The study examined the accuracy of stone casts produced from impressions taken in stock polycarbonate trays, some of which had been strengthened with autopolymerizing polymethyl methacrylate resin and four designs of custom-made trays. Impressions of an acrylic master model of the mandibular arch on which two extracoronal crown preparations and one intracoronal inlay preparation had been carried out were taken using a medium-bodied polyvinyl siloxane material. Each preparation had been intended with a reference point for later measurement. Five impressions were taken for each type of tray and these were cast in die-stone after 24 h. The distances between the points were measured with a reflex microscope and the means determined for each design of tray. The mean difference between casts produced from the various tray designs and the acrylic master model were determined for each of the distances between the three measuring points. Statistical analysis showed that, with the stock trays, there were no significant differences between any of the modifications, or between them and the acrylic model, for any of the three distances (P greater than 0.1). When the results from the custom-made trays were analysed there were statistical differences between the acrylic model and the casts from the unperforated tray with no stops (P = 0.02), and between the unperforated tray with no stops and the perforated with stops for the distance A-B (P less than 0.01). There were no other significant differences.(ABSTRACT TRUNCATED AT 250 WORDS)