Two strangers sit across from one another. Instead of exchanging a few facts about work and weather, they take turns answering prompts that become gradually more personal. After the conversation, both report feeling closer than they did at the start. What did the conversation change, and how long could that change last? Arthur Aron and colleagues’ 1997 paper, The Experimental Generation of Interpersonal Closeness: A Procedure and Some Preliminary Findings, gave researchers a way to ask the first question under controlled conditions.
This article is a critical guide to the study often called “Fast Friends.” It is distinct from a general guide to social connection or attachment. The original paper’s main contribution was a research method for generating and measuring initial closeness, not a recipe that guarantees love or lasting friendship. Later experiments help establish what the procedure can and cannot do. That difference matters when a laboratory task becomes a popular set of “questions to make someone fall in love.”
Why researchers needed a closeness task
Close relationships are difficult to manipulate experimentally. People choose partners, bring histories into an interaction and become closer through many events outside a laboratory. A comparison between existing friends and strangers can reveal differences, but it cannot by itself show which process caused the relationship to develop. Aron and colleagues wanted a repeatable procedure that would create some closeness among initially unacquainted people within a session.
Their method paired participants for roughly 45 minutes. Partners alternated increasingly personal self-disclosure and relationship-building tasks. The well-known set contains 36 prompts arranged in three sets that increase in intensity. The escalation is part of the design: an immediate demand for intimate disclosure would be a different intervention, with different risks and likely different results.
The paper’s authors were explicit that the procedure was intended for research. If it produced a reliable short-term difference in felt closeness, investigators could vary partner characteristics, expectations and social contexts while keeping the interaction structure more constant. That would make causal questions about initial connection more tractable. It would not make all relationships reducible to one conversation.
Even the phrase “generate closeness” needs definition. The study measured participants’ feelings after the interaction, not a lifelong obligation or a biological bond. A person can feel genuinely connected during a conversation and later decide not to continue it. That does not make the first feeling false; it locates it at the time and level the experiment measured.
The first comparison: more than small talk?
Study 1 compared the closeness-generating procedure with a task involving more superficial conversation. The structured, escalating task produced greater reported closeness after the interaction. This is the key result that supports using the procedure in later research. The comparison matters because simply spending time with another person and taking turns speaking might also increase familiarity. A small-talk condition helps ask what the gradual disclosure adds.
The outcome was still self-reported closeness shortly after a brief encounter. Participants knew they had shared personal material; that awareness could influence their report. That is not a flaw that makes the result meaningless. It is an interpretation limit. Stronger claims about enduring friendship would need follow-up contact, observed behavior and the preferences of both people, not just a post-session rating.
The original paper reports some follow-up observations from earlier classroom versions, including that some pairs spoke again or did something together. Those observations are interesting, but they do not prove that the procedure caused long-term friendship for the whole sample. The setting, selection of respondents and lack of a matched long-term control are relevant. The article’s own title calls the findings preliminary.
Later research by Sprecher compared Fast Friends with small talk and unstructured getting acquainted. In that study, zero-history dyads assigned to the structured procedure generally reported more closeness and other affiliative outcomes, especially relative to small talk. Video-chat and face-to-face modes did not differ significantly in the measured outcomes. This extends the laboratory evidence, while still studying initial encounters rather than relationships over months.
What did not add much in the original studies
Studies 2 and 3 tested several plausible ingredients beyond the conversation itself. Did partners need to be matched for agreement on important attitudes? Did being led to expect mutual liking matter? Did explicitly making closeness the goal increase the outcome? The authors reported no significant closeness effects for these manipulations in those studies, with sufficient power for the tested comparisons.
A null result is not proof that similarity, expectations or goals never matter in relationships. The manipulations may have been weak, the measured interval short or the procedure itself sufficiently powerful to make additional effects hard to detect. The result is narrower: under those particular conditions, the added manipulations did not yield detectable increases in postinteraction closeness. Reading the negative findings is as important as repeating the famous positive one.
