In this paper we describe an experimental comparison between the performance of listeners and the performance of an automatic method in an experiment of speaker identification. The database used in both, automatic and subjective experiments, was made up of 4 different speakers. Their voices were selected as belonging to a homogeneous group of subjects whose dialects and social habits were very similar. The speech material consisted in five phrases that have been recorded in three different sessions: the second session follows the first after about three months and the third session follows the first after about ten years. Each recording session uses the same telephone channel and the same recording apparatus. The subjective experiments consist in a discrimination test, each test item consisting in two speech samples. The listener decides whether the two samples were produced by the same or different speakers. The parametric method is a traditional automatic approach based on the evaluation of suitable acoustic parameters. The recognition test is performed with that parameter utilized in usual hypothesis test. The comparison of the test results obtained by automatic and subjective methods are discussed.