Triplett 1898 Social Facilitation: The First Experimental Psychology Study, Re-Examined

A few months ago a behavioral-design consultant walked me through her pitch deck for an enterprise client. Slide three opened with the canonical line: “Since Norman Triplett discovered in 1898 that cyclists ride faster when racing other cyclists, we’ve known that the presence of others boosts performance.” The slide then pivoted into a recommendation to redesign open-plan office layouts and remote-work norms around the premise that “people perform better around other people.”

Two of the three claims on that slide are inaccurate, and the third is at best a conditional truth. Triplett did publish a paper in 1898 that has been retrospectively crowned the first social-psychology experiment. He did report that cyclists rode faster when paced. But his statistical analysis was, by modern standards, not adequate to establish the claim he made — a fact that Wolfgang Stroebe documented in detail in a 2012 paper that almost nobody outside the history-of-psychology subfield has read. And the broader claim that “social presence boosts performance” is true under specific conditions and false under others — a reformulation that took Robert Zajonc until 1965 to articulate properly.

This article is about how a foundational empirical finding gets retrospectively cleaned up by the field that builds on it. The textbook story is wrong in the technical details. The underlying phenomenon is real. The interesting question is what we learn about evidence evaluation by looking at the actual gap between what Triplett reported in 1898, what Stroebe found when he re-examined the data over a century later, and what the contemporary social-facilitation literature actually supports.

The Textbook Canonical Story

Open almost any introductory social-psychology textbook and you will find some version of this narrative. In 1898, a graduate student at Indiana University named Norman Triplett noticed that cycling-race records were faster in races where cyclists raced against each other than in races against the clock. He hypothesized that the presence of competitors had a “dynamogenic” effect — that other people, by their mere presence, increased physical performance. To test this experimentally, Triplett designed a children’s reel-winding task. Children wound fishing reels alone and then in pairs. They wound faster in pairs. Social facilitation was born. The first experimental social-psychology study is in the books.

The story is appealing because it has a clean origin. It also serves the field’s institutional needs — social psychology gets to claim a deep empirical pedigree, the 1898 paper gets cited approximately forever as a foundational reference, and undergraduates get a memorable narrative that ties a concept to a person to a year. The cycling example is vivid, the reel-winding apparatus is endearingly low-tech, and the conclusion sounds intuitive. What’s not to like?

What’s not to like is that the canonical version smooths over several substantial problems with the original analysis, conflates the 1898 finding with a much later reformulation that is empirically stronger, and creates the impression that “social facilitation” was a single discovery rather than an 80-year process of figuring out what the actual effect is and when it appears.

What Triplett 1898 Actually Did

The paper is titled “The dynamogenic factors in pacemaking and competition,” published in the American Journal of Psychology in October 1898. It runs 27 pages. The full reference is Triplett, N. (1898), Volume 9, Number 4, pages 507-533, DOI 10.2307/1412188. The paper has two main empirical sections.

The cycling-records analysis. Triplett obtained records from the League of American Wheelmen for the 1897 season and earlier years. He distinguished three race types: unpaced against time (a solo cyclist racing the clock), paced against time (a solo cyclist with a paced rider or motor vehicle setting tempo), and actual competition (cyclists racing each other). He compared average speeds and reported that paced and competitive racing produced faster times than unpaced solo racing. The implied causal claim: the presence of a pacer or competitor exerts a “dynamogenic” influence — releasing latent energy, suggesting effort, or maintaining tempo.

The reel-winding experiment. Triplett’s experimental innovation was the apparatus. He built a setup with two fishing reels mounted on a wooden frame, connected by silk fishing line over pulleys to small flags. Children turned a crank to reel in the line; the flag’s position on a track marked progress. The dependent measure was time to complete a fixed distance (four laps). Children worked alone and in pairs (the “competition” condition, in which the two participants were aware of each other and could see the other’s flag). Forty children participated, ranging from 8 to 17 years old. Triplett reported that 20 of the 40 wound faster in the competition condition, 10 were slower, and 10 showed no clear effect or “overstimulation.”

The conclusion Triplett drew: the presence of a co-actor produced a measurable performance increase in roughly half the children, no effect in a quarter, and a counterproductive effect in the remaining quarter. He framed the overall pattern as supportive of the dynamogenic hypothesis.

