The activity of scientific research is a human construct, and so it is a social endeavour. Like many other social endeavours, it is natural for scientists to assess how their research affects other colleagues in the field and how it influences the course of further research. This is reflected in the history of scientific prizes awarded to prestigious researchers, from the Royal Society’s Copley prize to the relative newcomer that is the Nobel Prize, and to recently launched prizes such as the Breakthrough Prize. With the growth of scientific publications, and databases such as the Web of Science that record publications and citations, metrics such as the h-index have been developed to characterize the scientific output of a researcher1. Like all metrics, the h-index is narrow and can be misused, but certainly citation databases and analysis tools can enable an analysis of trends in citations as a measure of impact in scientific research over the last half century or so. Indeed, such an analysis has recently been published in Nature, and here the authors come to the conclusion that papers are becoming less disruptive over time2.

This publication analyses papers and citations data in the Web of Science, as well as additional abstracts in several patent databases, from 1945 to 2010. To calculate the disruption a paper produces the authors utilize an index, the CD5 index calculated for the 5 years following publication. This characterizes the degree by which a paper supersedes previous publications as later literature will cite that disruptive paper rather than previous literature. The opposite of a completely disruptive paper is a completely consolidating paper; later publications will cite a consolidating paper as well as the previous literature. Using this metric, the authors find that for all research disciplines, disruption has fallen (Fig. 1). Indeed, using other metrics such as the usage of words that evoke creation, discovery or perception as a proxy for disruption also show this drop, while words that evoke improvement or optimization become more frequent. Going beyond papers, the authors also see the same pattern with patents registered at the United States Patent and Trademark Office.

Fig. 1: The decline in disruptive index CD5 of papers published from 1945 to 2010.
figure 1

Where a larger y-axis value indicates a more disruptive paper and the shaded bands the 95% confidence intervals. Figure reproduced with permission from ref. 2, Springer Nature Ltd.

Full size image

What is driving this drop? The authors suggest that this decline is driven by changes in the scientific community; scientists are now focused on less diverse and/or narrow research areas, leading to a decline in the diversity of work cited, increased self-citation and an increase in the mean age of work cited. It is plain that the professional environment researchers work in has changed substantially from the 1940s.

In a Correspondence in this issue of Nature Materials, Andrew Akbashev and Sergei Kalinin discuss how efforts to target specific institutional or funder-required metrics can lead to lower-quality science. They note that these requirements can lead to a pressure to overpublish, leading to a scientific study being separated into several papers, or papers published that have little scientific value. They suggest in turn that this has a negative impact on science, as experts are asked to review more papers potentially leading to lower-quality reviews, that there are too many papers to read leading to fragmentation of scientific communities, and lower-quality papers may lead to confusion about the true state of the field. As they note, pure text documents can be generated easily, and in the age of ChatGPT and other large language models this is becoming even more so and care must be taken with these tools3. To counter these issues, the authors propose several solutions, such as moving to open-ended publications as those enabled in online notebooks for code storage and revision, or additional open-ended reviewing after publication.

One thing that both the Correspondence and the recent paper in Nature emphasize is that institutions should focus on scientific quality rather than on quantity, although how this should be judged is not stated. Both sets of authors claim that such a focus on quality would reduce pressure to overpublish, giving time to produce substantial disruptive work. As the authors of the Nature paper note, the number of truly disruptive papers published from 1945 to 2010 seems to be fairly constant, increases in this number would be welcomed.

This is not to say that research that consolidates previous work has little worth. Research that does involve optimization or improvement often has an applied basis, which we have previously noted is also of interest to the materials science community, especially as this can then drive further scientific research4. Both consolidating and disruptive papers, with conceptual novelty and substantial mechanistic insight, are always welcomed to be submitted to Nature Materials and, as always, we will aim to publish the most innovative and impactful studies while trying to balance the distribution and visibility of topics across our broad and multidisciplinary international readership.