Attention

4. Broadbent’s filter theory and the early/late selection debate

No treatment of attention has been more influential than that given in Donald Broadbent’s Perception and Communication (1958). This book had an important role in establishing ‘cognitive’ phenomena as reputable explananda for scientific psychology. Prior to Broadbent’s work, and that of his contemporaries, the influence of behaviourism was such that the postulation of mental entities not immediately connected with publicly observable stimuli or manifest behaviour was regarded with suspicion. Broadbent argued that behaviourist psychology’s picture of mental phenomena as arising from the passage of information between different processing channels, with the brain in the role of cognitive switchboard, was a picture of a system in which there would be a need for the management of bottlenecks in processing capacity. ‘The fact that any given channel has a limit is a matter of central importance to communication engineers, and it is correspondingly forced on the attention of psychologists who use their terms’ (Broadbent 1958: 5). Since the management of bottlenecks in information processing capacity was, according to Broadbent, the job of attention, the claim that psychology needed to take note of capacity had the effect of reintroducing the previously suspect phenomenon of attention to the psychologist’s agenda.

In the Broadbentian picture attention serves to manage a bottleneck that occurs when a large-capacity processing system feeds into a smaller-capacity processing system. This picture was common to almost all the theorists in the decades immediately following Broadbent’s work. For these theorists the first task for a theory of attention was to say which sorts of processing are handled by the large capacity system and which by the small capacity system, and thereby to locate the bottleneck of attention at a particular point in the brain’s information processing hierarchy.

Broadbent’s own view was that the large-capacity system, which operates before the attentional bottleneck, is responsible only for the detection of simple physical properties. The small-capacity system, operating after attention, is responsible for the detection of semantic features and for the recognition of stimuli with known identities. This was known as the ‘early selection’ view. Broadbent’s principal rivals, the ‘late selectionists’, held that almost all processing is done ‘pre-attentively’ by the large capacity system. According to this view the only processing that takes place after the attentional bottleneck is the processing by which already processed stimuli are brought to consciousness. In its extreme form the early selectionist view leads to a version of epiphenomenalism (see Epiphenomenalism), since it implies that the occurrence of conscious awareness is separate from and subsequent to the processing that makes a difference to the determinants of behaviour and thought.

The empirical study of attention in the decades following Broadbent’s work was dominated by the debate between those who followed Broadbent in his ‘early selectionist’ view and those who adopted the ‘late selectionist’ rival.

The early selectionists supposed that, since the capacity for extracting semantic information from stimuli was small and allocated by the paying of attention, subjects would be unaware of all of the semantic properties of those stimuli to which they were not attending. Evidence consistent with this hypothesis was easy to find. Cherry had shown, as early as 1953, that if headphones are used to present a subject with two streams of speech, one in each ear, and if the subject gives all of his attention to one of these streams, then the subject will be unable to report anything about the content of the unattended stream, and doesn’t even appear to know whether it is speech or nonsense. Non-semantic properties of unattended speech – properties such as pitch – can be detected for all speech whether attended to or not. All this is just as the early-selection theory would predict.

The late selectionists, meanwhile, supposed that since semantic processing went on whether or not one was paying attention, there would be processing of the semantic properties of stimuli that were not being attended to, and the representations resulting from that processing would have an effect on the subject’s subsequent performance. Evidence consistent with this hypothesis was also found. It was shown that subjects take longer to respond if the picture they are responding to shares a semantic property with a picture that has recently been ignored (Driver and Tipper 1989). It was also claimed, although this is more controversial, that semantic properties not attended to can elicit a slight change in skin conductivity (indicating fear or anticipation) in subjects who have been suitably conditioned (Corteen and Dunn 1974; Dawson and Schell 1982).

Neither Broadbent’s theory nor its rivals could accommodate all of the data concerning the fate of stimuli to which no attention is paid, and by the early 1990s the lack of resolution had come to be seen as evidence that the terms of the debate were somehow misconceived. There was little consensus, however, about where the early/late selection debate had gone wrong. Some psychologists took it that the post-behaviourist rehabilitation of attention had been a mistake (see Parasuraman 1998). Some, such as Alan Allport, thought that the mistake was in trying to provide a single unified theory of attention, and suggested that the explanation of attention needed to proceed in a more piecemeal fashion (Allport 1993). Others took the problems to be more specific to the terms of the early/late selection debate, arguing that the problem was not that there is no processing bottleneck that explains attention, but that the attempts to locate the bottleneck failed because it is a bottleneck that moves (Lavie 1995).