who discuss a shared experience later report details only another person observed. Gabbert, Memon and Allan (2003) found 71% of witnesses who discussed an event recalled items acquired in the discussion; Wright, Self and Justice (2000) found the more confident partner usually persuaded the other. In customer research, co-witnesses include focus groups, buying committees, churned accounts and communities. Collect individual accounts in parallel, ask the source question, and count independent groups rather than mentions.","content":"**Answer first:** memory conformity is what happens when people who experienced the same thing talk about it, and then each remembers details that only the other person saw. In the best-known study, **71 percent** of witnesses who discussed an event went on to report items they could only have learned from their partner. For customer research this matters because customers talk to each other constantly - in focus groups, in buying committees, in community forums and Slack channels, in advisory boards - and then give you accounts that sound independent but are not. The fix is to **collect each person's recall before they have compared notes**, ask where each detail came from, and count people who talked to each other as one source, not several.\n\nMost research teams know focus groups suffer from groupthink. Memory conformity is a different and less visible problem, and it follows participants out of the room.\n\n## Groupthink bends opinions. Memory conformity bends facts.\n\n| | Groupthink / opinion conformity | Memory conformity |\n|---|---|---|\n| **What changes** | What people say they prefer or believe | What people report actually happened |\n| **Where it shows** | In the group session | In later, individual accounts - after the group has dispersed |\n| **Does the participant know?** | Often, at some level (\"I went along with it\") | Usually not - the borrowed detail feels like their own memory |\n| **Typical research damage** | Inflated consensus on a concept | Inflated corroboration of an event, a workflow or a cause |\n\nThe existing guidance on [focus group disadvantages](/docs/focus-group-advantages-disadvantages) covers groupthink in the room. Memory conformity is worse in one respect: you cannot moderate it away inside the session, because the contamination has usually happened before the session starts.\n\n## The evidence\n\n### Gabbert, Memon and Allan (2003): the 71 percent study\n\nThe study that defined the modern research programme is \"Memory conformity: can eyewitnesses influence each other's memories for an event?\" (*Applied Cognitive Psychology* 17, 533-543, 2003). Its design is worth understanding because it removes any doubt about where the contaminated detail came from.\n\n- Pairs of participants watched a video of the same event, but **each person saw a different version**, containing unique items visible only to them.\n- In one condition, pairs were encouraged to discuss the event before each person individually completed a recall test. In the control condition, pairs did not discuss it.\n- In the discussion condition, **71 percent** of witnesses \"went on to mistakenly recall items acquired during the discussion.\"\n- There were **no age-related differences** in susceptibility between younger (18-30) and older (60-80) participants.\n\nBecause each unique item was physically unavailable to one of the two witnesses, any report of it by that witness could only have come from the conversation. That is the exact structure of a B2B customer study in which the admin and the end user saw different halves of the same rollout.\n\nLooking back on the paper in 2011, the authors described what it had demonstrated:\n\n> \"the opinions of others affect our behaviour but can also quite dramatically alter our reports as well as our beliefs and memories.\" - Gabbert, Memon and Allan, *Applied Cognitive Psychology* 25(S1), 2011\n\n### Wright, Self and Justice (2000): the confident one wins\n\nAn earlier study, \"Memory conformity: exploring misinformation effects when presented by another person\" (*British Journal of Psychology* 91(2), 189-202, 2000), found the same pattern and added a detail that research teams should find uncomfortable. Pairs viewed a crime in which half saw an accomplice and half did not. Initial memories were very accurate, \"but after discussing the crime with the other person in the pair (who saw a slightly different sequence), most pairs conformed.\" And: \"Confidence ratings strongly predicted which person in the pair persuaded the other.\"\n\nIn other words, when two accounts merge, the merged version is not the more accurate one. It is the one held by the more confident person. In a customer organisation, the most confident narrator of what went wrong is often the most senior person, the champion who bought your product, or the person who complained loudest - none of whom is guaranteed to have seen the most. The question of when confidence can and cannot be trusted is the subject of [Participant Confidence Is Not Accuracy](/docs/participant-confidence-vs-accuracy).