Three experiments investigated the role of working memory in various aspects of thinking in chess. Experiment 1 examined the immediate memory for brieflypresented chess positions from master games in players from a wide range of abilities, following the imposition of various secondary tasks designed to block separate components of working memory. Suppression of the articulatory loop (by preventing subvocal rehearsal) had no effect on measures of recall, whereas blocking the visuospatial sketchpad (by manipulation of a keypad) and blocking the central executive (by random letter generation) had equivalent disruptive effects, in comparison with a control condition. Experiment 2 investigated the effects of similar secondary tasks on the solution (i.e., move selection) of tactical chess positions, and a similar pattern was found, except that blocking the central executive was much more disruptive than in Experiment 1. Experiment 3 compared performance on two types of primary task, one concerned with solving chess positions as in Experiment 2, and the other a sentence-rearrangement task. The secondary tasks in each case were both designed to block the central executive, but one was verbal (vocal generation of random numbers), while the other was spatial in nature (random generation of keypresses). Performance ofthe spatial secondary task was affected to a greater extent by the chess primary task than by the verbal primary task, whereas there were no differential effects on these secondary tasks by the verbal primary task. In none of the three experiments were there any differential effects between weak and strong players. These results are interpreted in the context of the workingmemory model and previous theories of the nature of cognition in chess.Thinking in chess is of particular interest to the psychologist, in that it can be represented both visuospatially-as a sequence of moves on a chessboard-and propositionally--either according to various types of notation or as verbal protocols, such as those collected by early researchers (e.g., de Groot, 1965). Variation in these modes of representation could conceivably distinguish strong from weak players, and even underlie these individual differences in chess skill. Milojkovic (1982) has proposed that stronger players rely to a greater extent on propositional coding, supporting the claim with a demonstration that the time taken by a master to decide whether a particular piece was under attack is independent of the physical details of the problem, such as the spatial separation of the pieces, whereas the latencies ofa group of weaker players were affected by such spatial features.Holding (1985) has criticized the emphasis on spatial processing by pointing out the verbal components of chess and referring to evidence that top players have been reported to exhibit superior verbal skills, as manifested particularly by their competence in acquiring foreign languages and their prevalence in professional writing occupations.We thank Pertii Saariluoma for helpful discussions...