The concurrent-memory-load technique identifies attention demands with interactions between reaction-time-task parameters and the size of the load on memory. Three experiments are reported in which a multiple-choice, reaction time task involving two, four, and eig-ht stimulus-response (S-R) alternatives was performed alone and in the retention interval of a short-termmemory task involving ordered recall of eight digits. In Experiment 1, assignment of stimulus letters to response buttons (S-R mapping) was consistent for 6 days but varied on the seventh. Memory load and number of alternatives interacted early in practice, but the interaction diminished over days, and the effects were additive on Day 6. When the S-R mapping changed on Day 7, the interaction returned. In Experiment 2, S-R mapping varied daily for 6 days, and the interaction remained stable throughout practice. In Experiment 3, S-R mapping was consistent for 6 days and varied on the seventh, but the memory task was not introduced until Days 6 and 7. The interaction between memory load and number of alternatives was stronger on Day 7, after the mapping had changed, than it was on Day 6, after practice with consistent mapping. The discussion focuses on the validity of the concurrent-memory-load technique and its relation to another method for measuring attention and automaticity, and on the importance of practice with consistent mapping and practice with dual-task conditions in the development of automaticity.In general terms, attention refers to a cen-performance in some task environment. Pertral process that coordinates and controls formance is considered automatic to the extent that it is coordinated without attentional This research was supported by Grant AO 682and the development of automaticity from the National Research Council of Canada. I with practice refers to a transfer of control would like to thank Lee Smith for running the sub-from attention to reliable characteristics of jects. I am grateful to Ron Fisher, Paul Kolers, the task environment. The concurrent-memMorris Moscovitch, Ben Murdock Sandor Wise-or load technique has been offered as a man, and Jane Zbrodoff for their thoughtful comi • /• • ± *.-j^. ments and criticism throughout the development of method for measuring attention and automathis manuscript, and to Tom Carr, Michael Posner, ticity, and the experiments reported here and Anne Treisman for thorough reviews of an were designed to evaluate it. earlier version.The technique assumes that the capacity Requests for reprints should be sent to Gordon r .. 1 * • <• •,. i ,1 , • D. Logan, Department of Psychology, Erindale for attentional control is limited, that in pnnCollege, University of Toronto, Mississauga, On-ciple attentional capacity can be shared by all tario, Canada LSL 1C6.tasks and task components, and that atten-