In 1956, the psychologist George Miller published a paper called “The Magical Number Seven, Plus or Minus Two” in the Psychological Review. It became one of the most cited papers in the history of cognitive science. The claim was simple: the capacity of human short-term memory is approximately seven items, with most people falling between five and nine.
Forty-five years later, in 2001, Nelson Cowan revised the estimate downward. In a careful review of the experimental literature - “The magical number 4 in short-term memory” in Behavioral and Brain Sciences - he argued that when researchers controlled for chunking and rehearsal, the actual capacity of working memory was closer to four discrete items.
The number you actually run with, in real cognitive work, is between Cowan’s four and Miller’s seven. Juggling makes this visible in a way that few other tasks do, because each ball is one indivisible cognitive unit competing for the same attentional resource.
Why juggling tests working memory unusually cleanly
Most tasks that test working memory let you offload some of the load. You can write notes, group items into chunks, rehearse silently, or use external structure to extend your effective capacity. Juggling allows none of this. Each ball must be tracked in real time, in space, with no opportunity for chunking or external memory aids.
This means the number of balls a juggler can keep in the air is closer to a direct read of working memory capacity than most lab tasks produce. The numbers line up:
Three balls is the comfortable steady state for most adult jugglers after some practice. Four is harder, and the difficulty curve steepens noticeably. Five balls is the edge of what the typical practitioner can sustain at all. Seven is world-class. Nine is rare even among professionals.
These numbers are not coincidences. They track the working memory capacity range. Three balls fits comfortably below capacity, allowing some headroom for adjustment. Four to five sits at the limit. Anything beyond requires either exceptional baseline capacity or extensive automation of the lower-level processing - the same way expert chess players can hold more board positions than novices because chunked patterns count as single items.
What “drops” tell us about cognitive load
In the juggling literature, drops are a coarse but reliable indicator of cognitive overload. Studies by Kelso and colleagues, and later work by Beek’s group at Amsterdam, show that drop rates climb sharply as ball count approaches working-memory capacity, and even more sharply when secondary cognitive tasks are introduced.
The classic dual-task experiment: ask a juggler running a stable five-ball cascade to also count backwards from 100 by sevens. The drop rate increases significantly within seconds. The mental arithmetic competes for the same resource that the juggling tracking is using. There is no separate “juggling buffer” the brain can fall back on.
This is a useful demonstration of a property that is otherwise hard to see: cognitive load is fungible. Tasks that feel completely different - tracking moving objects and doing arithmetic - draw from the same finite pool. Adding any one task reduces capacity for any other.
The team analogue
The same finite pool applies to teams managing concurrent initiatives. Most engineering teams who try to track more than four major active workstreams report what looks structurally like overload: missed handoffs, dropped deliverables, slow context switches, accumulated debt that no one is tracking.
The number is not arbitrary. Four major initiatives is approximately Cowan’s working memory limit per coordination role. A team with one technical lead trying to keep four projects coordinated has loaded that role’s working memory to capacity. A fifth project does not get tracked - it gets dropped.
This shows up in retrospectives as “we forgot about X” or “no one was watching Y.” That is exactly the language of working memory failure. The team did not decide to drop the project. The coordination role’s working memory exceeded capacity, and one of the items fell out.
The number of balls you can juggle is approximately the number of concurrent projects your coordination role can actually track. Both numbers are limited by the same cognitive constraint.
What this implies for project portfolio limits
The implication for organisations that take working memory seriously is that there is a real ceiling on how many things a coordination role can usefully track, and it is not very high. Four is a reasonable target. Three is comfortable. Five and beyond requires either chunking - grouping projects into bundles that count as single items - or distribution of the coordination load across multiple roles.
Chunking works in some contexts. A “platform initiative” containing three coupled projects can be tracked as one item if the projects share enough state and timeline to be coherent. Chunking fails when the projects are unrelated and the bundle is just a label - the underlying load is still four items, even if the planning document shows one.
Distribution works better, but introduces a new coordination problem one level up. Two leads each tracking three projects is six projects total, but now there is a coordination cost between the two leads. The juggling parallel is exact: two jugglers passing balls between them can run more balls combined than either could alone, but the passing pattern is itself a new object that must be tracked, and the system can fail in new ways the solo cascade cannot.
What the juggling demonstration is good for
The reason juggling is useful as a teaching tool for cognitive load is not that it is a perfect analogy. It is that the failure mode is unmistakable. When you exceed capacity in a juggling pattern, the ball hits the floor. The signal is clear, immediate, and not deniable.
When teams exceed capacity on concurrent initiatives, the failure is delayed, distributed, and easy to attribute to other causes. The dropped ball might surface weeks later as a missed compliance deadline or a forgotten dependency. By then, the connection to the original overload has been lost.
A team that has held five balls in the air for thirty seconds, watched them fall, and felt the specific cognitive sensation of capacity overflow, has a reference experience that abstract slides do not provide. The metaphor stops being a metaphor and becomes an embodied calibration of what overload feels like before things start hitting the floor.
Further reading
- Miller, G.A. (1956). “The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information.” Psychological Review, 63(2), 81-97.
- Cowan, N. (2001). “The magical number 4 in short-term memory: A reconsideration of mental storage capacity.” Behavioral and Brain Sciences, 24(1), 87-114.
- Sweller, J. (1988). “Cognitive load during problem solving: Effects on learning.” Cognitive Science, 12(2), 257-285. The foundational paper on cognitive load theory.
- Beek, P.J., and Lewbel, A. (1995). “The science of juggling.” Scientific American, 273(5), 92-97. The cognitive demands of multi-ball patterns.
- Kelso, J.A.S. (1995). Dynamic Patterns: The Self-Organization of Brain and Behavior. MIT Press. Coordination dynamics under cognitive load.