RSET/ASET, a Flawed Concept for Fire Safety Assessment
Fire Science and Technology Inc., 9000 – 300th Place SE, Issaquah WA 98027
Joseph M. Fleming
Boston Fire Department, 115 Southampton Street, Boston MA 02118
B. Don Russell
Texas A&M University, Electrical Engineering Department, College Station TX 77843
For the evaluation of occupant safety in the case of building fires, the Required Safe Egress
Time/Available Safe Egress Time (RSET/ASET) concept has become widespread and is now commonly
used in the fire safety engineering profession. It has also become commonly used by smoke detector
(smoke alarm) manufacturers in assessing whether a particular detector technology is adequate. It is
shown in this paper that the concept is intrinsically flawed and its use promotes the diminishment of fire
safety available to building occupants. The concept innately ignores the wide variations in capabilities
and physical condition of persons involved in fire. It is based on implicitly assuming that, after a brief
period where they assess the situation and mobilize themselves, occupants will proceed to the best exit in
a robotic manner. This assumption completely fails to recognize that there are very few fires, especially in
residential occupancies, where occupants perished or were seriously injured who had endeavored to exit
in this robotic manner. Instead, in the vast majority of fire death and serious injury cases, the occupants
did not move in such a manner and their evacuation took longer than anticipated on the basis of robotic
movement. There is a wide variety of reasons for this, and these are well known in the profession. The
concept also ignores that there can be a wide variation in fire scenarios. The same building and the same
fire protection features can be evaluated, but both RSET and ASET can change drastically, depending on
the scenario used. The consequence of using the RSET/ASET concept for fire safety engineering or
product design purposes is that fire deaths and injuries are permitted to occur which are preventable.
Keywords: available safe egress time; escape behavior; fire escape models; human behavior; required
safe egress time; smoke alarms; smoke detectors; tenability in fire.
Most fires will become lethal to a building occupant, if the occupant remains exposed to the products of
combustion for a long enough time period. Since early and successful extinguishment cannot be ensured
for all cases, fire safety strategies for buildings are generally based on timely evacuation of occupants1.
Since a fire may not be obvious to the occupant, especially when sleeping, warning of the occupants by a
smoke detector2 is a necessary feature of fire safety.
The mother was woken by a smoke detector, which sounded, but sounded late. She found one of her
children and brought the child out of the burning building. She then went back in to get her other child.
1 This concept applies only to normal buildings. In some situations, evacuation is impossible and different fire
protection strategies are needed. Such exceptions includes submarines, aircraft in flight, and other specialized
2 We shall use the term smoke detector for all types of devices which are designed to cause an audible alarm to be
sounded upon the detection of smoke. Some authors distinguish between central-panel-connected ‘smoke detectors’
and single-station ‘smoke alarms.’ For the purposes of this paper, we do not distinguish between these types of
devices and will apply the term ‘smoke detector’ to either type of device.
But she was overcome by smoke while inside and perished. The second child, meanwhile, had exited by
himself through another exit and was no longer in the house. In another family, children were being taken
care by a babysitter. A fire started on the first floor of a house. There was a smoke detector response, but
it was late. The babysitter brought one child out successfully. Then she went to look for the second child.
The other child meanwhile had attempted to escape herself, but made the wrong choice of going upstairs,
instead of heading for the front door. By the time the babysitter reentered, the stairs were impassable and
the child died.
In another fire, the family was in one part of the house, while electrical wiring started a smoldering fire in
the living room. Eventually, the fire ignited the Christmas tree, which started burning rapidly. The smoke
detector sounded, but too late. The father thought that he could take the tree outside, but enough of it had
come alight that he got badly burned in the process. A smoldering fire started in the sofa. After a very
long time, the smoke detector sounded and the mother was woken. She rounded up her two children and
attempted to escape through the back door. Unfortunately, the back door had an inside-key deadbolt and
she could not get the key to work, despite fiddling with it. Eventually, she decided to exit through a
window and get a neighbor’s help to break the door down and rescue the children, who had meanwhile
run away from her. But when the neighbor arrived, he had to restrain her from reentering, since the fire
would have been lethal to anybody reentering. The two children perished. Fire histories of this type are
exceedingly common and are known to most fire safety practitioners. The common thread in all of them is
that the occupants were notified of the fire when the fire was at a late enough stage so that there would
probably have been enough time to move robotically to an exit, but not enough time to have an
intervening activity take place.
Another category of slow occupant responses is those where the occupants are moderately slowed down
due to their physical condition. Obviously, if occupants are bed-ridden and immobile, or so intoxicated
that waking is precluded, it is not expected that a smoke detector’s warning will benefit them, even if a
very early alarm were sounded. But when occupants are only modestly infirm, the fire safety profession
should not dismiss the potential for them to be saved. Examples here would be a person with a leg cast, or
one who has taken a tranquilizer, but is still able to wake to an alarm sound, albeit not to move rapidly
A number of specific cases have also been documented in the files of the Consumer Product Safety
Commission (CPSC). During the night, a student at the University of Miami of Ohio  was awoken by
the sound of a smoke detector. According to fire reports, the fire was started by smoking materials
igniting furniture. When he opened the door of his bedroom, the smoke was so thick that he had to jump
out the window to escape. He was joined by the two students in the adjacent bedroom, as well as several
students on the second floor, all of whom also jumped out windows. Three students died who did not
jump out of windows. Newspaper stories blamed the students for possibly being impaired, suggesting that
they took an especially long time to respond to the alarm and then commenced inefficient. But according
to official reports, the student who first awoke to the smoke detector was not impaired, yet at the time that
he awoke the egress paths were already blocked. The tragedy could have been avoided had the smoke
detector sounded earlier.