The researchers also examined attachment and introversion or extraversion as potential moderators. Such characteristics were not randomly assigned. An association involving them cannot establish that a trait caused the closeness outcome. The paper calls some of these findings tentative. A popular retelling that says the task “works for everyone” or “works best for one personality type” would overshoot the study.
This distinction between a manipulated condition and a measured characteristic is fundamental. Assignment to a conversation format supports a causal comparison about that format in the sample. A pre-existing trait can correlate with behavior while sharing causes with other unmeasured characteristics. The study is most persuasive where its experimental contrast matches the claim.
Closeness is not the same as intimacy, love or trust
The procedure invites reciprocal disclosure. Sharing something personal can feel intimate, especially when the other person responds attentively. But a short-term closeness rating cannot establish that the partner will be reliable, respect boundaries or remain available. Trust develops from repeated evidence across situations. Love and friendship involve commitments and histories far beyond a 45-minute task.
Aron, Aron and Smollan’s Inclusion of Other in the Self scale is one way researchers measure perceived closeness. Its overlapping-circle format asks people to indicate how connected they feel. It can be useful for a defined research question, but it is still a report of experience, not a direct measure of another person’s intentions or a relationship’s durability. Researchers should explain which outcome they use instead of treating “closeness” as self-evident.
When later studies adapt the procedure for different populations, they often change more than one feature. A 2024 online study with children aged 9–13 paired a question-and-answer task with collaborative play. The Fast Friends condition produced higher closeness than a comparison condition and more interest in future contact. Because both disclosure and play differed, the study cannot attribute the entire effect to the original adult question sequence. It does show that a structured, reciprocal interaction can be studied with younger participants in an online setting.
A German-language study examined a translated procedure and an extension designed to induce sexual intimacy, including participants with and without reported childhood maltreatment. It found increases in several intimacy and positive-affect measures. The sample and intervention differ from the original, and the authors called for further validation. It would be irresponsible to turn that result into a recommendation that vulnerable people be asked to disclose intimate material in any group setting.
How a later test should be read
Suppose a new study says Fast Friends increased connection. First check the comparator. Was the alternative small talk, silent reflection, unstructured conversation or another active exercise? Each contrast answers a different question. A benefit over no interaction may reflect conversation in general. A benefit over small talk offers more evidence that the structure or depth of exchange matters.
Next check what was held constant. Did both groups spend the same time together? Did they speak equally often? Did one group receive a game, a shared goal or a more responsive partner? If those features vary with the questions, the cause cannot be assigned solely to disclosure. The 1997 paper’s design and later adaptations are useful precisely because they make such distinctions visible.
Then check the outcome’s timing. A rating taken immediately after the task measures an immediate state. A desire to meet again is a prospective intention. Actual contact after a month is a different outcome. The longer the claim, the more follow-up evidence it needs. Attrition also matters: if only the pairs who liked each other answer later surveys, the follow-up picture may be optimistic.
Finally check consent and the social setting. Participants in a study can be briefed, decline, skip a prompt and be debriefed. An exercise imposed by an employer, school or dating platform may carry pressure that changes both its ethics and its effects. A person should be able to choose how much to disclose without being marked uncooperative.
Disclosure needs response and boundaries
The popular shorthand treats the 36 questions as a secret sequence. In practice, the human interaction matters. A prompt does not listen, notice discomfort or respond respectfully. A pair can answer the same questions and have different experiences because of attention, reciprocity, culture, power and prior expectations. The procedure standardizes part of the encounter, not every socially meaningful variable.
Reciprocity can help balance the task: both people have opportunities to speak and hear. But formal symmetry does not eliminate power differences. One participant may be a supervisor, teacher or service provider. Another may feel less able to decline a question. In such contexts, a “closeness” exercise may create unwanted exposure instead of connection. Voluntary participation, clear boundaries and easy exit are central if the method is adapted outside a research lab.