\n\n## Where co-witnesses hide in customer research\n\nAny time two participants share an experience and a channel to talk about it, you have a co-witness pair. The common ones:\n\n1. **Focus groups and workshops.** The obvious case: every participant hears every other participant's account before recalling their own in detail.\n2. **Buying committees.** You interview the economic buyer on Monday and the end user on Thursday. In between, they discussed your questions over lunch. Thursday's account now contains Monday's.\n3. **Win-loss and churn interviews inside one account.** The team that decided to leave you has usually agreed on a single story of why - often written up internally - before any of them speaks to you. See [churned customer interviews](/docs/churned-customer-interviews).\n4. **Customer communities and forums.** A thread in which one user describes a bug in vivid detail gives every reader a ready-made memory of that bug.\n5. **Advisory boards and repeat panels.** Members who meet quarterly converge on a shared account of your product's history.\n6. **Your own prompts.** \"Other customers have told us that...\" makes you the co-witness. The research version of this is covered in [the misinformation effect](/docs/misinformation-effect-user-interviews).\n7. **Public reviews read before the interview.** Participants who checked reviews to \"prepare\" arrive with other people's experiences available to borrow.\n\n## Five design rules for independent accounts\n\n1. **Individual recall before any group activity.** If you must run a group session, collect each person's written or recorded account first. The individual account is the evidence; the group is for generating ideas.\n2. **Interview co-workers in parallel, not in sequence.** Compress the window between the first and last interview in the same account so there is no time for the story to consolidate. Days between interviews in one company are days for conformity to work.\n3. **Ask the source question.** After the free-recall section, ask directly: \"Have you talked about this with anyone else - colleagues, other users, online?\" and \"Did you see that happen yourself, or did someone tell you?\" Tag every detail the participant did not observe first-hand.\n4. **Never cite one participant to another.** Do not share earlier interviews' findings during recall. If you need a reaction to another customer's view, ask for it *after* the participant's own account is recorded.\n5. **Count sources, not mentions.** Eight people from one Slack channel who agree are not eight independent confirmations. Group participants by who could have talked to whom and count independent groups, the same discipline described in [evidence corroboration from independent sources](/docs/evidence-corroboration-independent-sources).\n\n## Detecting conformity in data you already have\n\nYou can often find memory conformity after the fact by looking for three signatures.\n\n- **Shared unusual phrasing.** Two participants in the same account using the same distinctive phrase (\"the Tuesday meltdown\", \"the export wall\") almost certainly got it from one another or from an internal document.\n- **Details outside the participant's line of sight.** An end user describing the contract negotiation, or an executive describing a specific error message, is reporting something they probably did not see. Check each detail against the participant's role.\n- **Suspiciously tight agreement.** Real independent recall is messy: people disagree on order, dates and emphasis. Perfectly consistent accounts across a group are a warning sign, not a strong finding.\n\n## How Koji helps\n\nTraditional research makes memory conformity hard to avoid: human-moderated interviews have to be scheduled one after another, focus groups deliberately put co-witnesses in one room, and legacy survey tools capture no information about where a respondent's knowledge came from. Koji is built around individual, asynchronous conversations, which changes the structure of the problem.\n\n- **One participant, one conversation.** Every Koji interview is one-to-one with the AI interviewer, so nobody hears another participant's account during their own recall.\n- **Parallel instead of sequential.** Because AI-moderated interviews run asynchronously, everyone in a buying committee or a churned account can be invited at once and complete within hours instead of across a fortnight of calendar slots. The window for stories to consolidate shrinks accordingly.\n- **The source question built in.** Add a `yes_no` question (\"Have you discussed this with colleagues?\") and a `single_choice` question on how the participant learned of each issue (saw it myself / a colleague told me / read about it) to any study. Koji supports six structured question types - `open_ended`, `scale`, `single_choice`, `multiple_choice`, `ranking` and `yes_no` - and the [structured questions guide](/docs/structured-questions-guide) explains how to combine them with open conversation.