The two-point angular correlation function of galaxies, omega(theta), has been computed on equatorial fields observed with the Cerro Telolo Inter-American Observatory 4 m prime focus, within a total area of 2,31 deg(2). In the magnitude range 19 less than or equal to m(R) less than or equal to 21.5, corresponding to (z) approximate to 0.35, we find an excess of power in omega(theta) at scales 2 '' less than or equal to theta less than or equal to 6 '' over what would be expected from an extrapolation of omega(theta) measured at larger theta. The significance of this excess is approximate to 5 sigma. At larger scales, 6 '' < theta less than or equal to 24 '' the amplitude of omega(theta) is 1.6 times smaller than the standard nonevolutionary model. At these scales there is remarkable agreement between the present data and Infante & Pritchet (1995). At large angular scales (6 '' < theta less than or equal to 24 '') the data are best described by a model where clustering evolution in xi(r, z) has taken place. Strong luminosity evolution cannot be ruled out with the present data. At smaller scales, 2 '' less than or equal to theta less than or equal to 6 '', our data are formally fitted by models where epsilon = -2.4(Omega = 0.2, r(0) = 5.1 h(-1) Mpc) or r(0) 7.3 h(-1) Mpc (Omega = 0.2, epsilon = 0). If the mean redshift of our sample is 0.35, then our data show a clear detection of the scale (approximate to 19 h(-1) kpc) where the clustering evolution approaches a highly nonlinear regime, i.e., epsilon less than or equal to 0. The rate at which galaxies merge has been computed. If this rate is proportional to (1 + z)(m), then m = 2.2 +/- 0.5.