Cocktail Party Effect

selective attention

The cocktail party effect is the phenomenon of being able to focus one’s auditory attention on a particular stimulus while filtering out a range of other stimuli, much the same way that a party-goer can focus on a single conversation in a noisy room. This effect is what allows most people to ‘tune in’ to a single voice and ‘tune out’ all others. It may also describe a similar phenomenon that occurs when one may immediately detect words of importance originating from unattended stimuli, for instance hearing one’s name in another conversation.

The cocktail party effect works best as a binaural effect, which requires hearing with both ears. People with only one functional ear seems much more disturbed by interfering noise. However, even without binaural location information, individuals can selectively attend to one particular speaker if the pitch of their voice or the topic of their speech is sufficiently distinctive (albeit with greater difficulty). This phenomenon is still very much a subject of research, in humans as well as in computer implementations (where it is typically referred to as source separation or blind source separation).

The binaural aspect of the cocktail party effect is related to the localization of sound sources. Experiments have shown that the auditory system is able to localize at least two sound sources simultaneously and assign the correct sound source characteristics to these sound sources simultaneously too. In other words, as soon as the auditory system has localized a sound source, it can extract the signals of this sound source out of a mixture of interfering sound sources. It is assumed that the auditory system performs a kind of cross-correlation function between both ear signals.

Attempts have been made to simulate the cocktail party effect by technical means. ‘Cocktail party processors’ have been constructed which can extract the signal of a single sound source out of a mixture of sound sources. In the early 1950s much of the early work in this area can be traced to problems faced by air traffic controllers. At that time, controllers received messages from pilots over loudspeakers in the control tower. Hearing the intermixed voices of many pilots over a single loudspeaker made the controller’s task very difficult. The effect was first defined and named ‘the cocktail party problem’ by Colin Cherry in 1953. Cherry conducted attention experiments in which subjects were asked to listen to two different messages from a single loudspeaker at the same time and try to separate them. His work reveals that our ability to separate sounds from background noise is affected by many variables, such as the gender of the speaker, the direction from which the sound is coming, the pitch, and the rate of speech.

Some of the earliest work in exploring mechanisms of selective attention was performed by Donald Broadbent, who proposed a theory that came to be known as the filter model. This model was established using the dichotic listening task. In this type of experiment, a participant wears a pair of headphones and listens to two different auditory streams, one in each ear. The participant then pays attention to one stream while ignoring the other. After listening, the participant is asked to recall information from both the attended and unattended channels. Broadbent’s research using the dichotic listening task showed that most participants were accurate in recalling information that they actively attended to, but were far less accurate in recalling information that they had not attended to. This led Broadbent to the conclusion that there must be a ‘filter’ mechanism in the brain that could block out information that was not selectively attended to.

The filter model was hypothesized to work in the following way: as information enters the brain through sensory organs (in this case, the ears) it is stored in sensory memory. Before information is processed further, the filter mechanism allows only attended information to pass through. The selected attention is then passed into working memory, where it can be operated on and eventually transferred into long-term memory. In this model, auditory information can be selectively attended to on the basis of its physical characteristics, such as location and volume. Others suggest that information can be attended to on the basis of Gestalt features, including continuity and closure.

However, Broadbent’s model failed to account for the observation that words of semantic importance, for example one’s own name, can be instantly attended to despite having been in an unattended channel. Shortly after Broadbent’s experiments, Oxford undergraduates Gray and Wedderburn repeated his dichotic listening tasks, altered with monosyllabic words that could form meaningful phrases, except that the words were divided across ears. For example the words, ‘Dear, one, Jane,’ were sometimes presented in sequence to the right ear, while the words, ‘three, Aunt, six,’ were presented in a simultaneous, competing sequence to the left ear. Participants were more likely to remember, ‘Dear Aunt Jane,’ than to remember the numbers; they were also more likely to remember the words in the phrase order than to remember the numbers in the order they were presented.

In a later addition to this existing theory of selective attention, Anne Treisman developed the attenuation model. In this model, information, when processed through a filter mechanism, is not completely blocked out as Broadbent might suggest. Instead, the information is weakened (attenuated), allowing it to pass through all stages of processing at an unconscious level. Treisman also suggested a threshold mechanism whereby some words, on the basis of semantic importance, may grab one’s attention from the unattended stream. One’s own name, according to Treisman, has a low threshold value (i.e. it has a high level of meaning) and thus is recognized more easily. The same principle applies to words like ‘fire,’ directing our attention to situations that may immediately require it. The only way this can happen, Treisman argued, is if information was being processed continuously in the unattended stream.

In order to explain in more detail how words can be attended to on the basis of semantic importance, Deutsch & Deutsch and Norman later proposed a model of attention which includes a second selection mechanism based on meaning. In what came to be known as the Deutsch-Norman model, information in the unattended stream is not processed all the way into working memory, as Treisman’s model would imply. Instead, information on the unattended stream is passed through a secondary filter after pattern recognition. If the unattended information is recognized and deemed unimportant by the secondary filter, it is prevented from entering working memory. In this way, only immediately important information from the unattended channel can come to awareness.

Daniel Kahneman also proposed a model of attention, but it differs from previous models in that he describes attention not in terms of selection, but in terms of capacity. For Kahneman, attention is a resource to be distributed among various stimuli, a proposition which has received some support. This model describes not when attention is focused, but how it is focused. According to Kahneman, attention is generally determined by arousal; a general state of physiological activity. The Yerkes-Dodson law predicts that arousal will be optimal at moderate levels – performance will be poor when one is over- or under-aroused. Thus, arousal determines our available capacity for attention. Then, an allocation policy acts to distribute our available attention among a variety of possible activities. Those deemed most important by the allocation policy will have the most attention given to them.

The allocation policy is affected by enduring dispositions (automatic influences on attention) and momentary intentions (a conscious decision to attend to something). Momentary intentions requiring a focused direction of attention rely on substantially more attention resources than enduring dispositions. Additionally, there is an ongoing evaluation of the particular demands of certain activities on attention capacity. That is to say, activities that are particularly taxing on attention resources will lower attention capacity and will influence the allocation policy – in this case, if an activity is too draining on capacity, the allocation policy will likely cease directing resources to it and instead focus on less taxing tasks. Kahneman’s model explains the cocktail party phenomenon in that momentary intentions might allow one to expressly focus on a particular auditory stimulus, but that enduring dispositions (which can include new events, and perhaps words of particular semantic importance) can capture our attention. It is important to note that Kahneman’s model doesn’t necessarily contradict selection models, and thus can be used to supplement them.

Some research has demonstrated that the cocktail party effect may not be simply an auditory phenomenon, and that relevant effects can be obtained when testing visual information as well. For example, Shapiro et al. were able to demonstrate an ‘own name effect’ with visual tasks, where subjects were able to easily recognize their own names when presented as unattended stimuli. They adopted a position in line with late selection models of attention such as the Treisman or Deutsch-Normal models, suggesting that early selection would not account for such a phenomenon. The mechanisms by which this effect might occur were left unexplained. It has been suggested in brain imaging studies using PET that a variety of brain areas may be involved in selectively processing visual linguistic material (i.e. word form), including the inferior prefrontal and posterior insular cortices, the amygdala, caudate nucleus, and several areas of temporal cortex. It is currently unknown if these same brain areas are implicated in focusing attention for other visual or auditory stimuli.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.