Mutations that result in segregating sites (polymorphic sites) in a sa
mple of DNA sequences can be classified into different types. A patter
n of segregating sites is an array of the numbers of various types of
mutations. Using an urn model, the probability of a pattern of segrega
ting sites can be expressed as a recurrence equation and its value can
be computed sequentially. Among those that can be computed by this me
thod are the probability of obtaining k external mutations (mutations
that occur in external branches of the genealogy of a sample), the pro
bability of obtaining k internal mutations (mutations that occur in in
ternal branches), the probability of obtaining k singletons (segregati
ng sites at which one of the two segregating nucleotides is present in
on ly one sequence), and the probability of obtaining k non-singleton
s. Two applications of the method are discussed. One is a maximum like
lihood estimation of theta and another is a Bayesian statistical test
of the hypothesis of neutral mutations. (C) 1998 Academic Press.