In the heterothallic ascomycete Podospora anserina, the mating-type locus is occupied by two mutually exclusive sequences termed mat+ and mat-. The mat+ sequence contains only one gene, FPR1, while the mat- sequence contains three genes: FMR1, SMR1 and SMR2. Previous studies have demonstrated that FPR1 and FMR1 are required for fertilization. Further analyses have led to the hypothesis that mat+ and mat- genes establish a mat+ and mat- nuclear identity, allowing recognition between nuclei of opposite mating type within the syncytial cells formed after fertilization. This hypothesis was based on the phenotypes of strains bearing mutations in ectopic mat genes. Here we present an analysis of mutations in resident mat- genes which suggests that, unlike FMR1 and SMR2, SMR1 is not involved in establishing nuclear identity. In fact, mutations in these two genes impair nuclear recognition, leading to uniparental progeny, while mutations in SMR1 block the sexual process, probably at a step after nuclear recognition. The nuclear identity hypothesis has also been tested through internuclear complementation tests. In these experiments, the mat- mutants were crossed with a mat+ strain carrying the wild-type mat- genes. Our rationale was that internuclear complementation should not be possible for nuclear identity genes: the relevant genes should show nucleus-restricted expression, and diffusion of their products to other nuclei should not occur. This test confirmed that SMR1 is not a bona fide mat gene since it can fulfill its function whatever its location, in either a mat- or a mat+ nucleus, and even when present in both nuclei. SMR2, but not FMR1, behaves like a nuclear identity gene with respect to internuclear complementation tests. A model is proposed that tentatively explains the ambiguous behaviour of the FMR1 gene and clarifies the respective functions of the three mat- proteins.