That is the empirical core of what is now called the first social-psychology experiment.

What Stroebe 2012 Found When He Re-Examined the Data

In 2012, Wolfgang Stroebe published a short, devastating note in Perspectives on Psychological Science titled “The truth about Triplett (1898), but nobody seems to care.” Volume 7, Number 1, pages 54-57, DOI 10.1177/1745691611427306. The paper does what the title advertises: it goes back to Triplett’s actual data and runs the analysis that modern statistical standards would require.

The findings are unflattering to the canonical narrative.

On the cycling records. Triplett’s race-time comparison did not control for the confound that mattered most. Unpaced races against the clock were typically run on different courses, in different weather, at different points in the season, and often by different cyclists with different stakes than competitive races. The “effect” of pacing in Triplett’s data is confounded with virtually every other variable that affects cycling speed. A modern analysis would require either a within-cyclist comparison across race types or a much more careful statistical adjustment for confounders. Triplett did neither. The race-record evidence is suggestive at best.

On the reel-winding experiment. Stroebe ran the statistical analysis that Triplett did not. With 40 children split into three response categories (faster in competition, slower, no effect), the pattern Triplett reported is not statistically distinguishable from chance under modern testing standards. The “20 faster, 10 slower, 10 unclear” split does not yield a significant effect of the competition manipulation when analyzed properly. Triplett’s claim that the data support the dynamogenic hypothesis would, by 2012 standards, be considered overreach. The sample is small, the effect is small, the analysis is inadequate, and the conclusion is stronger than the data warrant.

On the broader pattern. Stroebe’s larger point is that the field has accepted a foundational narrative that the underlying paper does not robustly support. Triplett observed something. He pursued it with the tools available in 1898. He overinterpreted the results, as was common in early experimental psychology before the modern statistical apparatus existed. The cleanup happened retrospectively — later researchers found social-facilitation effects in better-designed studies, and the field credited the phenomenon back to Triplett. The 1898 paper functions in modern social psychology as an origin myth, not as a load-bearing piece of evidence.

Stroebe’s title carries the sting. “But nobody seems to care” is a comment on the field, not on Triplett. The historical record is what it is. The willingness of contemporary psychology to keep citing Triplett 1898 as the first proof of social facilitation, when a re-examination of the data shows the original analysis would not pass modern muster, is a statement about how disciplines handle their canonical findings.

Zajonc 1965: The Reformulation That Actually Works

The reason Stroebe’s critique does not invalidate the social-facilitation literature is that the modern version of the phenomenon traces not to Triplett 1898 but to Robert Zajonc’s 1965 Science paper, “Social facilitation,” Volume 149, Number 3681, pages 269-274, DOI 10.1126/science.149.3681.269.

Zajonc’s contribution was to resolve a paradox that had bedeviled the field for the prior several decades. Some studies reported that the presence of others improved performance. Other studies, using different tasks, reported that the presence of others impaired performance. The literature was incoherent. Zajonc proposed a unifying theoretical framework that organized the contradictions.

The reformulation, in plain language: the mere presence of others increases physiological arousal, and increased arousal facilitates the execution of well-learned (dominant) responses while impairing the execution of poorly learned (non-dominant) responses. On tasks where the correct response is the dominant one — simple, well-practiced behaviors — social presence improves performance. On tasks where the correct response is non-dominant — complex, novel, or poorly learned behaviors — social presence impairs performance. The same underlying mechanism produces opposite behavioral outcomes depending on the task structure.

This is a much sharper hypothesis than “the presence of others affects performance.” It is falsifiable in a way Triplett’s original claim was not. It makes specific predictions: for a simple motor task (winding a reel, running a familiar route), expect facilitation; for a complex cognitive task (learning a new maze, solving a novel problem), expect impairment. It also predicts the moderators — the size of the social-facilitation or social-inhibition effect should scale with the arousal-inducing properties of the audience situation.

Zajonc’s drive-theory formulation is what made social-facilitation research scientifically tractable. It told experimenters what to manipulate, what to measure, and what pattern of results to expect. The next two decades of empirical work largely confirmed the predictions.

