Protective action decision model - Revision history

Jan: Add 4 watermarked SVG illustrations from the Expert visual plan

2026-11-02T16:34:53Z

Add 4 watermarked SVG illustrations from the Expert visual plan

← Older revision		Revision as of 17:34, 2 November 2026
Line 7:		Line 7:

	The time from the cue to the start of movement is the pre-movement phase. It consists of a response period, until the cue is recognised, and a delay period after it. The movement itself ends when the occupant reaches a safe area.<ref name="NR-1.8" />		The time from the cue to the start of movement is the pre-movement phase. It consists of a response period, until the cue is recognised, and a delay period after it. The movement itself ends when the occupant reaches a safe area.<ref name="NR-1.8" />
			[[File:fse-protective-action-decision-model-v1.svg\|thumb\|upright=1.5\|alt=Horizontal timeline from fire initiation via cue, recognition and start of movement to a safe area, with the pre-movement phase bracketed.\|Timeline from fire initiation to a safe area: cue, response, delay and movement periods, with the pre-movement phase in which the protective action decision is made. Based on National Research Council of Canada et al. (2005), pp. 1.8-1–1.8-3.]]

	The movement chapter of the SFPE Handbook splits the delay time to start into three parts: perception, interpretation and action.<ref name="PR-348">Proulx (2002), p. 3-348.</ref> Occupants first perceive the alarm signal, a fire cue or a warning from others. They then interpret this information, often investigating or searching for more information before deciding to evacuate. Even after that decision, they usually carry out other actions first, such as getting dressed, gathering children or collecting valuables.<ref name="PR-348" /> ''Introduction to Fire Safety Engineering'' describes the same sequence as three phases: awareness, decision-making and execution.<ref name="VR-7.3" />		The movement chapter of the SFPE Handbook splits the delay time to start into three parts: perception, interpretation and action.<ref name="PR-348">Proulx (2002), p. 3-348.</ref> Occupants first perceive the alarm signal, a fire cue or a warning from others. They then interpret this information, often investigating or searching for more information before deciding to evacuate. Even after that decision, they usually carry out other actions first, such as getting dressed, gathering children or collecting valuables.<ref name="PR-348" /> ''Introduction to Fire Safety Engineering'' describes the same sequence as three phases: awareness, decision-making and execution.<ref name="VR-7.3" />
Line 18:		Line 19:

	In commitment, the person carries out the chosen response. If it fails, reassessment follows, which may be the most stressful stage.<ref name="BR-319">Bryan (2002), p. 3-319.</ref> With each failure, more energy goes into the response, the person becomes less selective about the risks taken, and the chance of success falls. At the Arundel Park fire, more people chose windows as an escape route when they were involved in a second attempt to get out.<ref name="BR-319" />		In commitment, the person carries out the chosen response. If it fails, reassessment follows, which may be the most stressful stage.<ref name="BR-319">Bryan (2002), p. 3-319.</ref> With each failure, more energy goes into the response, the person becomes less selective about the risks taken, and the chance of success falls. At the Arundel Park fire, more people chose windows as an escape route when they were involved in a second attempt to get out.<ref name="BR-319" />
			[[File:fse-protective-action-decision-model-v2.svg\|thumb\|upright=1.5\|alt=Flowchart of recognition, validation, definition, evaluation and commitment, with a loop from commitment through reassessment back to commitment.\|The decision processes of an individual in a fire: recognition, validation, definition, evaluation and commitment, with reassessment when a response fails. Based on Bryan (2002), pp. 3-317–3-319.]]

	== Cues and warnings ==		== Cues and warnings ==
Line 30:		Line 32:

	Seeking information is a typical response in sudden emergencies, because people need to make sense of an event before they act on it. This behaviour is known as milling, and cues were its strongest predictor.<ref name="AV-133" /> Pre-evacuation actions included making telephone calls, gathering belongings and finishing other tasks.<ref name="AV-133" /> The model explained between 49 and 56 per cent of the variation in evacuation delay, which its authors describe as extraordinarily high for a study of human evacuation.<ref name="AV-131" />		Seeking information is a typical response in sudden emergencies, because people need to make sense of an event before they act on it. This behaviour is known as milling, and cues were its strongest predictor.<ref name="AV-133" /> Pre-evacuation actions included making telephone calls, gathering belongings and finishing other tasks.<ref name="AV-133" /> The model explained between 49 and 56 per cent of the variation in evacuation delay, which its authors describe as extraordinarily high for a study of human evacuation.<ref name="AV-131" />
			[[File:fse-protective-action-decision-model-v3.svg\|thumb\|upright=1.5\|alt=Path diagram: cues, floor and information lead to perceived risk, information seeking, pre-evacuation actions and evacuation delay.\|The NIST model of evacuation delay in the World Trade Center towers, with the share of variation explained for each outcome in WTC 1 and WTC 2. Based on Averill et al. (2005), pp. 131 and 219.]]

