Hall was still frowning. "But why me?"

"Because you are single. We have to have one unmarried man."

Stone opened a briefcase and withdrew a file. He gave it to Hall. "Read that."

It was a Wildfire file.

"Page 255," Stone said.

Hall turned to it.

PROJECT: WILDFIRE

ALTERATIONS

1. Millipore® Filters, insertion into ventilatory system. Initial spec filters unilayer styrilene, with maximal efficiency of 97.4% trapping. Replaced in 1966 when Upjohn developed filters capable of trapping organisms of size up to one micron. Trapping at 90% efficiency per leaf, causing triple-layered membrance to give results of 99.9%. Infective ratio of .1% remainder too low to be harmful. Cost factor of four- or five-layered membrance removing all but .001% considered prohibitive for added gain. Tolerance parameter of 1/1,000 considered sufficient. Installation completed 8/12/66.

2. Atomic Self-Destruct Device, change in detonator close-gap timers. See AEC/Def file 77-12-0918.

3. Atomic Self-Destruct Device, revision of core maintenance schedules for K technicians, see AEC/Warburg file 77-14-0004.

4. Atomic Self-Destruct Device, final command decision change. See AEC/Def file 77-14-0023. SUMMARY APPENDED.

SUMMARY OF ODD MAN HYPOTHESIS: First tested as null hypothesis by Wildfire advisory committee. Grew out of tests conducted by USAF (NORAD) to determine reliability of commanders in making life/death decisions. Tests involved decisions in ten scenario contexts, with prestructured alternatives drawn up by Walter Reed Psychiatric Division, after n-order test analysis by biostatistics unit, NIH, Bethesda.

Test given to SAC pilots and groundcrews, NORAD workers, and others involved in decision-making or positive-action capacity. Ten scenarios drawn up by Hudson Institute; subjects required—to make YES/NO decision in each case. Decisions always involved thermonuclear or chem-biol destruction of enemy targets.

Data on 7420 subjects tested by H¹H² program for multifactorial analysis of variance; later test by ANOVAR program; final discrimination by CLASSIF program. NIH biostat summarizes this program as follows:

It is the object of this program to determine the effectiveness of assigning individuals to distinct groups on the basis of scores which can be quantified. The program produces group contours and probability of classification for individuals as a control of data.

Program prints: mean scores for groups, contour confidence limits, and scores of individual test subjects.

K.G. Borgrand, Ph.D. NIH

RESULTS OF ODD MAN STUDY: The study concluded that married individuals performed differently from single individuals on several parameters of the test. Hudson Institute provided mean answers, i.e. theoretical "right" decisions, made by computer on basis of data given in scenario. Conformance of study groups to these right answers produced an index of effectiveness, a measure of the extent to which correct decisions were made.

The data indicate that married men choose the correct decision only once in three times, while single men choose correctly four out of five times. The group of single males was then broken down further, in search of highly accurate subgroups within that classification.