Tower of London test: method

What it measures

Planning (look-ahead problem solving): solving in the minimum number of moves, with initial planning time (before the first move) and execution time.

Procedure

Three pegs that hold 3, 2 and 1 balls; three coloured balls. Each problem shows the start, the goal and the minimum number of moves. Only a top ball can move; a ball cannot go onto a full peg; lifting a ball and putting it back on the same peg is not a move. Input: tap a ball and then a peg, or drag; keys 1, 2 and 3 pick up and put down, Esc puts the ball back. One practice problem (2 moves) first; default for new players, skippable. Then 16 problems, from easy to hard by minimum moves (order shuffled within a level). A problem ends when solved, after 20 moves or after 60 s, and the run goes on. Ball colours are randomised each run, so a replay looks new but is just as hard. Leaving the browser tab ends the run. About 5 minutes.

Parameters

Scoring

Per problem: solved (goal reached within the move and time caps), number of moves, perfect (solved in the minimum), extra moves, planning time (problem onset to first ball touched) and total time. Primary: number of perfect solutions, 0 to 16; higher is better. Also: problems solved, total extra moves on solved problems, mean planning time, mean total time, results per level. 'How you solved': on problems of 4 or more moves, share of time before the first move = sum of planning times / sum of total times; 0.40 or more = 'planned first', otherwise 'worked it out as you went' (a description of this run, not a judgement). Compared with other players only as a band (below typical = under the 25th percentile, typical, above typical = over the 75th) from 100 comparable players, never as an exact rank, because one run has low retest reliability. Player comparisons use only first runs (no earlier run by the same visitor, account, or IP address and device) without quality flags, with the same scorer version and the same runner settings.

Valid-run rules

All 16 problems must be sent. The server rebuilds the problems from the run's seed and replays every move against the rules; the run is rejected if a move breaks a rule, move times go backwards, the first touch comes after the first move, or a problem ends before its last move. A solution after 61.5 s counts as unsolved. The server rebuilds every run from the raw log (and the run's seed where there is one) and computes the score itself; anything the page says about right or wrong is ignored. A log that breaks the game's rules is rejected, and nothing is stored. Stored runs can carry quality flags that are never shown and never block a result: 10 or more runs from one IP address, a missing visitor id, tab switches, and automation signals (automated browser, impossible speed, no input events, machine-regular timing).

Differences from the classic task

Shallice used a physical board and a smaller set of problems. This version has its own 16 problems from 2 to 6 moves, shows the minimum, and caps moves and time. Every minimum was found by a breadth-first search over all 36 positions, and each problem has exactly one shortest solution (re-checked for this text). Problem structure (goal shapes) is varied, following Kaller et al. (2011). No norms from other versions apply.

Limitations

Single runs of the Tower of London are noisy on retest (moves r = .15 over two weeks in Piper et al. 2015), so the result is shown only as a band. Tiredness, the device and how the controls feel affect the result. It does not measure intelligence and is not a clinical test.

Languages and review status

Language-independent stimuli, instructions in all site languages (English plus 99 other languages).