	The delays themselves were spread widely. In WTC 1 the mean delay was 5.61 minutes, the median 3.00 minutes and the most frequent value 1 minute. In WTC 2 these values were 6.04, 4.00 and 5 minutes.<ref name="AV-221">Averill et al. (2005), pp. 220–221.</ref> The reported delays peaked in the first or second minute and then tailed off. Their distribution was approximately lognormal or gamma-shaped, and the delay tended to increase with the height of the floor.<ref name="AV-127">Averill et al. (2005), pp. 127–128.</ref>		The delays themselves were spread widely. In WTC 1 the mean delay was 5.61 minutes, the median 3.00 minutes and the most frequent value 1 minute. In WTC 2 these values were 6.04, 4.00 and 5 minutes.<ref name="AV-221">Averill et al. (2005), pp. 220–221.</ref> The reported delays peaked in the first or second minute and then tailed off. Their distribution was approximately lognormal or gamma-shaped, and the delay tended to increase with the height of the floor.<ref name="AV-127">Averill et al. (2005), pp. 127–128.</ref>
Line 44:		Line 47:

	Data from drills and real fires show a wide range. In drills in seven residential buildings in Canada, announced weeks ahead without a date or time, the delay depended strongly on how well the alarm could be heard. In the three buildings with a clearly audible alarm, the mean delay to start was 2 min 49 s, about three-quarters of the total evacuation time.<ref name="PR-348b">Proulx (2002), pp. 3-348–3-349.</ref> In two buildings where more than a fifth of the occupants found the alarm too quiet, some occupants started within 2 minutes, but others took up to 25 minutes, with a mean of 9 minutes.<ref name="PR-348b" /> In three office buildings with trained occupants, the mean was 50 s, although many people still finished telephone calls, saved computer data or gathered belongings first.<ref name="PR-349" /> In real fires, delays of about 2.5 minutes were found in a high-rise office fire without an alarm, and about 10 minutes in a residential fire at night.<ref name="PR-349" />		Data from drills and real fires show a wide range. In drills in seven residential buildings in Canada, announced weeks ahead without a date or time, the delay depended strongly on how well the alarm could be heard. In the three buildings with a clearly audible alarm, the mean delay to start was 2 min 49 s, about three-quarters of the total evacuation time.<ref name="PR-348b">Proulx (2002), pp. 3-348–3-349.</ref> In two buildings where more than a fifth of the occupants found the alarm too quiet, some occupants started within 2 minutes, but others took up to 25 minutes, with a mean of 9 minutes.<ref name="PR-348b" /> In three office buildings with trained occupants, the mean was 50 s, although many people still finished telephone calls, saved computer data or gathered belongings first.<ref name="PR-349" /> In real fires, delays of about 2.5 minutes were found in a high-rise office fire without an alarm, and about 10 minutes in a residential fire at night.<ref name="PR-349" />
			[[File:fse-protective-action-decision-model-v4.svg\|thumb\|upright=1.5\|alt=Bar chart of delays before evacuation: under a minute in a store and offices, about 3 minutes with an audible home alarm, 9 with a quiet one.\|Typical delays before the start of evacuation in drills and real fires, as means or medians. Based on Proulx (2002), pp. 3-348–3-349, and Averill et al. (2005), pp. 220–221.]]

	These differences follow less from the type of occupancy than from the characteristics of the building and its occupants.<ref name="PR-350" /> The building characteristics include the type of warning system, the layout, visual access to other people, any focal point such as a stage, training and the frequency of false alarms. The occupant characteristics include familiarity with the building, responsibility, social ties, commitment to an activity, alertness and physical or mental limitations.<ref name="PR-350" /><ref name="PR-351">Proulx (2002), p. 3-351.</ref> Well-trained staff and wardens can shorten the delay to start dramatically.<ref name="PR-351" />		These differences follow less from the type of occupancy than from the characteristics of the building and its occupants.<ref name="PR-350" /> The building characteristics include the type of warning system, the layout, visual access to other people, any focal point such as a stage, training and the frequency of false alarms. The occupant characteristics include familiarity with the building, responsibility, social ties, commitment to an activity, alertness and physical or mental limitations.<ref name="PR-350" /><ref name="PR-351">Proulx (2002), p. 3-351.</ref> Well-trained staff and wardens can shorten the delay to start dramatically.<ref name="PR-351" />

Mark: FSE wiki: corrections after independent fact-check (written by Claude Opus 5.5)

2026-11-01T23:05:24Z

FSE wiki: corrections after independent fact-check (written by Claude Opus 5.5)

Mark: New FSE wiki article: Protective action decision model (written by Claude Opus 5.5)

2026-11-01T22:26:54Z

New FSE wiki article: Protective action decision model (written by Claude Opus 5.5)

New page

A '''protective action decision model''' describes how people who notice signs of a threat decide whether and how to protect themselves, for example by evacuating, moving to a refuge or staying where they are. In fire safety engineering, such models describe the steps between the first cue of a fire and the start of a protective action, and the factors that speed up or delay these steps.<ref name="BR-317">Bryan (2002), pp. 3-317–3-318.</ref><ref name="NR-1.8">National Research Council of Canada et al. (2005), pp. 1.8-1–1.8-2.</ref> The decision process takes place during the [[pre-movement time]], the period between the first cue and the start of purposeful movement to a safe place.<ref name="VR-7.3">Van Rossum (2027), §7.3.</ref>

