Division of Attention between Non-Verbal Signals: All-Or-None or Shared Processing?

Criteria proposed by Broadbent (1971) for deciding between “all-or-none” and “shared” models of divided attention are reviewed and modified. Of the modified criteria, the amount of the reduction in efficiency and the shape of the normalized ROC curve are identified as being diagnostic for deciding between the two models. When applied to the results of two experiments involving the recognition of difficult nonverbal signals, which showed performance on the signals presented together to be significantly worse than performance on the signals when presented alone, both the criteria favour the shared model rather than the all-or-none model. The results of a third experiment with easy signals indicated that artefacts unrelated to attention were unable to account for the reduction in efficiency with difficult signals. It is concluded that the shared model, which assumes some limit to processing capacity, has the widest application to these results and similar findings reported elsewhere. The role of the all-or-none model would appear to be restricted to conditions of extreme informational overload. Further, the movement of noncentral decision criteria and the unequal contribution of the individual dimensions to overall multi-dimension performance demonstrated in the first two experiments, suggest that a reduction in the quality of decision-making may accompany the division of attention, as well as the reduction of information derived from the signals.