On this page
The Allen Curve is a graph showing that the probability two colleagues communicate about work falls sharply as the physical distance between their desks increases, documented by Thomas J. Allen, a professor at the MIT Sloan School of Management, in research conducted through the 1960s and 1970s and published in his 1977 book Managing the Flow of Technology.
That one sentence is the whole finding. Everything else is interpretation, and a surprising amount of that interpretation is wrong.
The Allen Curve is one of the most cited and least checked studies in workplace literature. It gets deployed in return-to-office memos, office design pitches, and hybrid policy decks, usually attached to a specific number that turns out to be harder to trace than anyone citing it seems to realize. Here is what the research actually says.
Who Thomas Allen was, and what he was studying
Thomas J. Allen (1931 to 2020) joined the MIT faculty in 1966 and spent his career at the Sloan School of Management, where he held the Howard W. Johnson Professorship, according to MIT Sloan's obituary. His subject was not office design. It was how technical information moves inside research and development organizations.
He was answering a management question: why do some R&D labs turn information into good technical work while others sit on it? That project produced two durable ideas. One was the "technological gatekeeper," the small number of people who connect a lab to the outside world and redistribute what they learn. The other was the curve that carries his name.
The distance finding was a byproduct, not the point. This matters for how much weight the curve can carry.
What Allen actually measured
Allen studied engineers and scientists inside industrial R&D laboratories. He mapped who talked to whom about technical matters, using sociometric surveys, questionnaires, and communication logs, then plotted that against the physical distance separating their workstations.
The dependent variable is the part people forget. He was not measuring collaboration quality, output, or innovation. He was measuring the probability that a given pair of people communicated about technical work at least once in a week, against the straight-line distance between where they sat.
Managing the Flow of Technology was published by MIT Press in 1977 and summarized more than a decade of fieldwork, per MIT Press. The peer-reviewed foundation goes back further, to work including Allen and Cohen's 1969 study of information flow in two R&D laboratories in Administrative Science Quarterly. The commonly repeated claim that this was "late 1970s research" understates how long the underlying work ran.
The shape of the curve, and the number everyone repeats
The curve drops steeply and then flattens. Communication probability is highest for people sitting within a few meters of each other, falls fast over the next stretch, then levels off near zero. It does not decline gently. It collapses early and stops declining because there is almost nothing left to lose.
Now the number. Most articles about the Allen Curve cite 50 meters as the critical threshold for weekly technical communication. That figure appears on Wikipedia's Allen curve entry and in Elsevier's description of Allen's later book with architect Günter Henn, The Organization and Architecture of Innovation (2006).
Here is the problem. The peer-reviewed literature that cites Allen directly uses a different number. In Research Policy, Kabo and colleagues (2014) write that Allen "famously demonstrated that the probability of communication between pairs of engineers dropped precipitously at the 30 m mark," and elsewhere in the same paper cite the threshold as 30 meters alongside Kraut et al. (1988) and Monge and Kirste (1980).
Wikipedia's own talk page carries an unresolved 2012 objection from an editor working from the 1977 original, who reports that the graph on page 239 shows the probability of at least weekly communication is already around 0.05 at 50 meters, and that the curve reaches its low asymptote within roughly 25 to 30 meters.
Both readings point the same way. Fifty meters is not where communication starts to fail. It is well past the point where the curve has already bottomed out. Citing 50 meters as the threshold gets the direction right and the mechanism wrong, which leads teams to design for the wrong distances.
I was not able to open the 1977 edition directly to confirm the page 239 values, since the scanned copy on the Internet Archive is lending-restricted. So treat the 30 meter figure as the one supported by peer-reviewed citation, and the 50 meter figure as a widely repeated approximation that traces to publisher and encyclopedia summaries rather than to a verified reading of Allen's data.
One more figure to flag: the claim that people are "four times as likely to communicate with someone six feet away than someone 60 feet away" circulates constantly. I could not trace it to Allen's published work or to any primary source. It appears only in secondary commentary and vendor content. Do not put it in a deck.
What replication found
The core relationship has held up well, even as the specific numbers have not.
