This essay aims to discuss the degree to which Anne Treisman’s (1960, 1964) attenuation theory of selective auditory attention clarifies evidence admitting from the dichotic listening paradigm (e.g. Broadbent, 1957, 1958; Cherry, 1953; Conway, Cowan, and Bunting, 2001; Deutsch & Deutsch, 1963; Lavie, 1995; Lavie, Hirst, de Fockert, & Viding, 2004; Moray, 1959; Norman, 1969). Taken as a whole the early studies under discussion provide a telos of thought that may be said to form the foundation from which divergent understandings on contemporary attention research continue to find purchase.
How is it that humans come to selectively attend to what seems a bracketed amount of sensory information amid the other vast array of sensory information surrounding us? It might be said that the subject of attention has provided much grist for what one might call the mill of modern psychology. To William James’ (1920: pp. 160-163) functionalist thought and deft penmanship, consciousness was likened to a bird’s life on the wing or on the perch. That is to say, consciousness was viewed categorically as either occupied with attending to sensory images (as if perching), or filled with the relation of sensory images (as if in flight). James’ (1920: pp. 217-219) view of attention seems to extend, reasonably, from his view of consciousness, insofar as attention was either substantive (active) or transitive (passive). In contrast to James’ formulations of and for substantive or transitive modalities of attention, for Sigmund Freud (1912e: pp. 115-116) the famous injunction made plain to would-be physician analysts – in lieu of their charming later practices – demanded of them an “evenly hovering attention” (Ger. gleichschwebende Aufmerksamkeit). With the rising influence of John Watson, followed by Burrhus Skinner’s radical form of behaviourism, so too, came the hastening relegation of attention altogether. (As well as most objects of psychological concern deemed ‘mental functions’.) Psychological research centring on attention remained in a stasis for some thirty years or so – in equal measure unable to die yet unable to live.
However, concern and consideration of the active selection of certain stimuli over others (read attention) was once again to rise and find a place within common parlance through increased coverage in psychological journals with the shift in emphases favouring empirical laboratory research known as the cognitive psychology paradigm. Colin Cherry (1953) found himself fascinated with the so-named ‘cocktail party effect’ (i.e. how it was that we were able to follow one conversation amid the flood of many other people talking all at once). Cherry (1953) experimented with discrete audible messages given to one and then both ears – it is this method first employed by Cherry that became the basis for what psychologists now widely refer to as dichotic listening tasks. In addition to which, Cherry’s (1953) findings were compatible with the suggestion that the ability to discriminate between auditory stimuli relied in part on audible physical (e.g. gender) and spatial (e.g. location) differences. Cherry (1953) went on to test subjects’ ability with one auditory test message to be shadowed in one ear (i.e. repeated back verbatim) whilst, in coincidence, a second auditory test message was played to the other ear. Using this method Cherry noted that on the one hand very little information could be gained from quite pronounced verbal differences in the non-attended messages (e.g. reversed or foreign (German) speech), and on the other hand that physical differences (e.g. pure tonality) was for the most part quickly detected. Cherry’s interest in the unattended ear and its lack of auditory information was to leave him to conclude that the ear in question was in possession of practically no processing. It appeared that attention was something that could be directed to different locations; it also seemed as if attention were analogous to a spotlight beaming onto particular items and locations in an otherwise dark room cloaked in mystery.
Impressed by Cherry’s (1953) shadowing method, Broadbent (1954; see also 1952) set about testing participants’ ability to shadow in what are known as split-span experiments. Here, three pairs of digits (e.g. 37-42-12) are presented to each participant dichotically using headphones. In an initial experimental study, three pairs of digits (37-42-12) separated by a 500msecs interval are presented to participants in both ears. What Broadbent (1954) found (condition 1. 65% and condition 2. 20%, respectively) was that most participants were able to relay (1.) ear-by-ear (341-722 or 722-341) rather than (2.) pair-by-pair (37-42-12). In a second experimental study based on the example of the auditory characteristics of both stimuli, Broadbent (1954) presented participants with the task of recalling the paired-digits in their order of arrival (e.g. 37-42-12). The chosen dependent variable was time (in seconds) between the onset of each paired-digit with the range span of the ‘DV’ starting at 0.5 second rising to a maximum of 2 seconds. Here the participants’ performance suggested that task performance improved with the longer time interval. However, the performance was no match for results gained from presentations made to a single ear (rehearsal being the probable cause). Broadbent (1957) laid out his stall by arguing in favour of a limited capacity system requiring a process much like a filter system. Such a filter system for attention was of course not incompatible with Cherry’s (1953) own basic findings. Broadbent (1957) suspected that one stimulus might be selectively filtered, while the other residual stimulus might be, effectively, ignored. To Broadbent’s (1957) thought, both stimuli gained access in parallel (simultaneously) to a limited capacity sensory register (e.g. auditory, visual) where the stimuli then underwent a selective form of determination (preattentive analysis). Quite reasonably, Broadbent’s notional concept of twin parallel channels converging into a single limited information channel led him to devise a Y-shape mechanism. In a mechanical representation of his concept, Broadbent constructed some tubing into a Y-shape and sourced some balls that could travel freely down within the tubes. Within the upper branches of the Y-shape (the sensory store) the capacity was such that a number of balls could be housed. In the narrow stem of the Y-shape (the limited perceptual channel) only a single ball (serial mode) could pass freely one-at-a-time. At the connecting joint of the branches and the stem, Broadbent placed a hinged flap (the selective filter). Here then, two balls serve to represent two digits, and the branches represent the ears; thus, as the flap (selective filter) moves from one position to the other, so too a ball (digit) is able to flow freely across the flap (selective filter) into the narrow stem below (the limited perceptual channel). One can say therefore that the model serves to demonstrate the selective filtering mechanism as acting to protect a limited capacity information store (i.e. short-term memory) positioned beyond the filter. Moreover, the brief memory buffer afforded by the sensory store Broadbent considered sufficient to provide a determination (preattentive analysis) of physical characteristics (e.g. intensity or pitch) which determined whether information might be selected for additional attentive-processing. None the less, in accordance with the digital nature of the selective filter, various data unable to progress for attentive-processing are reasonably inferred as being in a state of arrest at the preattentive-processing stage.
