101 
It is currently impossible to calculate accurately the overall 
probabilities for most experiments. But crude estimates can often 
be made for the specific events. The probability of event (b) can be 
estimated from the nature of the physical containment practices and 
procedures and from specific knowledge of accidental release due to 
failure of equipment or technique. An idea of the probability of events 
(c) and (d) may be obtained from the known properties of the host- 
vector system in the particular experiment. Some estimate of event 
(e) may be made from the known properties of the host cell. For events 
(f) the details of the particular experiment--for example, whether it 
is a shotgun experiment or not --will be helpful in obtaining a range 
of probabilities. As more data are accumulated, some idea of the 
probabilities of events (g) and (h) will be possible. Again the details 
of the particular experiment, and the relation between the foreign 
DNA and the new host, will be relevant. (See Section IV-C-l-a of this 
EIS for a discussion of expression of foreign gene in recipient host 
cells. ) It may also be possible to make some estimate of the probability 
of event (i) in any given case. Estimates of the likelihood of event 
(j) will be among the most difficult, given the complex and poorly under- 
stood nature of what makes any agent a producer of disease. Useful 
though approximate estimates of events (k), (1), and (m) can be obtained 
from prior knowledge of the properties of the vector, including its 
transmissibility, stability, and ability to enter other cells and establish 
itself therein. (See Section VI-C-3-b of the EIS and Section III-B-1 
of the Guidelines. ) 
It is not necessary to know the probability of each event to make 
useful estimates. If the probability of a few of the necessary events 
is known with some confidence, useful numbers can be obtained. In 
such a case, even if the probability of all the other events is assumed 
to be high --even equal to 1 (that is, a certainty) --an upper limit of 
the overall probability can be obtained. It must be recognized that 
the description of an overall probability as very low, or low, or high 
is relative and depends on how many organisms are present. Thus, 
the overall probability of a detrimental effect will vary directly with 
the number of organisms prepared. 
The only article of which we are aware that has been published 
in the scientific literature attempting to assess numerically the 
probability of a man-made epidemic resulting from recombinant DNA 
research was written by Dr. Robin Holliday, FRS, Head of the Division 
of Genetics, National Institute of Medical Research, London (16). 
Excerpts from Dr. Holliday's article appear as Appendix P. 
Virologists of the National Institute of Allergy and Infectious 
Diseases are preparing to conduct experiments at the Frederick 
Cancer Research Center, Frederick, Maryland, which should provide 
data to help determine if recombinant DNA research could result in 
real hazards. 
