The influence of surface topography on overlayer thickness determination by angular dependent X-ray photoelectron spectroscopy (ADXPS) was investigated by a geometrical model with a surface topography consisting of hexagonal close-pack hemispheres or spherical segments.Calculations consider all tilting and shadowing effects including shadowing by neighboring segments. The results are compared to ADXPS experiments of monolayers of n-octanethiol and PTCDA (3,4,9,10-perylene-tetracarboxylic-dianhydride) on gold. Rough surfaces were prepared by hexagonal close-pack polystyrene latex beads of 500nm diameter covered with 50nm gold on a silicon wafer. Flatter topographies consist of a 200nm evaporated gold layer on Si(100) and of Au(111) surfaces.Experimental data and the theoretical results of the model agree for the different topographies. Small deviations in the order of =<10% between experimental data and model calculations, based on the straight line approximation (SLA) of electron emission, can be explained by elastic electron scattering. The model for hemispherical surfaces can also be extended to rough surfaces with a similar, but less defined, corrugation.Thickness determination is greatly facilitated at an off-normal 'magic angle' of approximately 55 o , where the ratio of the observed and true overlayer thickness (d o b s and d 0 , respectively) is almost constant, i.e. d o b s /d 0 =1.6 for a large range of topographies. This value for the 'magic angle' is higher than the angle reported in the literature, i.e. approximately 35 o obtained from a simple two-dimensional model [W.S.M. Werner, Surf. Interf. Anal. 23 (1995) 696] and 40-45 o from a three-dimensional model of a spherical topography without neighbor-shadowing effects [P.L.J. Gunter, O.L.J. Gijzeman, J.W. Niemantsverdriet, Appl. Surf. Sci. 115 (1997) 342].