This study looks at the real and perceived effects of inter-candidate variation in linguistic ability during pair-tested oral examinations in a southern Italian context. Pair testing has not been taken up in schools and universities in southern Italy despite its wide use in international testing systems. Colleagues and prospective candidates often report to teachers their belief that test results can be adversely or positively affected by candidate pairing. To investigate the existence and validity of these beliefs, a series of tests was devised consisting of paired oral tests and pre- and post-test questionnaires. It was found that while the great majority of candidates believed that their test scores could be, and in many cases had been, affected by their candidate pairing, no evidence came to light to support these beliefs. In fact the linguistic ability of a candidate's partner appeared to have little or no effect on his or her performance.