In a fire in Georgia , a 57 year old woman and her 96 year old mother died in a fire, thought to be
electrical in nature, that smoldered for a period of time. Initial reports from fire investigators stated that
“smoke detectors were operational but were not in the area where the fire started,” implying a delayed
detection time. Investigators also noted that “the 57 year old did not know the layout of the home and
wasn’t able to break a window,” implying a slow egress time. Both assumptions were attempts by the
investigators to explain why two people died despite working smoke alarms. However, after further
analysis, investigators came to the conclusion that the smoke had plenty of time to reach the detector and
that the most likely reason for the occupants being trapped was the delayed response of the smoke alarm.
Around 6:00 a.m., two adults awoke to the sound of a smoke alarm and realized that it was coming from
the apartment next door . Despite heroic attempts, thick smoke prevented them from rescuing the 23
year old mother and her small child. According to investigators this was an electrical fire that started
adjacent to the mother’s bed. The child was trapped because she was incapable of self-rescue; yet she
could have been evacuated had her mother been able to act. But the mother was likely incapacitated by
high levels of CO, yet high levels of concomitant smoke did not activate the smoke alarm.
Development of the RSET/ASET concept
In the evaluation of fire safety for buildings or other places of human occupancy, the modern trend has
been to seek to establish quantitative performance metrics. Quantification is always desirable, but
especially so when a comparison of two or more alternative strategies is to be made. The most important
area where this has been of issue is the provision for adequate escape potential for the occupants of a
place where a fire has started. Over the last few decades, a metric has arisen where adequate consideration
has not been given to its correctness, since superficially it seems to be a sensible concept. The concept is
that occupants of a place undergoing a fire will require a fixed, calculable (or prescribable by fiat) time to
make their escape. This time is called RSET, the Required Safe Egress Time. A parallel calculation is
then made of ASET, the Available Safe Egress Time. If the simple relation ASET > RSET is fulfilled,
then it is deemed that the building’s fire safety is fully adequate, at least in regards to the safety of
escaping occupants. The calculations involved are usually simple to make and the arithmetic relationship
is trivially easy to evaluate.
The first significant example of use of the RSET/ASET was the 1975 “Indiana Dunes I” smoke detector
study sponsored by NIST . At that time, the terms RSET and ASET were not yet developed, but the
evaluation of occupant fire safety in this research study was done using the concept, albeit without the
In the first phase of the Indiana Dunes I study, there was surprisingly no account taken of the time the
occupants will consume to make their ultimate escape. The authors simply assumed that, if there was any
time, regardless how slight, after the alarm was sounded and before untenable conditions were reached on
the escape route, “success” was achieved. In other words, the assumption was made that the occupant will
need zero time to make the escape. In 1976, Phase 2 of Indiana Dunes I testing was reported , and the
analysis was also identical. In 1983, Waterman, one of the authors of the Indiana Dunes I studies,
published a follow-on paper  which stated that, of course, people will need more than zero time to
make their escape, but he did not reanalyze the data with that in mind. Instead, he referred to an
unpublished study by Rexford Wilson, who had reanalyzed the data on the basis that it would take the
occupants 1, 2, 3, 4, or 5 minutes to effectuate their escape. Wilson then recommended that, of these
choices, 3 minutes should be selected. This was taken to be conservative, since tests conducted by the
City of Los Angeles Fire Department showed that, in a staged exercise, all occupants would escape in less
than 1 minute. The staged exercise involved persons who were woken from sleep by a simulated fire
alarm and were to make their exit upon being woken.
In fairness, it must be pointed out that, while it is obvious today that, in many cases, people perish in fires
since they do not move robotically toward a viable exit, this was not necessarily widely known in the
mid-1970s. In 1977, Stahl and Archea  published a state-of-the-art review of the literature on the
emergency evacuation of buildings. The paper discusses the fact that occupants may not proceed to an
exit upon being notified by an alarm signal, if they perceive that this is unlikely to be a real fire. It also
establishes that it was already known in the mid-70s that an occupant is likely to “investigate” the status
of the fire before commencing the escape. But the authors only make reference to only a single,
unpublished paper which discusses the possibility that an occupant may reenter a burning building, and
no references to any other behaviors which would cause a significant delay in the person’s exiting the
building. No instances of attempts to fight a fire that have gone badly wrong are cited. Furthermore, the
majority of studies to that date focused on commercial or institutional occupancies and on high-rise
buildings, rather than single-family homes, even though the latter are where the overwhelming majority of
fire deaths occur. It is the view of the present authors that evacuation studies of institutional buildings
should not be used as the basis for assessing fire safety in residences. Apart from the basic fact that fire
deaths in commercial/institutional occupancies are comparatively rare, human behavior in private homes
tends to be different than in public occupancies. It is extremely unlikely, for example, that a person would
delay the escape from an office building due to searching for their pet dog. More important, people are
rarely asleep in a commercial building when a fire occurs. Even studies on high-rise housing tend not be
helpful towards understanding fires in single-family homes. In 1981, a more comprehensive review paper
 was able to identify three studies where reentry of buildings was mentioned. But it again demonstrated
that the published studies on human behavior in fires remained highly focused on commercial,
institutional, and high-rise buildings, devoting but little effort to examine common behavioral aspects
associated with fire deaths in single-family houses.
