ON THE TRUE SHAPES OF GALAXIES

We analyse the distribution of axial ratios of images in the APM Bright Galaxy Survey (APMBGS). The sample analysed contains a total of 20 399 galaxies (2135 ellipticals, 4782 SOs, 13 482 spirals). For these objects the APM survey provides homogeneous measurements of image characteristics with great accuracy. We apply analytic inversion to infer the true distribution of axial ratios for oblate and prolate models. We also apply Monte Carlo techniques to explore triaxial models. For elliptical galaxies, we find that neither pure oblate nor pure prolate models can satisfactorily account for the observed distribution of apparent axial ratios, since the derived solutions for the distribution of true axial ratios become significantly negative for q > 0.9. We consider triaxial models in which the true axial ratios b/a and c/a of galaxies are selected from Gaussian distributions. Models with [(b - c)/(a-c)] = 0.6 provide a good fit to the observed distribution of axial ratios. We analyse possible variations of the intrinsic flattening of ellipticals depending on low and high local density environments. We find that the projected distributions of axial ratios of high and low local density environment ellipticals are remarkably similar. We therefore see no evidence in the data that the environment has significantly affected the shapes of early-type galaxies. For spiral galaxies, pure oblate models fail to reproduce the distribution of observed axial ratios, but consistent fits can be obtained with nearly oblate models with thickness c/a congruent-to 0.2 and ellipticity b/a congruent-to 0.9. The distributions in this model are very similar to the classical oblate solution of Sandage, Freeman & Stokes. The inferred slight triaxiality could be due to the presence of substructure in discs, or may reflect truly triaxial potentials in spirals.