In a novel choice attention-gating paradigm, observers monitor a stream of 3 ϫ 3 letter arrays until a tonal cue directs them to report 1 row. Analyses of the particular arrays from which reported letters are chosen and of the joint probabilities of reporting pairs of letters are used to derive a theory of attention dynamics. An attention window opens 0.15 s following a cue to attend to a location, remains open (minimally) 0.2 s, and admits information simultaneously from all the newly attended locations. The window dynamics are independent of the distance moved. The theory accounts for about 90% of the variance from the over 400 data points obtained from each of the observers in the 3 experiments reported here. With minor elaborations, it applies to all the principal paradigms used to study the dynamics of visual spatial attention.We explored a method of measuring the trajectory of spatial attention that is analogous to measuring the trajectory of subatomic particles in a Glaser bubble chamber (Gray & Isaacs, 1975). In the bubble chamber, a three-dimensional space is filled with a superheated liquid. A particle traveling through the liquid causes rapid localized boiling-microscopic bubbles-along its path. The bubbles can be photographed to indicate the particle's track and the tracks of decay and reaction products that it might produce. The complete tracks of all the bubbles define the trajectory of the particle (see Figure 1). In our procedure, a two-dimensional 3 ϫ 3 array is filled with letters. These letters change 7 to 10 times per second. Thus each letter occupies a little cube in a threedimensional space-time volume. During a movement of voluntary attention, some of the letters along the attention trajectory are entered into memory. The track of the remembered letters defines the trajectory of attention through the three-dimensional volume.Once attention trajectories have been measured in a threedimensional array of letters, it would be desirable to know whether these trajectories are the same trajectories as have been inferred from other procedures, such as partial report, which involves transfer from iconic memory (Neisser, 1967) to durable storage (Coltheart, 1980) of letters in a single two-dimensional array. A strong test of the possibility of equivalent attention trajectories in different experimental paradigms requires that all paradigms be tested with the same observers and with similar stimulus materials. Therefore, in addition to the main experiment, which measured attention shifts in three-dimensional letter arrays, two supplementary experiments were conducted with two-dimensional letter arrays: partial-report and whole-report procedures, each with and without poststimulus masks, all with the same 3 ϫ 3 letter arrays.Whereas estimating an attention trajectory from the letter-filled three-dimensional displays is relatively straightforward, estimating attention trajectories from partial-report experiments is complicated, requiring a full model of information decay, information transfer rates fro...