Kabo and colleagues (2014) studied researchers in two academic buildings and introduced a measure called path overlap, the extent to which two people's routine walking routes through a building coincide. Researchers whose paths overlapped more were more likely to form new collaborations and more likely to win funding on joint grant applications. Proximity was not just about how far apart people sat but about whether the building routed them past each other.
Allen himself revisited the question with Henn in 2006 and reached a conclusion that surprises people who expect technology to solve the problem. Reviewing email data and his own later work on dispersed organizations, he concluded that decay with distance applies to all communication media, not only face-to-face contact. Frequent in-person contact goes with more phone and email contact, not less.
Independent work supports that. Lekse and Recker (2014), studying enterprise microblogging across 897 Australian supermarket stores, found distance decay persisted on an internal social network where every participant was technically equidistant and all conversations were publicly visible.
And a counterintuitive result worth holding onto: when Bernstein and Turban (Philosophical Transactions of the Royal Society B, 2018) tracked two firms moving to open-plan offices with wearable sensors, face-to-face interaction fell roughly 70 percent while electronic messaging rose. Removing walls did not produce the proximity effect. People withdrew.
What remote work did to the curve
The obvious question after 2020: does digital proximity substitute for physical proximity? The evidence says it substitutes partially, and unevenly by activity.
Yang and colleagues (Nature Human Behaviour, 2021) analyzed communication data from 61,182 Microsoft employees across the shift to firm-wide remote work. Collaboration networks became more static and more siloed, cross-group collaboration time fell about 25 percent, and communication shifted from synchronous to asynchronous. The bridges between disparate parts of the organization thinned out.
Brucks and Levav (Nature, 2022) ran a lab study and a field experiment across five countries and found videoconferencing reduced the generation of creative ideas, apparently because a screen narrows cognitive focus. Notably, videoconferencing groups were just as good, possibly better, at selecting which ideas to pursue. Distance hurt divergent work and not convergent work.
Emanuel, Harrington and Pallais studied software engineers at a Fortune 500 firm from 2019 to 2024, using office closures and return-to-office mandates as natural experiments. Sitting near teammates increased the coding feedback engineers received by 18.3 percent, with gains concentrated among younger and less-tenured engineers. Senior engineers sitting near teammates wrote less code. Proximity bought development and cost short-term output.
The pattern across all three: physical proximity mostly affects the unplanned, low-stakes, developmental interactions. Scheduled work travels over video fine. Chance contact does not.
Distance bands and what they mean for team design
| Distance band | What the research shows | Practical implication for team design |
|---|---|---|
| Same room, under about 10 m | Highest communication probability; the steepest part of Allen's curve sits here | Reserve true adjacency for pairs who need constant unplanned contact, such as a new engineer and their reviewer |
| About 10 to 30 m, same floor | Probability falls sharply across this band; Kabo et al. (2014) place the precipitous drop at the 30 m mark | This is the band you can actually design. Shared routes and shared amenities inside it change who meets whom |
| Beyond about 30 m, same building | Curve has flattened near its floor; Allen's 1977 graph is reported at roughly 5% weekly contact by 50 m | Do not expect spontaneous contact. Anything crossing this band needs a scheduled reason to exist |
| Different floor or building | Kabo et al. (2014) found path overlap and co-location predict new collaborations and funded grants | Treat cross-building work as cross-site work. Assign explicit connectors rather than hoping for overlap |
| Different city or fully distributed | Distance decay persists in digital channels (Allen and Henn, 2006; Lekse and Recker, 2014); cross-group collaboration time fell about 25% at Microsoft (Yang et al., 2021) | Bridging ties must be built deliberately. They will not form as a side effect of shared tooling |
What this means for hybrid design now
The Allen Curve does not tell you whether to bring people back. It tells you what proximity buys and what it costs, so that whatever call you make is a call you can execute well.
If you are mandating five days in office, the curve says your gain is concentrated in unplanned developmental contact, and it only materializes inside the first 30 meters or so. A full building with teams scattered across four floors reproduces the distance problem indoors. Seating adjacency, shared routes, and where the coffee sits will do more for your stated goal than the mandate itself. Our piece on the relationship infrastructure RTO needs works through that build.
