CONTEXT: Manganese is an iron-peak element and although the nucleosynthesis path that leads to its formation is fairly well understood, it remains unclear which objects, SN II and/or SN Ia, that contribute the majority of Mn to the interstellar medium. It also remains unclear to which extent the supernovae Mn yields depend on the metallicity of the progenitor star or not. AIMS: By using a well studied and well defined sample of 95 dwarf stars we aim at further constraining the formation site(s) of Mn. METHODS: We derive Mn abundances through spectral synthesis of four Mn I lines at 539.4, 549.2, 601.3, and 601.6 nm. Stellar parameters and data for oxygen are taken from Bensby et al. (2003, 2004, 2005). RESULTS: When comparing our Mn abundances with O abundances for the same stars we find that the abundance trends in the stars with kinematics of the thick disk can be explained by metallicity dependent yields from SN II. We go on and combine our data for dwarf stars in the disks with data for dwarf and giant stars in the metal-poor thick disk and halo from the literature. We find that dwarf and giant stars show the same trends, which indicates that neither non-LTE nor evolutionary effects are a major concern for Mn. Furthermore, the [Mn/O] vs [O/H] trend in the halo is flat. CONCLUSIONS: We conclude that the simplest interpretation of our data is that Mn is most likely produced in SN II and that the Mn yields for such SNae must be metallicity dependent. Contribution from SN Ia in the metal-rich thin disk can not, however, be excluded.