We set up a model for the evolution of the galaxy luminosity function, taki
ng advantage of recent work that brought in some better understanding of th
e mass function for gravitationally condensed objects. We add to this a sim
ple model of star formation that reproduces the behaviour of the Tully-Fish
er relation, in order to attach a luminosity to a massive halo with a given
velocity dispersion. The physics of cooling of the gravitationally heated
baryonic component allows us to distinguish halos that become groups or clu
sters from those that eventually form galaxies (possibly within the former
objects). With our new mass function and our new application of the cooling
criteria - which motivated this paper - we get a satisfactory and natural
cutoff at the bright end of the luminosity function, the needed flat slope
for faint magnitudes and the correct trend in colors (brighter galaxies are
redder) within the framework of the hierarchical clustering picture. This
infirms earlier claims that the latter was inadequate to reproduce the form
er observations. We find the velocity dispersion to be a much better parame
ter than mass or radius to characterize galaxies. This model of the salient
features that may describe galaxies allows one to discuss galaxy evolution
as a function of redshift, in number as well as in luminosity. We find tha
t bright galaxies form at z similar to 2 from mergers with a rather quiet e
volution afterwards, whereas small galaxies are the result of a continuous
merging process active up to the present epoch. The transition is found to
occur at the observed transition between bright spirals and small dwarf ell
ipticals or irregulars. The galaxy luminosity was larger in a recent past f
or bright galaxies, as has been observed in the CFRS survey. This is becaus
e the mass of gas in a typical L* galaxy such as the Milky Way is a small f
raction of the total baryonic mass and thus star formation is already slowi
ng down. The evolution in number, which is quite well controlled in our mod
el, agrees reasonably well with the counts as a function of apparent magnit
ude. The quasar multiplicity as a function of redshift is also discussed.