The first important study to focus on non-robotic behavior of people was not available until 1987. In that
year, Levin  described his EXITT evacuation model, which was the first computer model to consider
that human movements in fire are different from water flowing through a channel or marbles sliding
downhill3. It explicitly described that the first action of a person becoming aware of a fire is to
“investigate” the fire, which will usually involve moving towards the fire, and not away from it. It also
considered that in residences (unlike in public occupancies), upon learning of a fire, a person is likely to
first look for, and attempt to rescue family members. Once these preliminaries are completed, however,
the occupant was not allowed further counterproductive actions in Levin’s model.
The actual RSET/ASET terminology itself and the mathematic relationship “ASET > RSET” were first
explicitly set forth by NIST’s Cooper in 1983 . Interestingly, Cooper was a fluid mechanics specialist
and did no work on the RSET part of the equation. For ASET, he proposed a fluid-mechanical program
that calculates when the upper gas layer descends down to a person’s head, at which point the ASET time
period is terminated. The calculation of RSET he left to others.
Adoption of the RSET/ASET concept
The early history of the RSET/ASET concept was a NIST development. In more recent times, the concept
has become widespread, not only throughout the US design community, but also internationally. Barely
two years after NIST proposed the concept, the late Jonathan Sime, who was a specialist on human
behavior in fires, proposed a revised RSET scheme . Instead of assuming that the only time the
individual will need is to react to the alarm signal and march robotically to the optimal exit, he set forth
that RSET must consist of three components: RSET = Tr + Tc + Te, where Tr = recognition phase, which
includes acts such as investigating; Tc = coping phase, which includes acts such as firefighting; and Te, =
escape phase, which involves all activities that transpire thereafter, until the occupant actually exits the
building. Conceptually, Sime’s scheme would be perfectly satisfactory, since the defined periods are
elastic enough to accommodate the activities that realistically may transpire. Practically, however, the
scheme is unsound, since it implies that a fixed, specified amount of available time is “good enough.” By
contrast, a safety-oriented methodology should deliver the maximum escape time that can be physically
and economically provided.
It is also common to find the RSET/ASET concept used as an integral part of the performance-based
option in the NFPA 101 Life Safety Code , although the latter is not widely used in connection with
single-family housing. NFPA’s Fire Alarm Code, NFPA 72 , also contains a performance-based
3 Even today, it is rare to find any evacuation model which considers the actual behavior of the occupants and is not
simply a mechanistic exercise in mathematics.
option but, again, this option is generally used for occupancies other than single-family housing. The
Society of Fire Protection Engineers has issued a Guide  on human behavior in fire, where they
recommend the use of the RSET/ASET concept. The Guide briefly refers to some studies documenting
counterproductive activities, but then SFPE makes no recommendation that time be allowed for such
activities in doing calculations.
Interestingly, not long ago NFPA’s John Hall published a paper entitled “How many people can be saved
from home fires if given more time to escape?” . Hall endeavors to use existing statistics to answer
this question, even though he recognizes that these are problematic. In the course of analyzing the
statistics, Hall makes some decisions which are inexplicable. He considers that some victims are
described as “acting irrationally” and these are unlikely to benefit from having additional escape time. He
cites as an example of irrational activity the common situation where a victim ends up in a closet and
perishes there. These are invariably either small children, or else adults who got lost due to thick smoke.
Thus, in fact, the opposite is true: these victims would almost assuredly have been saved had there been
enough warning. With enough warning, escaping occupants will not encounter smoke which is extremely
thick and will see the correct egress path. Small children will need adults’ help in any case to exit. But if
time is sufficient, there will be time for adults to round up and escort the children and there will be no
reason for them to meander off by themselves into a closet. It is also generally considered an irrational
activity to go back into a burning house to remove possessions or search for pets. But, again, if there is
sufficient time provided between when an individual is first warned of a fire and when conditions get
terminally bad, the likelihood is increased that individuals performing such counterproductive acts will
still be saved. Hall also does not adequately address the reality that some ‘irrational’ behavior is caused
by loss of mental and decision-making capacities due to CO and toxic gas exposure.
Hall also considers that individuals who are “unable to act,” since they are too young or are physically
impaired, are unlikely to benefit from having additional escape time. This is an overgeneralization. If such
individuals are alone in the house, then this is likely to be true. But in many cases there are competent
adults in the same household. It will take much more time to assist an impaired individual, but if a
competent adult is available, there is no reason to believe “unable to act” individuals will not benefit from
having increased escape time. It is important to note that, despite these problems with the approach, Hall
concluded that, for individuals capable of reacting to an alarm sound, roughly half of the deaths and 2/3 of
the injuries could be prevented if more time was available between alarm and the point when conditions
become untenable. This ultimate conclusion, of course, precisely supports our thesis. Curiously, despite
emphasizing agreement  with Hall’s conclusion, NIST remains  an advocate of the kind of
RSET/ASET methodology which is detrimental towards saving lives.
The National Research Council Canada (NRCC), where much valuable research on human behavior in
fires has been done, recently published a study  where the authors accede to the RSET/ASET concept.
Yet, at the same time, they point out4 that, for normal, healthy individuals in a Canadian single-family
house at night, as much as 11 min can be required after sounding of the alarm before the occupant has
finally exited the premises. This is a helpful antidote to Wilson’s notion that 3 min should be perfectly
In recent years, the International Organization for Standardization (ISO) has taken the lead in promoting
the RSET/ASET concept. The first-generation ISO document on this topic was ISO 13387-8 . It
instructs the reader to use the RSET/ASET concept, but without giving any details how to do it.
