California State University, Northridge

Selective Attention and Arousal

By Iris Beneli
(spring 1997)

Selective attention theories have suggested that individuals have a tendency to orient themselves toward, or process information from only one part of the environment with the exclusion of other parts. There is an abundant amount of evidence which supports that selective attention is governed by our arousal level. The most persistent question in the literature has been whether the shifts in attention that accompany changes in the arousal level are automatic, or deliberate. One segment of the research community has dealt with this issue via the capacity models. These theories propose that we all have a limited amount of mental capacity to allocate to various tasks, at any given time. The number of tasks referred to here can also be named chunks. Chunks are units which have already been stored in long term memory, items which we are familiar with, or are structurally similar to items we are familiar with. It is difficult to normalize the number of items or chunks any individual can store in short term memory, or retrieve from long term memory. There are many individual differences due to prior experience and perception of the material being handled.

The initial evidence for this phenomenon was derived from work by a Dutch psychologist named de Groot during the 1940's. His experiments required chess players of different abilities to reproduce chessboards mid-game. Each participant was given 5 seconds to view the board, after which all the pieces were taken off the board. The participant was then asked to reconstruct the game. The initial findings showed that the more proficient players could reproduce 90% of the pieces on their first attempt, while the weaker players only reproduced 40% of the pieces. To assess whether the findings were due to each player's ability or familiarity with the board, the players were then exposed to a board with various pieces placed randomly upon it. In this experiment the position of each piece did not hold a personal meaning to the player, since the participants had not been stopped playing a game of chess. This experiment showed no significant differences between the strong players and weaker players.

The models proposed by Broadbent (1957), Treisman (1960), Deutsch and Deutsch (1963), and Norman (1968), were the more prominent theories in the field, to elaborate on de Groot's memory model. All of these models attempted to explain the process by which we attend to certain information, but not all information available to us. One may identify one's environment by means of sound, smell, taste, visual and tactile stimuli as the information referred to above. The issue all of these theories had to resolve was the location of selection to the stimuli. More specifically, the models had to explain the process by which we are able to make sense of our environment, given that we are constantly bombarded with information.

The initial model was termed the bottleneck theory of attention, since information could only be attended to from one source at any given time. The shape of the model is similar to the letter 'Y', symbolizing two incoming sources of information with the two arms of the Y and one final source being recognized, via a filter. This filter symbolizes the location of selection to attention. Broadbent (1957) developed the filter model to explain the proposition that a bottleneck occurs before pattern recognition, and that attention determines what information reaches the pattern recognition stage. This model asserts that the selective filter allows information to come in from only one channel at a time, into working memory.

Studies utilizing the dichotic listening task, help to illustrate Broadbent's filter model. In these studies the participant is asked to put on a set of headphones, and requested to listen to only one ear, and report that information. The information presented to the participant is different between the two ears, and therefore fits the filter model perfectly. Participants tend to lack awareness of the unattended ear's content, or even language. Those who do know that the other ear's information varies from the attended ear's information can only report whether it was a human voice, and whether it was a male's voice or a female's voice.

An extension on Broadbent's Dichotic listening task is the Shadowing task, which requires participants to repeat the attended ear's information out loud. Shadowing a message provides proof that the participant is following instructions, and is attending to the correct ear. During a shadowing task, subjects are completely unaware of the unattended ear's message. Furthermore, when information is switched from one ear to the other in a contextual flow, the participant follows that switch. Treisman (1960), found that the contextual effects of language, which are the influences of the surrounding context on the recognition of patterns, would often cause participants to report information from the unattended ear.

Example: Your name is Bill, and while at a public place you hear a familiar voice call out your name, and at the same time an unfamiliar voice call out your name. Bill is a very common name, therefore you would most likely attend to the familiar voice.
Example: Your name is Ido, and while at a public place you hear a familiar voice call out your name, and an unfamiliar voice call out your name. Ido is a very uncommon name (in certain parts of the world), and therefore you are very likely to attempt to attend to both speakers. Preference for which voice you would attend to first would depend on personal choice and prior experience.

The questioning of Broadbent's selection filter's location arises, since the participant is able to follow the switch between ears in continuing a message. Treisman proposed a model which consists of two components, each relying on the other to function properly, named the attenuation model. One already established component of this model is the selective filter, and the newly proposed element is a ‘dictionary'. This dictionary symbolizes information, or words, which require a very low threshold in order to be recognized. The threshold may also be conceptualized as volume required to hear certain words or information. In describing this phenomenon the cocktail party effect is often used as an example:

You are at a party, and are speaking to a friend when across the crowded room you hear someone say your name. How is that possible?

