Lesson 9 · ACT-R: declarative vs. procedural memory

Knowing That vs. Knowing How

Lesson 3 built a production system from IF–THEN rules over a working memory of facts. ACT-R (Adaptive Control of Thought—Rational) is what happens when that skeleton grows into a full theory of human cognition, built by John Anderson starting in the 1980s — and its first big move is splitting memory into two kinds that lesson 3 left undifferentiated:

  • Declarative memory — “knowing that”: facts, retrievable as chunks. A chunk is a small, structured unit of knowledge (think: a record with a few named fields) — “the capital of France is Paris,” “the knight on d5 is pinned,” “my opponent is rated 2400.” Chunks are things you can state.
  • Procedural memory — “knowing how”: production rules, exactly lesson 3’s IF–THEN form, that fire on the contents of declarative memory (and the current goal) to produce behavior. Productions aren’t things you consciously state — you can’t fully narrate the rule that lets an experienced driver brake at the right instant; you just do it.

Every chunk in declarative memory carries an activation value — a number capturing how easily and quickly it can be retrieved right now. High-activation chunks surface fast and reliably; low-activation ones are slow, error-prone, or simply fail to come to mind. (This is lesson 2’s “chunk” — Miller’s 7±2 — made formal and numeric: a unit worth holding onto, now with a quantity attached to how held-onto it is.)

A chess player’s mind holds all four items above. Which one lives in procedural memory rather than declarative?

Which of these is stored in PROCEDURAL memory (a production rule), not declarative memory (a chunk)?