Functionally, the document espouses Sime’s rewrite of the RSET concept, except that ‘recognition phase’
and ‘coping phase’ are lumped into a single ‘pre-movement time.’ This term is unfortunately misleading,
4 The NRCC report states that an RSET up to 16 min can be needed, of which 5 min is the time between the ignition
and the sounding of the alarm, while 11 min is the post-alarm time.
however, since ‘pre-movement time’ actually includes the time when the occupant is in motion, but not
yet on the ultimate path towards the exit. Even more confusingly, Cooper, Sime and the earlier
researchers assumed that, for RSET and ASET, t = 0 corresponds to the time of the sounding of the alarm.
ISO redefined these times to begin at ignition, and not at the time of the alarm. Thus, at the present time,
if RSET and ASET terms are used, the start time is ambiguous, unless the author makes this explicitly
clear. ISO 13571  explicitly instructs the reader to make use of the RSET/ASET concept, although it
fails to give any useful guidance for actually doing it. ISO 19706  mandates the use of the
RSET/ASET concept, but refers to other ISO standards for the actual details. Finally, ISO is in the
process of developing ISO 16738 , which is devoted solely to giving guidance on doing RSET/ASET
calculations. The current draft discusses at length various human factors, but in the final analysis, ends up
wholeheartedly recommending RSET/ASET calculations as a mathematical exercise in physics. The
concept that as much escape time as physically and economically viable should be provided does not
enter into it, and it continues the RSET/ASET orthodoxy that the minimum calculated RSET is sufficient,
and the benefits of providing more are not considered. Disappointingly, the document explicitly points
out that counterproductive, time-consuming behavior may be encountered in a fire, but then illogically
assumes that the analyst can successfully do some mathematical calculation to account for all that. This
document keeps the ISO 13387-8 notion that there will be some time elapsed prior to movement, but the
‘pre-movement time’ is renamed ‘pre-travel activity.’ Following ISO’s lead, various countries have
subsequently taken on the RSET/ASET concept. For instance, a recent paper from the Peoples’ Republic
of China  explains that Chinese fire evacuation provisions for subway trains are based on
Despite the use of quantitative variables, the evaluation of both RSET and ASET is highly subjective. To
calculate ASET, one must quantify the conditions (and therefore, the time) when occupants will no longer
be able to move safely through the exit path and into the outdoors or similar safe area. This is a two-part
problem: setting ‘tenability’ or ‘incapacitation’ criteria, and then assessing the results of experimental
fires or computer calculations against these criteria. With criteria in hand, the assessment is trivial, but the
criteria themselves are not amenable to any sort of rigorous study. The basic problem is that there is a
complex interaction between physical and psychological variables and these are not suitable for
experimental study (one generally cannot perform experiments that would endanger human volunteers,
while animals do not constitute a suitable surrogate where human intelligence is involved). Thus, various
criteria have been proposed [e.g., 20], but they are highly arbitrary and have little basis in either physics
or physiology. There is even no agreement in the profession as to the nature of the population to be
protected: should it be the average individual? or should it be a deliberately-selected case of infirmity? or
should it be based on a normal population distribution, but taken at some level much below the mean? if
so, how much?
Even though ASET has been ill-defined and lacking objectivity, the problems are much worse with
RSET. The basic RSET concept is that human beings act like robots and will proceed to march to the
correct exit in a linear and straightforward manner. This time to accomplish this has acquired the name
‘movement time’ and represents simply extrapolations primarily from fire drills. Some actual fire
evacuations have been studied, but these invariably have been of successful evacuations, i.e., dead victims
were not studied. To this robot-like ‘movement time’ is added a ‘pre-movement time,’ with the sum of the
two comprising RSET. The pre-movement time attempts to take into account the psychological fact that
victims are not athletes waiting for the starter’s signal, but rather will require a certain time before they
proceed to move anywhere, and an even further time before they decide to go towards the exit. As
mentioned above, the latest Canadian study pointed out that up to 11 min might be invested in different
actions by the occupants, once the alarm sounds. This finding appears to have been totally ignored,
presumably either due to conflict with preconceptions or possibly due to a refusal to recognize the facts of
human behavior. We can find no other papers espousing the need for realistic times to be used. Instead, in
the latest NIST study on this topic , pre-movement times in the range of 0 to 80 s (this is not a
misprint) were used. NIST further elaborated that the maximum 80 s pre-movement time should pertain
only to occupants identified as ‘elderly,’ while the remaining population was expected to not consume
more than 50 s. Neither the elderly nor the non-elderly were presumed to be consuming time to search for,
or assist others. In actual fact, a mother with four small children would be expected to be able to safely
rescue each and every one of them, despite the fact that this is a four-fold burden.
The RSET concept is fundamentally flawed because both the movement and the pre-movement times are
viewed as fixed numbers that can be adequately obtained and that will provide adequate safety for the
intended occupants. As explained above, reality is very different from this mechanistic view and an
individual may require what somebody unacquainted with the circumstances might judge to be an
“unreasonable” amount of time. Even individuals who are normal and without the responsibilities of
caring for others may take a long time if they had taken some medicine which made them sleepy and slow
to react. Thus, RSET is innately a stochastic distribution and it is improper to reduce it to a single
number, and flagrantly improper if the chosen number is not at the high-end tail of the distribution.
The solution to this situation does not lie in re-computing movement and pre-movement times on a more
realistic basis. This is because the basic RSET/ASET concept converts a quantitative question into a
categorical one. The correct question to ask when comparing fire safety strategies is: “What is the
available safe egress time with Strategy B, compared to Strategy A?” This is to be answered
quantitatively. If Strategy B offers significantly more egress time than Strategy A, then it is the one that
should be chosen, assuming that the implementation is affordable. The RSET/ASET scheme, however,
converts this pivotal safety problem into a triviality: If RSET = 100 s, ASET for Strategy A = 105 s, while
ASET for Strategy B = 1000 s, then both strategies fulfill the requirement that ASET > RSET and,
consequently, both are deemed to be equally acceptable. In actual practice, no fire safety strategy can be
100% successful, since for some individuals RSET → ∞. But the likelihood of saving lives will be
increased if the assessment strategy used recognizes that as ASET is progressively increased, more and
more lives will be saved.
