How do galaxies acquire their mass?
We introduce a toy model that describes (in a single equation) the mass in stars as a function of halo mass and redshift. Our model includes the suppression of gas accretion from gravitational shock heating and AGN jets mainly for M_halo > M_shock ~ 10^12 M_Sun and from a too hot IGM onto haloes with v_circ < 40 km/s, as well as stellar feedback that drives gas out of haloes mainly with v_circ < 120 km/s. We run our model on the merger trees of the haloes and subhaloes of a high-resolution dark matter cosmological simulation. The galaxy mass is taken as the maximum between the mass given by the model and the sum of the masses of its progenitors (reduced by tidal stripping). Designed to reproduce the present-day stellar mass function of galaxies, our model matches fairly well the evolution of the cosmic stellar density. It leads to the same z=0 relation between central galaxy stellar and halo mass as the one found by abundance matching and also as that previously measured at high mass on SDSS centrals. Our model also predicts a bimodal distribution (centrals and satellites) of stellar masses for given halo mass, in good agreement with SDSS observations. The relative importance of mergers depends much more on stellar than halo mass. Galaxies with m_stars > 10^11 M_Sun/h acquire most of their mass through mergers (mostly major and gas-poor), as expected from our model's shutdown of gas accretion at high M_halo. However, mergers are rare for m_stars < 10^11 M_Sun/h (greater than our mass resolution), a consequence of the curvature of the stellar vs. halo mass relation. So gas accretion must be the dominant growth mechanism for intermediate and low mass galaxies, e.g. dwarf ellipticals in clusters, except that gas-rich galaxy mergers account for the bulk of the growth of ellipticals with m_stars ~ 10^10.5 M_Sun/h, which we predict must be the typical mass of ULIRGs.