In 2013 Todd Rose gave a TEDx talk that opened with a story about the Air Force:

It’s 1952, and the Air Force has a problem. They’ve got good pilots flying better planes, but they’re getting worse results. And they don’t know why. For a while, they blamed the pilots. They even blamed the technology. They eventually got around to blaming the flight instructors. But it turned out that the problem was actually with the cockpit. Let me explain.

Imagine you’re a fighter pilot. You’re operating a machine that in some cases can travel faster than the speed of sound, and where issues between success and failure, sometimes life and death, can be measured in split seconds. If you’re a fighter pilot, you know that your performance depends fundamentally on the fit between you and your cockpit. Because after all, what good is the best technology in the world, if you can’t reach the critical instruments when you need them the most? But this presents a challenge for the Air Force. Because obviously, pilots are not the same size. So, the issue is: how do you design one cockpit that can fit the most individuals?

For a long time, it was assumed that you could do this by designing for the average pilot. That almost seems intuitively right. If you design something that’s fit for the average sized person, wouldn’t it fit most people? It seems right but it’s actually wrong. And 60 years ago, an Air Force researcher, Gilbert Daniels, proved to the world just how wrong this really is, and what it was costing us. He studied over 4,000 pilots and he measured them on ten dimensions of size, and he asked a very simple question: how many of these pilots are average on all ten dimensions? The assumption was that most of them would be.

Do you know how many really were? Zero. The Myth of Average

Rose goes on to claim that Daniels’ study begat adjustable seats and controls in fighter planes. He even coined the extremely quotable line “if you design a cockpit to fit the average, you’ve literally designed it for nobody”. Absolutely top tier content.

The problem? Its most accurate claims are oversimplied, and the least accurate are fabrications¹. Adjustable seats were common in military aircraft before the end of WWII, there was broad interest in their continued improvement, and there was no large-scale body measurement study designed in response to planes crashing more often².

But there was a study that measured 4063 pilots. It covered 132 dimensions (not 10), and did absolutely no editorializing. Daniels did subsequently publish a tech note that showed the entire population could be ruled “not average” using as few as 10 measurements, but they had to be carefully chosen.

Luckily the primary sources have been declassified and are publicly available, so let’s look at them.

The Study

Searching for Air Force studies published by G. S. Daniels quickly turns up an obvious match, which can introduce itself:

Body size data for 132 measurements of over 4,000 Air Force flying personnel are presented. Organization of the survey is briefly discussed and the techniques of measurement are illustrated by photographs for the benefit of other anthropologists. Both diametral and surface measurements are included. Dimensions are given in both centimeters and inches.

A description of the statistics and an explanation of their use are given with some discussion of certain statistical shortcuts employed in the reduction of the data. The tabulations include range, mean, standard deviation, coefficient of variation, and twenty-five selected values from the first to the ninety-ninth percentile. Means and standard deviation values for each dimension are also given for nine subgroups based on flight duties.

A Glossary and Bibliography are included.

These data are presented for use by the designers of aircraft, clothing and equipment. Hertzberg & Daniels, 1952, p. iii

This is obviously the right study, but it stays strictly on-topic: statistical data for each measurement, with certain cohorts broken out. It also makes clear that cockpit design was not its prime motivator:

Many persons asked why was this new survey necessary? … The main reason lay in the immediate need for entirely new information about bodily dimensions. Rapid developments in aircraft had resulted in physiological stresses on crewmen undreamed of only ten years before. To counteract these stresses a skin-tight garment had to be worn. The loosely-fitted flying clothing of World War II had been adequately designed by the use of conventional body diameters and segment lengths taken at that time—stature, shoulder breadth, chest depth, sitting height, arm length, etc.—but obviously a skin-tight garment would demand something further. Hertzberg & Daniels, 1952, p. 1

To editorialize: what the author calls “skin-tight garments” would probably be referred to as a “compression suit” today. The first military jets arrived in 1942 and represented a step function increase in aircraft capabilities. Suddenly, the system’s³ performance was limited by the pilot’s ability to endure g-forces instead of the plane’s ability to generate them.