Smoke detectors as a case example
The RSET/ASET analysis originated with research on smoke detectors. Not surprisingly, this is the one
area where use of the concept has led to the greatest problems in its application. Use of the RSET/ASET
concept has formed the basis for various published studies giving the conclusion that photoelectric and
ionization smoke detectors are ‘equally’ competent at saving lives. But examination of the actual test data
then shows that the conclusions would be very different if the RSET/ASET concept had not been used.
Both types of detectors ‘save lives,’ but a proper analysis of the large body of collected data very clearly
indicates that these technologies are not equivalent, and that use of photoelectric smoke detectors in the
home would result in a notable increase in ‘saving lives.’ Extensive performance testing of photoelectric
and ionization detectors by Texas A&M University in full-scale house fires, revealed very delayed
response of ionization detectors to certain fires even though conditions were rapidly becoming untenable.
RSET/ASET analysis of such data for an ‘average’ detector response gives a false conclusion that
photoelectric and ionization detectors are both adequate when, in reality, ionization detectors frequently
only sounded so late as to be ineffective .
NIST’s latest series of tests (sometimes called “Indiana Dunes II” to indicate that it is a modern
revisitation of the original 1970s work)  can perhaps be best used to illustrate why the RSET/ASET
concept is inimical to life safety. The NIST Press Release  on the study stated that: “Smoke alarms
are of two types—ionization and photoelectric. Some combination models are sold. According to the two-
year NIST home smoke alarm performance study, ionization smoke alarms respond faster to flaming
fires, while photoelectric smoke alarms respond quicker to smoldering fires. The report concluded that,
despite these differences, the placement of either alarm type on every level of the house provided the
necessary escape time for the different types of fires examined. The researchers determined the necessary
escape times [i.e., RSET] by considering the time that the alarms sounded in various locations and the
development of untenable (unsurvivable) conditions [i.e., ASET].”
The problem with this NIST conclusion is that it is not supported by the data obtained in the study. Table
1 is part of the analysis of the data that NIST published  in response to questions  regarding the
study. But such a simple RSET/ASET analysis using averages fails to bring out several important
In the smoldering scenarios, the ionization detectors often did not provide the necessary escape
In the most common flaming scenario (cooking fires), the photoelectric detectors, although a
little slower, provided at least 8-9 minutes for most fires.
In the Fast/Ultrafast Flaming scenario, neither type of smoke detector may provide sufficient
warning if the occupants are asleep.
These results have to be put in the context that most flaming fires occur while occupants are
awake (very small RSETs), while the vast majority of smoldering fires occur while occupants
are asleep (potentially long RSETs).
Table 1 Average ASET times, along with their standard deviations, as compiled by NIST 
2136 ± 1001
129 ± 74
739 ± 148
* Fast/Ultra-Fast fires
** Medium/Fast fires
276 ± 331
177 ± 69
796 ± 241
In the Report  itself, NIST provides a more accurate statement, “Both common residential smoke
alarm technologies (ionization and photoelectric) provided positive escape times in most fire scenarios.”
NIST is correct in stating that there were positive escape times in most scenarios (57/64) but they fail to
point out that the most common smoke detector in use, i.e., the ionization type, fails to provide a positive
escape time in a large fraction of the smoldering fires which could be the most common type of fatal fire
that occurs while people are sleeping (Table 2).
Table 2 Tests with positive ASET results
Number of tests with
“positive escape times”
Smoldering (12 fires)
Flaming (16 fires)
Cooking (4 fires)
Focusing now specifically on smoldering fires, Table 3 gives the ASET results for the smoldering fire
tests, as reported by NIST . For the 30 cases (10 tests, 3 variants) considered (tests where data were
not successfully collected for both detector types are excluded here), the average ASET = 1794 s for the
photoelectric detectors and 160 s for the ionization. It can easily be seen that 1794 >> 160 and that,
consequently, the photoelectric technology is the one that is more likely to save lives. This, however, is
not how NIST’s interpretation was made. Again, it should also be noted that NIST’s conclusions were
based on ‘average’ smoke detector performance, even though in certain test fires, ionization detectors
failed to sound at all.
Table 3 Available Safe Egress Time, ASET (s) for smoldering fires in the NIST Indiana Dunes II tests
1085 SDC01 Smoldering chair in living
SDC04 Smoldering mattress in
SDC06 Smoldering mattress in
SDC08 Smoldering mattress in
SDC11 Smoldering chair in living
SDC23 Smoldering chair in living
SDC27 Smoldering chair in living
room (air conditioning)
SDC31 Smoldering chair in living
SDC34 Smoldering chair in living
SDC37 Smoldering mattress in
2290 2290 2290 95 105 105
2650 2650 2650 65 70 70
18 1432 1432 22 74 74
92 92 3458 100 100 378
3298 3298 3298 16 16 16
2772 2800 2800 -54 -54 -54
270 270 1076 230 230 416
26 26 2254 26 26 374
568 568 568 298 298 298
In 2000, Fleming  proposed a way of restructuring the RSET/ASET concept for smoke detectors, so it
would no longer be inimical to life safety. It requires defining:
Margin of Safety = ASET – RSET
The Margin of Safety variable, which is a quantitative variable, is to be maximized in order to improve
life safety. By comparison, the RSET/ASET scheme is a categorical assignment:
Margin of Safety ≥ 0 → Pass
Margin of Safety < 0 → Fail
As explained above, one of the major faults of the RSET/ASET scheme is that RSET values, far from
being some simple calculation or measurement, are actually a poorly-defined stochastic distribution. But
by using Fleming’s Margin of Safety variable, this would be less of an obstacle. For the purposes of
making a comparative design, one could fairly arbitrarily select an RSET value, and it would still be clear
which of two alternate designs provides better life safety. It bears emphasizing that only comparative, not
absolute, designs can ever be rationally made. Since RSET is a stochastic distribution, it has neither a
design value nor a fixed upper limit. Thus, one cannot ever conclude that a design is “good,” but only that
design A is “safer” than design B. However, it should be clear that such a solution is practical and is not
an obstacle to competent design.
