We present clustering measurements for samples of galaxies selected by morphological type and luminosity from the recently completed Stromlo-APM redshift survey. We find very different results between real and redshift-space estimates of the correlation function. The real space correlation function for the all-galaxies sample is well fitted on scales 0.2-20h(-1) Mpc by a power law with slope gamma(r) = 1.71 and correlation length r(0) = 5.1h(-1) Mpc. In redshift space the slope is shallower, gamma(s) = 1.47 and the correlation length is slightly higher, s(0) = 5.9h(-1) Mpc. Early-type galaxies are clustered more strongly by a factor 3.5-5.5, than late-type galaxies. In real space the slope of the correlation function for early-type galaxies is gamma(r) 1.85, slightly steeper than for late types, gamma(r) 1.64. In redshift space, however, early-type galaxies have a very shallow correlation function slope, gamma(s) = 1.25. This implies that these early-type galaxies suffer from enhanced redshift-space distortions compared to late-type galaxies. Low-luminosity galaxies are clustered more weakly by a factor of similar to 2 than L* and brighter galaxies on scales greater than or similar to 1h(-1) Mpc. Also the slope of the correlation function is steeper for low-luminosity galaxies, so that the amplitude is a factor 4 lower at 10h(-1) Mpc. No difference, however, is seen between the clustering of L* and more luminous galaxies, an observation which may be hard to reconcile with some theories of biased galaxy formation. Both redshift-space and real-space clustering estimates show a similar dependence on luminosity. Our results hint that luminosity segregation may be a purely primordial effect, due to a lower bias factor for low-luminosity galaxies, whereas morphological segregation, being most apparent on scales less than or similar to 1h(-1) Mpc, may be enhanced by environmental factors.