Situation Awareness¶

Status: established
Last updated: 2026-06-01
Sources: 9781119636113.Ch17.Pdf, 001872095779049543.Pdf, 154193128803200221.Pdf
Tags: [situation-awareness, SA, cognitive-engineering, decision-making, human-factors, team-performance, system-design, automation]

Summary¶

Situation awareness (SA) is a person's understanding of what is happening in the current situation and is critical for performance in complex, dynamic domains including aviation, air traffic control, driving, military operations, emergency management, health care, and power grid operations. SA can be thought of as an internalized mental model of the current state of a person's environment, forming the central organizing feature from which all decision making and action takes place (Endsley, 2021).

Problems with SA were found to underlie human error in 88% of accidents by commercial airlines (Endsley, 1995b). Reviews of errors and accidents in air traffic control, nuclear power, health care, and driving show that problems maintaining accurate SA are responsible for the majority of human error in many complex systems. People's primary struggle is most often not in determining the correct thing to do, nor in physically performing their tasks, but in fundamentally understanding what is going on in the situation (Endsley, 2021).

Body¶

Context¶

Endsley (2021), in her handbook chapter on situation awareness, examines SA as a person's understanding of what is happening in a current situation — an internalized mental model of the environment's state that organizes decision making and action in dynamic domains such as aviation, air traffic control, driving, military operations, emergency management, health care, and power-grid operation. The chapter sets out a definition and three-level model, a cognitive process model, the factors that undermine SA, team SA, and approaches to training, design, and measurement. Within this knowledge base SA is the integrative cognitive-state construct that sits between perception and action: it draws on the information-processing account in Information Processing and the perceptual coding in Sensation And Perception, shares attentional limits with Mental Workload, names display design (the concern of Representation Design) as a primary lever for support, and frames the out-of-the-loop problem central to Supervisory Control Of Automation.

Key Points¶

The model summarised here from the 2021 handbook chapter originates in two earlier primary sources now held in the corpus. Endsley (1988) first set out the three-level definition of SA — perception of elements (Level 1), comprehension of their meaning (Level 2), and projection of their future status (Level 3) — and introduced design and measurement approaches for it (PDF pp. 1–3, orig. pp. 97–99). Endsley (1995) then developed the full theory of SA in dynamic systems, presenting SA as a state distinct from the decision making and performance it informs, and identifying attention and working memory as the limiting factors and mental models, schema, and goal-directed processing as the mechanisms that support it (PDF pp. 1–4, orig. pp. 32–35). The 88% figure and the SAGAT measure described below trace to this body of Endsley's own work.

Endsley's most widely used definition treats SA as "the perception of the elements in the environment within a volume of time and space, the comprehension of their meaning and the projection of their status in the near future" (Endsley, 1988). This yields three ascending levels: Level 1, perceiving the status, attributes, and dynamics of relevant elements (other aircraft, warning lights, vital signs, aircraft positions); Level 2, comprehending their significance in light of goals (a warning light's seriousness, trampled grass signalling a recent camp, a rash indicating shingles); and Level 3, projecting future status (anticipating occurrences, disease prognosis, the actions of pedestrians). The levels are not strictly sequential — higher-level SA can supply default values for unperceived Level 1 elements and drive the search for them. SA covers only the dynamically changing situational information that dictates when background knowledge applies; its elements are domain-specific, illustrated in detail for air traffic control across all three levels. Endsley stresses that information available from observation, system and interface knowledge, or teammates is not SA until the person who needs it has acquired and understood it (PDF pp. 2–3, orig. pp. 435–436).

The cognitive model rests on information-processing theory (Endsley, 1988, 1995c, 2015a). Pre-attentive sensory stores detect properties such as proximity, colour, shape, and movement that cue focal attention; limited attention is the major constraint on perceiving multiple items in parallel. Working memory is significantly related to SA in novices but not experts, who offload onto long-term memory. Long-term memory supports SA through mental models — mechanisms for describing system purpose, explaining states, and predicting future states (Rouse & Morris, 1985) — and through schema, prototypical system states (engine failure, attack formation) that enable one-step retrieval and carry associated action scripts, reducing working-memory load. Effective processing alternates between goal-driven (top-down) and data-driven (bottom-up) modes; purely data-driven operators react only to salient cues, while purely goal-driven operators miss signals that goals should change. Expertise has a large effect: one study found a tenfold difference in SA between the lowest and highest pilots (test-retest reliability above 0.94), with better SA linked to attention sharing, pattern matching, spatial ability, perceptual speed, and working memory (PDF pp. 5–7, orig. pp. 438–440).

Eight factors undermine SA (Endsley et al., 2003; Endsley & Jones, 2012): attentional tunneling (locking onto part of the environment and dropping scanning); the requisite memory trap (overloading working memory); workload, anxiety, fatigue and other stressors (WAFOS), which narrow attention and prompt premature closure; data overload (intake outpacing assimilation, often from ineffective presentation); misplaced salience (salient features misdirecting attention); complexity creep (features outrunning operators' mental models, harming Levels 2 and 3); errant mental models, including mode errors (Sarter & Woods, 1995); and out-of-the-loop syndrome, where automation lowers awareness of its own state so failures go undetected (Endsley & Kiris, 1995) (PDF pp. 8–9, orig. pp. 441–442).