The original authors addressed this concern in their debriefing. They told participants the task was an unusual way to form relationships and that no friendship obligation followed. That detail is easy to lose when the method becomes an icebreaker. It recognizes that an experimentally induced feeling should not be turned into a social debt.
A well-designed program could use low-stakes reciprocal questions and opportunities for participants to opt in, but that would be an Institute application requiring its own evaluation. Success would include whether people felt respected, retained control over disclosure and could choose further contact. A higher immediate closeness score alone would not establish that the program improved a community.
Digital conversations and new questions
The 2021 video-chat comparison suggests that the procedure’s measured short-term effects need not depend on physical co-presence in every study. That finding does not mean every online setting is equivalent. Camera use, privacy, disability access, language, latency and the ability to leave all change an interaction. A text chat with an unknown participant poses another set of questions.
AI-mediated versions raise still more. An interface can ask increasingly personal questions and produce a feeling of closeness, but that feeling does not prove reciprocal human concern. If a person believes a system is a human partner, the meaning of disclosure and consent changes. The original Fast Friends paper did not test AI systems and should not be cited as proof that a chatbot can form a friendship. Research must separate reported closeness from transparency about the partner and from actual support.
The useful research inheritance is methodological: define the interaction, vary one feature, measure a named outcome and follow up when making a long-term claim. The ethical inheritance is equally important: a participant’s willingness to disclose is not a resource an institution may extract merely because the procedure produces a compelling score.
What the evidence does not show
The 1997 study does not show that 36 questions reliably make strangers fall in love or form lasting friendships. It found a short-term closeness difference relative to a particular small-talk comparison. Later studies support the procedure’s use for some initial-connection questions, but designs and populations differ. A single conversation cannot certify compatibility, safety or commitment.
It does not show that deeper disclosure is always better. Disclosure can be valuable when chosen and met with care; it can be intrusive when required or poorly received. The negative results for attitude matching and expected liking also do not prove those factors never matter. They place limits on the specific manipulations tested.
Finally, the evidence does not justify using a closeness task as therapy, screening or a mandatory workplace exercise without separate evidence and safeguards. The original authors described a research method and included caution about creating relationships that might not naturally arise. That caution remains relevant when the method is used at scale.
Return to the two strangers
After the session, both people may say they feel closer. The experiment can tell us something about the structured interaction that preceded that report. It cannot tell us whether they will exchange messages next week, whether either wants a friendship or what a later conflict will reveal. Those are questions for follow-up and for the people themselves.
Fast Friends is valuable because it made initial closeness experimentally tractable. The careful use of it preserves the modesty of that achievement. A structured conversation can open a connection; whether a relationship grows depends on what happens next.
Research references
- Aron, A., Melinat, E., Aron, E. N., Vallone, R. D., & Bator, R. J. (1997). The experimental generation of interpersonal closeness: A procedure and some preliminary findings. Personality and Social Psychology Bulletin, 23(4), 363–377. View source
- Sprecher, S. (2021). Closeness and other affiliative outcomes generated from the Fast Friends procedure: A comparison with a small-talk task and unstructured self-disclosure and the moderating role of mode of communication. Journal of Social and Personal Relationships, 38(5). View source
- Swerbenski, K. L., Barnett, K. C., Devine, P. G., & Shutts, K. (2024). Making “Fast Friends” online in middle childhood and early adolescence. Social Development, 33(1), e12708. View source
- Weilenmann, S., von Luckner, A., Peyk, P., Huber, C., Schnyder, U., & Pfaltz, M. C. (2022). Experimental induction of emotional and sexual intimacy: Exploring the validity of the German Fast Friends procedure in individuals with and without childhood maltreatment. Archives of Sexual Behavior, 51(3), 1703–1719. View source
- Aron, A., Aron, E. N., & Smollan, D. (1992). Inclusion of Other in the Self Scale and the structure of interpersonal closeness. Journal of Personality and Social Psychology, 63(4), 596–612. View source