In the 1940s, a Dutch chess master named Adriaan de Groot ran an experiment that changed how we think about expertise.
He showed chess positions to players of different skill levels — grandmasters, international masters, club players, beginners — and asked them to remember and recreate the positions after brief exposure.
Grandmasters were dramatically better at remembering realistic game positions. Their performance seemed to confirm the common assumption: stronger players have better memories, which is why they’re stronger players.
But de Groot added a crucial control: he also showed players random positions, with pieces distributed across the board in ways that would never occur in a real game.
For random positions, grandmasters were no better than beginners. The memory advantage completely disappeared.
This finding, later confirmed and extended by Herbert Simon and William Chase, revealed something fundamental about expertise: grandmasters don’t remember chess positions through superior general memory. They remember them through recognition — they see familiar patterns and store them as chunks. Patterns that appear in real games are stored as patterns. Random arrangements of pieces aren’t, because they don’t correspond to anything in the grandmaster’s library of chunks.
This is the concept of mental representations, and Anders Ericsson placed it at the center of his theory of expertise. Deliberate practice doesn’t build skills in the way we usually imagine — through muscle memory or repetition-induced automaticity alone. It builds sophisticated mental representations that allow experts to perceive, process, and respond to their domain in fundamentally different ways than novices.
What Mental Representations Are
A mental representation is an internal model — a structured pattern of knowledge that lets you perceive and process information efficiently.
At the most basic level: a beginner looking at a chess position sees thirty-two pieces on sixty-four squares. A grandmaster looking at the same position sees maybe five or ten meaningful patterns. The grandmaster’s perception is organized at a higher level of abstraction. They’re not processing pieces; they’re processing relationships between pieces and their strategic implications.
This is chunking. Individual elements are grouped into higher-order units that can be processed as single entities. Novices process at the element level; experts process at the chunk level.
The cognitive benefit is enormous. Working memory is limited — we can hold roughly seven items in active attention at once. If each item is an individual chess piece, we can track seven pieces. If each item is a meaningful pattern involving multiple pieces, we can track seven patterns, which might represent fifty pieces’ worth of strategic complexity.
Experts aren’t processing more information. They’re processing the same information at a higher level of abstraction, which means they can handle more complexity within the same cognitive constraints.
How This Works in Magic
When I started learning card effects, I was processing at the element level.
Each sequence was a series of discrete moves. Step one: this. Step two: this. Step three: this. The sequence existed in my mind as a list of steps, not as a coherent whole. Performing it required conscious step-by-step execution. Attention was consumed by “what’s next?” rather than available for the audience.
This is the novice state. The elements haven’t chunked into patterns.
Over time, with sufficient correct practice, the sequence chunked. The individual steps became a single unit — a routine — that could be executed at a higher level of abstraction. Instead of “step one, step two, step three,” the internal representation became “this routine” that executed as a whole. Attention freed up for what matters: the audience.
But mental representations in magic go beyond technical execution.
Expert performers have mental representations of audience states — patterns they recognize in how a group is responding, that trigger automatic adjustments. They have representations of effect structure — the underlying logic of why a type of effect works, which lets them generate new effects that share the relevant structure. They have representations of performance dynamics — how energy flows, what creates momentum, where the danger points are.
These higher-level representations are what distinguish a technically proficient performer from an expert performer. Both might execute the same moves equally well. The expert perceives the performance situation at a higher level of abstraction and responds with more sophisticated judgment.
The Difference This Made for Me
I can trace a specific moment when I felt a mental representation shift.
For a long time, when I was performing and something unexpected happened — an audience member said something I didn’t anticipate, a prop wasn’t where it should be, someone walked past at a critical moment — I experienced it as disruption. The sequence I was executing got interrupted, and I had to stop, assess, figure out what to do, restart.
The processing was at the element level. Sequence got interrupted. Elements scattered. Have to reassemble.
