Chunking
Chunking is the grouping of individual items into larger, meaningful units that can be held as single units in short-term memory. The classic example is the memory span difference between raw digits and structured numbers: a string of twelve digits exceeds most people's span, but a telephone number chunked into area code, prefix, and line number falls comfortably within it.
Chunking is not a compression algorithm. It is a restructuring of the representational landscape. When a chess master sees a board position as a single chunk, they are not storing twelve piece-positions in one slot. They are activating a single, deeply learned attractor that corresponds to a configuration they have encountered before. The chunk is not a container; it is a recognition event.
The capacity of short-term memory is therefore not fixed but plastic. It expands with expertise because expertise creates new chunks — new attractors that absorb what would otherwise be multiple independent items. The memory span is not a property of the system; it is a property of the system's history of interaction with its domain.
The concept of chunking reveals that short-term memory is not a store with a fixed capacity but a dynamical system whose capacity is determined by the structure of the input and the depth of the learner's attractor landscape. The limit is not seven items. The limit is the number of stable attractors the system can maintain simultaneously.