The virulence of a suspension of particles, typically viruses or bacteria, can be assayed by administering several dilutions of the suspension into host organisms and observing the number of hosts that become infected at each dilution. The usual analysis uses likelihood methods to estimate a parameter that depends on both the density of particles in the suspension and on the chance that each particle will cause an infection. In the simplest model, each particle acts independently of the others and each gives rise to an infection with the same probability
p
. In these circumstances, the probability that a host becomes infected can be found using the Poisson distribution. More elaborate models allow
p
to vary from host to host according to some distribution. Methods for detecting such
host variability
are available. Some discussion is given to the effects of failure of the assumption of independence.