The Parasuraman Vigilance Taxonomy

Converging evidence strongly supports that vigilance tasks impose substantial demands on observers’ information-processing resources. Using the attentional resource model, Parasuraman and Davies (1977), viewed vigilance decrement as a result of a lack of available information-processing resources. The Parasuraman taxonomy of vigilance categorizes tasks by type of discrimination in addition to event rate.  Parasuraman proposed that type of discrimination is distinguished by whether signal detection requires successive or simultaneous discrimination. Successive discrimination tasks are those in which observers must rely on their working memory for critical information to compare current stimuli against in order to discriminate signal from nonsignal stimuli.  In simultaneous discrimination tasks, a comparative judgment is made in which all of the information needed to make a discrimination between signal and nonsignal stimuli is provided in the stimuli presented.  The important difference between these two types of tasks is the presence (in successive tasks) or absence (in simultaneous tasks) of the mental workload imposed on the observer as a result of relying on his or her working memory. The other significant variable is event rate, or the rate of the background, nonsignal stimulus events, which evidence shows is event inversely related to detection performance.  More specifically, vigilance tasks with event rates of 24 events/minute and higher are classified as having a high event rate, while any task with a rate below this previously researched and supported cut-off value is considered to have a low event rate (See et al._1995.pdf)

Parasuraman_1979.pdf found that a sensitivity decrement was only found when the observer was undergoing a successive discrimination task when stimuli were presented at a rapid pace.  However, such an effect was not found in tasks where either memory was not needed for discrimination (i.e. simultaneous tasks) or when event rate was low.  According to this model, sensitivity decrements occur in tasks that are high-event rate/successive-discrimination tasks, as the combination of the associated increase in mental workload in addition to a fast rate of stimuli presentation progressively makes it harder to maintain as time lapses during the vigil.  In order to verify this argument, which had been based on results of an experiment using only auditory stimuli, he ran a second experiment in which three visual vigilance tasks at a high event rate were compared. The results of this study confirmed his original findings, showing effects on sensitivity decrement only in the successive-discrimination task, and also supported the distinction between successive and simultaneous discrimination during vigilance tasks in the presence of both auditory and visual stimulation. Additionally, general analysis of the findings of 27 previous studies, offered strong support that this taxonomy could be generalized across a variety of vigilance tasks. 

Warm et al._2008.pdf provide evidence showing that the NASA-TLX, the Multiple Resource Questionnaire, and neuroimaging studies support the attentional resource model, which was used as the foundation of the Parasuraman taxonomy.  As mentioned in the work of Warm et al. (2008), these measures show that the workload imposed by vigilance tasks reflects the associated mental effort and depletion of information-processing resources.  Use of the transcranial Doppler sonography (TCD) as a means of monitoring human-system performance in jobs that require vigilance tasks is an interesting and innovative idea, especially when relating to a vigilance task in which a sensitivity decrement can have detrimental effects.  Warm et al. (2008) suggest that using such a tool could then be used to help decide when task aiding is needed or when observers need to rest or be removed from the given task due to their sensitivity decrement.  However, say individuals in such positions were regularly measured using TCD, assuming that there are no harmful health effects of such frequent exposure to ultrasound signals, I am somewhat confused about how this would be a practical and realistic way to address this problem.  See et al. (1995) reported that in a study by Teichner (1974) it was found that resulting decrement in performance is generally complete after only 25-30 minutes after the beginning of the vigil, with a minimum of 50% of such loss occurring within the first 15 minutes of the watch.  Given this evidence, one would believe that the TCD-measured temporal declines in hemovelocity, which as mentioned by Warm et al. (2008), have been found to parallel vigilance decrements, would occur around similar time intervals during a vigilance task.  This would be an extremely short period of time to constantly have workers rest or be removed from their posts.  However, on the other hand, when it is something like military surveillance or a job of equivalent stature in which missing a signal could have disastrous effects, the cost might be worth the added expense of hiring two men for every job in order to allow the observer to rest when there is a recorded temporal decline in hemovelocity.  Even then however this seems like an extreme measure, given the frequency of need to rest after such a short lapse in time. 

There are many strengths of the Parasuraman taxonomy.  I believe the most important being its practical and everyday implications, as Parasuraman’s looked to define specific variables that cause vigilance failures in certain tasks as opposed to in others.  This in itself is a crucial area of concern, as vigilance is increasingly becoming a component of life in the workplace, as the use of semiautomated systems increases with the prominence of technology within our society. Such vigilance tasks, as briefly discussed above, include jobs such as long-distance driving and military surveillance, where error in signal detection can have disastrous, if not fatal, consequences.  In light of this research, thinking about just how commonplace vigilance tasks are really sheds light on the importance of this research.  Overall, the Parasuraman taxonomy has expanded our knowledge of vigilance, specifying conditions that have since been used as a basis from which to further test and explain the decrement found in an observer’s ability while performing a vigilance task. 

However, there are weaknesses of the Parasuraman taxonomy that must be acknowledged and addressed. Through their meta-analysis, See et al. (1995) found that many studies have since found both the presence of sensitivity decrement during tasks that are not consistent with the model in addition to the absence of sensitivity decrement during tasks where the model suggests should occur. Such findings have caused researchers to question the results of the Parasuraman taxonomy, as they have found that this sensitivity decrement occurs as a task becomes more demanding, and it may therefore be a result of the total demand from the combination of task characteristics.  The meta-analysis conducted by See et al. (1995) supported the importance of type of discrimination and event rate significantly affected sensitivity decrement, however, they also found that type of stimuli (sensory or cognitive) and the average level of sensitivity associated with the task were also two characteristics that significantly affected sensitivity decrement.  Therefore, this sensitivity decrement in vigilance tasks may occur more frequently than the Parasuraman taxonomy suggests, as vigilance decrement occurs in a variety of situations, and not just in high-event rate/successive discrimination tasks.  This work specifically suggests that a sensory-cognitive dimension should be added to the vigilance taxonomy, as this research has found that the impact that both the type of discrimination and the event rate have on sensitivity decrement will change depending on the type of stimuli presented.  In addition, while the Parasuraman taxonomy classifies tasks depending on whether or not they result in a sensitivity decrement, these later findings suggest that instead, sensitivity should be classified in terms of the magnitude of change over time.  In addition, the Parasuraman taxonomy classifies event rate as either high or low, however, recent work suggests that it is important to take a more continuous outlook on the task, in order to measure for changes in decrement of sensitivity as a function of event rate, type of discrimination, and type of presented stimuli (See et al., 1995).  This makes sense as the Parasuraman taxonomy groups events by such finite categories, however, something such as the set event rates conducted in laboratory experiments might not be so translatable to real-world situations where nonsignals may not have any pattern.  Future research should look into the effect of event rate on signal detection in situations where there are multiple, differing signals that the observer must detect, in addition to situations where there are many different types of nonsignals within the same vigil, in order to see how such conditions affect mental workload and sensitivity decrement.