The Ringelmann Effect: The 1913 Rope-Pulling Study That Predicts Your 10× Team Won't Be 10× Anything

60-Second Summary

Ringelmann (1913): individuals pull less hard in a group than alone. Two people: ~93% of individual force each. Three: ~85%. Eight: ~49%. Effect is real, replicated widely, and mostly forgotten in modern management.
The effect has two components: coordination loss (mechanical — bodies interfering with each other) and social loafing / motivation loss (Latané, Williams & Harkins, 1979). Coordination loss dominates in physical tasks; social loafing dominates in knowledge work.
Directly predicts why doubling headcount does not double output — and why the productivity curve of knowledge-work teams flattens sharply beyond 5–7 people. Related to but distinct from Brooks' Law.
Individual identifiability is the strongest counter-lever: when contribution is visible, social-loafing loss collapses. This is why small teams, clear ownership, and public artefacts beat headcount every time.
Ringelmann-aware team design: keep teams small (Amazon's two-pizza rule is a modern restatement), make individual contribution identifiable, resist the reflex to solve output problems with headcount, and be honest with the CFO about the shape of the curve.

In 1913, a French professor of agricultural engineering named Maximilien Ringelmann attached a dynamometer to a rope and asked people to pull. He tested them alone and in groups of two, three, and eight. He was studying farm productivity, not psychology. The finding was so counter-intuitive that he almost didn't publish it, and when he did, in an obscure French agricultural journal, it sat unread for sixty years before social psychologists rediscovered it. It is now one of the most robustly replicated effects in group behaviour, and it is nearly absent from modern management writing. Which is a shame, because it predicts, with unnerving accuracy, why your engineering VP's plan to solve delivery problems by hiring six more people will produce approximately zero additional output.

What Ringelmann actually did

“L'homme moyen tire à peu près 85 kilogrammes; deux hommes en tirent à peu près 165 kg au lieu de 170; trois hommes tirent à peu près 218 kg au lieu de 255; huit hommes tirent seulement 392 kg au lieu de 680.”
— Ringelmann (1913), Recherches sur les moteurs animés — the original numbers

Translated: the average man pulls about 85 kg alone. Two together pull 165 kg, not 170 (each performing at 97% of solo). Three pull 218 kg, not 255 (85% each). Eight pull only 392 kg, not 680 (58% each). The larger the group, the lower each individual's contribution. Ringelmann attributed it initially to coordination loss — bodies literally interfering with each other's pulling angles. Sixty years later, Latané, Williams & Harkins (1979) ran the classic follow-up: they asked people to shout as loudly as possible, alone and in perceived groups (with confederates instructed to be silent). Even with no coordination possible — because the confederates weren't actually pulling or shouting — individual output dropped as perceived group size rose. This second effect was named 'social loafing'.

The two mechanisms decomposed

Coordination loss (mechanical)

Bodies, tools, or workflows interfere with each other.
Dominant in physical tasks (rope-pulling, factory lines).
Fixed by better process, tooling, and role clarity.
Reducible in principle to near-zero with sufficient investment.
Modern equivalents: merge conflicts, meeting overhead, handoff friction.

Social loafing (motivational)

Individual effort drops when contribution is not identifiable.
Dominant in knowledge work, brainstorming, committees.
Fixed by identifiability, ownership, public artefacts.
Not fixable by process alone — it is motivational, not procedural.
Modern equivalents: 'the team' owned that decision, 'we all contributed', diffusion of responsibility in incident response.

93%

individual effort in a 2-person group vs solo (Ringelmann's original data)

Ringelmann (1913)

49%

individual effort in an 8-person group vs solo

Ringelmann (1913); replicated in Ingham et al. (1974) and dozens since

~40%

typical reduction in individual output in unidentifiable knowledge-work groups

Meta-analysis by Karau & Williams (1993), Journal of Personality and Social Psychology

Ringelmann in knowledge work

The intuition-breaking finding is that the effect is not mostly physical. Karau & Williams's 1993 meta-analysis of 78 studies found robust social loafing across every task type: brainstorming, decision-making, vigilance tasks, writing, negotiation, and computer-mediated group work. Effect sizes were moderate to large (Cohen's d = 0.44). The strongest moderator was individual identifiability. When contributions were visible and attributable, loafing shrank toward zero. When contributions were merged into a group output, loafing rose sharply.

