Dr. Donald S. Fredrickson 
May 16, 1978 
Page 2 
This type of question immediately suggests that a stochastic model of escape, 
establishment, and harmful consequences ought to be considered. It also 
begins to suggest that a series of uncertain events will have associated proba- 
bility distribution of occurrences, and that someone is going to have to decide 
what an acceptable value of P is. 
An alternative question is: 
2. What set of guidelines can be developed with sufficient flexibility 
to encompass the fast-changing field of recombinant DNA? 
This is a difficult question to evaluate, perhaps even more so than question 
1, because there are no operational definitions for a number of words used 
in the question. First one would have to seek a resolution on terms like "suffi- 
cient," and also discover the parameters which describe the characteristics 
which are changing so rapidly. 
Another question might be: 
3. What is the escape risk from PI through P4 facilities? 
Dr. Roy Curtiss ( Annual Review of Microbiology , 30 : 507-333, 1976) attempts 
to answer this question, but the basis for his judgments is unclear. 
Fault tree analysis, a techniques applied to systems, such as missies 
and airplanes, is an excellent quantitative tool for assessing the likelihood 
of failure modes of components which might lead to system failure. In this 
case, the system failure is "organism escape" and component failures might 
include fans and filters in air purification systems, or glove box leaks. 
But a broader tool which would be applicable for questions 1 or 2 is 
decision analysis. It attempts to apply logical, mathematical and scientific 
procedures to decision problems which have the following characteristics: 
uniqueness, importance, uncertainty, long run implications and complex 
preferences. All of these characteristics appear in the recombinant DNA 
issue. Decision analysis is a systematic procedure to identify and define 
the decision, the decision-maker, the criteria for a "good" decision, and 
to evaluate the alternatives. It has been used in the past on such diverse 
problems as the decisions: 1) to seed hurricanes, 2) to provide adequate reserve 
for electric utilities, 3) to make corporate resource allocations, and 4) to 
plan wildfire protection in a mountain range. 
The basic point is that there are several evaluation tools available 
which have not been applied to the recombinant DNA issue, but which if applied 
would systematize, discover and evaluate the important decision variables 
in the problem. Once the controlling variables are identified, and even if 
[Appendix A — 317] 
