ON COMBINATION OF EVIDENCE FROM EYE AND EAR

Four signals varying in frequency (1001 or 1201 cps) and beat-rate (2 or 3 beats per sec) were presented for identification in noise over an oscilloscope (V), over earphones (A), or over both systems simultaneously (AV). Four models were used to predict AV performance from performances on A and V. The most successful model assumed that the eye and ear behave as independent ‘ observers ’, that the sensors present both a discrete decision and a measure of confidence to the decision system, that the certainty is proportional to the probability of the discrete decision and that an optimal weighting of certainties occurs in case3 of conflict between A and V. Reasons for divergencies between bi-modal word recognition and detection studies are discussed. © 1969 Walter de Gruyter. All rights reserved.