New airplanes have to confirm that all passengers have to be evacuated within 90 seconds or they wouldn't be certified by aviation authorities. But where does the 90 seconds limit come from? Why isn't it 60 or 120 seconds? Is it based on a series of studies or is it just a nice round number that was chosen based on a rule of thumb?
Early research on post-crash fires focused on their causes and development, but soon it became obvious that the best way of reducing fatalities from post-crash fires lay in speedy evacuation of crew and passengers. Quoted from DOT/FAA/AR-95/84:
In April of 1964, the FAA crash-tested a Douglas DC7 transport aircraft to examine the postcrash causes of fatalities. This test was followed by a crash test of a Lockheed L1649 transport in September.
Here is the link to a 1964 FAA documentary movie on the DC-7 test, and this links to a short excerpt from a documentation on the L-1649 test. Actually, the test was performed by the Flight Safety Foundation in Phoenix, AZ, under contract from the FAA.
During the same period the FAA initiated research to prevent two types of postcrash fires. […] The second is a "flash over" condition which occurs when the cabin interior reaches a high enough temperature that the entire inside instantly ignites. […] Much of the FAA's Research and Development (R&D) has been dedicated to gaining the time by preventing or inhibiting the fires and by faster evacuation procedures.
It was observed that a structurally sound cabin remained habitable with the fuel fire raging outside until at about two minutes into the fire, when heat in the cabin became so intense that a flashover condition developed. This flashover was seen as the limit to the time available for evacuation.
Amendment 121-2, issued on March 3, 1965 and titled "Regulations, Procedures, and Equipment for Passenger Emergency Evacuation; Flight Attendants; and Assignment of Emergency Evacuation Functions for Crew Members" mandated rules requiring that passenger evacuation had to be accomplished in two minutes, using only half the available exits. It also marked the start to mandatory pre-flight briefings on emergency procedures. It in turn was based on the 1963 FAA notice of proposed rulemaking NPRM 63-42.
NPRM 63-42, which preceded the aforementioned tests by one year, proposed an evacuation limit of two minutes and the use of ropes to speed up evacuation:
In determining an equitable time factor applicable to all accidents requiring emergency evacuation of the airplane, the Agency recognizes there are many variables which can have a direct effect on the success of the evacuation. However, in order to assist the air carriers in having a means by which they can establish and subsequently measure their ability to execute the evacuation procedures, the Agency proposes that a time factor of 2 minutes to evacuate the airplane be specified
The evaluation of the 1964 tests and later research led the FAA to shorten the evacuation time to 90 seconds. Also, it was found out that ropes actually delay evacuation and slides were mandated to replace them. DOT/FAA/AR-95/84 again:
In 1967, the FAA initiated research to develop safety fuels to reduce postcrash fires and other requirements that addressed improved interior lighting, evacuation slides deployable in 10 seconds or less, improved exit distribution and excess exits, self-extinguishing interior materials, and protection of fuel and electric lines. These changes were coupled with a reduction of the required time for evacuation from 2 minutes to 90 seconds.
And that has stayed the same ever since.
90 seconds comes from a safety aspect, specifically a fire safety aspect
Aviation Safety Study SA9501 Footnote 3 says:
Internationally, industry accepts 90 seconds as a reasonable estimate of the survivable time in an evacuation where fire is present.
Which means that it is predicted than a burning aircraft is not sustainable for life beyond that period, but interestingly this is not the condition that the aircraft is certified under.
FAA Chapter 77 Section 1.7 says:
A full-scale emergency evacuation demonstration simulates an aborted takeoff. The operator must show that the aircraft, emergency equipment, and emergency procedures allow the evacuation of the aircraft at full seating capacity, including crewmembers, in 90 seconds or less.
So the 90 second rule comes from a fire safety aspect - as humans are not expected to survive in those conditions for longer than that; but the aircraft is only tested on whether it can evacuate on an aborted takeoff (not specifying that a simulated fire must be the cause of this aborted takeoff).
It comes from previous accidents and improvements in auto-deploying/inflating slides.
One airline wanted the FAA to impose a 90-second limit in the 60s, but the FAA opted for a two-minute limit based on previous accidents. But in 1966 the limit was changed to 90 seconds with improvements in auto-deploying slides.
One air carrier proposed a 90-second maximum time period for the demonstration of emergency evacuation of passengers in a survivable accident. The Agency has considered the relative speed in which fires have developed in accidents, and the practical limitations imposed by existing aircraft configurations, and has concluded that the two-minute maximum time period is reasonable for the aborted takeoff and gear-up crash landing evacuation demonstration. No maximum time period has been provided for the ditching demonstration, since experience shows that passengers generally are alerted prior to, and fire rarely occurs in, actual ditching.
CFR Final Rule, Docket No. 2033, March 9, 1965
Down to 90 seconds
(...) the Agency now proposes to require emergency evacuation demonstrations under both Parts 25 and 121 to be conducted within 90 seconds. The decrease of 30 seconds from the present requirement of Sec. 121.291 is made possible by equipment advances (primarily the improved automatically deployed and inflated slides) that have occurred since that standard was adopted.
Notice No. 66-26, 1966
I'm still looking for the accidents the FAA took as examples, but I'm having no luck. At least now we know the 90 seconds come from improvements in slides.