SA extends to teams. Team SA is the degree to which every member has the SA needed for their job, and shared SA the degree to which members hold the same SA on shared requirements (Endsley, 1995c; Endsley & Jones, 2001); both predict team performance. Beyond shared environmental information, teams need awareness of interdependencies between members' tasks, communicated through verbal and nonverbal communication, shared displays, and a shared environment. Shared displays tailored to each member's explicit SA requirements improve performance under high workload, whereas untailored displays that repeat everything do not. Shared mental models aid shared SA and reduce communication needs — better aircrews communicated less — while crews new to working together are more accident-prone (44% of aviation accidents occur on a new crew's first leg, 73% on the first day). Effective teams plan contingencies, share information, self-check, and prioritise goals (Orasanu & Salas, 1993; Taylor, Endsley & Henderson, 1996); poor teams withhold information and lack error-detection processes (PDF pp. 9–11, orig. pp. 442–444).

Training and design follow from the model. Tools such as ISAT (rapid experiential exposure), VESARS (SA feedback via behavioural, communication, and query measures), and SAVI (peer-instruction practice against expert assessments) build mental models and schema. For design, Endsley distinguishes the information gap — adding data without aiding integration overloads operators — from genuine SA support, and prescribes an SA-oriented design process of requirements analysis, design principles, and measurement. Goal-Directed Task Analysis identifies goals, subgoals, decisions, and the Level 1–3 SA each decision needs, is goal-based and technology-free, and has been completed for many domains. Fifty design principles address general support, uncertainty and confidence, complexity, alarm management, automation, and team operations (PDF pp. 11–15, orig. pp. 444–448).

Measurement gives sensitivity and diagnosticity beyond performance and workload measures alone. Process measures (eye tracking, communications, physiological) are objective and continuous but miss whether information is understood; performance measures confuse SA with performance; subjective measures (SART, Likert) capture confidence rather than accuracy and correlate poorly with objective SA (Endsley, 2020a); objective measures (SAGAT, SPAM) query SA knowledge directly. SAGAT, the most widely used objective measure (Endsley, 1988, 1995a, 2000a), freezes a simulation, blanks displays, and compares operators' answers — drawn from GDTA requirements across all three levels — against the simulated truth; across studies it shows 94% sensitivity (68 studies), r = .46 prediction of performance (35 studies), high test-retest reliability, and no performance effect from freezes (11 studies) (PDF pp. 16–17, orig. pp. 449–450).

Conclusion¶

Endsley (2021) concludes that SA is a distinct stage from decision making and action but a key input to both: people's primary struggle is usually not choosing or executing the right action but understanding what is happening, with SA problems underlying human error in 88% of commercial aviation accidents (Endsley, 1995b) and the majority of errors across air traffic control, nuclear power, health care, and driving. Because accurate decisions depend on good SA even where procedures are well established, the chapter argues for designing, training, and measuring systems explicitly around SA requirements rather than data volume.

References¶

Endsley, M.R. (1988) 'Design and evaluation for situation awareness enhancement', Proceedings of the Human Factors Society 32nd Annual Meeting, 32(2), pp. 97–101. doi: 10.1177/154193128803200221. endsley1988design

Endsley, M.R. (1995) 'Toward a theory of situation awareness in dynamic systems', Human Factors, 37(1), pp. 32–64. doi: 10.1518/001872095779049543. endsley1995theory

Endsley, M.R. (2021). Situation awareness. In G. Salvendy & W. Karwowski (Eds.), Handbook of Human Factors and Ergonomics (5th ed., pp. 434-455). John Wiley & Sons. endsley2021situationawareness

Orasanu, J., & Salas, E. (1993). Team decision making in complex environments. In G. A. Klein, J. Orasanu, R. Calderwood, & C. E. Zsambok (Eds.), Decision making in action: Models and methods (pp. 327–345). Norwood, NJ: Ablex. To be validated.

Rouse, W. B., & Morris, N. M. (1985). On looking into the black box: Prospects and limits in the search for mental models (DTIC #AD-A159080). Atlanta, GA: Center for Man-Machine Systems Research, Georgia Institute of Technology. To be validated.

Sarter, N. B., & Woods, D. D. (1995). "How in the world did I ever get into that mode": Mode error and awareness in supervisory control. Human Factors, 37(1), 5–19. To be validated.

Taylor, R. M., Endsley, M. R., & Henderson, S. (1996). Situational awareness workshop report. In B. J. Hayward & A. R. Lowe (Eds.), Applied aviation psychology: Achievement, change and challenge (pp. 447–454). Aldershot: Ashgate Publishing Ltd. To be validated.

Open Questions¶

How can SA be maintained during rapid transitions between automation control and manual intervention?
What methods best support shared SA in distributed teams with limited communication bandwidth?
How should SA measurement adapt to increasingly automated systems where the human role shifts from operator to supervisor?
What are the most effective approaches for training projection (level 3) SA in complex domains?
How can system designs better support SA for information uncertainty and confidence calibration?