Table 4 shows the values for the Margin of Safety in the NIST tests considered above. These correspond
to RSET = 65 s, which is the value specified in NIST’s spreadsheet ; the published NIST report
considered RSET values ranging from 5 s to 140 s without adopting a unique value. Using the categorical
RSET/ASET concept, “success” is found for 27 out of 30 cases for the photoelectric detectors and for 21
out of 30 cases for the ionization detectors. But using Margin of Safety, it can be seen that the average
margin of safety is 1606 s for photoelectric detectors, and 95 s for ionization detectors. Again, we
emphasize that an unrealistic RSET = 65 s value was specified. If Wilson’s 3 min value were used, the
margins of safety would become 1491 s for the photoelectric detectors and –19 s (i.e., occupants became
incapacitated and failed to exit) for the ionization detectors. It must be emphasized that this stark
conclusion of “occupants got incapacitated and failed to exit” applies now to the average result for the
ionization detectors, and not just a few unfortunate cases. But, as described above, NRCC’s 11 minute
value is a much more realistic example to consider, if saving occupants is the objective. In such case, the
average margin of safety becomes 1011 s for the photoelectric detectors and –500 s for the ionization. The
evidence is clear which is the preferred solution, and which is a failure.
Table 4 Margin of Safety (s) for smoldering fires in the NIST Indiana Dunes II tests
1020 SDC01 Smoldering chair in living
SDC04 Smoldering mattress in
SDC06 Smoldering mattress in
SDC08 Smoldering mattress in
SDC11 Smoldering chair in living
SDC23 Smoldering chair in living
SDC27 Smoldering chair in living
room (air conditioning)
SDC31 Smoldering chair in living
SDC34 Smoldering chair in living
SDC37 Smoldering mattress in
2225 2225 2225 30 40 40
2585 2585 2585 0 5 5
-47 1367 1367 -43 9 9
27 27 3393 35 35 313
3233 3233 3233 -49 -49 -49
2707 2735 2735 -119 -119 -119
205 205 1011 165 165 351
-39 -39 2189 -39 -39 309
503 503 503 233 233 233
The results of Table 4 can be used to illustrate two main points: (1) in smoldering fires, photoelectric
detectors provide vastly more escape time than ionization detectors; and (2) ionization detectors often are
not providing even minimally sufficient ASET for the occupants. The NIST data (as opposed to the NIST
conclusions) generally agree with previous research [e.g., 31] where long-smoldering synthetic materials
were tested. However, some of the results of this study are confusing or misleading and were not
resolved despite specific notification :
(1) For various bedroom fires (SDC04, SDC06, SDC23, etc.), the ASET did not change when a
detector is located in the room of origin, as opposed to in the hallway. This is probably due to the
fact that NIST did not have a smoke detector located in the room itself.
(2) For some fires (e.g., SDC34), the difference between the photoelectric and the ionization
detectors in the burn room was only a few seconds, while the difference in the hallway right
outside the room was more than 1000 seconds. No other researcher studies [e.g., 32] found this
type of variance.
(3) For one living room fire (SDC23) the detectors on the 2nd floor landing responded several
thousand seconds before the detectors in the foyer on the 1st floor, directly outside the living
It is disturbing that the NFPA 72 Technical Committee  had a task group consider the RSET/ASET
concept in great detail, as propounded in the latest NIST study, and ended up wholeheartedly endorsing
both the RSET/ASET concept and the fallacious notion that photoelectric and ionization detectors are
equally suited for residential fire detection purposes. The Committee explicitly considered the various
factors involved in defining the ASET, but peremptorily accepted NIST’s unrealistic RSET values and
did not in any way consider the life safety of individuals who behave in a non-robotic fashion. It may be
noted that the only two fire service personnel assigned to the task group—individuals who would most
likely be most familiar with occupant behavior in residential fires—recorded their dissent from the
Summary and conclusions
Human behavior in fires is not mechanistic or robotic. It is common to find that individuals engage in
actions which are counterproductive, unsafe, or seemingly unreasonable. A robot could evacuate a house
in a very short time. Yet people encountering fires in their homes often behave much less efficiently and
become trapped in a fire. Consequently, engineering strategies which ignore these realities are flawed and
will necessarily give misleading conclusions.
The RSET/ASET concept was originally developed in 1975/76, even though it did not acquire the
terminology until somewhat later. Most of the early literature on evacuation was focused solely on fire
drills, and none of it involved single-family houses. During the 1970s, a few research studies began to be
done where single-family house occupants were interviewed. But these few studies focused on examining
the evacuation of occupants who had successfully evacuated, and not on determining the activities of
individuals who perished.
