Bibliometric methods for scientific evaluation are examined. The citation frequency for scientific publications, as given in the Science Citation Index, is influenced by various forms of citation bias as well as by the characteristic dynamics and citation practices of the various scientific fields, and therefore cannot be taken as an unequivocal measure of scientific quality. Even for single authors the citation frequency is very variable; therefore this parameter should not be used for graded evaluation of individual scientists or research groups. At higher aggregate levels (large institutes etc) bibliometric indicators may, as part of a more general evaluation, give relevant information about research activity, provided due corrections are made for variable research field effects. The journal impact factor, i e the mean citation frequency of all articles in a journal, has been suggested as a rapid indicator of article quality. However, the distribution of citation frequency values within a journal is extremely broad and skewed; therefore assigning the same value to all articles would not seem to serve the purpose of evaluation particularly well. Furthermore the citation frequencies of articles published by individual authors or research groups are found to correlate extremely poorly with the corresponding journal impact factors. The latter parameter would thus appear to be unsuitable as an indicator of scientific quality.