A model of this kind was used by the National Institute of Standards and Technology (NIST) to analyse the evacuation of the World Trade Center towers in 2001. It was based on a review of about 300 studies of how people respond to warnings of natural and technological hazards, including events that called for protective actions other than evacuation.<ref name="AV-218">Averill et al. (2005), p. 218.</ref> The model explained about half of the variation in the time occupants waited before starting to evacuate.<ref name="AV-131">Averill et al. (2005), p. 131.</ref> Research on [[human behaviour in fire]] shows that a fire alarm on its own rarely triggers evacuation. The delay to start depends on the information and cues people receive, the risk they perceive and the behaviour of others.<ref name="PR-349">Proulx (2002), p. 3-349.</ref><ref name="PR-350">Proulx (2002), p. 3-350.</ref><ref name="AV-131" />

== From cue to protective action ==
The International Fire Engineering Guidelines (2005) divide the time from ignition to safety into a sequence of events.<ref name="NR-1.8" /> A cue first signals the fire. It may come from an automatic alarm, from the fire itself or from other people. The occupant then recognises the cue as a sign of fire, and after a delay, during which other actions or no action may take place, starts to move.

The time from the cue to the start of movement is the pre-movement phase. It consists of a response period, until the cue is recognised, and a delay period after it. The movement itself ends when the occupant reaches a safe area.<ref name="NR-1.8" />

The movement chapter of the SFPE Handbook splits the delay time to start into three parts: perception, interpretation and action.<ref name="PR-348">Proulx (2002), p. 3-348.</ref> Occupants first perceive the alarm signal, a fire cue or a warning from others. They then interpret this information, often investigating or searching for more information before deciding to evacuate. Even after that decision, they usually carry out other actions first, such as getting dressed, gathering children or collecting valuables.<ref name="PR-348" /> ''Introduction to Fire Safety Engineering'' describes the same sequence as three phases: awareness, decision-making and execution.<ref name="VR-7.3" />

Evacuation is not the only possible protective action. The guidelines mention progressive evacuation in high-rise buildings, in which floors far from the fire are first placed on alert. Other options are a fire-safe refuge where occupants wait for the fire service, and the newer concept of protection in place, in which occupants stay where they are.<ref name="NR-1.8b">National Research Council of Canada et al. (2005), p. 1.8-2.</ref> In the protect-in-place concept, staying is itself the intended protective action, chosen to avoid smoke-filled corridors and stairs.<ref name="NR-1.8b" />

== Stages of the decision process ==
The behaviour chapter of the SFPE Handbook describes the decision process of an individual in a fire as six linked processes, based on work by Withey.<ref name="WI">Withey (1962), as cited in Bryan (2002), p. 3-317.</ref> In recognition, a person identifies ambiguous cues as a sign of a fire. Many people first interpret such cues in terms of the most familiar and harmless explanation, partly because they feel personally invulnerable to risk.<ref name="BR-317" /> In validation, the person tries to confirm the first impression, often by seeking reassurance or more information from others who have noticed the same cues.<ref name="BR-318">Bryan (2002), p. 3-318.</ref>

In definition, the person relates the information to the situation, such as the location and size of the fire. Stress and anxiety rise fastest before the ambiguous cues have been given a clear meaning.<ref name="BR-318" /> In evaluation, the person weighs possible responses, such as escaping, fighting the fire or raising the alarm. Because fires develop quickly, this evaluation may have to be completed within a few seconds.<ref name="BR-318" /> The location of exits, the presence of other people, the perceived effects of the fire and the actions of others all influence this choice.<ref name="BR-318" />

In commitment, the person carries out the chosen response. If it fails, reassessment follows, which is the most stressful stage.<ref name="BR-319">Bryan (2002), p. 3-319.</ref> With each failure, more energy goes into the response, the person becomes less selective about the risks taken, and the chance of success falls. At the Arundel Park fire, more people chose windows as an escape route when they were involved in a second attempt to get out.<ref name="BR-319" />

== Cues and warnings ==
People become aware of a fire in many ways. In a study of 569 participants in residential fires, 26.0 per cent first noticed the fire by smelling smoke, 21.3 per cent were told by others and 18.6 per cent heard noises. Another 13.4 per cent were told by family members, 9.1 per cent saw smoke and 8.1 per cent saw the fire.<ref name="BR-316">Bryan (2002), p. 3-316.</ref> The way people are alerted shapes how serious they think the threat is.<ref name="BR-315">Bryan (2002), p. 3-315.</ref>

The alarm signal is probably the least reliable cue, because false alarms, test alarms and pranks have reduced its credibility.<ref name="PR-349" /> Three or more false alarms in six months can be regarded as a large number, and in such buildings the delay to start may be extended almost without limit.<ref name="PR-350" /> Smelling or seeing smoke is ambiguous and tends to start an investigation rather than an evacuation. A warning from another person is a better sign of a real problem. Messages from a voice communication system, or directly from staff, are taken most seriously.<ref name="PR-349" /> An [[evacuation alarm system]] with live, informative voice messages is probably the best way to start a fast response, while prerecorded messages are rarely as effective.<ref name="PR-350" />