At some point, and I can’t pinpoint exactly when, the experience of performing shifted. The unexpected things stopped feeling like disruptions to a sequence and started feeling like events within a performance. The representation had shifted from “sequence of steps” to “performance as a whole.” Unexpected events within a performance are normal; they don’t interrupt the sequence because there’s no sequence at the level I’m now processing.
I noticed this first when I stopped losing my place. When something unexpected happened, I no longer felt the cognitive jolt of “where was I?” I was already somewhere larger. The unexpected event was within that larger something, not disrupting it.
This is what mental representations feel like from the inside when they’ve developed sufficiently. Not a new skill. A change in the level at which you’re perceiving.
What Practice Is Actually Doing
Understanding mental representations changes the goal of practice.
The naive view is that practice builds skills — that the mechanism is repetition creating muscle memory creating automaticity. Repeat enough times and the skill becomes automatic.
This is partially right but misses the more important process. Repetition does create automaticity. But deliberate practice does something more: it builds and refines mental representations.
When you practice an effect correctly, you’re not just training the hands. You’re building an internal model of the effect — what it should look and feel like at each moment, what the audience’s experience should be, what the indicators are that it’s going well versus poorly, what the intervention points are if something goes wrong.
This internal model is what lets you perform reliably across varying conditions. The hands do what they’ve been trained to do. But the monitoring and adjustment that keeps performance on track is the mental representation doing its work in real time.
This is also why just doing effects many times isn’t sufficient. You can repeat an effect a thousand times without building a sophisticated mental representation if the repetitions don’t include feedback, reflection, and refinement. The repetitions build automaticity in whatever pattern you’re repeating — including the wrong pattern, if you haven’t identified the wrong pattern as wrong.
Deliberate practice, in Ericsson’s framework, is specifically the kind of practice that builds and refines mental representations: practice that includes clear targets, immediate feedback, and deliberate adjustment.
The Expert’s Perception Problem
There’s a specific challenge in mentoring and teaching that mental representations explain.
Expert performers often give advice that’s accurate but impossible for novices to act on. “Feel the flow of the routine.” “Read the audience.” “Trust your instincts.” “Know when to slow down and when to push.”
These aren’t meaningless platitudes. They’re accurate descriptions of what experts actually do, described from inside the expert’s mental representation. The expert isn’t trying to unhelpfully mystify. They’re describing genuine experience.
But the novice can’t act on the advice because they don’t have the mental representations the advice presupposes. “Read the audience” means something to someone who has a sophisticated mental model of audience states and what each state implies. It means nothing useful to someone who has never built that model.
The transmission of expertise isn’t description. It’s building mental representations in the learner. Which requires specific, deliberate experience designed to develop those representations.
This is why the most valuable mentoring involves not just telling but structured experience — performance situations designed to build specific representations, with feedback designed to correct and refine them.
A Personal Reflection
The model of expertise that Ericsson offers — expertise as accumulated mental representations rather than raw skill or logged hours — is simultaneously more demanding and more hopeful than the alternatives.
More demanding because it means the goal is not to reach 10,000 hours but to build the specific representations that expertise actually consists of. That requires knowing what those representations are, designing practice to build them, and honestly assessing whether they’re developing. This is much harder work than just putting in time.
More hopeful because it means the ceiling is not fixed. The grandmaster doesn’t have better general memory than the club player — they have better chess-specific mental representations. Those representations were built, not given. What was built through deliberate practice by one person can be built through deliberate practice by another.
The difference between where I was in that Innsbruck hotel room — holding a deck and feeling hopelessly behind — and where I am now is not ten years of time. It’s ten years of accumulated mental representations: internal models of effects, audiences, routines, performance dynamics, and my own particular intersection of all of these.
Those representations are why I can do things now that I couldn’t do then, beyond just the technical execution. I perceive the performance situation differently. The patterns are available. The chunks are there.
That’s what practice actually builds.
Not hours. Not muscle memory, exactly. Mental representations.
Everything else follows from those.