Working memory — the cognitive system that holds information actively in mind during an experience — has a fundamental capacity limit. In 1956, psychologist George Miller published what became one of the most cited papers in cognitive science, establishing that this capacity is approximately seven items, plus or minus two. This has become known as Miller’s Law, or, in Miller’s own phrase, “the magical number seven.”
The implication for performers is uncomfortable: your audience cannot hold more than about seven discrete pieces of information in active working memory at any moment. When you add an eighth, something drops. Usually the first thing in.
I understood this intellectually before I understood it in practice. The gap between those two kinds of understanding took me an embarrassingly long time to close.
The Multi-Phase Problem
My early mentalism routines were long. I thought longer meant more impressive. I thought adding phases — additional demonstrations, additional moments of apparent impossibility — accumulated impact. Each new phase added to the total effect, compounding the experience.
This is wrong, and the working memory research explains why.
When you run a four-phase routine, you are asking the audience to maintain cognitive tracking of: the premise established in phase one, the specific conditions of phase two, the escalation in phase three, and whatever you’re asking them to notice in phase four. Plus the social context, your instructions, their own emotional responses, the names of participating spectators. You’re well past seven pieces of information before phase three begins.
What actually happens in this situation is that the audience starts dropping material. They cannot hold all of it simultaneously, so the working memory system automatically prioritizes. Recent information stays. Older information decays. By the time you reach the climactic moment of your carefully constructed phase four, the audience may have dropped the specific conditions from phase one that made the climax meaningful. The thing that should have been the payoff lands flat because the setup has been forgotten.
I experienced this directly. There was a routine I was proud of — three connected effects that built on each other, each one establishing a condition that made the next one more impressive. In practice, running it for myself, the architecture was elegant. The third phase was made more astonishing by the conditions established in the first. Running it for actual audiences, the third phase consistently failed to land with the impact I expected.
Watching a recording, I figured out what was happening. By phase three, the audience had largely released their grip on phase one. The callback that should have landed with accumulated weight was landing in isolation.
What Working Memory Actually Means for Performance
The seven-item limit is not a hard wall. It’s a useful approximation of a much more complex system. Working memory doesn’t simply hold seven discrete facts — it holds “chunks,” which can be small or large depending on how meaningfully they’re organized. An expert chess player can hold far more about a board position in working memory than a novice, because the patterns they recognize are larger chunks.
This suggests the performer’s job is partly about chunking: organizing information in ways that make multiple related pieces pack together into a single memorable unit. If phases one and two are presented as a unified idea — “this is about the relationship between an object and a word” — the audience holds them as a chunk rather than two separate items. More cognitive space becomes available for phase three.
Narrative is the primary chunking mechanism available to performers. When multiple moments in a routine are connected by a clear story arc, they compress into fewer working memory units. The audience is holding a story, not a list of events. Stories have internal logic that helps maintain coherence through time.
This is one of the reasons that a strong narrative frame transforms how audiences experience even technically simple material. It’s not just aesthetically satisfying — it’s cognitively efficient.
The Simplification Problem
Once I understood the working memory constraint, I went through a period of aggressive simplification. I cut phases. I removed anything that felt like elaboration. I stripped routines down to their essential moment.
This went too far. A routine with no texture, no development, no shape — just a single moment of surprise — doesn’t accumulate any emotional weight. It reads as a trick rather than an experience. Working memory limits don’t mean less is always more. They mean that what you include needs to earn its place by serving a specific function.
The question I learned to ask about each element of a routine: what does this put into the audience’s working memory, and does that addition make what comes later more powerful, or does it simply replace something that was already there?
If adding a phase to a routine means a different phase gets dropped, the question becomes which one is more worth holding. If the new phase merely creates information that competes with more important information already in play, it shouldn’t be there.
The Recency and Primacy Effects
Working memory research also reveals two relevant patterns: the primacy effect (the first items encountered in a sequence are remembered better) and the recency effect (the last items encountered are remembered better). Items in the middle of a sequence are most vulnerable to being dropped.
For performance, this has direct structural implications. The most important information should appear either at the beginning or at the end of a routine, not in the middle. If there’s a condition that needs to be established and then recalled, establish it very early, with emphasis, and call it back at the climax. Don’t bury it in the middle.
This is also why the structure of any performance is important beyond just pacing. The beginning and end hold special cognitive authority. What you put there will be remembered. What you put in the middle is competing for limited resources and should either be interesting enough to maintain attention actively or be brief enough to not crowd out the things that matter.
A Practical Audit
After working through this material, I developed a rough practice for auditing routines. I run through a routine mentally and list, at each moment, what a typical audience member is holding in working memory. Not just the explicit information I’ve given them — also the emotional context, the instructions I’ve issued, the social dynamics I’ve established with participating spectators.
The list quickly becomes illuminating. Most of the time, by the climactic moment, the list is too long. Something needs to be cut or compressed. The question is what has been earning its place and what is simply occupying space.
What I’ve consistently found is that the cuts I was most reluctant to make — the clever elaborations, the extra conditions, the additional phases that showed off the range of what the routine could do — were often exactly the cuts that made the remaining material land harder. The routine got shorter and better at the same time.
This is counterintuitive if you’re coming from a mindset that equates quantity with value. It aligns perfectly with what cognitive science tells us about how minds actually work under the conditions of live performance.
The Seven as a Design Constraint
I’ve started treating Miller’s seven as a design constraint rather than a limitation — the way an architect treats load-bearing walls. You don’t resent the constraint. You design within it, and you find that the constraint produces cleaner, stronger work than you would have produced without it.
A routine designed to fit within working memory limits is a routine in which every element serves the climax. There’s no room for elaboration for its own sake. The design question becomes: what is the minimum information the audience needs to hold in order for the payoff to land with full force? Build exactly that, in that order, and trust that the audience’s memory will take care of the rest.
Because here’s the other thing the research shows: information that is emotionally significant gets preferential treatment from memory systems. What audiences genuinely feel lands in long-term memory far more reliably than what they intellectually processed. If you’ve given them an experience, they’ll remember it. If you’ve given them a list of impressive things, they’ll remember the list incompletely and the experience not at all.
Every time I’m tempted to add one more phase to a routine because it’s clever and I can do it, I think about Miller and his seven items. Usually I cut instead. And usually the routine is better for it.