Bond & Titus 1983: The Meta-Analysis That Confirmed Zajonc

The definitive empirical test of Zajonc’s drive theory came in Charles Bond and Linda Titus’s 1983 meta-analysis, “Social facilitation: A meta-analysis of 241 studies,” Psychological Bulletin, Volume 94, Number 2, pages 265-292, DOI 10.1037/0033-2909.94.2.265.

The numbers: 241 studies, 24,000-plus participants, every published study of social facilitation Bond and Titus could find from the post-Zajonc era. The meta-analysis tested both whether the basic pattern held and whether Zajonc’s specific predictions about task complexity were supported.

The headline result: social presence reliably affected performance, in the directions Zajonc predicted, with effect sizes that varied with task type as the drive-theory account would expect. On simple tasks, the mere presence of others produced modest but consistent performance facilitation. On complex tasks, social presence produced performance impairment. The interaction between social presence and task complexity was robust across studies, methods, and populations.

The Bond and Titus paper is the load-bearing empirical foundation of contemporary social-facilitation theory. It is rarely cited in popular treatments of the phenomenon, which prefer the cleaner Triplett origin story. But the relationship between citation prominence and evidentiary weight is, as ever, the inverse of what a careful reader should want. Triplett 1898 is the famous reference; Bond and Titus 1983 is the reference that actually establishes the claim.

The History-of-Psychology Lesson

The pattern in social facilitation is not unique. It is recognizable across the empirical sciences. A foundational early study makes a claim. The claim is overinterpreted by the standards of its time. The phenomenon is real but the original evidence is inadequate. Later researchers, with better tools and better statistical methods, do the work to establish the phenomenon properly. The field then retroactively credits the discovery to the original, partly out of historical convenience and partly because origin stories function differently in scientific narratives than they do in scientific evidence.

This is fine, as long as you understand what is happening. The pathology arises when the textbook story gets imported into applied work. A practitioner who cites “Triplett 1898” as evidence for an intervention is citing a 19th-century pilot study with an inadequate analysis. A practitioner who cites “Zajonc 1965 and the subsequent literature, including Bond & Titus 1983” is citing the actual body of evidence. The former sounds more authoritative because it is older and more famous. The latter is the citation a careful evaluator should want.

The other lesson — and this is the one I find more useful for strategists — is that “the textbook version is cleaner than the actual data” is a generalizable hypothesis. When you encounter a tidy historical narrative attached to a foundational empirical claim, the prior probability that the narrative oversimplifies should be high. The right reflex is to read the original paper, then read whatever modern re-examination exists, and notice the gap. The gap is almost always there. Sometimes the gap is small and the textbook version is roughly accurate. Sometimes the gap is large and the textbook version is misleading. You will not know which without checking.

This is the same epistemic hygiene that the replication crisis hub is built around. The point is not that early studies are uniformly unreliable. The point is that the social processes by which findings get canonized in a field are not the same as the social processes by which findings get rigorously tested. The former rewards memorable stories with clear attributions. The latter requires statistical adequacy, replication, and willingness to abandon hypotheses that do not hold up.

What Strategists Should Take Away

A few practical implications.

First, when someone cites a foundational study from before approximately 1960, treat the citation as a flag to check rather than as load-bearing evidence. Early experimental psychology preceded the modern statistical apparatus, and many foundational findings have been re-examined with results unflattering to the original. This is not a blanket dismissal — some early findings hold up beautifully — but the prior should be skepticism, not deference.

Second, when an intervention is justified by a single eponymous study, ask what the meta-analytic evidence shows. If the meta-analysis exists and confirms the effect, the intervention has a real empirical foundation. If the meta-analysis exists and shows the effect is small, conditional, or absent, you have learned something important. If the meta-analysis does not exist, you should treat the intervention as speculative regardless of how famous the original study is.

Third, for the specific case of social facilitation: the effect is real, but the moderation by task complexity is the operationally important part. “People perform better around other people” is wrong as a blanket claim. “People perform better around other people on well-practiced tasks and worse on novel cognitive tasks” is closer to what the evidence supports. The implication for workplace design is not “more open-plan offices everywhere” but rather “the right physical and social setup depends on what the work actually requires.” Routine, well-learned work might genuinely benefit from co-presence. Novel cognitive work — the kind that requires deep concentration and produces most of the value in modern knowledge work — is likely to be impaired by it.

