Source
Kirsh and Maglio (1994), “On distinguishing epistemic from pragmatic action”, LIT-1054, read in full in NOTE-878 from the scan of the version of record, with Kirsh (1995), LIT-1076, read in NOTE-871, which restates the Tetris case (§5.1).
What this asserts
Two things, as the authors do. First, that players rotate each piece more often than plan-then-act would need (Figure 6: about 1.7 to 1.75 rotations against an expected 1.5 for four-orientation pieces, read off the figure), start rotating within a few hundred milliseconds, and use the translate-to-wall check more the higher the drop (Figure 15). Second, that these are epistemic actions in the authors’ sense, “a physical action whose primary function is to improve cognition” by reducing the memory, the number of steps or the probability of error of mental computation, taken because a physical rotation (as little as 100 ms) is faster than a mental one, at the moment of use and with nothing stored.
What the record holds toward it
THEORY-360 (Active) holds the timing (“rotation begins in earnest by 400 to 600 ms”) and Table 1’s longer drops after the wall check at every skill level. The physical-versus-mental comparison comes from a pilot with three players, and on its figure 800 to 1200 ms is total reaction time, not the increment for a 90° rotation (about 300 ms). No score or placement was compared with and without the moves.
What would confirm or reject it
The promote_when, carried over unchanged from THEORY-360: an experiment that varies what the physical action costs and measures the computation it is said to replace directly.
What this does not say
- That a result is stored for later; the paper sets itself against caching.
- That offloading pays wherever action is possible; the paper names speed and a determinate local effect as the conditions.