''Introduction to Fire Safety Engineering'' reports the same pattern: a slow-whoop tone is less effective than spoken words, and live messages are followed faster than automatic alarms.<ref name="VR-7.3" /> A message only works if it agrees with what people see. In a fire in the World Trade Center in 1975, occupants on smoke-logged floors were told to return to their offices but stayed in the core lobbies. An evacuation message followed six minutes later.<ref name="BR-316" />

== Perceived risk and information seeking ==
In the NIST model, the delay before evacuation depended on six factors. These were environmental cues, the starting floor, information received without asking for it, perceived risk, the search for more information and other actions taken before leaving.<ref name="AV-131" /> Perceived risk was preferred to actual risk, because people act on what they believe to be true.<ref name="AV-132">Averill et al. (2005), p. 132.</ref> In both towers, perceived risk rose mainly with the starting floor and with exposure to cues such as smoke, damage or fire. Information that people received had a weaker effect.<ref name="AV-133">Averill et al. (2005), pp. 132–133.</ref>

Seeking information is a typical response in sudden emergencies, because people need to make sense of an event before they act on it. This behaviour is known as milling, and cues were its strongest predictor.<ref name="AV-133" /> Pre-evacuation actions included making telephone calls, gathering belongings and finishing other tasks.<ref name="AV-133" /> The model explained between 49 and 56 per cent of the variation in evacuation delay, which its authors describe as extraordinarily high for a study of human evacuation.<ref name="AV-131" />

The delays themselves were spread widely. In WTC 1 the mean delay was 5.61 minutes, the median 3.00 minutes and the most frequent value 1 minute. In WTC 2 these values were 6.04, 4.00 and 5 minutes.<ref name="AV-221">Averill et al. (2005), pp. 220–221.</ref> The reported delays peaked in the first or second minute and then tailed off. Their distribution was approximately lognormal or gamma-shaped, and the delay tended to increase with the height of the floor.<ref name="AV-127">Averill et al. (2005), pp. 127–128.</ref>

== Social influences ==
Other people strongly shape each step of the decision. In experiments by Latane and Darley, smoke was let into a room where students were filling in a questionnaire. Of the students who were alone, 75 per cent reported the smoke, but only 10 per cent did so when two passive companions were present.<ref name="LD">Latane and Darley (1968), as cited in Bryan (2002), p. 3-317.</ref> Social inhibition, diffusion of responsibility and imitation of others explain this effect.<ref name="LD" /> It can make people in restaurants, hotels and other public buildings ignore early cues.<ref name="BR-317" /> ''Introduction to Fire Safety Engineering'' links this to the conformity experiments of Asch and notes that group behaviour can also help, when a well-trained part of the group takes the right decision and others follow.<ref name="VR-7.3" />

Staff and wardens have a large effect. In an unannounced drill in a department store, the fast response of staff led customers to start moving after 25 s on average, with a maximum of 55 s.<ref name="PR-349" /> In an underground station, passengers did not respond to the fire alarm at all, but moved at once when staff prompted them, and within 15 s of a precise live message.<ref name="PR-349" /> People also follow authority: some wait for permission from a manager before leaving.<ref name="VR-7.3" />

Responsibility and social ties play a part as well. People at home react quickly to a smoke alarm because they know nobody else will deal with it, while visitors to a public building assume that someone will tell them if something is really wrong.<ref name="PR-350" /> Occupants also tend to gather the people to whom they have emotional ties, such as family members, before they leave. This [[affiliative behaviour]] takes time, especially when the members of a group are not together at the start of the incident.<ref name="PR-350" />

== Use in fire safety engineering ==
The decision process enters a design through the pre-movement time in the comparison of available and required escape time. In the past, engineers assumed that occupants would start to evacuate as soon as they perceived the alarm or a fire cue. Reports by fire victims have always challenged this assumption.<ref name="PR-348" /> Some designers used to double the calculated movement time, while others added 15 to 30 s. The SFPE Handbook argues that values should instead be based on research findings.<ref name="PR-348" />

Data from drills and real fires show a wide range. In seven residential buildings in Canada, occupants started to evacuate between 2 and 25 minutes after the alarm, with a mean of 9 minutes.<ref name="PR-349" /> In three office buildings with trained occupants, the mean was 50 s, although many people still finished telephone calls, saved computer data or gathered belongings first.<ref name="PR-349" /> In real fires, delays of about 2.5 minutes were found in a high-rise office fire without an alarm, and about 10 minutes in a residential fire at night.<ref name="PR-349" />

These differences follow less from the type of occupancy than from the characteristics of the building and its occupants.<ref name="PR-350" /> The building characteristics include the type of warning system, the layout, visual access to other people, any focal point such as a stage, training and the frequency of false alarms. The occupant characteristics include familiarity with the building, responsibility, social ties, commitment to an activity, alertness and physical or mental limitations.<ref name="PR-350" /><ref name="PR-351">Proulx (2002), p. 3-351.</ref> Well-trained staff and wardens can shorten the delay to start dramatically.<ref name="PR-351" />

