We reinvestigate the validity of the limiting similarity principle via numerical simulations of the Lotka–Volterra model. A Gaussian competition kernel is employed to describe decreasing competition with increasing difference in a one-dimensional phenotype variable. The simulations are initiated by a large number of species, evenly distributed along the phenotype axis. Exceptionally, the Gaussian carrying capacity supports coexistence of all species, initially present. In case of any other, distinctly different, carrying capacity functions, competition resulted in extinction of all, but a few species. A comprehensive study of classes of fractal-like carrying capacity functions with different fractal exponents was carried out. The average phenotype differences between surviving species were found to be roughly equal to the competition width. We conclude that, despite the existence of exceptional cases, the classical picture of limiting similarity and niche segregation is a good rule of thumb for practical purposes.