Adjustable Seats

One thing the study did not do was lead to the introduction of adjustable seats. In fact, the Air Force even got some help from the auto industry:

In order to study the fundamentals of comfortable cockpit seating, there was made available to the Air Technical Service Command, through the courtesy of the Murray Corporation of America, the so-called Universal Test Seat, which is a piece of laboratory apparatus designed by a group of automotive engineers at the University of Michigan for studying seating requirements in automobiles. The original was loaned this Command by the Murray Corporation, and a copy was made of it, incorporating certain modifications which were adapted to the study of seating in aircraft. Randall et al., 1946, p. 79

Fighters may have had adjustable seats well before the study, but that doesn’t mean they were good:

It has been the common practise in fighter type aircraft to install the seat on slide tubes which lie at or about the 13-1/2 degrees from the vertical. It has been found geometrically that a seat mounted in this manner and being raised from its neutral position 3-1/2” will be moved posteriorly .9 inch, and also will be moved anteriorly if lowered the same amount. This is completely contrary to the proper utilization of the variables in the sizes of men. A seat is raised to permit a shorter man to maintain visual requirements and certainly he should not be moved aft from the rudder pedals: the same holds for the accommodation of the taller man in the reverse direction. Randall et al., 1946, p. 83

A well-intentioned focus on “reference points” in cockpit design not only allowed manufacturers to ignore some important facts about human physiology, but also the variations in space requirements created by the various pieces of equipment commonly used by pilots:

However, there has been little regard paid to the construction and use of items of personal equipment with which the pilot must operationally be concerned, and the functional end product is a great variety of levels at which the pilot is held seated on his personal equipment, all of which, except for the simple seat-type parachute and pack pad, will contribute more or less to a mal-function of the pilot and his cockpit in flight. … If for some reason or other the pilot is forced to use the seat type parachute, the one-man life raft, and a cushion, he may be raised as much as 5” above the original reference point. Randall et al., 1946, p. 82

At the beginning of the war, pilots had a height restriction of 5’4” to 5’8”. Combined with ⁺⁵⁄₋₀ inches of variability introduced by their equipment, it’s not surprising that adjustability had already become a well-established requirement by the jet age.

But what about the increase in incidents and accidents? While I didn’t come across any evidence of an increase associated with jet engines⁴, I did stumble across another reasonable explanation:

For some period the fighter stature was held at 5’8” instead of the 5’10” recommended by Armstrong. This acceptance limit was adequate so long as peacetime requirements remained. However, with the advent of stepped-up military requirements in 1942, such a large number of men was required for pilot training that a 5’8” limit actually prevented full use of the potentials available. The greatest defect which appeared in this regard was due to the fact that the fighter-type aircraft available for military use at that time had been designed around the 5’8” [limit] and, without due regard to this fact, the limits were stepped up to 5’10” again, irrespective of the abilities of the planes to accommodate these higher statures.

This situation would not have been too disastrous had the original design requirements remained in use. That is to say, that these aircraft had been designed to fly not more than 3 and 1/2 hours. However, it is easily recognizable that this situation did not remain, inasmuch as long range requirements entered in and wing tanks and belly tanks were added to these same aircraft to enable them to fly as much as seven to eleven hours. There could be no modifications of the cockpit to provide any comfortable conditions for the pilots of the large stature who would be trained to fly these planes. This situation subsequently developed into probably the most difficult problem from the human operational standpoint encountered in World War II. Randall et al., 1946, p. 5

By the end of WWII the height limit for fighter pilots had expanded to at least 6’. Sending pilots four inches outside of their aircraft’s design requirements on missions two to three times longer than their aircraft’s design requirements obviously invites an increase in incidents and accidents, but none of that was a surprise, and solving it was already very much a priority.

