We report on the results of R band observations of the error box of the gamma-ray burst of August 28, 1997, made between 4 hours and 8 days after this burst occurred. No counterpart was found varying by more than 0.2 magnitudes down to R = 23.8. We discuss the consequences of this non-detection for relativistic blast wave models of gamma-ray bursts, and the possible effect of redshift on the relation between optical absorption and the low-energy cut off in the X-ray afterglow spectrum.