Understanding the clustering of galaxies has long been a goal of modernobservational cosmology. Redshift surveys have been used to measure thecorrelation length as a function of luminosity and color. However, whensubdividing the catalogs into multiple subsets, the errors increaserapidly. Angular clustering in magnitude-limited photometric surveys hasthe advantage of much larger catalogs but suffers from a dilution of theclustering signal because of the broad radial distribution of thesample. Also, up to now it has not been possible to select uniformsubsamples based on physical parameters, such as luminosity andrest-frame color. Utilizing our photometric redshift technique, avolume-limited sample (0.10=5.77+/-0.10h-1Mpc. We find that r0increases with luminosity by a factor of 1.6 over the sampled luminosityrange, in agreement with previous redshift surveys. We also find thatboth the clustering length and the slope of the correlation functiondepend on the galaxy type. In particular, by splitting the galaxies infour groups by their rest-frame type, we find a bimodal behavior intheir clustering properties. Galaxies with spectral types similar toelliptical galaxies have a correlation length of6.59+/-0.17h-1Mpc and a slope of the angular correlationfunction of 0.96+/-0.05, while blue galaxies have a clustering length of4.51+/-0.19h-1Mpc and a slope of 0.68+/-0.09. The twointermediate color groups behave like their more extreme ``siblings''rather than showing a gradual transition in slope. We discuss thesecorrelations in the context of current cosmological models for structureformation.