Visual Short Term Memory for Location: Does Objecthood Matter?

VSTML has a high capacity.1 We further document its capacity while testing for object-based constraints on capacity. Sample displays had 36 possible element locations (18 filled), followed by a 1 sec interval, and then a test display in which one element moved (a sample element offset and a new test element onset). Capacity was measured with Cowan\u27s K (most conservative, reported here), Pashler\u27s K, and a new K2 to be presented, adapted for the dual events of location-change. Is VSTML capacity fixed by object? In Experiment 1, capacity was reliably higher for displays of one big, 36-location object (K=11.1 elements; 11.1 per object) than for displays of 3 12-location objects (K = 9.8, or 3.3 per object). Clearly, capacity was not fixed by object. Are object representations independent from each other? If so, then capacity could depend on internal object structure but not on relations between objects. Displays were 3 12-location objects. In Experiment 2, element size was constant throughout a display or varied between objects in a display. Variation disrupted symmetry and alignment between objects, and caused capacity to be reduced reliably. In Experiment 3, element shape was constant or varied between objects; variation reduced symmetry, but did not reduce capacity. In Experiments 4 and 5 we also varied elements within object; internal object variation of size and shape reduced capacity more than between-object variation. Kranged from 10.8 to 6.4 elements in these experiments. Object-based models are supported by the greater importance of within-object relative to between-object relations (Experiments 4 and 5) but wounded by flexibility of capacity (Experiment 1) and effects of size relations that span separate objects (Experiment 2)