Vigilance and Sustained Attention¶

Status: established
Last updated: 2026-06-01
Sources: 17470214808416738.Pdf, Science.472714.Pdf, 001872089606380211.Pdf
Tags: [vigilance, sustained-attention, vigilance-decrement, monitoring, signal-detection-theory, automation-monitoring, complacency, attention, cognitive-foundations]

Summary¶

Vigilance is the capacity to sustain attention and detect infrequent, unpredictable signals over a prolonged watch. Mackworth (1948) first demonstrated experimentally that detection efficiency declines after about half an hour on watch — the "vigilance decrement" — using his Clock Test. Parasuraman (1979) used signal detection theory to show that the decrement reflects a loss of perceptual sensitivity only when the task loads memory and presents events rapidly; otherwise it reflects a shift in response criterion. Molloy and Parasuraman (1996) extended the problem to automation, showing that operators monitoring an automated system are slow to detect even a single failure, and that detection falls over time much as in classic vigilance tasks. Together the three papers establish that humans are poorly suited to passive monitoring, a finding central to Supervisory Control Of Automation, Mental Workload, and Aviation Human Factors.

Body¶

Context¶

This article draws on three primary sources spanning the history of the field. Mackworth (1948), in The breakdown of vigilance during prolonged visual search, reports the laboratory study that founded vigilance research. Parasuraman (1979), in a Science report, analyses the decrement with signal detection theory to separate perceptual from decisional causes. Molloy and Parasuraman (1996), in Human Factors, carry the question into the monitoring of automation. Within this knowledge base vigilance is the sustained-attention counterpart to the momentary attention treated in Information Processing: it shares its signal-detection method with Psychophysics And Signal Detection Theory, its capacity concerns with Mental Workload, and its monitoring problem with Supervisory Control Of Automation and Trust In Automation, where it appears as automation-induced complacency.

Key Points¶

Mackworth (1948) established the vigilance decrement. His Clock Test placed a single observer alone in a cabin watching a black pointer that moved in small regular steps around a blank dial; at long irregular intervals the pointer made a "double jump" twice the usual distance, which the observer had to detect by pressing a key (PDF pp. 3–4, orig. pp. 7–8). Over a two-hour watch, detection efficiency was definitely lower after the observer had been watching for about half an hour, and one-hour watches showed the same advantage of the first half-hour over the second (PDF p. 16, orig. p. 20). A half-hour rest after a half-hour watch restored performance for a further spell, and a sudden telephone message in mid-watch produced a temporary improvement lasting about half an hour, whereas prior instruction that a period would contain more signals had no demonstrable effect. Mackworth interpreted the main downward trend as the partial experimental extinction of a conditioned voluntary response, produced by the absence of knowledge of results, and the telephone effect as disinhibition (PDF p. 16, orig. p. 20). He adopted the term vigilance for a readiness to perceive and respond that, unlike attention, need not be consciously experienced (PDF p. 2, orig. p. 6).

Parasuraman (1979) identified the conditions under which the decrement is perceptual rather than decisional. Using signal detection theory, he separated sensitivity (d′) from response criterion (β) and distinguished two task types: successive discrimination, where the target is a change relative to an absent standard and so imposes a memory load, and simultaneous discrimination, where target and comparison are present together and impose none (PDF p. 1, orig. p. 924). Across auditory and visual experiments, a true sensitivity decrement appeared only when the task both loaded memory (successive discrimination) and presented events at a high rate; under low event rates or simultaneous discrimination the decrement instead reflected a more cautious response criterion with no loss of detectability (PDF pp. 1–3, orig. pp. 924–926). Parasuraman concluded that memory load and event rate are the two critical factors controlling sensitivity decrements, a result he linked to an effort theory of attention and to performance evaluation in radar monitoring and inspection (PDF pp. 3–4, orig. pp. 926–927).

Molloy and Parasuraman (1996) extended vigilance research to the monitoring of automation. Using a multi-attribute flight-simulation task, they compared participants who monitored an automated engine-status function while also performing tracking and fuel-management tasks (multicomplex), those who performed the monitoring task alone (single-complex), and those who performed a simple visual vigilance task (PDF pp. 1–2, orig. pp. 311–312). Detection of a single automation failure under the multicomplex load was poorer than detection of the same malfunctions under manual control, and more participants detected the failure in the first 10 minutes of a 30-minute session than in the last 10 minutes, for both the simple and the multicomplex tasks; the single-complex group detected the failure equally well in both periods (PDF p. 1, orig. p. 311). The authors concluded that automation-related monitoring inefficiency — overreliance, or complacency — arises even for a single failure and is worsened by concurrent task load, with direct implications for vigilance theory and automation design (PDF p. 1, orig. p. 311).

Conclusion¶

The three papers trace one finding across half a century: humans sustain attention poorly during passive monitoring. Mackworth (1948) showed the decrement exists and is sensitive to rest and stimulation; Parasuraman (1979) showed it has two distinct causes — a genuine loss of perceptual sensitivity under high memory load and high event rate, and a shift of response criterion otherwise — so that not every vigilance decrement is the same phenomenon; and Molloy and Parasuraman (1996) showed the problem persists and intensifies when the human supervises automation, where it manifests as complacency. The practical lesson, recurring through the applied articles in this knowledge base, is that monitoring roles should be designed against the limits of sustained attention rather than assuming the watchkeeper stays reliably alert.

References¶

Mackworth, N.H. (1948) 'The breakdown of vigilance during prolonged visual search', Quarterly Journal of Experimental Psychology, 1(1), pp. 6–21. doi: 10.1080/17470214808416738. mackworth1948vigilance

Molloy, R. & Parasuraman, R. (1996) 'Monitoring an automated system for a single failure: Vigilance and task complexity effects', Human Factors, 38(2), pp. 311–322. doi: 10.1518/001872089606380211. molloy1996monitoring

Parasuraman, R. (1979) 'Memory load and event rate control sensitivity decrements in sustained attention', Science, 205(4409), pp. 924–927. doi: 10.1126/science.472714. parasuraman1979memory

Open Questions¶

How do the perceptual (sensitivity) and decisional (criterion) accounts of the decrement map onto monitoring of modern automated displays, where event rate and memory load are jointly engineered?
Can the rest-pause and stimulation effects Mackworth (1948) observed be used to schedule watchkeeping in remote-monitoring and control-room work (see Supervisory Control Of Automation)?
What design and trust interventions most reduce automation-induced complacency (Molloy & Parasuraman, 1996) without inducing distrust or disuse (see Trust In Automation)?