== Limits and uncertainty ==
Human behaviour is one of the least predictable aspects of fire safety. Although many data from fires are available, they have not yet led to generally valid models for predicting pre-evacuation times.<ref name="VR-7.3" /> Researchers hesitate to suggest numbers because the research findings are limited.<ref name="PR-351" /> Drill data may not apply directly to real fires. Some argue that real fires should cause faster responses, but interviews with victims show that the first signs of a real fire are usually judged extremely ambiguous, which lengthens the delay.<ref name="PR-351" />

Statistical models of the decision process also have limits. Even the NIST model left about half of the variation in delay unexplained.<ref name="AV-131" /> Its data came from survivors, so occupants of the upper floors of the south tower who waited too long were missing from the sample.<ref name="AV-221" /> Comparisons between the towers need care, because the cues that led people to evacuate differed between them.<ref name="AV-127" /> Night-time fires, sleeping occupants and people with limited mobility or alertness can lead to much longer delays than daytime drills suggest.<ref name="PR-349" /><ref name="PR-351" />

== Common misconceptions ==
A widespread belief is that people in a fire tend to [[panic]]. Research shows the opposite. Deliberate, purposeful behaviour is the most frequent response, and panic is an infrequent or unusual reaction in most fires.<ref name="BR-315" /> ''Introduction to Fire Safety Engineering'' adds that a lack of awareness of the danger is usually a greater problem than panic, and that people find it hard to judge how fast a fire develops.<ref name="VR-7.3" />

A second misconception is that an alarm leads straight to evacuation. A fire alarm on its own rarely triggers evacuation unless training and other information are provided. It works best as a way to alert occupants and prepare them for further information.<ref name="PR-350" /> A third misconception concerns the choice of route. People prefer the way by which they entered the building and attractive, transparent routes. Exit signs in their usual form have little influence on this choice, and emergency doors meant only for incidents are hardly used at all.<ref name="VR-7.3" />

== See also ==
* [[Evacuation modelling]]
* [[Required safe egress time]]
* [[Stay-put strategy]]
* [[Phased evacuation]]
* [[Escape signage]]

== Notes ==
<references />

== References ==
* Averill, J. D., Mileti, D. S., Peacock, R. D., Kuligowski, E. D., Groner, N., Proulx, G., Reneke, P. A., & Nelson, H. E. (2005). ''Occupant behavior, egress, and emergency communications'' (NIST NCSTAR 1-7). National Institute of Standards and Technology.
* Bryan, J. L. (2002). Behavioral response to fire and smoke. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), ''SFPE handbook of fire protection engineering'' (3rd ed., pp. 3-315–3-341). National Fire Protection Association; Society of Fire Protection Engineers.
* National Research Council of Canada, International Code Council, Department of Building and Housing, & Australian Building Codes Board. (2005). ''International fire engineering guidelines'' (2005 ed.). Australian Building Codes Board.
* Proulx, G. (2002). Movement of people: The evacuation timing. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), ''SFPE handbook of fire protection engineering'' (3rd ed., pp. 3-342–3-366). National Fire Protection Association; Society of Fire Protection Engineers.
* Van Rossum, E. (2027). ''Introduction to Fire Safety Engineering''. Fire Safety Academy.

[[Category:Human Behaviour & Evacuation]]

← Older revision		Revision as of 00:05, 2 November 2026
Line 1:		Line 1:
	A '''protective action decision model''' ~~describes how~~ people who notice signs of a threat ~~decide whether and how~~ to ~~protect themselves, for example by~~ evacuating, moving to a refuge or staying where they are. ~~In fire safety engineering~~, such ~~models describe~~ the ~~steps between~~ the ~~first cue~~ of ~~a fire~~ and the ~~start of a~~ protective action~~, and the factors that speed up or delay these steps~~.<ref name="BR-317">Bryan (2002), pp. 3-317–3-318.</ref><ref name="NR-1.8">National Research Council of Canada et al. (2005), pp. 1.8-1–1.8-2.</ref> The decision process takes place during the [[pre-movement time]], the period between the first cue and the start of purposeful movement to a safe place.<ref name="VR-7.3">Van Rossum (2027), §7.3.</ref>		A '''protective action decision model''' is a model of the decision process that leads people who notice signs of a threat to a protective action. Such an action may be evacuating, moving to a refuge or staying where they are.<ref name="NR-1.8b">National Research Council of Canada et al. (2005), p. 1.8-2.</ref> The SFPE Handbook of Fire Protection Engineering (2002) and the International Fire Engineering Guidelines (2005) describe such a protective-action decision process. So does a study of the 2001 World Trade Center evacuation by the National Institute of Standards and Technology (NIST). None of them uses the name protective action decision model.<ref name="BR-317">Bryan (2002), pp. 3-317–3-318.</ref><ref name="NR-1.8">National Research Council of Canada et al. (2005), pp. 1.8-1–1.8-2.</ref><ref name="AV-218">Averill et al. (2005), p. 218.</ref> The decision process takes place during the [[pre-movement time]], the period between the first cue and the start of purposeful movement to a safe place.<ref name="VR-7.3">Van Rossum (2027), §7.3.</ref>