TODO

But Scott, what about this (from 99pi)?

This worked more or less up until World War II, when the Army began recruiting hundreds of new pilots to expand its air forces (which became a separate branch of the military in 1947). But with the birth and expansion of the Air Force came a decline in performance and a rash of deaths. Even with no war, pilots continued to die during training, as they were unable to control their planes.

The Average Man

So the study was intended to aid to cockpit design, and probably helped improve adjustable seats, but what about that “myth of average”? Well, this was an overconstrained outcome for a study so proud of measuring more human features than any that had come before:

The authors believe this to be the greatest number of measurements ever taken on so large a series of subjects. Although most of the measurements are considered conventional by physical anthropologists, approximately 35 are believed to be entirely new. Hertzberg & Daniels, 1952, p. 1

Even accounting for correlations⁵ and ignoring “jagged” profiles, statistical theory predicts that every additional measurement will reduce the number of people who fall within a standard deviation of the mean. Daniels breaks this down nicely in his tech note:

of the original 4063 men
    1055 were of approximately average stature

of these 1055 men
    302 were also of approximately average chest circumference

of these 302 men
    143 were also of approximately average sleeve length

of these 143 men
    73 were also of approximately average crotch height

of these 73 men
    28 were also of approximately average torso circumference

of these 28 men
    12 were also of approximately average hip circumference

of these 12 men
    6 were also of approximately average neck circumference

of these 6 men
    3 were also of approximately average waist circumference

of these 3 men
    2 were also of approximately average thigh circumference

of these 2 men
    0 were also of approximately average crotch length Daniels, 1953, p. 3

Being able to disqualify an entire population by using less than 8% of a set of highly correlated metrics is impressive,but Daniels explicitly selected these ten measurements to minimize the number needed to get to zero, which is not the same as saying “nobody fell within the average across ten measurements”. In fact, the most average man⁶ fell outside of one standard deviation in only 38% of studied measurements, so a cockpit designed for an “average man” across 81 carefully chosen measurements would have actually fit someone!

Regardless, this is a super useful lesson for anyone looking at multivariate systems and the primary sources are well worth a look for anyone who finds themselves in that space on a regular basis.

The lesson for the rest of us? Stay skeptical of secondary sources⁷.

References

Hertzberg, H. T. E. and G. S. Daniels. Air Force Anthropometry in 1950. American Journal of Physical Anthropology, n.s. V. 10, No. 2. 1952.
Randall, F. E., A. Damon, R. S. Benton, and D. I. Patt. Human Body Size in Military Aircraft and Personal Equipment. AAF Technical Report No. 5501, Wright Field, 1946.
Daniels, G. S. The "Average Man"?. Technical Note WCRD 53-7, Wright Air Development Center, Wright-Patterson Air Force Base, Ohio, 1953.

Which is too bad because, while less dramatic, the truth would have served the thesis just as well ↩
The video says “good pilots flying better planes, but they’re getting worse results”, but an excerpt of Rose’s book by the Toronto Star claims more spefically that the problem was an increase in incidents and accidents (both technical terms) that he attributes to the transition to jets. ↩
The Army Air Force obviously considered the pilot and plane as a system by this time and seemed rather cranky that aircraft manufacturers were not keeping up.

It should be quite apparent that the operational behaviour of an airplane of unlimited potentialities is actually no better than the behaviour characteristics imposed upon it by the physiological capabilities of the human being involved. It has been the experience of the Army Air Forces during the progress of World War II that many problems relating to inefficiencies on the part of the flight personnel could have been eliminated had the designers of the planes been fully cognizant of some of the implications of human biology. Randall et al., 1946, p. 4

↩
To be fair, I didn’t specifically look for any ↩
Taller people tend to be heavier, for example ↩
One of the two men who made it to step 10 ↩
Regrettably, one of the pop sci books I enjoyed most in 2019 also turned out to be oversimplified at the best of times ↩