Human behaviour in fire

Human behaviour in fire is the way people perceive, interpret and respond to a developing fire and its warnings. It includes recognition of cues, decisions about action, interaction with other people and movement towards safety. Behaviour can also include investigating, warning others, collecting belongings or attempting another protective action.

People do not always evacuate immediately after an alarm. Initial information can be ambiguous, and occupants may seek confirmation before deciding what to do. Familiarity, alertness, responsibilities, social relationships and the building layout influence the response.

Research challenges the assumption that fire behaviour is dominated by irrational panic. Proulx reports that reasonable and altruistic actions are common, while panic is rare in the reviewed fire incidents. Engineering models should represent this range of behaviour instead of treating occupants as either passive objects or reckless crowds.

Perception and interpretation

A fire cue can be an alarm, smoke odour, visible smoke, flame, heat, noise or information from another person. Each cue supplies different evidence about the event. The observer must notice it and decide whether it indicates an emergency.

Bryan describes recognition as the identification of ambiguous cues as a fire incident. Recognition can be delayed when the signs are weak or consistent with a normal event. The observer's prior experience and expectation of a favourable outcome can influence this interpretation.

Validation follows when a person seeks confirmation that the initial interpretation is correct. This may involve looking for additional cues or asking other people. The process can reduce uncertainty but also consumes time before evacuation begins.

Warning systems provide different amounts of information. A bell announces a signal but not its cause or the required action. Directive voice messages can explain the incident and the response, although conflicting visual cues can undermine their credibility.

Repeated unwanted alarms can reduce the credibility of a warning. Proulx describes occupants becoming less willing to investigate or respond when false alarms are frequent. The alarm system's operational history therefore becomes part of the behavioural context.

Decision processes and protective action

After recognition and validation, a person defines the situation and evaluates possible responses. Bryan's process model includes commitment to an action and reassessment if it is incomplete or conditions change. Behaviour is thus a sequence rather than one irreversible choice.

Human response includes decisions and reassessment over time. Based on Bryan (2002), pp. 3-318–3-319.

Bryan organises the response around six connected processes: recognition, validation, definition, evaluation, commitment and reassessment. This is a conceptual account of how decisions can change as a fire develops. It does not assign a universal duration to each process or establish that every person follows one fixed sequence.

Possible actions are not limited to flight. Bryan describes decisions that can include obtaining an extinguisher, activating an alarm, seeking refuge or evacuating. The chosen response reflects the person's location, perceived threat and assessment of available options.

A person can also change actions after receiving new information. Smoke along a route, a staff message or the behaviour of companions may trigger reassessment. Static calculations that assign one action at the start cannot capture those changes.

Perceived time available is part of evaluation. Sime links this perception to information about fire location and development, as discussed by Bryan. The time a person believes remains for escape can therefore differ from a calculated tenability interval based on the actual fire.

Social influence and affiliation

The presence of other people changes cue interpretation. Bryan describes experiments in which passive companions inhibited reporting of introduced smoke. These specific findings illustrate social influence, not a universal delay for every public occupancy.

The experiment involved students completing a questionnaire while smoke entered the room. Bryan reports smoke reporting by 75% of subjects tested alone, compared with 10% when two passive participants were present. In groups of three uninformed subjects, someone reported the smoke in 38% of groups during the six-minute observation period.

The outcome measured was reporting smoke, not successful evacuation from a real building fire. The group conditions also differed deliberately in whether other participants remained passive. These distinctions matter because the experiment demonstrates social influence on interpretation, without supplying an evacuation-time distribution for a public building.

Noticing a cue and acting on it were separate observations. Bryan reports median smoke-noticing delays of 5 s for subjects alone and 20 s in the group conditions. The much longer observation period for reporting shows why noticing smoke cannot simply be equated with deciding to leave.

People can mimic adaptive as well as nonadaptive responses. Observing others investigate, warn or evacuate can supply information about what is happening. Staff behaviour can consequently provide a strong cue when an alarm alone is ambiguous.

