The contribution of social factors to seasonal plasticity in singing behavior and forebrain nuclei controlling song, and their interplay with gonadal steroid hormones are still poorly understood. In many songbird species, testosterone (T) enhances singing behavior but elevated plasma T concentrations are not absolutely required for singing to occur. Singing is generally produced either to defend a territory or to attract a mate and it is therefore not surprising that singing rate can be influenced by the sex and behavior of the social partner. We investigated, based on two independent experiments, the effect of the presence of a male or female partner on the rate of song produced by male canaries. In the first experiment, song rate was measured in dyads composed of one male and one female (M-F) or two males (M-M). Birds were implanted with T-filled Silastic capsules or with empty capsules as control. The number of complete song bouts produced by all males was recorded during 240 min on week 1, 2, 4, and 8 after implantation. On the day following each recording session, brains from approximately one-fourth of the birds were collected and the volumes of the song control nuclei HVC and RA were measured. T increased the singing rate and volume of HVC and RA but these effects were affected by the social context. Singing rates were higher in the M-M than in the M-F dyads. Also, in the M-M dyads a dominance-subordination relationship soon became established and dominant males sang at higher rates than subordinates in T-treated but not in control pairs. The differences in song production were not reflected in the size of the song control nuclei: HVC was larger in M-F than in M-M males and within the M-M dyads, no difference in HVC or RA size could be detected between dominant and subordinate males. At the individual level, the song rate with was positively correlated with RA and to a lower degree HVC volume, but this relationship was observed only in M-M dyads, specifically in dominant males. A second experiment, carried out with castrated males that were all treated with T and exposed either to another T-treated castrate or to an estradiol-implanted female, confirmed that song rate was higher in the M-M than in the M-F condition and that HVC volume was larger in heterosexual than in same-sex dyads. The effects of T on singing rate and on the volume of the song control nuclei are thus modulated by the social environment, including the presence/absence of a potential mate and dominance status among males.