Selective attention is when we only respond to some aspects of the environment and ignore others - we select some parts of information and filter out the rest. We can selectively attend to auditory information or visual information.
Selective Auditory Attention
Dichotic listening paradigm
Just for the record, a 'paradigm' is a fancy word given to a research technique that can be applied to a variety of different experiments
In dichotic listening experiments participants listen to a headset with 2 different channels; one in each ear. They are asked to shadow (ie repeat back) the message reaching one ear and report it while ignoring the other. Dichotic listening experiments were used to demonstrate that people can only do one thing at a time, even when they appear to be doing two.
Most people can do a dichotic listening task without significant difficulty. Cherry (1953) found that we use the physical characteristics of the stimuli, e.g. the intensity, pitch, loudness and source of message to separate the wanted and unwanted messages. For example, we typically have no trouble when told "only listen to the female voice/the low pitched voice/quieter voice". However when the different messages played in each ear were spoken by the same voice (they had the same physical characteristics), participants found it almost impossible to distinguish between the two. For example, people have difficulty when told to "listen to the guy side talking about cheese/elephants/cheese and elephants" Cherry et al concluded that we can focus our attention using phonological cues, but not semantic cues. This could reflect how the Phonological Loop represents new information acoustically rather than semantically, hence acoustic cues such as tone and pitch are attended to as opposed to information concerning the nature of the message/conversation.
It is thought that people with a greater working memory span are less likely to hear their name in a dichotic listening experiment. This is probably because they are better at focusing their attention on the voice they are instructed to listen to (Moray, 1959).
Cocktail Party Phenomenon (Cherry 1953)
The cocktail party phenomenon (CPP) is an example of selective auditory attention and is based on Cherry's (1953) experiments on dichotic listening. In this phenomenon, people can focus their attention on a single conversation even whilst many others are going on around them. For example, when conversing in a noisy crowded party, most people can still listen and understand the person they are talking with, while ignoring background noise and other conversations.
The second aspect of the CPP is the finding that salient information breaks through the attentional filter meaning that we may hear someone mention our name even though we ignored the rest of the conversation (Moray, 1959). This phenomenon is unexpected, as according to the Selective Attention theory we should only register information based on physical characteristics, whereas our name is a completely arbitrary label (with semantic value). In this instance we can also see attention switching, where the individual is able to keep focus on an immediate conversation and then "switch" to engage with hearing their own name, which seems to be a somewhat conditioned response. This could be explained in terms of Automaticity; we learn to react unconsciously to a stimulus (normally our name).
Thus, the cocktail party phenomenon, according to Broadbent's theory, suggests that all messages are processed first and then only the ones that are important are filtered through, therefore processing is delayed slightly after all the messages have been received. However, his model stimulated tests that later falsified it (see Underwood below).
Broadbent (1954) Split Span Experiment
Broadbent used a split-span experiment to test people’s ability to ‘shadow’. In this experiment three digits were presented to the left ear (7,2,6) and three digits were presented to the right ear (1,9,5) dichotically using headphones. Then pairs of digits were presented to both ears separated by a 500ms interval (7,1)(2,9)(6,5) and participants were asked to recall the digits in temporal order. It was found that participants could only repeat the digits in ear order e.g 7,2,6 and 1,9,5. This suggests the participants could only do one thing at a time and therefore only attend to one ear at a time. Broadbent claimed that two stimuli or messages that were presented at the same time gained access in parallel to a sensory buffer. One of the inputs is then allowed through a filter on the basis of its physical characteristics with the other remaining in the buffer for later processing. The filter prevents overloading of the limited capacity mechanisms beyond the filter which process the input thoroughly (e.g. in terms of its meaning).
Broadbent's Human information processing theory (1958)
This states that there is a limited capacity channel for access to consciousness so it is necessary to 'filter out' most input. The filter can only be tuned to physical characteristics of input. This filter is similar to Baddeley's concept of the Episodic Buffer. It is only through the filter that input can be transferred to the LTM. All processing systems must be able to do go through these stages with information: register --> encode --> store --> retrieve --> recode --> output. Information must be processed then recoded for the next stage.