	A model of ~~this kind was used by the National Institute of Standards and Technology~~ (~~NIST~~) ~~to analyse the evacuation of the World Trade Center towers in 2001~~. It was ~~based~~ on a review of about 300 studies of how people respond to ~~warnings of~~ natural and technological hazards~~, including~~ events ~~that called for~~ protective actions other than evacuation.<ref name="AV-218"~~>Averill et al. (2005), p. 218.<~~/~~ref~~> The model explained about half of the variation in the time occupants waited before starting to evacuate.<ref name="AV-131">Averill et al. (2005), p. 131.</ref> Research on [[human behaviour in fire]] shows that a fire alarm on its own rarely triggers evacuation. The delay to start depends on the information and cues people receive, the risk they perceive and the behaviour of others.<ref name="PR-349">Proulx (2002), p. 3-349.</ref><ref name="PR-350">Proulx (2002), p. 3-350.</ref><ref name="AV-131" />		NIST calls its own model a model of evacuation delay.<ref name="AV-fig">Averill et al. (2005), pp. 132 and 220.</ref> The model was built on a review of about 300 studies of how people respond to risk events and risk information for natural and technological hazards. These studies included events in which protective actions other than evacuation were warranted, and research on evacuation in high-rise fires was used as well.<ref name="AV-218" /> The model explained about half of the variation in the time occupants waited before starting to evacuate.<ref name="AV-131">Averill et al. (2005), p. 131.</ref> Research on [[human behaviour in fire]] shows that a fire alarm on its own rarely triggers evacuation. The delay to start depends on the information and cues people receive, the risk they perceive and the behaviour of others.<ref name="PR-349">Proulx (2002), p. 3-349.</ref><ref name="PR-350">Proulx (2002), p. 3-350.</ref><ref name="AV-131" />

	== From cue to protective action ==		== From cue to protective action ==
Line 10:		Line 10:
	The movement chapter of the SFPE Handbook splits the delay time to start into three parts: perception, interpretation and action.<ref name="PR-348">Proulx (2002), p. 3-348.</ref> Occupants first perceive the alarm signal, a fire cue or a warning from others. They then interpret this information, often investigating or searching for more information before deciding to evacuate. Even after that decision, they usually carry out other actions first, such as getting dressed, gathering children or collecting valuables.<ref name="PR-348" /> ''Introduction to Fire Safety Engineering'' describes the same sequence as three phases: awareness, decision-making and execution.<ref name="VR-7.3" />		The movement chapter of the SFPE Handbook splits the delay time to start into three parts: perception, interpretation and action.<ref name="PR-348">Proulx (2002), p. 3-348.</ref> Occupants first perceive the alarm signal, a fire cue or a warning from others. They then interpret this information, often investigating or searching for more information before deciding to evacuate. Even after that decision, they usually carry out other actions first, such as getting dressed, gathering children or collecting valuables.<ref name="PR-348" /> ''Introduction to Fire Safety Engineering'' describes the same sequence as three phases: awareness, decision-making and execution.<ref name="VR-7.3" />

	Evacuation is not the only possible protective action. The guidelines mention progressive evacuation in high-rise buildings, in which floors far from the fire are first placed on alert. Other options are a fire-safe refuge where occupants wait for the fire service, and the newer concept of protection in place, in which occupants stay where they are.<ref name="NR-1.8b"~~>National Research Council of Canada et al. (2005), p. 1.8-2.<~~/~~ref~~> In the protect-in-place concept, staying is itself the intended protective action, chosen to avoid smoke-filled corridors and stairs.<ref name="NR-1.8b" />		Evacuation is not the only possible protective action. The guidelines mention progressive evacuation in high-rise buildings, in which floors far from the fire are first placed on alert. Other options are a fire-safe refuge where occupants wait for the fire service, and the newer concept of protection in place, in which occupants stay where they are.<ref name="NR-1.8b" /> In the protect-in-place concept, staying is itself the intended protective action, chosen to avoid smoke-filled corridors and stairs.<ref name="NR-1.8b" />

	== Stages of the decision process ==		== Stages of the decision process ==
Line 17:		Line 17:
	In definition, the person relates the information to the situation, such as the location and size of the fire. Stress and anxiety rise fastest before the ambiguous cues have been given a clear meaning.<ref name="BR-318" /> In evaluation, the person weighs possible responses, such as escaping, fighting the fire or raising the alarm. Because fires develop quickly, this evaluation may have to be completed within a few seconds.<ref name="BR-318" /> The location of exits, the presence of other people, the perceived effects of the fire and the actions of others all influence this choice.<ref name="BR-318" />		In definition, the person relates the information to the situation, such as the location and size of the fire. Stress and anxiety rise fastest before the ambiguous cues have been given a clear meaning.<ref name="BR-318" /> In evaluation, the person weighs possible responses, such as escaping, fighting the fire or raising the alarm. Because fires develop quickly, this evaluation may have to be completed within a few seconds.<ref name="BR-318" /> The location of exits, the presence of other people, the perceived effects of the fire and the actions of others all influence this choice.<ref name="BR-318" />

