The influence of different representations of the unresolved resonances of 238U on the computed self-shielding factors is examined. It is shown that the evaluated infinitely diluted average capture cross section does not provide sufficient information to determine a unique set of unresolved resonance parameters; different sets of unresolved resonance parameters equally consistent with the evaluated average capture cross section yield significantly different computed self-shielding factors. In the conclusion it is recommended that the resolved resonance description of the evaluated 238U cross sections be extended to higher energies and that thick sample transmission data and self-indication data be used to improve the evaluation of the unresolved resonance region.