Two point lod scores are widely used in pedigree analysis as they prov
ide a fast and efficient method of establishing linkage. Groups of mar
kers that lie in close proximity to one another can be formed by admit
ting any locus that is linked to at least one existing member of the g
roup with lod score greater than some predetermined value. It seems na
tural to extend this technique to Radiation Hybrid Mapping both for co
nstructing groups of tightly linked loci that may then be analysed usi
ng more powerful statistics and as a method of ordering in its own rig
ht. A general extension of two point analysis is derived and the probl
ems associated with radiation hybrid data are discussed. In particular
, the additional parameters representing the probabilities of differen
t fragments being retained (which have no parallel in classical linkag
e analysis) lead to a range of estimators of the breakage probability,
theta, which have equal and maximal likelihood. Ways of circumventing
this problem are discussed along with the potential errors they intro
duce. Importantly the ambiguity in estimation of theta is not carried
through to the lod score as this depends only on the maximum value of
the likelihood and not on the particular value of theta at which it oc
curs. Thus even though two point analysis fails to provide robust esti
mates of either breakage probabilities or the distance between loci, i
t represents a simple and effective method of constructing linkage gro
ups that may be analysed with more powerful statistical methods. This
is particularly important given the number of microsatellites, ESTs an
d candidate genes currently being typed on radiation hybrids.