In Treisman's attenuation model, the selective filter distinguishes between two messages on the basis of their physical characteristics, such as location, intensity and pitch. The ‘dictionary' in Treisman's model allows for selection between messages on the basis of content. Certain information requires a very low threshold in activating awareness of a stimulus, such as our name in the cocktail party example. The attenuation model therefore proposes that there is a decrease in the perceived loudness of an unattended message. This message will usually not be loud enough to reach its threshold unless it has a very low threshold to begin with (your name), or there is a general momentary decrease for all messages. An example of a general momentary decrease for all messages can be illustrated with the following example:

You are at the airport, and you are searching for a friend who just arrived from London. Your friend has a small frame and dark hair. You will be searching the airport gate for all individuals who have a small frame and dark hair, and quickly eliminate all those who do not.

Broadbent and Treisman's models proposed that the selection filter in attention occurs prior to selection, or pattern recognition stage. Later models by Deutsch and Deutsch (1963), and Norman (1968), attempted to merge growing information regarding memory and the selection process of attention. These more recent models claimed that selection occurs after the pattern recognition stage. In these models attention is equivalent to the selection stage.

Deutsch and Deutsch suggested that both channels of information are recognized but are quickly forgotten unless they hold personal pertinence to the individual. In shadowing experiments, the participant is asked to repeat a certain message, that would create the personal significance needed in attention. Norman elaborated on Deutsch and Deutsch's model by suggesting that selection is determined not only by the pertinence of the sensory input but also the strength of the input.

Strength of the input may be explained by stimulation of any of the sensory systems: visual, auditory, olfactory, and tactile. This stimulation is accompanied by an increase in arousal. From the Darwinian perspective, this arousal increase is necessary for survival. Perhaps an increase in olfactory stimuli could save lives during a fire. This increase in stimuli would then increase activity in the brain, not just the area involved in processing incoming olfactory stimuli since the entire brain is needed in dealing with the situation. There are two primary systems involved in the biological mechanisms of arousal; the first being the reticular activating system (RAS), and the second is the autonomic nervous system. The reticular activating system is often associated with general arousal, which moderates sensory thresholds, muscle tonus, and various other responses. For example, the RAS may relay a message to the heart requesting that more blood be pumped in preparation for this fight or flight scenario. At the same time the autonomic nervous system produces a number of bodily changes that prepares the individual to utilize large amount of energy.

The cognitive interpretation of any given scenario influences the anticipatory arousal reaction. Given that we interpret an environment as threatening or potentially exciting, we would be likely to experience an increase in arousal, both cortical (RAS), and autonomic (heart rate). This perspective allows for the merging of both a positive and a negative situation to elicit the same physiological reaction. For example, during a tennis match both our mental as well as our physical arousal are necessary for a successful game. The situation need not be a life threatening one in order to produce arousal, rather, many activities are associated with arousal. Social psychologists have maintained that cognitive dissonance may produce arousal. Cognitive dissonance proposed by Leon Festinger (1957), has been defined as the following: "two elements [beliefs and/or behaviors] are in a dissonant relation if, considering these two alone, the obverse of one element would follow from the other" (Brewer & Crano). That is, when an unexpected event or piece of information is unpleasant, it requires alleviating, which produces arousal. Arousal here is interpreted as both the cognitive and the physiological reaction to the stimuli.

It has been implied that a certain level of arousal is required for all activities. In 1955 Hebb proposed a relationship between arousal and performance which could be represented as a normal bell curve. As illustrated above, arousal is necessary for functioning in a fight or flight scenario, Hebb suggested that arousal is necessary for behavioral efficiency in our everyday lives. Hebb hypothesized that low arousal levels would produce negative behavioral responses, and support was found for this in a study by Heron (1957). In this study, Heron paid male college students a substantial sum of money to lie on a bed for as many days as they could endure. Various measures were taken to produce a restricted stimulation environment. After an average of a day, results began to surface, with subjects indicating that they were having trouble thinking clearly. After about 48 hours, most of the participants were unable to complete basic mathematical computations (ex: 12 x 5 = ?). Hallucinations became common, and concentration was low.

Attempting to unite selective attention and arousal, one may find that it is a circular model. An optimal level of arousal allows information to be received by choosing the stimuli threshold necessary for recognition, while the selection filter of information relies on various aspects of our cognitive make and our experiences in choosing the stimuli we attend to. Prior experience, as well as our perception of the stimuli influences our arousal level as well as the selection process. Stimuli may then elicit or reduce arousal, sending the process into a new cycle.



Resources on the World Wide Web

A brief overview of attention and automaticity is offered at this site. It describes the dichotic listening task, the bottleneck model, attenuation model, and late selection models, with illustrations.

This site explains Broadbent's filter model, and it's problems.

This site offers an extensive list of references in the implications of arousal on memory.

The model of arousal (as proposed by Apter in 1982) is illustrated here. This model accounts for arousal coupled with hedonic tone, our drive to experience pleasure.

A general overview of arousal is given here with a list of references. A connection is made between Hull's theory and Berlyne who attempted to explain the connection between arousal and curiosity.


References