Michael Posner’s (1987: p. v, p. x) thirtieth anniversary foreword introduction to Broadbent’s classic (1958) Perception and Communication overflows with the “genuine pleasure” and “exhilaration” that the nascent science of cognitive psychology might possibly have shared upon the reception of this “landmark” work. It is here that Broadbent (1958: p. 299ff) provided his now-famous arrow-and-box schematic in order to point up his putative component processing stage model. Moreover, Broadbent’s wider cognitive vision held these component stages as intervening between a stimulus and its corresponding response (i.e. stimulus → processing → response).
That having been noted, Neville Moray (1959) showed that approximately one third of the time a participant detected their name in the non-shadowed messages of the unattended channel. Moray’s (1959) incongruous findings gravely undermined Broadbent’s (1957, 1958) filter model, whilst also opening the door to the suggestion that a degree of semantic filtering appeared necessary. Also in line with Moray’s incongruent findings, Anne Triesman (1960) presented findings indicating that ’breakthrough’ (where a word is spoken from the unattended channel as if on the attended channel) occurred in 6% of her dichotic listening trials. Her findings urgently required an interpretation of a pressing question: How could ‘breakthrough’ be possible if Broadbent’s cognitive filter model was an accurate account? Gray and Wedderburn (1960) added to the uncertainty with findings which suggested that ‘meaning’ selection in dichotic information trials (e.g. who 6 there – 4 goes 1) could override the previously conceived inflexibility of the filter model. Taken together these findings, presumably, added to the mounting number of problematic questions in Broadbent’s in-tray.
J. Anthony Deutsch and Diana Deutsch (1963) espoused a theory stating that the meaning of all incoming items of information could be determined in coincidence (parallel mode) including those presented to the unattended ear in the dichotic listening task. The provocative suggestion put forward by Deutsche and Deutsche hinged upon the interesting concept that the processing functionality required of the selective filter in Broadbent’s (1958) model was itself so complex a processor that no such selective filter appeared necessary if indeed its main function was to protect a limited capacity store (i.e. STM) beyond the filter. Put simply, why have two complex processors (i.e. one early and one late) when one late selection processor would suffice?
Enter Anne Treisman (1964) with a workable solution to her problematic question (Triesman, 1960; see also Gray and Wedderburn, 1960) concerning ‘breakthrough’ from unattended channels in dichotic listening tasks. Treisman (1964) had closely examined her own and other evidence (e.g. Cherry, 1953; Deutsch and Deutsch, 1963; Gray and Wedderburn, 1960; Treisman, 1960) and concluded that, in contradistinction to Broadbent (1957, 1958), there was reasonable support for the theoretical proposal that unattended messages might well be analysed at both the physical and semantic levels of analysis (see also Deutsch and Deutsch 1963). That is, to Treisman’s (1964) thought, in contrast to the somewhat binary modalities available to Broadbent’s (1958) selective filter model, unattended messages were not merely processed fully nor ignored completely. Instead unattended messages were dampened by a process known in electronics as attenuation. One might say that the foundation of Treisman’s (1964) attenuated filter theory turns on this quite reasonable difference from Broadbent’s (1958) own less flexible filtering account. However, should the unattended message contain information of sufficient intensity or meaning (e.g. physical difference, name, or danger signal) beyond a certain threshold then Treisman’s (1964) attenuation model accounted for the forthcoming ‘breakthroughs’ into consciously attended processing. Moreover, a reasonable feature of the threshold activation proposal is that liminality might be lowered momentarily for certain audible stimuli (e.g. name, certain insults, danger signals, etc.) by the listener’s own level of expectation.
