Span of apprehension was measured for stimulus ensembles of various sizes, using letters, digits and an arbitrarily selected set of simple lines. Span was significantly decreased when letters and digits were presented as mixed sequences, in proportion to the number of letter-digit juxtapositions. Span also varied with the serial position at which a single juxtaposition occurred, being greatest where this was near the mid-point. These results were thought to arise from the higher transitional probabilities between items of the same class than of different classes, and the greater ease in grouping such items. No effect of stimulus uncertainty was found. Random letter span was significantly less than digit span, but span for sequences reflecting letter frequencies in the language did not differ significantly from digit span (at 100 msec exposure duration).

The unfamiliar line material yielded relatively lower spans, varying inversely with ensemble size (two, four and eight choice). This was determined not by overall stimulus uncertainty but by the degree of heterogeneity of the individual stimulus arrays. The relatively less heterogeneous stimulus assays were thought to lend themselves more readily to verbal recoding. With very familiar material transitional probability, in relation to previous experience of the language, determines span. With the less familiar material not subject to such response bias, it was the recodability of the individual display which determined the span. With only a single exception, comparable results were obtained at moderately and at very brief exposure durations.