Early "bottleneck" theories of selective attention allowed for only one channel of input to be semantically analysed, other information being discarded. Later modifications to attention theory proposed that all inputs were analysed but that much of this is unconscious and automatic. However, automatic processes are difficult to unlearn and control. This paper reports a study of theStroop effect, that these overlearned, automatic processing could intrude on a colour identification task. It was found that ink colour identification was slower for a list of colour names than when neutral words were used and that unconscious semantic processing was taking place.
Models of selective attention deal with processes that occur when multiple streams ofinformation is input. They have been conceptualised as an information processing structure; information stored in memory, or input from the senses, is processed by a limited-capacity cognitive system. One of the earliest models was developed by Broadbent[ref.1] who considered that incoming information on different sensory "channels" was selectively filtered and that only one stream could besemantically processed at a time. A channel in this context was a stream of information, distinguishable by a physical feature such as timbre, loudness or spatial location. Broadbent showed that attention could be switched between channels, a conclusion he drew from a series of dichotic listening tests, where subjects were able to recall separate numerical information being simultaneously fed to eachear. He proposed that input from one ear was temporarily stored, being processed when attention could be switched. This model typified the "bottleneck" approach to selective attention.
However, Cherry[ref.2], had already described the "cocktail party" effect, that it was possible for other input to intrude if relevant information was detected. Cherry's observations were confirmed by Gray andWedderburn[ref.3] in another dichotic experiment. They refuted Broadbent's model by showing that subjects could switch attention unconsciously if the dichotic content was meaningful. If this happened on the basis of meaning, some level of semantic processing must have occurred on an unattended channel.
Treisman[ref.4] produced a modified model that allowed for several channels to be analysed;information that was not immediately relevant was not lost but attenuated. A semantic "dictionary" then decoded meanings according to importance and pertinence. It was implicit to this theory that all input was semantically analysed in order that its relevance could be evaluated.
These models were problematic however, particularly in explaining how a physical filter can act upon meaning. Thiswas addressed by Kahneman's capacity model[ref.5]. Kahneman redefined attention as "mental effort", limited resources being allocated according to momentary intentions (tasks related to current goals) and enduring dispositions (responses to other important stimuli). Furthermore, Kahneman believed that some tasks required little processing as they were overlearned, automatic skills. Thus, severalactivities might share limited cognitive resources.
In a series of studies, Shiffrin and Schneider[ref.6] identified certain basic properties of these automatic processes: they are relatively free of capacity limitations, operate in parallel, require considerable training, are difficult to "unlearn", and are unconscious actions. Reading is such a task; it takes a significant amount of practicebut eventually becomes automatic. As an indicator, Healy[ref.7] found that subjects had difficulty in counting the frequency of letters in phrases containing common words such as "the". This recognition of whole words, rather than individual letters, is a result of an overlearned, automatic process.
J. Ridley Stroop[ref.8] had described the difficulty in naming ink colours of words that were...