Initial processing occurs on the basis of physical properties (e.g. pitch). It is stored in immediate memory and due to a low semantic capacity, a filter is required to get rid of irrelevant stimuli. People attend to a physically defined stream known as a channel, and physical properties are used to see whether information fits into the channel. Top-down and bottom-up influences are relevant.
However, this doesn't tell us how a person can hear their name being said when mentioned in another unattended conversation. If we are filtering out other conversations, how can we hear our name being mentioned? We must be aware of it to some extent.
Underwood (1974) showed that with practice and dichotic listening, people may become much better at reporting details of the unattended message; practice reduces the amount of processing that must be devoted to the shadowing of the task, therefore Underwood's findings suggest that filtering can occur at different stages depending on how much processing capacity is available.
Gray and Weddurburn (1960) showed that attention can be switched from channel to channel in a dichotic listening experiment. They presented a mixture of digits and phrases to each ear and found that participants found it easier to recall the objects category-by-category, e.g. phrases in a row and then digits, therefore suggesting that people can pay attention to two attention channels at once.
Treisman's attenuation model
On the basis of this, Treisman (1964) suggested the attenuation model which incorporated three processing levels: physical, linguistic and semantic. There is a flexible bottleneck in processing, depending on available processing capacity. Information is only filtered when there are insufficient resources to process it further and other information has priority. The filtered information is attenuated, that is it receives less processing, rather than being completely unprocessed. Some words of higher value have a lower threshold (e.g. name), as a result, these words may be processed despite a lack of attention (could potentially explain CPP).
Interference in Selective Auditory Attention
Changing State Hypothesis
Jones et al. (1999) suggested that type of noise distraction influenced our ability to selectively attend to particular aspects of our environment. Jones formulated the changing state hypothesis which states that the degree of change of background noise determines the extent of interference on an STM task. Research in support of Jones' theory has demonstrated how unchanging background noise such as a constant tone or rhyming words had no interference on STM task performance, whereas changing sounds (e.g. pulsed) or non rhyming words did cause interference.
Jones argues that selective attention involves streaming of both the target and distractor stimuli, as in they are both perceived and maintained in memory as two separate events. This can explain why an unchanging distractor stimulus causes less (or no) interference: the distractor is easier to maintain separate from the stream of target information. Streaming is said to happen unconsciously at the level of sensation and perception (i.e. unconscious distribution of attention), as well as consciously at the level of deliberate maintenance of correct serial order of items (i.e. conscious selective attention).
These theories and findings link with Divided Attention and Dual Task Performance (see below).
The Sternberg Task (1966)
Sternberg proposed the 'High speed memory scanning paradigm' - a method of how people search their short term memory. He was initially interested in whether we search our STM serially (one after the other) or in parallel (everything at once).
His experiment consisted of a number of trials where participants were displayed with a 'positive set' of numbers (the amount of numbers varied between 1-6). They were instructed to rehearse the positive set in their STM. PPS were then shown a 'test stimulus' (a number) and they were asked whether that number had appeared in the positive set. Their reaction time to these questions was measured.
Sternberg found that the participants' mean reaction time increased linearly with the amount of numbers in the sequence. This suggested that participants were exhaustively scanning their short term memories for the test stimulus (i.e. serially) rather than all at once (in parallel).
The results indicated that the average rate of this internal serial search process is between 25 and 30 symbols per second.
However, Neisser (1964) conducted an experiment that at first seemed to contradict Sternberg. Neisser’s experiment involved searching a column of letters for digits. It was found that participants were able to search for 9 digits just as quickly as when searching for just 1 digit. This suggests that participants were able to search for all targets at the same time in a parallel search.
The important difference between these tasks, however, is the nature of them. Sternberg's task of searching STM for numbers was a serial search because it is a novel task - one that we are not used to, so we must do it in a controlled manner. However, the task in Neisser's experiment was searching for digits among letters - we have had years of discriminating between digits and letters, which means we can carry out the task more automatically and thus, in parallel.
