The luminosity dependence of clustering in galaxy distributions is analysed using the CfA redshift catalogue, where the magnitude correction for both Galactic and internal extinction is taken into account. It is found that, contrary to previous results, the luminosity dependence of clustering is very weak when the magnitude correction is made, showing no evidence of a positive dependence (luminosity segregation) in the magnitude ranges of interest. The results are not consistent with the luminosity segregation predicted in standard biased cold dark matter scenarios. The morphological dependence is also investigated further. The late-type galaxies cluster in the same way for most magnitude ranges after the magnitude correction. There is even some anti-segregation, whereby faint galaxies cluster more strongly than luminous ones. When early-type galaxies are considered, it is found that luminous ones cluster more strongly than faint ones even after the magnitude correction, that is, the luminosity segregation still exists for early-type galaxies, although the degree of luminosity segregation is slightly reduced after the magnitude correction. Combined with the morphological segregation, these facts are responsible for the absence of luminosity segregation in the total sample. The diffusely distributed late-type galaxies are subject to strong internal extinction, and therefore they are incorporated into brighter samples after the magnitude correction. The clustering amplitudes of brighter samples are then reduced in the mixture of ellipticals and spirals. The effects of various errors included in the catalogue are also investigated, and it is shown that these results are not severely affected by random peculiar velocities of galaxies, nor by random errors in the magnitude and velocity measurements.