Concentration on certain stimuli in the environment and not on others, enabling important stimuli to be distinguished from peripheral or incidental ones (APA n.d.)
Theories and models of attention differ in a number of ways, as they attempt to explain how and when an individual processes the physical and semantic aspects of their environment. Early selection models of attention suggest that stimuli pass through a filter before they can be attended to and processed for meaning. Broadbent (1958) investigated this idea of early selection by using a split-span, or dichotic listening paradigm. In the dichotic-listening paradigm, participants are presented with two different auditory stimuli (one in each ear) at the same time and are instructed to attend to one or both of the stimuli. Afterward, they are asked about the content of either the message they were asked to attend to or the message that they were told to ignore. Broadbent found that participants had increased accuracy when asked to recall stimuli presented in the same ear than when asked to recall stimuli presented in each ear, suggesting that the information was filtered based on physical characteristics of the stimuli (such as pitch, direction, or the ear in which the information was presented). According to this model, stimuli that are presented are held in a preattentive state and must be either selected for relevance and processed for meaning, or filtered out as irrelevant. All semantic processing of the stimuli takes place after being filtered and selected for attending. In essence, the information selected has no meaning during the selection process.
However, this model must be an incomplete explanation. Cherry’s (1953) work found that participants receiving stimuli in both ears, attending to one message in one ear, were able to detect their name from the unattended ear. This is defined as the “cocktail party effect”: the ability to detect significant stimuli in a multi-talker situation. This unattended-to information gained attention and was processed for meaning, which contradicts the idea of selection based on basic characteristics, rather than by meaning.
Late selection models of attending indicate that information is selected after processing for meaning. Gray and Wedderburn (1960) hypothesized that if early selection models were correct, then the presence of other, conflicting information presented at the same time in the same ear would be able to disrupt recall of information. In their study, participants were given a mixture of words and numbers presented in each ear. For example, in the left ear, participants heard mice – 5 – cheese and in the right ear, 3 – eat – 4. If the early selection model was correct, then participants should have reported all stimuli presented, grouped by ear. However, they found that participants reported information grouped by type: mice, eat, cheese and 5, 3, 4. This suggests that stimuli were selected according to meaning, not basic characteristics.
Memory selection models of attending merge the early and late selection models. They follow the principle of early selection that stimuli are selected based on physical characteristics, but also that of late selection: that attended and unattended information both pass through the filter to be processed for selection based on meaning. Meaning, not the filter, decides the focus of attention. Deutsch and Deutsch (1963) suggested that meaning is assigned based on importance, and that important stimuli will be processed, placed into memory, and then on to awareness thus, only important stimuli will be remembered. In this model, all information is processed, but only information deemed important will be available for later recall.
The attenuation model of attention disagrees with the idea that stimuli must pass through a filter and that the filter decides what stimuli pass into awareness. Treisman (1969) proposed that while an early selection filter exists, the filter attenuates, or weakens, stimuli that are unattended. If stimuli are to be attended to, they must surpass a threshold to leak through the filter. This threshold itself acts as a filter, in that important words (such as an individual’s name) pass through relatively easily, whereas unimportant words (such as chair) have greater difficulty in passing through the filter.
The multimode model of attention adheres to the notion of a flexible filter, whereby what kind of selective attending is used is dependent on what is needed at the time. In this model, both physical features and meaning can be used for selection, based on their costs and benefits. Johnston and Heinz (1978) found that as the perceptual-processing system shifts from early to late, it collects more information from unattended sources but requires more capacity to focus on an attended source. In their study, participants performed a listening task and a reaction-time (RT) task at the same time. The listening task required participants to either attend to a single list or to a target list of two or three lists. The RT task required the detection of light signals and was used to measure the amount of capacity devoted to the listening task. Longer RTs implied more capacity expended on listening. In the first condition, the target and nontarget lists differed in voice (a male or a female), but were similar in meaning. In this condition, faster response times were recorded than those in the second group, implying that early selection was able to be used due to processing based on physical characteristics (i.e., voice, direction). In the second condition, the two lists were spoken in the same voice but differed in meaning. In this condition, response time was longer, implying that early selection could not be used and late selection was necessary due to the difference in meaning.
Each of these models differs slightly in when exactly meaning is derived and attending occurs. To review, in early selection models, information must pass through a filter before either meaning or attending. In late selection models, information is processed for meaning and then attending may follow. Memory selection models hold that all information, regardless of attention, will pass through the filter to be processed for selection based on meaning. In the attenuation model, information is filtered, but is attenuated instead of completely filtered in or out. If information is perceived despite being attenuated, then it may be attended to and continue to have meaning extracted. In the multimode model, selective attention varies according to the individual’s needs at the time, and can use the processes outlined in early, late, memory, and attenuation models of selective attention. Each of these models, however, retains the idea that some mechanism restricts the flow of all information being attending to at the same time. In addition, regardless of when meaning is eventually attached, all information requires meaning to differentiate it and activate perceptual processing.