The formation of tidal dwarf galaxies (TDGs) is triggered by the encounters of already existing galaxies. Their existence is predicted from numerical calculations of encountering galaxies and is also well documented with observations. The numerical calculations on the formation of TDGs furthermore predict that TDGs cannot contain significant amounts of non-baryonic dark matter. In this paper, the first exhaustive sample of TDG-candidates from observations and numerical calculations is gathered from the literature. These stellar systems are gas-rich at the present, but they will probably evolve into gas-poor objects that are indistinguishable from old dwarf elliptical galaxies (dEs) based on their masses and radii. Indeed, known gas-poor TDGs appear as normal dEs. According to the currently prevailing cosmological paradigm, there should also be a population of primordial galaxies that formed within haloes of dark matter in the same mass range. Due to their different composition and origin, it would be expected that objects belonging to that population would have a different structure than TDGs and would thus be distinguishable from them, but such a population cannot be identified from their masses and radii. Moreover, long-lived TDGs could indeed be numerous enough to account for all dEs in the Universe. Downsizing, i.e. that less massive galaxies tend to be younger, would then be a natural consequence of the nature of the dEs. If these claims can be kept up in the light of future observations, the presently prevailing understanding of galaxy formation would need to be revised.

18 pages, 2 figures, accepted by MNRAS