Models of attention and memory dating from the 1950s are reviewed with particular emphasis on dichotic listening experiments. Structural models, such as that of Broadbent (1958) are compared with, what are termed functional models eg. those of Neisser (1967), Shiffrin and Schneider (1977). Methodological and scoring problems in split span experiments are examined. Preliminary experiments were designed to show that the Gray and Wedderburn (I960) effect reflects a perceptual process which is largely unconscious rather than report preferences or strategies. The following four experiments investigated the effects of presentation rate, delayed recall and priming on responses to lists of differing semantic complexity. The faster presentation rate was found to encourage responses based on context rather than spatial location, as did delayed recall and priming of contextual lists by the contiguous presentation of similarly structured lists. Interference effects were found on lists with primes similar in content rather than structure. Evidence was found that context and category lists have different quantitative effects on responses. The interaction of variables in these experiments is emphasised and the results are interpreted in terms of functional models, with particular emphasis on components of analysis-by-synthesis (Neisser, 1967) levels of processing (Craik and Lockhart, 1972) and automatic processing (Shiffrin and Schneider, 1977).