Аннотация
In the last 60 years, the science of scientometrics, the quantitative analysis of science as a communication system, has developed from an esoteric interest in citation indexing to a key tool for research policy and evaluation. The aim of this review is to trace and summarise that trajectory under the single argument that scientometric indicators are only as meaningful as the data infrastructures on which they are built, and the evaluative contexts in which they are applied. We present the literature in five methodological layers, linked by a common thread on research evaluation and policy, and spanning bibliographic data sources, citation indicators and normalisation, science mapping and visualisation, altmetrics and the widening of impact, as well as the recent computational turn based on machine learning and natural language processing. Along with this narrative review, we present a companion bibliometric analysis, mapping the field by topic within the OpenAlex database, which reveals an analytic corpus of 37,166 works, whose growth we illustrate, whose core journals we identify, and whose thematic structure we discuss as a characterization of the field from a methodological perspective. The review finds that the best lasting gains in the field have been in the quality of data coverage, data normalisation and in the responsible interpretation of indicators, and the current transition from counting to understanding indicators should be subject to the same standards.