An associative memory model HASP that consists of hetero-associative and auto-associative networks, can treat one-to-many associations. By using a simple model, which seems to capture characteristics of the HASP, we analyzed its recalling process. In order to treat one-to-many associations, we introduced an external input to break the symmetry between the associative vectors. We analyzed how the recalling process depends on to which one of the two subnetworks the external input is injected. From the theoretical analysis and simulation studies, it has shown that the associative vector can be recalled by injecting the external input into the auto-associative networks better than by injecting it into hetero-associative networks.