Attention - Bottleneck Theories of Attention PDF

Preview

Summary

Theories of attention...

Description

Bottleneck Theories of Attention WHAT IS ATTENTION? 







William James (1980): Attention is taking possession of the mind, in clear and vivid form, of one out of hat may seem several simultaneously possible objects or trains of thoughts. It implies withdrawal from some things in order to deal effectively with others’ We need attention for selection: Our cognitive system is limited in capacity and resources o Selection for perception o Information selection for response/action Different types of attention o Selective attention (selecting the relevant information and filtering out irrelevant information) o Space-based and object-based attention o Exogenous vs. endogenous attentional systems Attention is important for other cognitive functions o Learning: From effortful (attention demanding) to automatic processing or behaviour o Problem solving: Attended information can be manipulated in working memory o Memory: We better recall attended information than information that as not attended o Language: Comprehension and production o Perception: Different processing for attended and unattended visual stimuli o Awareness: The relationship between attention and consciousness

EARLY SELECTION THEORIES OF ATTENTION First evidence of Bottleneck theories 



Broadbent (1958): o Air-traffic controllers during WW2 (competing messages from departing and arriving aircrafts) o People were severely limited in their ability to act on multiple signals arriving on different channels o Pilots had to try to monitor several sources of concurrent information, which might include numerous visual displays inside the cockpit, the visual environment outside the plane and auditory messages coming in over the radio o Researchers were motivated to discover more about the limitations of human performance Bottleneck theories of attention o Limited capacity requires a bottleneck in the flow of information processing o Different theories on how the information that passes through the bottleneck is selected o Early studies on attention focus on dichotic listening and shadowing effect

Shadowing effect   

Repeat relevant message (in one channel/ear) as it arrives. Virtually no memory of the unattended messages Physical attributes of unattended channel were detected e.g. male vs. female voice, human vs. musical instruments, bleep vs. tone Semantic attributes of unattended channel were missed e.g. do not notice foreign language or repeated items

Cognition – Attention: Lecture 1



General problems o Inability of naïve participants to shadow successfully due to their unfamiliarity with the shadowing task rather than an inability of the attentional system (Eysenk & Keane, 1990)

Dichotic listening 







In a selective attention task, the instruction is to attend to the message presented to one ear and to ignore the other message which is simultaneously presented to the other ear – this simulates the cocktail party situation (Cherry, 1953) Effective cues in dichotic listening for message selection o Physical separation e.g. in which ear the message was presented (most effective cue) o Physical acoustic of the voices e.g. pitch, male/female Most of the first studies were of selective attention. Results from studies by Broadbent (1952; 1954), Cherry (1953) and Poulton (1953; 1956) showed that both the physical acoustic differences between voices and the physical separation of locations were helpful for message selection. The most effective cue was physical separation o Confirms that a listener can selectively attend to stimuli that possess some common physical feature and can reject stimuli that do not possess that feature o Broadbent (1958) interpreted the data as demonstrating that stimuli that do not need responses are discarded before they have been fully processed. Because physical features of the input are effective cues for separating messages, there is a filter which operates at the level of physical features, allowing the information characterised by that feature through the filter for further processing. In unattended messages, only physical properties of the input seemed to be detected and it is these properties that can guide the setting of the filter General problems o We can never be sure that the participants have not actually switched attention to the unattended channel o Memory for unattended channel may depend on familiarity, importance or similarity to the attended channel o There are effects of practise on unattended information o There is implicit memory for the unattended channel even when there is not explicit memory  Experiments about GSR (Corteen & Wood, 1972; Corteen & Dunn, 1973)

Filter theory (Broadbent)  







Concerned about transmission of information within nervous system Information transmission is maximal when a given stimulus always gives rise to the same response – there is no uncertainty between the stimulus input and the response output If a different response were to occur on some occasions, the amount of information transmitted would be reduced – if the amount of information transmitted is calculated and divided by the time taken to make the response, the rate of information transmission can be found Led to conclusion that as a communications system, the whole nervous system could be regarded as a single channel which was limited in the rate at which information could be transmitted The limited capacity section of the nervous system would need to be preceded by a selective filter, or switch, which protected the system from overload and passed

Cognition – Attention: Lecture 1



 





on only some small, selected portion of the incoming information. All other information is blocked These conclusions are combined with the idea of a short-term buffer store that precedes the selected filter. This buffer is a temporary memory store in which the unselected information can be held in parallel for short periods of time Selection from the parallel input is made at early levels of processing and is an early selection model In basic terms... o Short-term store holds all incoming physical information from different channels e.g. ears, eyes o Filter selects relevant information. Selection on the basis of physical features (channel, voice) e.g. the effective cues o All unselected information is lost and not processed further

P system uses a different form of memory system to that of the original sensory buffer. Sensory information is initially registered in the S system in some raw form that codes the physical nature of the stimulus e.g. large brown shape, while outputs from the P system code the categorical nature of the stimulus e.g. bear – semantic information is now made explicit o The representation of the stimulus entering the P system from the S system simply codes physical characteristics of the stimulus e.g. size, shape, position, colour but does not code its identity – at this stage, the commitment to ‘bear’ has not been made o On leaving the P system, the stimulus has been identified, hence ‘bear’ is now coded o In summary… The P system affects perceptual-semantic categorisation – items (sensory representations) in the P system are assigned to their corresponding categories e.g. their identity CHALLENGES o Treisman (1960): Relationship between attended and unattended information  Attended channel: Passage from a novel  Unattended channel  A different passage from the same novel  Technical passage about biochemistry book, a message in a foreign language or nonsense syllables  Attended channel interfered more with recall than unattended channel o Moray (1959): Listeners reported own name when presented in unattended ear o Gray & Wedderburn (1960): Participants heard Dear-7-Jane in the left ear and 9-Aunt-6 in the right ear. They reported hearing Dear Aunt Jane and 9-