If you are running hybrid or staying distributed, the curve says the specific thing at risk is weak-tie contact across groups, not core team delivery. That is consistent with Bloom and colleagues (Nature, 2024), whose randomized trial of 1,612 employees at Trip.com found hybrid work cut quit rates by about a third with no measurable effect on performance grades over the following two years. Retention held. What needs deliberate design is the bridging.
Either way, the deciding factor is whether you can see what is happening. Proximity is a lever with a delayed, uneven payoff, and every study above found the effect concentrated in specific groups: junior engineers, cross-group ties, divergent creative work. Averages hide all of it. A mandate that lifts one team and quietly buries another looks flat in aggregate for two quarters, then shows up in attrition.
This is where daily measurement earns its place. A team-level engagement signal like DEBI, read weekly rather than annually, tells you within weeks whether a seating change or a policy change is landing. A hotspot map tells you which team is absorbing the cost. Manager scorecards tell you whether the managers closest to the change are the ones struggling with it. None of that argues for or against your policy. It is the instrumentation that makes your policy correctable while correction is still cheap. The same logic applies to closing the alignment gap on team focus.
Where the Allen Curve gets overstated
Be honest about the study's boundaries, because they are real.
It comes from a specific era and a specific industry. Allen studied engineers and scientists in industrial R&D laboratories in the 1960s and 1970s, before email, mobile phones, shared documents, and persistent chat. The communication he measured had few substitutes. Distance was a harder constraint then than it is now.
The work he studied was unusually interdependent. Technical problem solving in an R&D lab involves frequent, short, unpredictable exchanges. A sales team, a finance function, or a support queue does not have the same interaction profile. Generalizing a curve derived from R&D engineers to all knowledge work overstates it.
Frequency is not value. The curve measures how often people talk, not whether the talking helps. Our deeper look at why distance and depth both matter covers that distinction.
Correlation, not a clean causal design. Allen observed existing labs. People who work together are often seated together in the first place. The modern studies with natural experiments and randomization, particularly Emanuel et al. and Bloom et al., carry more causal weight than the original, and they find smaller, more conditional effects.
The Allen Curve is a good directional heuristic and a bad universal law. Use it to ask better questions about your own building and your own teams. Do not use it as evidence for a policy you already decided on.
FAQ
What is the Allen Curve in simple terms? It is a graph showing that the farther apart two coworkers sit, the less often they talk about work, and that the drop is steep rather than gradual. Communication probability falls fast within the first tens of meters, then flattens near zero. MIT professor Thomas J. Allen documented it in R&D laboratories and published it in 1977.
Does the Allen Curve still apply to remote teams? Partially. Allen and Henn concluded in 2006 that distance decay applies to all communication media, including email, not just face-to-face contact. Later work found the same pattern in internal social networks. What remote work changes is which activity suffers: research at Microsoft found cross-group collaboration time fell about 25 percent under firm-wide remote work, while scheduled team delivery held up.
Is the 50 meter figure in the Allen Curve accurate? It is the most repeated number and the least well supported. Peer-reviewed papers citing Allen directly, including Kabo et al. (2014) in Research Policy, place the precipitous drop at about 30 meters. The 50 meter version traces to publisher and encyclopedia summaries. Readers working from the 1977 original report that weekly communication probability is already near 5 percent by 50 meters, meaning the curve has flattened well before that point.
Who was Thomas Allen and when did he do this research? Thomas J. Allen (1931 to 2020) was a professor at the MIT Sloan School of Management who joined the MIT faculty in 1966. The distance research was part of more than a decade of fieldwork on information flow in R&D organizations across the 1960s and 1970s, published as Managing the Flow of Technology (MIT Press, 1977).
Does the Allen Curve mean companies should mandate return to office? No. The curve describes what physical proximity does to communication frequency. It does not price the tradeoffs your company faces on retention, hiring reach, or real estate. It is most useful after the decision, as a guide to executing it well: if you bring people back, seat interdependent people within about 30 meters of each other, and measure whether the intended effect is showing up.
