Measures of association of genes at different loci (linkage disequilibrium) are widely used to
determine whether the structure of natural populations is clonal or not, to map genes from
population data, or to test for the homogeneity of response of molecular markers to background
selection, for example. However, the usual definitions of parameters for gametic associations may
not be suitable for all these purposes. In this paper, we derive the recursion equations for one- and
two-locus identity probabilities in an infinite island model. We study the role of drift, gene flow,
partial selfing and mutation model on the expected association of genes across loci. We define the
‘within-subpopulation identity disequilibrium’ as the difference between the joint two-locus
probability of identity in state and the expected product of one-locus identity probabilities. We
evaluate this parameter as a function of recombination rate, effective size, gene flow and selfing
rate. Within-subpopulation identity disequilibrium attains maximum values for intermediate
immigration rates, whatever the selfing rate. Moreover, identity disequilibrium may be very small,
even for high selfing rates. We discuss the implications of these findings for the analysis of data
from natural populations.