Affiliative behaviour is the tendency to seek or remain with people who have emotional ties. Family members may gather before leaving, and groups may travel together. This can increase preparation time while also providing assistance and shared information.

Evacuation is often a social process rather than isolated individual movement. Based on Proulx (2002), pp. 3-344–3-347.

Group movement can be limited by a slower member. Proulx gives the example of a family moving with a child or older person. An individual walking-speed value may therefore be inappropriate for the group as a whole.

Responsibility also affects response. Proulx contrasts occupants who expect to manage their own household alarm with visitors who wait for instructions from building staff. The distribution of expected responsibility should be considered when assigning warning and management roles.

Pre-movement activities

Pre-movement time is the period before directed evacuation travel begins. It includes perception, interpretation and preparatory action, and it may involve movement within the current space. It should not be imagined as everyone waiting motionless for one fixed duration.

Observed activities include dressing, collecting children or pets, securing work, gathering belongings and contacting staff. These tasks respond to personal circumstances and the information available. Their duration can vary between occupants in the same building.

Proulx's residential drills showed different response times between buildings and warning conditions. Poor alarm audibility and the arrival of other cues affected when people left. The values belong to those case studies and should not be used as universal delays.

The contrast between reported case means is substantial. In three Canadian office-building drills, Proulx recorded more than 1,000 individual start times and reported a mean of 50 s. In three residential buildings with good alarm audibility, the reported mean was 2 min 49 s.

A separate winter residential drill had a mean start delay of 5 min 19 s. Occupants spent time putting on outdoor clothing before leaving. The office occupants also undertook preparatory actions, including finishing calls, saving computer work and gathering belongings, despite their training.

These means describe different buildings and circumstances, rather than controlled estimates of one isolated cause. They also do not represent the last person to leave or an upper design percentile. Using a mean as though every occupant starts at that time would remove the variation described in the observations.

New Zealand's C/VM2, second edition, 2025, assigns pre-travel times by occupant state, familiarity, alarm signal and location relative to the fire. These are regulatory design inputs for that method. They do not eliminate the need to establish the relevant population and cue conditions.

Direct cues such as seeing or smelling smoke may be available near the fire, while remote occupants may rely on indirect information. C/VM2 limits when indirect cues can be assumed in its method. Social interaction and spatial arrangement therefore matter to timing analysis.

The New Zealand method specifically distinguishes socially connected spaces from separate tenancies. Its 2025 commentary allows some indirect cues within a shared retail area or single-management office. It does not assume the same interaction between separate apartments, hotel rooms or office tenancies, even when occupants are on the same floor.

Panic and cooperative behaviour

Panic is often used as a general explanation for crowd emergencies, but fire research does not support treating it as the usual response. Proulx's summary identifies panic as rare and altruistic, reasonable behaviour as common. People often act on limited and ambiguous information.

Competitive rushing in some crowd incidents should not automatically be labelled as irrational panic. Proulx distinguishes different crowd phenomena and discusses density, blocked movement and communication failures. The physical conditions can create forces and risks without requiring a crowd-wide loss of reason.

Cooperation can include warning others, assisting with movement and sharing responsibility. Bryan reports purposeful, altruistic responses as frequent in the reviewed incidents. Such behaviour can improve some outcomes while adding actions that a simple evacuation model omits.

Assuming universal panic can lead to the wrong engineering question. Instead of asking only how to suppress irrational behaviour, the analysis should examine whether information, routes and assistance match occupants' needs. Many delays arise from uncertainty and social commitments rather than uncontrolled emotion.

This does not mean that dangerous crowding never occurs. It means that each incident requires examination of density, route capacity, information and group interaction. A single emotional label cannot replace that analysis.

Movement and changing conditions

Walking speed is one part of evacuation movement, but it changes with density, route geometry, occupant characteristics and smoke. Proulx describes speed decreasing as crowding constrains movement. Clear-air unrestricted walking is not a universal emergency value.

