Variations In Citation Practices Across The Scientific Landscape: Analysis Based On A Large Full-Text Corpus

Publication full text is a potent new source of data for scientometric research and offers a unique look into how practices of citation differ across scientific disciplines. We present an analysis of citation label styles, the use of cited author surname as an integral part of the narrative of citing sentences, and the use of reporting verbs, across a corpus of 58 million citing sentences. Results are both aggregated over five broad scientific disciplines and visualised across a map of science for greater detail.The occurrence of author-year style citation labels overlaps with integral citations in social sciences and humanities, but mathematics and computer science uses integral citation without author-year labels, and life and earth sciences shows the opposite pattern. Reporting verbs are slightly more common in social sciences and humanities, especially cognition verbs, which also feature often in mathematics and computer science.Individual verbs show great differences in their usage across science. Several seemingly similar verbs favour almost entirely distinct areas of research. Overall, we observe remarkably gradual changes in these phenomena across the scientific landscape, suggesting that disciplinary conventions in writing styles overlap and combine in areas between prominent disciplines.