	In commitment, the person carries out the chosen response. If it fails, reassessment follows, which is the most stressful stage.<ref name="BR-319">Bryan (2002), p. 3-319.</ref> With each failure, more energy goes into the response, the person becomes less selective about the risks taken, and the chance of success falls. At the Arundel Park fire, more people chose windows as an escape route when they were involved in a second attempt to get out.<ref name="BR-319" />		In commitment, the person carries out the chosen response. If it fails, reassessment follows, which may be the most stressful stage.<ref name="BR-319">Bryan (2002), p. 3-319.</ref> With each failure, more energy goes into the response, the person becomes less selective about the risks taken, and the chance of success falls. At the Arundel Park fire, more people chose windows as an escape route when they were involved in a second attempt to get out.<ref name="BR-319" />

	== Cues and warnings ==		== Cues and warnings ==
	People become aware of a fire in many ways. In a study of 569 participants in residential fires, 26.0 per cent first noticed the fire by smelling smoke, 21.3 per cent were told by others and 18.6 per cent heard noises. Another 13.4 per cent were told by family members, 9.1 per cent saw smoke and 8.1 per cent saw the fire.<ref name="BR-316">Bryan (2002), p. 3-316.</ref> The way people are alerted ~~shapes~~ how serious they think the threat is.<ref name="BR-315">Bryan (2002), p. 3-315.</ref>		People become aware of a fire in many ways. In a study of 569 participants in residential fires, 26.0 per cent first noticed the fire by smelling smoke, 21.3 per cent were told by others and 18.6 per cent heard noises. Another 13.4 per cent were told by family members, 9.1 per cent saw smoke and 8.1 per cent saw the fire.<ref name="BR-316">Bryan (2002), p. 3-316.</ref> The way people are alerted may influence how serious they think the threat is.<ref name="BR-315">Bryan (2002), p. 3-315.</ref>

	The alarm signal is probably the least reliable cue, because false alarms, test alarms and pranks have reduced its credibility.<ref name="PR-349" /> Three or more false alarms in six months can be regarded as a large number, and in such buildings the delay to start may be extended almost without limit.<ref name="PR-350" /> Smelling or seeing smoke is ambiguous and tends to start an investigation rather than an evacuation. A warning from another person is a better sign of a real problem. Messages from a voice communication system, or directly from staff, are taken most seriously.<ref name="PR-349" /> An [[evacuation alarm system]] with live, informative voice messages is probably the best way to start a fast response, while prerecorded messages are rarely as effective.<ref name="PR-350" />		The alarm signal is probably the least reliable cue, because false alarms, test alarms and pranks have reduced its credibility.<ref name="PR-349" /> Three or more false alarms in six months can be regarded as a large number, and in such buildings the delay to start may be extended almost without limit.<ref name="PR-350" /> Smelling or seeing smoke is ambiguous and tends to start an investigation rather than an evacuation. A warning from another person is a better sign of a real problem. Messages from a voice communication system, or directly from staff, are taken most seriously.<ref name="PR-349" /> An [[evacuation alarm system]] with live, informative voice messages is probably the best way to start a fast response, while prerecorded messages are rarely as effective.<ref name="PR-350" />

	''Introduction to Fire Safety Engineering'' reports the same pattern: a slow-whoop tone is less effective than spoken words, and live messages are followed faster than automatic alarms.<ref name="VR-7.3" /> A message ~~only works if it agrees~~ with ~~what people see~~. In a fire in the World Trade Center in 1975, occupants on smoke-logged floors were told to return to their offices but stayed in the core lobbies. An evacuation message followed six minutes later.<ref name="BR-316" />		''Introduction to Fire Safety Engineering'' reports the same pattern: a slow-whoop tone is less effective than spoken words, and live messages are followed faster than automatic alarms.<ref name="VR-7.3" /> If a message conflicts with other cues, such as the smell or sight of smoke, its credibility may be questioned and the information disregarded.<ref name="BR-316" /> In a fire in the World Trade Center in 1975, occupants on smoke-logged floors were told to return to their offices but stayed in the core lobbies. An evacuation message followed six minutes later.<ref name="BR-316" />

	== Perceived risk and information seeking ==		== Perceived risk and information seeking ==
Line 41:		Line 41:

	== Use in fire safety engineering ==		== Use in fire safety engineering ==
	The decision process enters a design through the pre-movement time in the comparison of available and required escape time. In the past, engineers assumed that occupants would start to evacuate as soon as they perceived the alarm or a fire cue. Reports by fire victims have always challenged this assumption.<ref name="PR-348" /> ~~Some designers used to double~~ the calculated movement time, while ~~others added~~ 15 to 30 s. The ~~SFPE Handbook~~ argues that values should instead be based on research findings.<ref name="PR-348" />		The decision process enters a design through the pre-movement time in the comparison of available and required escape time. In the past, engineers assumed that occupants would start to evacuate as soon as they perceived the alarm or a fire cue. Reports by fire victims have always challenged this assumption.<ref name="PR-348" /> Earlier editions of the handbook chapter suggested doubling the calculated movement time, while many fire safety engineers preferred to add 15 to 30 s. The 2002 edition argues that values should instead be based on research findings.<ref name="PR-348" />

