Summary. Statistical selection procedures are discussed in general terms. Statistical selection procedures have been designed specifically to answer questions like ‘Which treatment or variety can be considered to be the best?’. In a certain sense statistical selection procedures are more realistic in answering such a question than the usual testing and multiple comparisons procedures. The statistical procedures of Bechhofer and Gupta are considered. Some practical applications are given. Finally, attention is paid to the relatively new combined procedure of Hsu (1981, 1984) and the procedures of Somerville (1985).