We have used the projected two-point correlation function, $w(r_p)$, to investigate the spatial distribution of the 591 galaxies with secure redshifts between $0 \leq z \leq 1.3$ in the five CFRS fields. The slope of the two-point correlation function for the sample as a whole is $\gamma=1.64\pm0.05$, very similar to the local slope, and $\gamma$ is therefore not strongly evolving with redshift. However, the amplitude of the correlation function decreases strongly with increasing redshift, so that at $z\approx0.6$ it is a factor of 10 lower (for $q_0=0.5$) than for a similarly-selected local galaxy population, on scales $0.1<r<2h^{-1}$ Mpc ($q_0=0.5$). As a whole, the CFRS data is adequately represented by $r_0(z=0.53)=1.33 \pm 0.09 h^{-1}$Mpc for $q_0=0.5$, and $r_0(z=0.53)=1.57 \pm 0.09 h^{-1}$Mpc for $q_0=0$. Unless the galaxy population at high redshift is quite different from any population seen locally, an unlikely possibility, then this implies growth of clustering as described by the evolutionary parameter $\epsilon$ to be between $0 < \epsilon < +2$. No difference in the clustering of red and blue galaxies is seen at $z \geq 0.5$, although at lower redshifts, $0.2 \leq z \leq 0.5$, blue galaxies are somewhat less strongly correlated than the redder galaxies, as seen in local samples. This effect could be the equivalent for field galaxies to the Butcher-Oemler effect seen in clusters of galaxies. The cross-correlation functions between red and blue samples have comparable amplitudes to the auto-correlation functions of each. The distribution and power spectrum of pair separations does not indicate