Cognition – Attention: Lecture 1

o

7-6. Stimuli are not selected based on physical characteristics determined by the filter but according to meaning instead Corteen & Wood (1972); Corteen & Dunn (1973)  Experiments using Galvanic Skin Response: When participants expect to get a small electric shock, their skin resistance changes  The authors conditioned participants to expect a shock in association with specific words  Participants were given dichotic listening tasks and asked to attend to one ear while ignoring information on the other ear. Every so often, one of the shock-associated words were presented on the unattended channel and the participants showed a clear GSR, although they claimed that they did not detect any of those words  Subjects also showed a GSR to other words semantically related to the shock-associated words that had not been included in the training set – suggests that not only was there semantic access of the unattended words, but also semantic generalisation  Experiments can be interpreted as providing more evidence that the selective processes in attention come after the meaning of words has been accessed

Attenuation Model (Treisman, 1960) 





Following occurrence of word A, the thresholds of words B and C are lowered because they are very probable following word A. If word C is activated by the attenuated signal from the rejected ear, it may be heard The filter is not such an all-or-nothing affair as Broadbent said – rather than blocking out all information that does not meet the selection criterion for attention, the filter attenuated or reduced the strength of the unattended channels (1964) o If incoming information is not totally blocked off, then partial information that is consistent with current expectation might be sufficient to raise the activation of those words above the threshold of consciousness General problems o Does not explain how semantic analysis words o The nature of the attenuation process has never been precisely specified

LATE SELECTION THEORIES OF ATTENTION Deutsch & Deutsch (1963)  

All stimuli are fully analysed, with the most important or relevant stimulus determining the response Places the Bottleneck in processing much nearer the response end of the processing system than Treisman’s attenuation theory

Pertinence model & context effects (Norman, 1986; 1976)

Cognition – Attention: Lecture 1



 





Selection comes after semantic memory has been accessed and both attended and unattended sources of information have automatic and effortless access to semantics Once semantics have been activated, pertinence values are assigned to current ongoing cognitive activities Pertinence is determined by the context of the inputs and activates nodes/locations in semantic memory. Selection is based on the summation of activation in semantic memory from both sensory input and context Example o Attended ear e.g. they were standing near the bank o Unattended ear e.g. one of the following – river, money o Participants interpreted bank as a riverbank if they heard river and a financial bank if they heard money General problems o Even if pertinence is controlled for: Participants are more likely to detect information in the attended channel and less likely in the unattended channel and participants find it very difficult to recall semantic content o Solso (1979): Model is uneconomical – all information is analysed for meaning, even the information that is not relevant to the task; this would require a large processing capacity

IS SELECTION EARLY OR LATE? 









The early vs. late discussion in early theories of attention was influenced by the assumption that the brain was a serial processor (e.g. the P system) of limited capacity Other proponents of late selection put forward their own versions of attentional theory – these theories are still structural and are concerned with the location of a bottleneck, where parallel processing stops and serial processing begins EARLY o Johnston & Heinz (1978): Allow Bottleneck to move in a flexible way – selection will take place as early as possible in processing, but exactly where selection happens depends on current task demands and circumstances. The more stages of processing that are required prior to selection, the greater the demands on processing capacity o Duncan (1980): Selection involved passage between two levels, where that passage is controlled by a selector. Only information which is passed through the selector gains awareness or the ability to control a response, but all stimuli are fully identified at level 1. The limit in such model is entry to level 2, where awareness can deal efficiently with only one stimulus at a time LATE o Bottleneck metaphor running thin – as soon as it can be moved around or hypothesised to be located at either end or almost anywhere in the processing continuum, it ceases to be a bottleneck at all o A different metaphor from the single channel processor was needed. Norman (1968): There is no fixed, structural point where the system is limited – the system is limited by having only a fixed amount of processing resources o The concept of resource limitation began to gain popularity – began experimenting on the limits of task combination and examine ng how these how these resources could be shared between tasks BOTH – Perceptual load theory (Lavie, 1994; 1995) o Perceptual load theory stipulates that both early and late selection occur, depending on the processing demands of the task

Cognition – Attention: Lecture 1

Perception has limited capacity but operates in an automated, involuntary manner  In tasks involving large amounts of information (high perceptual load), capacity is fully exhausted by the processing of the taskrelevant stimuli – this results in no perception of task-irrelevant information (early selection)  If information doesn’t exceed the brain’s processing capacity, it will be perceived. In tasks of low perceptual load, not all processing capacity is taken up by the task. This results in the spare processing capacity spilling-over, resulting in the perception of task-irrelevant information (late selection) and increased distractor interference This theory resolves the early vs. late selection attention debate by explaining that tasks of low perceptual load result in late selection, whereas tasks of high perceptual load result in early selection attention effects 

o

SUMMARY  





All information gets into a sensory register and is processed in parallel (all the sensory information is processed at once) Somewhere along the way, information is filtered or selected either… o Early  At perceptual level (Broadbent, 1958; Treisman, 1960) o Late  Near the response level (Deutsch & Deutsch, 1963)  After semantic has been assigned (Deutsch & Deutsch, 1963; Norman, 1968; 1976) Only selected information makes it into awareness and long-term memory Perceptual load theory: The attentional system is limited in capacity but operates in an automated and involuntary manner. Tasks of low perceptual load result in late selection, whereas tasks of high perceptual load result in early selection (Lavie, 1995; Lavie & Cox, 1997; Lavie & Tsal, 1994).

Cognition – Attention: Lecture 1...