Notwithstanding, interest for researchers in the field of auditory attention (e.g. Deutsch and Deutsch, 1967; Treisman and Geffen, 1967; Treisman and Riley, 1969; see also Norman, 1968) continued for the remainder of the nineteen-sixties to veer in emphasis toward the precise nature of the early-or-late analysis problem (i.e. where to locate the ‘bottleneck’ in attentional processing). A notable exception among these researchers looking into the so-named bottleneck problem was Donald Norman. For Norman (1968) there simply is no bottleneck problem worthy of concern; the only problem, to his thought, lay in the imposition of a structural constraint within the memory and attention modelling of his time. In his scheme Norman (1968) views each stimulus as undergoing sensory encoding (i.e. physical and semantic) on the basis of its contextual degree of pertinence. Here then, for Norman (1968) the basis of a stimulus’ physical and/or semantic properties above threshold, and thus pertinence, might be suggested to correspond to the given stimulus gaining greater degrees of access to stored long-term knowledge.
Nilli Lavie (1995, 2000; see also Lavie, Hirst, de Flockert, & Viding, 2004) argues that there are occasions when early selection (following Broadbent 1958) and when late selection (following Deutsche and Deutsche 1963, 1967) may be required to take place. Lavie (1995) does not dispute the importance of selective mechanisms of filtering in cognitive processing however she does propose a limited intermodal attentional capacity where full-capacity effectively spans across the intermodalities of the attentional system. That is, Lavie (1995: p. 452) holds that total attentional capacity, though limited, might be allocated such that excess capacity extends across attended and unattended intermodal registers: “spare capacity beyond that taken for high-priority relevant stimuli is automatically allocated to the irrelevant stimuli.” In her study Lavie (1995) focused her concern on the interaction between neutral and incompatible distractor types under high and low loads. Lavie’s (1995) findings appear to suggest that under high perceptual load the distractor type (i.e. a letter on a screen in one of six positions) was not a significant factor, whereas under low perceptual load the incompatible distractor (e.g. a large /X/ to a small /z/) the mean target identification rose significantly from approximately 460msecs to 510msecs indicating that the ability to dampen down irrelevant information decreased as perceptual load decreased. In other words, Lavie (1995) seems to invite the reasonable interpretation that as fewer demands are made on our attention, so too the likelihood of lapses of attention increase.
In the above we saw that the reportage aimed to highlight a small part of the degree to which Anne Treisman’s (1960, 1964) attenuation theory of selective auditory attention clarifies evidence admitting from the dichotic listening paradigm. After some brief initial historical observations, the second paragraph centres on the early shadowing method employed by Colin Cherry (1953). Cherry’s preferred method of shadowing is introduced and his interest in the cocktail party effect pointed to as formative in the future methods employed by other researchers. The third and fourth paragraphs report on some of Donald Broadbent’s (1954, 1957, 1958) classic cognitive psychological studies: the 1954 split-span experiments; the 1957 Y-shape mechanism constructed as a consequence of his earlier conceptual notion of and for a limited perceptual channel; and, the 1958 landmark figure of the schematized component stage model. Note was made of Broadbent’s wider vision for these component stages as intervening between a stimulus and its correspondent response. Neville Moray’s (1959) findings suggesting problems with Broadbent’s (1958) filter model and Anne Treisman’s (1960) study of ‘breakthrough’ are reported as well as other problematic evidence in the fifth paragraph. Diana and J. Anthony Deutsch’s (1963, 1967) proposal for semantic level parallel processing of all incoming attentional data is outlined in the sixth paragraph. In the seventh paragraph a short account of Treisman’s (1964) attenuated filter theory extending Broadbent’s (1957, 1958) inflexible filter theory is offered with brief descriptions of how her findings might account for breakthrough (e.g. name, danger signals, etc.). The eighth paragraph introduces Donald Norman’s (1968) proposal for a non-contentious processing system for memory and attention. (Here non-contentious is used purely in the sense that Broadbent’s (1958) filter theory, Treisman’s (1964) attenuated selective filter model, and Lavie’s (1995) perceptual load theory are contentious systems). By way of conclusion the ninth paragraph reports on Nilli Lavie’s (1995) perceptual load study and introduces her concept that the full capacity of attentional resources are constantly dedicated to attended stimuli. Mention is made of the suggestion that the excess capacity is allocated to unattended stimuli. Whilst Treisman’s (1960, 1964) contributions represent an extension to Broadbent’s (1958) original filter model, a modest submission might be ventured that threshold activation might also have some bearing on the intermodal allocation of attentional resources, perhaps even at the neurophysiological level (see Lavie, 1995, 2000; Posner, 1980; Treisman, 1988, 1993; Wolfe, 1994, 1998).
- Auditory Cortex Tracks Both Auditory and Visual Stimulus Dynamics Using Low-Frequency Neuronal Phase Modulation (plosbiology.org)
- Short review: Zawidzki, T., Bechtel, W. (2005), ,,Gall’s Legacy Revisited: Decomposition and Localization in Cognitive Neuroscience”, [in:] Erneling, Ch., Johnson, D. (eds.) The Mind as a Scientific Object, Oxford: Oxford University Press, pp. 293-316. (biolawgy.wordpress.com)
- Making the Invisible Visible: Verbal Cues Enhance Visual Detection (socialpsychology.org)
- Making the invisible visible: Verbal — not visual — cues enhance visual detection, says Penn researcher (esciencenews.com)