	Data from drills and real fires show a wide range. In seven residential buildings in Canada, occupants started to ~~evacuate between 2 and~~ 25 minutes ~~after the alarm~~, with a mean of 9 minutes.<ref name="PR-~~349~~" /> In three office buildings with trained occupants, the mean was 50 s, although many people still finished telephone calls, saved computer data or gathered belongings first.<ref name="PR-349" /> In real fires, delays of about 2.5 minutes were found in a high-rise office fire without an alarm, and about 10 minutes in a residential fire at night.<ref name="PR-349" />		Data from drills and real fires show a wide range. In drills in seven residential buildings in Canada, announced weeks ahead without a date or time, the delay depended strongly on how well the alarm could be heard. In the three buildings with a clearly audible alarm, the mean delay to start was 2 min 49 s, about three-quarters of the total evacuation time.<ref name="PR-348b">Proulx (2002), pp. 3-348–3-349.</ref> In two buildings where more than a fifth of the occupants found the alarm too quiet, some occupants started within 2 minutes, but others took up to 25 minutes, with a mean of 9 minutes.<ref name="PR-348b" /> In three office buildings with trained occupants, the mean was 50 s, although many people still finished telephone calls, saved computer data or gathered belongings first.<ref name="PR-349" /> In real fires, delays of about 2.5 minutes were found in a high-rise office fire without an alarm, and about 10 minutes in a residential fire at night.<ref name="PR-349" />

	These differences follow less from the type of occupancy than from the characteristics of the building and its occupants.<ref name="PR-350" /> The building characteristics include the type of warning system, the layout, visual access to other people, any focal point such as a stage, training and the frequency of false alarms. The occupant characteristics include familiarity with the building, responsibility, social ties, commitment to an activity, alertness and physical or mental limitations.<ref name="PR-350" /><ref name="PR-351">Proulx (2002), p. 3-351.</ref> Well-trained staff and wardens can shorten the delay to start dramatically.<ref name="PR-351" />		These differences follow less from the type of occupancy than from the characteristics of the building and its occupants.<ref name="PR-350" /> The building characteristics include the type of warning system, the layout, visual access to other people, any focal point such as a stage, training and the frequency of false alarms. The occupant characteristics include familiarity with the building, responsibility, social ties, commitment to an activity, alertness and physical or mental limitations.<ref name="PR-350" /><ref name="PR-351">Proulx (2002), p. 3-351.</ref> Well-trained staff and wardens can shorten the delay to start dramatically.<ref name="PR-351" />

	== Limits and uncertainty ==		== Limits and uncertainty ==
	Human behaviour is one of the least predictable aspects of fire safety. Although many data from fires are available, they have not yet led to generally valid models for predicting pre-evacuation times.<ref name="VR-7.3" /> Researchers hesitate to suggest numbers because the research findings are limited.<ref name="PR-351" /> ~~Drill data may not apply directly to real fires~~. Some argue that real fires should cause faster responses, but interviews with victims show that the first signs of a real fire are usually judged extremely ambiguous, which ~~lengthens~~ the delay.<ref name="PR-351" />		Human behaviour is one of the least predictable aspects of fire safety. Although many data from fires are available, they have not yet led to generally valid models for predicting pre-evacuation times.<ref name="VR-7.3" /> Researchers hesitate to suggest numbers because the research findings are limited.<ref name="PR-351" /> The SFPE Handbook regards unannounced drills as perfectly representative of a fire in which the only sign is the sudden sound of the alarm.<ref name="PR-348" /> Some argue that real fires should cause faster responses, but interviews with victims show that the first signs of a real fire are usually judged extremely ambiguous, which further increases the delay.<ref name="PR-351" />

	Statistical models of the decision process also have limits. Even the NIST model left about half of the variation in delay unexplained.<ref name="AV-131" /> Its data came from survivors, so occupants of the upper floors of ~~the south tower~~ who waited too long were missing from the sample.<ref name="AV-221" /> Comparisons between the towers need care, because the cues that led people to evacuate differed between them.<ref name="AV-127" /> Night-time fires, sleeping occupants and people with limited mobility or alertness ~~can lead~~ to ~~much~~ longer ~~delays than daytime drills suggest~~.<ref name="PR-349" /><ref name="PR-351" />		Statistical models of the decision process also have limits. Even the NIST model left about half of the variation in delay unexplained.<ref name="AV-131" /> Its data came from survivors, so occupants of the upper floors of WTC 2 who waited too long were missing from the sample.<ref name="AV-221" /> Comparisons between the towers need care, because the cues that led people to evacuate differed between them.<ref name="AV-127" /> Night-time fires, sleeping occupants and people with limited mobility or alertness are likely to need a longer time to start.<ref name="PR-349" /><ref name="PR-351" />

	== Common misconceptions ==		== Common misconceptions ==