Citation metrics are statistical measures of scientific output that draw on citation indexes. They purport to capture the impact of scientific articles and the journals in which they appear. As evaluative tools, they are mostly used in the natural sciences, but they are also acquiring an important role in the humanities. While the strengths and weaknesses of citation metrics are extensively debated in a variety of fields, they have only recently started attracting attention in the philosophy of science literature. This paper takes a further step in this direction and presents an analysis of citation metrics from the perspective of a Kuhnian model for the development of science. Starting from Gillies’ argument against the use of citation metrics for scientific research (2008), this paper shows that citation metrics interfere with the development of normal science in certain fields or subfields. The main issue is that citation metrics do not take field-specific differences into account, thereby selectively favoring some fields and arbitrarily hindering the development of others. In other words, this paper shows that current citation metrics fail to “carve science at its joints”. In light of this, the paper cautions against their use for evaluative purposes.