This is the highest level of attention and refers to the ability to attend to multiple tasks at the same time, for instance a mother keeping an eye on her children whilst shopping.
Dual Task Performance
The three main variables that influence dual task performance are:
1) Task Similarity
Task similarity makes dual task performance more difficult. Allport et al (1972) asked participants to shadow an auditory message given to one ear and then measured their recall of words from the message given to the other ear. Memory for the non-shadowed words was poor. However, if the non-shadowed words are presented visually, rather than in an auditory form, then recall is much better. Recall was even better if the non-shadowed words were presented as pictures. This shows that task similarity can impair performance compared to tasks that are slightly different so tap different mental resources.
Low level tasks interfere with each other if they require processing at the same modality such as remembering digits and letters. Low level tasks in different modalities can be combined much more easily. High level tasks that tap the Central Executive resources of planning or retrieval from long term memory are difficult to combine. There is a small detrimental effect of combining any task, even if they are low level tasks using different modalities, which is known as the "dual task effect". Believed to be caused by the extra load put on the central executive.
2) Task Difficulty
Individual task difficulty affects dual task performance. In general, a difficult task is one that is novel and unpracticed. Sullivan (1976) used the tasks of shadowing an auditory message and detecting target words on a non-shadowed message at the same time. When the shadowing was made harder fewer targets were detected on the non-shadowed message
Practice makes tasks easier. Spelke et al (1976) found that practice can produce dramatic improvement. They trained participants to read stories and type words from dictation simultaneously. Initially this was hard to combine but, with practice, both were as good when combined.
General Resources Models of Divided Attention
General resource models of divided attention propose that there is a general pool of cognitive resources that can be shared between tasks. Resources are viewed as types of mental energy or mental capacity of the central processor. The more resources that are required for one task, the fewer are available for another task. Thus, general resource models predict that dual task performance decreases as individual task difficulty increases.
Kahneman (1973) proposed a single capacity model. It assumed a single pool of resources that are shared amongst competing tasks. If the available capacity exceeds the performance there is no detriment to the tasks at hand. Performance only suffers when the pool of resources is not big enough for all of the tasks involved.
Johnston & Heinz (1978) proposed a model of selective attention that has a limited pool of resources. If one task requires many resources then stimuli related to another task will be filtered out earlier. However, if the primary task requires few resources selection will happen at a later stage and attention can be divided between tasks.
Psychological Refractory Period and Bottleneck Theory
Most dual task studies show an impressive ability to combine tasks providing they are not too similar or difficult. There is always a cost of doing 2 things at once. In any study where participants are asked to respond to 2 stimuli there is usually a ‘hangover’ when responding to the second stimuli.This is called the psychological refractory period.
Welford (1952) argued that this reflected a processing bottleneck that makes it impossible to select or initiate 2 responses simultaneously.
Pashler (1990) showed that the psychological refractory period persisted even when stimuli and response were very different and when excessive practice had happened.
Studies of the PRP suggest that our ability to do two things at once may reflect our ability to switch our attention or resources between tasks rather than a genuine ability to divide attention.
Yerkes-Dodson Law (1908)
The Yerkes-Dodson law considers the relationship between arousal and performance. Too much arousal (stress) leads to poor performance through narrowing of vision - 'tunelled vision'. Too little arousal also leads to poor performance. However, there is an optimum level of arousal for facilitating peak performance. If a task is particularly hard, the level of arousal needed will be lower. For example, if during a skydive someone's parachute fails to open first time, their level of arousal is likely to massively increase. However, in order to pull the safety cord and open the parachute properly a low level of arousal is necessary i.e. high arousal leads to panic, low arousal leads to calmly opening the safety parachute.
Overtraining and relaxation techniques can be used combat the problem of having a high level of arousal when a low level is needed to carry out the task successfully. This can help make the sub-skills more 'automatic' which gives you more 'conscious capacity' to do the task. For example, soldiers are overtrained in combat to reduce arousal caused by gunfire etc from the opposition. They must have a clear head in order to plan their counter attack.
