This study aimed to ascertain the ability of the microbiology laboratory to detect and identify catalase-negative Gram-positive cocci with particular reference to vancomycin-resistant enterococci (VRE). Twenty-seven reference strains and 42 prospectively collected catalase-negative Gram-positive cocci were screened by agar dilution breakpoint susceptibility and linked biochemical methods in routine use. Ability to speciate organisms was then compared using: (i) a multiplex polymerase chain reaction, designed to detect gene sequences specific to Enterococcus faecalis and E. faecium, and vancomycin resistance (van) genes; (ii) a commercial "API 20 strep" (iii) an algorithm using individual tests from a commercial API 20 strep strip; and (iv) the same algorithm utilising traditional phenotyping methods. All vancomycin resistant catalase-negative Gram-positive cocci were detected by an agar dilution screening plate containing 4 micrograms/ml of vancomycin. Polymerase chain reaction (PCR) detected all enterococci with van genes, speciated all vancomycin-sensitive E. faecalis and E. faecium isolates and excluded non-enterococcal vancomycin-resistant catalase-negative Gram-positive cocci. Algorithm-based methods speciated 41 of the 42 study isolates (98%). The API 20 strep correctly identified only 25 (60%) of these organisms, 38 of which were vancomycin-sensitive E. faecalis. VRE are detected by current screening methods for vancomycin-resistant catalase-negative Gram-positive cocci in this laboratory. API 20 strep, currently used to speciate catalase-negative Gram-positive cocci, is less reliable and should be replaced. Algorithm-based phenotyping by either method tested is more reliable for speciation than API 20 strep in its recommended form. Compared to the other methods tested, PCR is a rapid, accurate and inexpensive method of detecting and speciating vancomycin-resistant enterococci and it provides important extra information impacting on clinical therapy and infection control.