This paper examines a number of the criticisms that citation analysis has been subjected to over the years. It is argued that many of these criticisms have been based on only limited examinations of data in particular contexts and it remains unclear how broadly applicable these problems are to research conducted at different levels of analysis, in specific fields, and among various national data sets. Relevant evidence is provided from analysis of Australian and international data. Citation analysis is likely to be most reliable when data is aggregated and at the highly-cited end of the distribution. It is possible to make valid inferences about individual cases, although considerable caution should be used. Bibliometric measures should be viewed as a useful supplement to other research evaluation measures rather than as a replacement.