\n- **Per-participant evidence.** Koji's automatic thematic analysis links every theme to the specific participants and quotes behind it, so you can see at a glance whether a theme rests on five independent accounts or on five people from one team.\n- **Voice or text.** Participants can respond by voice or text on their own schedule, which is what makes fast parallel collection practical.\n\nThe result is a study where the independence of each account is designed in, rather than hoped for.\n\n## Frequently asked questions\n\n### What is memory conformity?\n\nMemory conformity is when people who discuss a shared experience later report details that only another person observed, as if those details were their own memories. It was demonstrated by Gabbert, Memon and Allan (2003), where 71 percent of witnesses who discussed an event recalled items acquired during the discussion.\n\n### How is memory conformity different from groupthink?\n\nGroupthink changes what people say they believe or prefer, usually inside the group. Memory conformity changes what people report actually happened, and it persists in later individual accounts after the group has dispersed. Participants usually do not realise the detail was borrowed.\n\n### Are some people more susceptible to memory conformity than others?\n\nAge did not matter in the 2003 study: younger (18-30) and older (60-80) participants were equally susceptible. Confidence matters more. Wright, Self and Justice (2000) found that confidence ratings strongly predicted which person in a pair persuaded the other.\n\n### Should I stop running focus groups?\n\nNot necessarily. Focus groups remain useful for generating ideas and hearing reactions. Just do not treat group sessions as evidence about what happened to each person. Collect individual accounts first, and use the group for what it is good at.\n\n### How do I interview several people from the same customer account?\n\nInvite them all at the same time and collect their accounts in parallel, before they can compare notes. Ask each person whether they have discussed the topic with colleagues, and tag details they did not see first-hand. Count the account as one source when their stories converge.\n\n### How can I tell if my existing data is affected?\n\nLook for shared unusual phrasing among people who know each other, details reported by participants who could not have observed them, and agreement that is tighter than independent memory normally produces. Any of the three suggests the accounts were not independent.\n\n## Related Resources\n\n- [The Misinformation Effect](/docs/misinformation-effect-user-interviews) - when the post-event information comes from the researcher\n- [Participant Confidence Is Not Accuracy](/docs/participant-confidence-vs-accuracy) - why the confident account is not the accurate one\n- [Focus Groups vs. Interviews](/docs/focus-groups-vs-interviews) - choosing the format for the question you have\n- [Evidence Corroboration From Independent Sources](/docs/evidence-corroboration-independent-sources) - counting sources instead of mentions\n- [Hearsay and Secondhand Customer Evidence](/docs/hearsay-secondhand-customer-evidence) - weighing what a participant was told rather than saw\n- [Structured Questions Guide](/docs/structured-questions-guide) - the six question types, including source-tracking questions\n","category":"Research Methods","lastModified":"2026-09-19T03:20:54.397985+00:00","metaTitle":"Memory Conformity: When Participants Share Memories","metaDescription":"71% of witnesses who discussed an event recalled details only their partner saw. How to keep customer accounts independent.","keywords":["memory conformity","memory conformity focus groups","social contagion of memory","co-witness contamination","participants influencing each other","independent customer accounts","buying committee interviews"],"aiSummary":"Memory conformity is when people who discuss a shared experience later report details only another person observed. Gabbert, Memon and Allan (2003) found 71% of witnesses who discussed an event recalled items acquired in the discussion; Wright, Self and Justice (2000) found the more confident partner usually persuaded the other. In customer research, co-witnesses include focus groups, buying committees, churned accounts and communities. Collect individual accounts in parallel, ask the source question, and count independent groups rather than mentions.","aiPrerequisites":["Experience running interviews or focus groups","A study involving several people from the same organisation or community"],"aiLearningOutcomes":["Distinguish memory conformity from groupthink","Identify co-witness pairs in a customer study","Apply five design rules for independent accounts","Detect conformity in existing transcripts"],"aiDifficulty":"intermediate","aiEstimatedTime":"12 min"}],"pagination":{"total":1,"returned":1,"offset":0}}