On stairs, movement depends on step geometry and crowding as well as individual ability. Queues can form when demand exceeds the capacity of a route element. The time spent waiting is different from the time spent walking towards that restriction.

Smoke can reduce visibility and slow movement. The Australian Building Codes Board's (ABCB) 2020 review found a lack of reliable general correlations for predicting walking speed in smoke. Its numerical modification factors are tied to the Australian verification method and should retain that context.

Smoke can alter walking speed and route choice. Based on Australian Building Codes Board (2020), pp. 280–282.

Irritant smoke can affect the eyes and willingness to continue. ABCB notes that occupants may seek an alternative path or wait for assistance under its stated conditions. Route choice can therefore change after movement has begun.

Evacuation joins these movement processes to the earlier response stages. Total time includes warning, interpretation, preparation and travel to the place of safety. Treating walking speed as the entire evacuation would omit important sources of delay and variation.

Building and occupant characteristics

Familiarity with the building affects both route choice and information seeking. Proulx reports a preference for familiar entry routes in many situations. Emergency routes that occupants have never experienced may need clearer guidance and staff support.

Alertness influences the response to an alarm. Sleeping occupants can require longer to recognise and prepare for action, while people focused on an activity may initially disregard other cues. C/VM2 reflects some of these differences in its method-specific inputs.

Physical, sensory or cognitive limitations can change preparation and movement. Proulx includes these among occupant characteristics that can extend response time. The analysis should represent the people expected to use the building rather than an assumed uniformly able population.

Door operation provides a concrete example of this variation. Bryan reviews a study of ambulant participants with disabilities and a separate group of seven manual-wheelchair users negotiating a door. The latter group generally needed longer, and some attempts failed. Replacing these interactions with one reduced walking speed would omit the effect of opening and passing the door.

Visual access to other occupants can provide information. Open spaces allow people to observe group responses, while separate rooms can limit that cue. Building layout thus affects both direct wayfinding and social interpretation of the event.

Staff training and availability can alter the timing. Proulx describes staff prompting occupants and providing building-specific directions. Their effectiveness depends on recognition, authority, coverage of the occupied space and consistency with other cues.

Modelling and evidence limits

Evacuation models often simplify human behaviour. The 2005 International Fire Engineering Guidelines (IFEG) describe assumptions of simultaneous departure, uninterrupted movement and excellent knowledge of the shortest route. Those simplifications can help examine capacity but do not provide a complete behavioural prediction.

A model should be suited to the occupancy and layout. A simple approach may fail to identify local queues or group decisions in a complex building. Expensive software does not guarantee better input data or a more faithful account of human response.

Case studies and drills provide evidence about particular populations and conditions. Warning credibility, staff action, notice and weather can influence the observations. Results should retain those circumstances instead of being transferred as universal occupant constants.

Interview evidence also requires careful interpretation. Recalled timings can be uncertain, and people may have experienced only part of the event. Proulx presents such studies as valuable context rather than exact measurements applicable to every incident.

See also

• Escape signage
• Voice alarm system
• Evacuation modelling

• Protective action decision model

References

• Australian Building Codes Board, National Research Council of Canada, International Code Council, & Department of Building and Housing New Zealand. (2005). International fire engineering guidelines (2005 ed.). Australian Building Codes Board.
• Australian Building Codes Board. (2020). Fire safety verification method data sheets (Version 1.1).
• Bryan, J. L. (2002). Behavioral response to fire and smoke. In P. J. DiNenno (Ed.), SFPE handbook of fire protection engineering (3rd ed., pp. 3-315–3-341). National Fire Protection Association.
• Ministry of Business, Innovation and Employment. (2025). C protection from fire: Verification method C/VM2 (2nd ed.). New Zealand Government.
• Proulx, G. (2002). Movement of people: The evacuation timing. In P. J. DiNenno (Ed.), SFPE handbook of fire protection engineering (3rd ed., pp. 3-342–3-366). National Fire Protection Association.