Cover of the 2021 edition

Imagine you are a researcher, studying, say, bird migration. How do you find out about new papers on your topic? You have written a PhD on the subject, which enabled you to become familiar with all the recent research on the topic. You regularly read papers and look up the citations. You attend conferences and meet other researchers in the area. In the library, you regularly check the contents list of your favourite journals. Finally, you have colleagues in your institution whose subject interests overlap with yours.

By the time you have followed up all these leads, what need is there to discover more? And yet, a new survey has found, some 45% of all academics’ content reading is found via discovery, by searching for new content. It seems an amazingly high proportion for any subject. Asked to guess a figure, I would have suggested perhaps 10%.

Yet the statistic is believable, because it is derived from How Readers Discover Content in Scholarly Publications. The latest (2021) edition of this report, by Renew Consultants (Simon Inger and Tracy Gardner), must be the most statistically valid survey ever carried out of academic discovery – based on the data of over 15,000 researchers around the world. Now the report has been running for 16 years, with a new edition published every three years, its value lies not just in what it can tell us about search and discovery today, but in how that user behaviour has changed over time.

Probably the best way to review the report is to form a list of, firstly, what might be expected, and secondly, what is a surprise. First, what is to be expected as typical of researchers:

  • Many academics start a search using Google (either Google Scholar, or just Google)
  • In physics, academics start with preprint repositories for discovery.
  • In countries with lower GDP, people search for OA content more than in richer countries. That’s probably more an indication of lower rates of higher education than a lower wealth.
  • Related articles are the most popular feature of a publisher website; the TOC is declining in importance (p35). Saved search alerts are being used less and less.
  • Use of social media for article discovery is steadily increasing.
  • In the humanities, researchers use the library as much as they use Google Scholar
  • Journal aggregations are most common in North America (presumably this means EBSCO, ProQuest, JSTOR, and similar)
  • A researcher typically finds content via searching for articles on a subject more than by any other means (p25). This is similar in most countries.
  • Users search for OA content in all regions – I don’t know why this should be so surprising. Renew’s recommendation is simply “Publishers need to ensure their OA content is as discoverable in all the usual discovery resources as their non-OA content” (p20). There is a clear indication here that corporates search for OA content, which is simply because they typically don’t have access to subscription-based content via an institution. Paying for access to an individual article is an expensive way to discover the article isn’t really relevant! So corporate searchers will be in favour of better discovery tools.

What is surprising to discover:

  • A&I databases are still widely used as the starting point for discovery in life sciences and medicine (although not in other areas). In fact they have grown in importance for discovery since 2012. I suppose if PubMed is considered as an A&I database, it is understandable that it is widely used by clinicians (p18).
  • Social media is not (yet) a common method of discovery for academic content.
  • Not only do researchers discover content via search around 45% of the time, but this proportion is increasing, rather than by recommendations.

Search and Open Access
People actively search for open access articles, and unsurprising the wealth of their country has an impact on this. People in poorer countries actively search for OA content more than people in richer countries. People in the corporate sector search for OA content more often than their counterparts in the academic and medical sectors

Search and Discovery

We should be careful about a report entitled “How Readers Discover Content”. Presumably this does not include readers looking up a paper for which they have all the relevant details already? Does this count as discovery? “People discover articles through search around 45% of the time. However, discovery via search has increased over time.” Presumably, the only way you might discover content is by recommendations. It would be interesting to measure (albeit difficult) what percentage of content is identified by personal recommendation, or by reading a paper and looking up the citations. Even so, 45% is an astonishingly high figure for an academic who is in principle simply keeping up with a subject they are already familiar with. I did chat with one Irish researcher who estimated he sifts through 50 or more new articles on arXiv each week, and then uses various means to skim those articles to determine the ten or so he will actually read in full. As a researcher into AI topics, he was, perhaps unsurprisingly, looking to use AI to manage that weekly trawl more effectively.

A&I databases

One of the most interesting observations is the survival of A&I databases – not the most obvious starting point for a search, since they only comprise abstracts. For years, Google has indexed the full text of articles, even if access to the full text was not directly available. Perhaps what we can look forward to in years to come is the replacement of A&I databases by collections of full text.