Since the majority of studies on human factors in fire historically focused on occupants who were
successful in escaping from fires, not ones who were unsuccessful and perished, this created the
misleading impression that designers can solely consider behaviors which are goal-oriented and ignore
behaviors which are counterproductive or otherwise detract from expeditious exiting. These priorities
should be reversed—to improve fire safety, it is much more important to study the failures than the
successes. This, of course, is more difficult since decedents cannot be interviewed nor can they fill out
questionnaires. But, in most cases, facts can be gleaned by interviewing firefighters, neighbors, and
The RSET/ASET concept is highly simplistic and offers no incentive for improvements in fire safety so
that more potential victims could be saved. It is a simplistically deterministic scheme improperly imposed
upon a stochastic reality. The time period required for individuals to escape from fire cannot sensibly be
expressed by a single number. Instead, there is a distribution. A fraction of persons encountering fire will
indeed respond as well as a robot and athletically and single-mindedly propel themselves outdoors. At the
other extreme are persons who would take forever, e.g., a bed-ridden invalid with no available rescuers.
But in between these two extremes is a very wide range of behaviors and required egress times. Any
rigidly-set criterion number, unless so large as to capture everything but the extreme tail of the
distribution, will unnecessarily sacrifice individuals who could otherwise be saved.
A stochastic distribution cannot be properly represented by a single value picked from the population. In
the RSET/ASET scheme, the situation is actually even worse, since the final results are presented as
categorical (i.e., yes/no) rather than quantitative. If RSET is assumed as = 100 s and there are two
alternative design choices, one giving ASET = 105 s and the second ASET = 1000 s, under the
RSET/ASET scheme these two designs are deemed identical, since in both cases ‘ASET > RSET’ is Yes.
Thus, two designs that are obviously exceedingly different in practice get treated incorrectly as identical.
It is not difficult to report the actual test result numbers and to allow a comparison to be made of
It is especially misleading when RSET/ASET analyses are reported using low RSET values. Canadian
researchers have shown that RSET = 11 min (using the definition that RSET starts at time of alarm) may
sometimes need to be considered even if individuals are healthy and not handicapped. It also bears
emphasizing that, in this research, they did not include counterproductive behaviors, which would greatly
increase this time period.
The RSET/ASET concept ignores that the same building + fire protection features may experience vastly
different RSET and ASET values, simply because a different fire scenario is used, indicating that these
variables have no true or unique value. A fire may occur when occupants are awake (typically a small
RSET), or when they are asleep (potentially a large RSET). If the victim is intimate with a flaming fire,
ASET might be zero, while with a smoldering fire, ASET may be 30 minutes or more. Even flaming fires
vary greatly in their characteristics. In an ultra-fast fire, untenable conditions might be reached in 60-90
seconds while in a moderately fast growing fire, untenable conditions might not occur for 6-9 minutes.
The consequence of using the RSET/ASET concept for fire safety engineering or product design purposes
is that fire deaths and injuries are permitted to occur which are preventable. The evaluation of all life
safety warning systems, including smoke detectors, should be based on the earliest possible warning of
the presence of a fire. The use of RSET/ASET analysis will not achieve this objective; therefore, such
analysis should not be used as a design methodology by design professionals, nor by detector
It is recommended that the RSET/ASET concept be abandoned and that egress analyses be properly
reported on a comparative ‘Margin of safety’ basis. This applies not just to design work, but also to
experimental research projects. The research required to obtain the data is invariably extensive and costly,
so it is inappropriate to take shortcuts and oversimplify the findings so that the benefits are lost and fire
safety is needlessly sacrificed. A ‘Margin of safety’ analysis constitutes a safety-conscious methodology
which aims to deliver the maximum escape time that can be physically and economically provided.
Standards and guidance documents which are based on the RSET/ASET concept should be revised to
provide adequate life safety for individuals who do not respond to fire circumstances in a robotic fashion.
This process is already starting to happen. Utilizing the ‘Margin of Safety’ concept and realistic
assumptions regarding occupant behavior, and fire scenarios several governmental bodies 
and fire safety organizations  are starting to espouse the use of photoelectric technology, as
opposed to ionization. It is hoped that the present paper stimulates this effort.
 Consumer Product Safety Commission Complaint filed by Joseph M. Fleming: Oxford, Ohio Fatality (April
 Consumer Product Safety Commission Complaint filed by Joseph M. Fleming: Riceboro, Georgia Fatality
(November 30, 2007).
 Consumer Product Safety Commission Complaint filed by Joseph M. Fleming: Lincoln, Nebraska Fatality
(December 26, 2006).
 Bukowski, R. W., Waterman, T. E., and Christian, W. J., Detector Sensitivity and Siting Requirement for
Dwellings (NBS-GCR-75-51), [U.S.] Natl. Bur. Stand., Gaithersburg MD (1975).
 Harpe, S. W., Waterman, T. E., Christian, W. J., Detector Sensitivity and Siting Requirements for Dwellings:
Phase 2 (NBS GCR 77-82), [U.S.] Natl. Bur. Stand., Gaithersburg MD (1977).
 Waterman, T. E., Detector Response vs. Available Escape Time in Residences, pp. 25-50 in Proc. Applications
of Fire Technology Workshop, Society of Fire Protection Engineers, Boston, MA (1983).
 Stahl, F. I., and Archea, J., An Assessment of the Technical Literature on Emergency Egress from Buildings
(NBSIR 77-1313), [U.S.] Natl. Bur. Stand., Gaithersburg MD (1977).
 Paulsen, R. L., Human Behavior and Fire Emergencies: An Annotated Bibliography (NBSIR 81-2438), [U.S.]