This directly maps onto engineering-team design. A team of three where each engineer owns a service, ships their own PRs, and is visibly on-call for what they built produces near-full individual output. A team of ten where 'the team' owns the codebase, PRs are collectively reviewed, and features are attributed to 'the team' produces per-head output at roughly half of what those same ten engineers would produce distributed into two teams of five with clear ownership.

The formula and the curve

Ringelmann-derived approximation for individual effort as a function of group size (physical task): E_i ≈ 100% × N^(-0.15) where N is group size. For knowledge work with strong identifiability, the exponent is smaller (~-0.05). For knowledge work with no identifiability, the exponent approaches -0.20. The practical takeaway: total group output plateaus rapidly beyond 5–7 people for most knowledge-work tasks, and can decline in absolute terms beyond ~12 in coordination-heavy work (Brooks' Law adjacent).

The 'add headcount' reflex

When output is under-target, the reflex is to add headcount. Ringelmann says: unless you also change identifiability, total output rises by less than proportional to headcount, and per-head output falls. The right first move is almost always to split, not add — because splitting restores identifiability, and identifiability restores individual output. Two teams of five typically out-produce one team of ten by 30–50%.

Ringelmann-aware team design

Keep teams small. Amazon's two-pizza rule (~6–8) is a Ringelmann-aware heuristic. Bezos was applying the effect without naming it.
Make individual contribution identifiable. Public PRs, named ownership, visible decision logs, individual on-call rotations. This is the highest-leverage move.
Resist the reflex to solve output problems with headcount. Splitting existing teams often out-performs adding to them.
In meetings: use round-robin, not open floor. Round-robin restores identifiability of individual input; open floor produces social loafing on ideas and dominance by extroverts.
In decision-making: named accountability with a single DRI. 'The team decided' is a Ringelmann signature. 'X decided, informed by A, B, C' is Ringelmann-resistant.
In incident response: named incident commander, named owners of workstreams. Diffusion of responsibility is the acute form of social loafing and it kills incident response time.
In performance calibration: measure individual contribution to team output, not team output attributed to all members. This is unpopular and correct.

The CFO conversation

'Adding 6 engineers to this team of 10 will not produce 6/10 more output. It will produce approximately 2/10 more output because per-head productivity drops. If we split the current 10 into two teams of 5 and then add 3 to each, we will get closer to 6/10 more output, and per-head productivity will actually rise. The delta is real money.' This conversation is more useful than any tooling investment most orgs make.

FAQ

Frequently asked questions

Doesn't this contradict the value of collaboration?

No. It says the value of collaboration is bounded by identifiability. Collaboration between two identifiable specialists produces near-full individual output plus interaction value. Collaboration in a large unidentifiable group loses individual output faster than it gains interaction value.

How does this relate to Brooks' Law?

Brooks' Law (adding manpower to a late project makes it later) is a specific case of Ringelmann plus coordination-onboarding cost. Ringelmann is the underlying mechanism; Brooks named the delivery consequence.

Isn't remote / async work supposed to reduce this?

Partly. Async work with strong written artefacts increases identifiability, which reduces social loafing. But it can increase coordination loss (merge conflicts, decision latency). Net effect is empirically mixed and depends heavily on artefact discipline.

What about generative AI reducing the coordination-loss half?

Plausible for merge/handoff friction. Does nothing for the social-loafing half. Identifiability of human decision remains the lever.

Takeaways

Ringelmann (1913): individuals pull less hard in a group than alone. Eight-person groups produce ~49% per-head of solo output. Replicated for 100+ years.
Two mechanisms: coordination loss (mechanical, fixable with process) and social loafing (motivational, fixable with identifiability).
Social loafing dominates in knowledge work. It is present in brainstorming, decision-making, code review, and incident response.
The highest-leverage counter-move is identifiability: named ownership, public artefacts, single DRIs, round-robin input.
Adding headcount without changing identifiability produces sub-linear output gains. Splitting is often the higher-leverage move.