However, Hanoch & Vitouch (2004) argue that the Yerkes-Dodson law is too simplistic to discuss the relationship between cognitive processes and emotional arousal. They believe arousal is a subjective term and with no single definition, it can be measured in different ways. It may not be a bad thing that people remember less when they have too little or too much arousal, less is often more especially as what is normally remembered is the most important factor.
Reason (1979) asked participants to record their action slips (ie unintended action or failures to act). 60% of people reported that action slips fell into 2 categories:
Storage failures: slips resulting from forgetting actions that has already been performed. Storage failures arise when carrying out habitual action schemas; they are unintended and leave little or no memory because they are automatic - meaning that they use very few cognitive resources and thus scarcely leave a memory trace. e.g. forgetting if you locked the front door when leaving the house or forgetting that you had just made a cup of tea and proceeding to fill the kettle as though to make one.
Test Failures: slips resulting from a failure to monitor the progress of an action, such as performing a task and then lapsing into habitual behaviour. This type of slip arises because automatic processes happen when an appropriate stimulus is present. e.g. picking up your tooth brush to put away and then beginning to brush your teeth or intending to make a friend a black coffee but accidentally putting milk into it because that's how you usually have it.
The Supervisory Attentional System
Norman & Shallice (1980 and 1986) proposed a three component model of action control made up of cognitive subsystems, schemas and contention scheduling; overseen by a supervisory attentional system.
When a stimulus triggers only one particular action schema or response that action is performed automatically. When a stimulus triggers several actions schemas the strongest inhibits the rest in a process called contention scheduling, overseen by the control process SAS. When a habitual action is triggered but it is inappropriate SAS increases the activation of a more appropriate competing action schema so that action ‘wins the competition’ to be performed. The SAS can override contention scheduling when we need to make novel responses to familiar stimuli or stop a habitual action routine part way through. BUT this doesn't explain how things move from controlled to automatic processing. This is explained by the ACT* (Adaptive Control of Thought).
Exam Question Plans
2010/2011 What is the Cocktail Party Problem? Explain how the distinction between automatic and controlled processing may be involved in its resolution?
Define the CPP. How do we hear what we want to hear when there are lots of concurrent noises/sounds? - Can only use physical cues (location, pitch, vol), not semantic cues to distinguish between messages
so how do you recognise your name in a different conversation, when this isn't your focus of attention? - it's become an automatic process due to years of conditioning to respond immediately to it
"Describe controlled and automatic processing - and the differences between them"
- controlled processing: slow and attention demanding, costly in terms of cognitive resources, enables flexible behaviour, used when responding to novel situations or overriding habitual responses, associated with decision making and choosing - result in explicit memory (conscious process).
- automatic processing: fast and efficient, makes minimal demands on cognitive resources, leaves plenty of processing capacity for performing other tasks, unconscious, triggered when an appropriate stimulus is presented, hard to change/stop - inflexible, acquired through consistent mapping.
- example of difference = Stroop Effect (reading coloured words; saying colour of ink = controlled, saying the word = automatic)
- mention difference between Sternberg (1966) and Neisser's (1967) tasks - idea of parallel and serial searching reflects controlled and automatic processing
- bring in how practice has an influence - in a novel situation we may have to use controlled processing but with practice, the behaviour may become automatic.
- Shneider & Shiffrin (1977) - proposed that an automatic process is the activation of an established sequence of nodes within our long term memory. A controlled process is the temporary activation, by attention, of a novel sequence of nodes, requiring STM.
- Instance theory (1988) - alternative theory to the above. Emphasizes the importance of knowledge in performing a successful task - novices are slow because they have little knowledge about the task and therefore have to use effortful, controlled processes to reach solution.
- in the CPP: using controlled processing to focus attention on current conversation, as this is a novel task. Attending to your name is an automatic process, so causes you to switch the focus of your attention. (how the two interact)