Fourth, the gap between Triplett 1898 and the actual modern literature is a useful prompt for self-audit. How many other foundational findings are you carrying around in your mental model based on the textbook version rather than the actual evidence? The answer is almost certainly “more than you think.” The remedy is to pick a few that you rely on most heavily and check.

Sources

• Triplett, N. (1898). The dynamogenic factors in pacemaking and competition. American Journal of Psychology, 9(4), 507-533. DOI: 10.2307/1412188
• Stroebe, W. (2012). The truth about Triplett (1898), but nobody seems to care. Perspectives on Psychological Science, 7(1), 54-57. DOI: 10.1177/1745691611427306
• Zajonc, R. B. (1965). Social facilitation. Science, 149(3681), 269-274. DOI: 10.1126/science.149.3681.269
• Bond, C. F., & Titus, L. J. (1983). Social facilitation: A meta-analysis of 241 studies. Psychological Bulletin, 94(2), 265-292. DOI: 10.1037/0033-2909.94.2.265

Related

• Asch Conformity Cross-Cultural Variation: Bond & Smith 1996 — a foundational finding that does replicate, but with substantial cultural moderation that the popular framing skips.
• Festinger Social Comparison Theory: What Replicates and What Doesn’t — another mid-20th-century social-psychology framework with mixed modern empirical support.
• Milgram Obedience Experiments: What Holds Up — the most famous early-social-psychology study, and what the post-2000 re-examinations show.
• Hawthorne Effect: The Original Study and What Actually Happened — another canonical social-influence finding whose actual evidence base is much weaker than the textbook story.
• Mere Exposure Effect: When It Works and When It Doesn’t — Zajonc’s other major contribution, with a more replicable empirical record than 1898 social facilitation.

FAQ

Was Triplett 1898 actually the first social-psychology experiment?

Historically, yes, in the sense that it is the earliest published paper that is recognizable as an experimental social-psychology study. Several earlier studies in Germany and the US examined related phenomena (suggestibility, group judgment) but did not have the experimental structure Triplett used. The “first experiment” claim is defensible as a matter of historical chronology. The separate question of whether the experiment was statistically adequate is what Stroebe addressed.

Does this mean social facilitation is not real?

No. Social facilitation, as reformulated by Zajonc 1965 and confirmed by Bond and Titus 1983, is one of the more robustly established phenomena in social psychology. What is unreliable is the specific 1898 evidence base, not the modern phenomenon. The lesson is about citation hygiene, not about whether the underlying effect exists.

Why does the field keep citing Triplett if the analysis was inadequate?

Several reasons. Origin stories serve narrative and pedagogical functions that do not require evidentiary load-bearing. The 1898 paper is a useful historical marker even if its statistics are weak. And once a citation pattern is established, it perpetuates by inertia — every new textbook cites Triplett because every previous textbook did. Stroebe’s 2012 critique is now widely available, but the canonical narrative has institutional momentum.

What is the strongest contemporary evidence for social facilitation?

The Bond and Titus 1983 meta-analysis remains the definitive synthesis. Subsequent work has refined the moderators — particularly the role of evaluation apprehension (whether the audience is evaluative rather than just present) and the specific physiological mechanisms — but the core pattern Zajonc predicted has held up across four decades of additional research.

How should I cite social-facilitation evidence in my own work?

If you want to be historically complete, cite Triplett 1898 as the original observation, Zajonc 1965 as the theoretical reformulation, and Bond and Titus 1983 as the meta-analytic confirmation. If you only have space for one citation, Bond and Titus 1983 is the right choice. Citing Triplett alone, without the later reformulation and meta-analysis, gives the misleading impression that the phenomenon was established in 1898 — which the Stroebe re-examination shows it was not.

What other foundational psychology findings have been re-examined like this?

Many. Stanley Milgram’s obedience studies have been re-examined by Gina Perry and others with substantially more critical results than the textbook version suggests. Philip Zimbardo’s Stanford Prison Experiment has been comprehensively criticized as more theatrical than scientific. The Hawthorne studies have been re-analyzed with the original data and the famous “Hawthorne effect” largely disappears. The pattern — textbook story cleaner than the actual evidence — is common enough that “check the original paper” should be the default reflex for any famous finding.