Natl. Bur. Stand., Gaithersburg MD (1981)
 Levin, B. M., EXITT—A Simulation Model of Occupant Decisions and Actions in Residential Fires: Users
Guide and Program Description (NBSIR 87-3591), [U.S.] Natl. Bur. Stand., Gaithersburg MD (1987).
 Cooper, L. Y., Concept for Estimating Available Safe Egress Time in Fires, Fire Safety J. 5, 135-144 (1983).
 Sime, J. D., Perceived Time Available: The Margin of Safety in Fires, pp. 561-570 in Fire Safety Science—
Proc. 1st Intl. Symp., Hemisphere, Washington DC (1986).
 Life Safety Code (NFPA 101), National Fire Protection Assn., Quincy MA (2009).
 National Fire Alarm Code (NFPA 21), National Fire Protection Assn., Quincy MA (2007).
 Human Behavior in Fire—Engineering Guide, Society of Fire Protection Engineers, Bethesda MD (2003).
 Hall, J. R. jr., How Many People Can Be Saved from Home Fires if Given More Time to Escape? Fire
Technology 40, 117-126 (2004).
 Gann, R. G., On the Use of Bench-Scale Smoke Toxicity Data in Fire Hazard and Risk Assessment, pp. 1421-
1429 in Interflam 2004, Interscience Communications Ltd., London (2004).
 Gann, R. G., International Study on the Sublethal Effects of Fire Smoke on Survivability and Health (SEFS),
pp. 427-439 in Fire Risk and Hazard Research Application Symp., Fire Protection Research Foundation,
Quincy MA (2003).
 Proulx, G., Cavan, N. R., and Tonikian, R., Egress Times from Single Family Houses (IRC-RR-209), National
Research Council Canada, Ottawa (2006).
 Fire Safety Engineering – Part 8: Life Safety – Occupant Behaviour, Location and Condition (ISO/TR 13387-
8), International Organization for Standardization, Geneva (1999).
 Life-threatening Components of Fire — Guidelines for the Estimation of Time Available for Escape Using Fire
Data (ISO CD 13571), International Organization for Standardization, Geneva (2007).
 Guidelines for Assessing the Fire Threat to People (ISO 19706), International Organization for
Standardization, Geneva (2007).
 Fire Safety Engineering: Evaluation of Behaviour and Movement of People (ISO 16738), International
Organization for Standardization, Geneva (to be published).
 Zhong, M., Shi, C., Tu, X., Fu, T., and He, L., Study of the Human Evacuation Simulation of Metro Fire Safety
Analysis in China, J. Loss Prevention in the Process Industries 21, 87-298 (2008).
 Bukowski, R. W., Peacock, R. D., Averill, J. D., Cleary, T. G., Bryner, N. P., Walton, W. D., Reneke, P. A.,
and Kuligowski, E. D., Performance of Home Smoke Alarms—Analysis of the Response of Several Available
Technologies in Residential Fire Settings (NIST Tech. Note 1455), Nat. Inst. Stand. and Technol.,
Gaithersburg MD (2003); revised and reissued in 2008 as Tech. Note 1455-1.
 Database of Full Scale Smoke Detector Performance Tests, Texas Engineering Experiment Station, Texas
A&M University, 2008
 Commerce’s NIST Reports Current Smoke Alarms Save Lives if Properly Used, NIST News Release, Nat.
Inst. Stand. and Technol., Gaithersburg MD (Feb. 26, 2004).
 Questions and Answers Clarifying Findings of NIST Home Smoke Alarm Study, Nat. Inst. Stand. and
Technol., Gaithersburg MD (2008).
 Fleming, J., Letter to Charles Dickerson, Acting Director, Unites State Fire Administration and James Hill,
Acting Deputy Director, National Institute of Standards and Technology (October 30, 2006).
 NIST analysis spreadsheet for Technical Note 1455-1, Nat. Inst. Stand. and Technol., Gaithersburg MD
 Fleming, J. M., Smoke Detector Technology and the Investigation of Fatal Fires, Fire & Arson Investigator
50:3, 35-40 (Apr. 2000).
 Schuchard, W. F., Smoldering Smoke, Fire J. 73:1, 27-29,73 (Jan. 1979).
 Su, J. Z., Crampton, G. P., Carpenter, D. W., McCartney, C., and Leroux, P., Kemano Fire Studies—Part 1.
Response of Residential Smoke Alarms (Res. Rep. 108), Natl. Res. Council, Ottawa (2003).
 NFPA 72 Technical Committee on Single- and Multiple-Station Alarms and Household Fire Alarm Systems,
Task Group on Smoke Detector Technology, Task Group Report, “Minimum Performance Requirements for
Smoke Alarm Detection Technology,” National Fire Protection Assn. (22 Feb. 2008).
 The State Massachusetts Building Code, 7th ed., Single- and Multiple-station Smoke Alarms, 780 CMR
907.2.10, effective 9/1/2008.
 An Act Relating to Requiring the Installation of Photoelectric-Only Smoke Detectors, Act of the General
Assembly, S.226, State of Vermont (2007-2008).
 Boston City Council, Petition for a Special Law re: Smoke Detectors in the City of Boston (15 Sept. 2007).
 Australasian Fire Authorities Council, Position on Smoke Alarms in Residential Accommodation (1 June
 International Association of Fire Fighters, Photoelectric Smoke Alarms – Resolution No. 15, Las Vegas
Convention (Aug. 2008).
FIRE AND MATERIALS
Eleventh international conference
San Francisco, California
26th – 28 th January