The technique for obtaining radial velocities of g ?Taxies from UK Schmidt telescope objective-prism spectra, first developed by Cooke (1980), is used here to investigate the possibility of using these radial velocities to produce a large-scale survey of the Universe in three dimensions. A 4.8° x 5.2° region of the sky in ESO/SERC survey field 145 is examined, which contains part of the Indus supercluster. Computer programs are used to obtain a large sample of spectra from a measurement of the plate of this region by the Edinburgh COSMOS machine in its mapping mode. The Indus region surveyed is found to contain no obvious large-scale features, but several small ones. The reliability of objective-prism redshifts is tested by measuring some of the areas a second time, and the objective-prism redshifts are also calibrated by comparing them with redshifts obtained independently from slit spectra. The effective "redshifts" of stellar spectra are measured to check the zero point of the redshift scale. The results indicate that the accuracy of the redshifts in the COSMOS sample is not as good as expected, and that there is a systematic drift in redshift zero point with magnitude. This situation may be improved by using a magnitude-independent method of wavelength calibration and by limiting the sample to images brighter than B ~ 18. It is found to be difficult to determine redshifts reliably using only the one standard feature used by Cooke (1980), and around 5% of the spectra this feature is mis-identified, causing a redshift discrepancy of around ±0.1. A feature in some stellar spectra, at 4470A, causes them to appear like galaxies on the objective-prism plate, and these are found to be contaminating the sample, causing an excess of redshifts in the range 0.11-0.13. The objective-prism sample is paired with a sample of images from the direct plate of the region which have been automatically separated into stars and galaxies. It is found that redshifts are attained for about 20% of galaxies, but that this fraction decreases with magnitude. A large fraction of objects fainter than B ~ 18.5 are found to be compact. Most of these are contaminating stars, but there is marginal evidence for the existence of a small population of compact galaxies in the rich clusters of the Indus supercluster.