A prediction based on the model of attention advanced by Dixon (1971) wm tested: namely, that responses to homophones presented to one ear, at supraliminal intensities, would be influenced by subliminal cue words presented to the other ear. Considerable support was found for the hypothesis in terms of response latencies, but not in terms of verbal content. It is suggested that these data make it possible to reconcile apparent discrepancies between the results of other recent studies of dichotic listening. The view of those attention theorists (Deutsch & Deutsch, 1963; Norman, 1968; Dixon, 197 1 ) who maintain that selection among inputs follows a preliminary semantic analysis of all signals, both attended and unattended, has received support from recent studies of dichotic listening (Lewis, 1970; Corteen & Wood, 1972; Mackay, 1973; Henley & Dixon, 1974), and dichoptic viewing (Somekh & Wilding, 1973).Data from these studies have suggested not only that material, of which subjects remain unaware, on an unattended channel, is analysed for meaning, but moreover that it may be integrated with primary material on the attended channel when it is relevant to the ongoing task. Mackay, for example, has shown that a word on the unattended channel in dichotic listening will serve to reduce the ambiguity of an attended sentence.According to the model of attention put forward by Dixon (1971), a supraliminal stimulus, having passed through analysers to dictionary units, ultimately activates conceptually associated units while at the same time setting in motion a central control mechanism which, acting upon information from various inputs (relating to past and ongoing events, motivational states, etc.), selectively inhibits such conceptual associates as are not relevant to the task in hand. A subliminal stimulus, on the other hand, activates the conceptual associates, but is of insufficient strength to set the central control mechanism in operation. Under these conditions, sensory data are analysed and classified, but nothing is blocked.The present experiment tested a prediction from this theory that responses to homophones presented to one ear, a t supraliminal intensities, should be affected by subliminal cue words on the other ear. It was postulated that a supraliminal homophone would activate associates belonging to two spheres of meaning. Assuming responses to either meaning of the homophone to be equiprobable, there would be no a priori reason why any one set of conceptual associates should be inhibited more than the other. Under these circumstances, the ambiguity should be resolved by a subliminal cue word activating its associates. The latter would, presumably, overlap one set of those associates activated by the supraliminal stimulus, and as a result of the 'extra strength' thus gained, it would be to this meaning of the homophone that the subject should respond. By the same token, if the subliminal stimulus did reduce the ambiguity facing the central control a t the level of the 37-2 upon personal preferences. J. Pers...