Results from a second empirical test of a theory of intergenerational family relations, as an example of empirically driven theory construction in social gerontology, are presented. The theory we tested reflects an attempt to explain later-life patterns of affection, association, and consensus between parents and children as interdependent components of a meta-construct--family intergenerational solidarity. The theoretical model also identifies variables that are expected to predict variations in levels of solidarity. A recent empirical test (Atkinson, Kivett, and Campbell, 1986) offered little support for the central proposition of the model: that solidarity can be indexed by a linear-additive composite of affection, association, and consensus. The replication analysis reported here is based on a different sample, as well as alternate measures and statistical methods. The convergent results suggest that (a) later-life intergenerational solidarity is not a unidimensional construct, and (b) different variables predict each component. This exercise of theory statement, testing, and retesting leads us to suggest several refinements to the theory acknowledging a more complex relationship between dimensions of intergenerational relations in old age.