Disseminating scholarly discoveries outside of academia is important to ensure that they are widely used, to justify public financing of academic research, and to inspire future scholars. Although scientific findings are usually published in articles and books for other scholars, those that are relevant to the public may attract mass media attention (Weigold, 2001; Dudo, 2015). The number of science articles in the news has been increasing (Pellechia, 1997), perhaps reflecting the growing number of people having a degree. For example, a series of 39 surveys in the United States from 1996 through 2002 (showed that 42% of adults claimed to follow health news stories in the mass media (Brodie, 2003). It is therefore reasonable for universities and scholars to seek press coverage to promote their work. From a scientometric perspective, newspaper coverage of research may also be used as evidence of the wider impact of the cited papers (Fogg-Rogers, Grand, & Sardo, 2015) in addition to reflecting or causing (by publicity) its greater value within academia (e.g., Fanelli, 2013). It is important for research institutions to know which types of research attract press attention so that they can (a) focus marketing efforts on media-friendly types of research and (b) interpret press coverage impact evidence in the context of the type of research covered. For example, a newspaper story about a pure mathematics article might be a bigger achievement than one for an article about an astronomical discovery.

Prominent academic journals are a regular source for many newspapers (Entwistle, 1995; Bartlett, Sterne, & Egger, 2002; Lewison, 2002; Weitkamp, 2003). Prestigious health journals are frequently covered (Conrad, 1999; Lewison et al., 2008; Van Trigt et al., 1994), because they may have more relevance for the public. Major UK newspaper stories about biomedical research in 2001 cited New Scientist (7% - a magazine rather than a journal), British Medical Journal (6%), Lancet (4%) and Nature (3%) (Lewison, 2002). In 2008-9, the UK-based weekly science magazine New Scientist drew mainly from Proceedings of the National Academy of Sciences of the USA (7%), Nature (6.5%), Science (4.6%), Lancet (3.9%), and New England Journal of Medicine (2.6%) (Lewison & Turnbull, 2010). News stories about cancer in major US newspapers in 2003 mostly cited New England Journal of Medicine, Journal of the National Cancer Institute, and Journal of the American Medical Association (Moriarty, Jensen, & Stryker, 2010). Most articles in prestigious health journals attract no press attention, however. For example, only 7% of 1,193 Lancet and British Medical Journal articles (1999-2000) were reported in two UK newspapers (the Times and the Sun), mostly for women’s health issues and cancer (Bartlett, Sterne, & Egger, 2002). Nevertheless, many news stories reporting research findings do not directly cite academic articles or journals and instead name authors, researchers, research institutions or other secondary sources. It is therefore difficult to be sure about the extent to which academic articles or journals are reported by the press without reading a large sample of scientific news stories (De Semir, Ribas, & Revuelta, 1998; Lewison, Tootell, Roe, & Sullivan, 2008; Mellor et al., 2011).

Health research is a common theme in science news, whereas the social sciences attract little coverage. The weekly science section of the New York Times 1980-2000 covered medicine and health (39%), physical, earth and life science (31%), technology and engineering (21%), and history and culture (6%) (Clark & Illman, 2006). However, health news may also convey incorrect information or contain omissions or misinterpretations of published research (Chang, 2015). Science news in three US newspapers (New York Times, Washington Post, Chicago Tribune) 1966-1990 was predominantly (70%) about medicine and health (Pellechia, 1997). In the UK, most of the science news in five newspapers 2000-2001 was about health and medicine, compared with biology (18%), social sciences (11%), environment (8%) and physical sciences (8%) (Weitkamp, 2003). Social science has a similarly low share of coverage in the New York Times (11%) and Los Angeles Times (8%) weekly science sections (Evans, 1995). There is also periodic mass media coverage of individual narrow scientific areas with high levels of public interest (Schäfer, 2012).

Despite the above focused investigations into the types of research reported in the press, there is no quick method to estimate the number of citations to academic journals on a large scale from full-text newspaper articles. Previous studies have employed either content analysts or coders to investigate small random or complete samples of news stories from a narrow set of sources (e.g., Clark & Illman, 2006; Lewison & Turnbull, 2010; Mellor, Webster, & Bell, 2011; Weitkamp, 2003), or have read complete sets of news stories to identify mentions of collections of
journal articles (e.g., Bartlett, Sterne, & Egger, 2002). Some previous studies have automatically identified relevant articles using keyword queries in news databases (e.g., Lexis-Nexis) to identify mentions of named medicines (Moynihan et al., 2000) but this method only works for topics with a finite set of keywords and a more general automatic method is needed. In response, this article introduces for the first time an automatic method to find press coverage of almost any area of research. This article also uses the method to analyse mentions of a substantial number of science (n=9,639) and social science (n=3,412) academic journals in a set of digitised newspaper stories (eight UK daily newspapers during 2006-2015), to update previous manual studies in a faster and easily replicable way. Previous studies have reported which journals are frequently covered in small sets of newspaper stories. Moreover, there have been initiatives by Altmetric.com¹(¹https://www.altmetric.com/about-our-data/our-sources/news) and PlumX²(²https://plumanalytics.com/tag/news-mentions/) to capture mentions of scholarly articles in news sources. However, it seems that article metadata (e.g., title, journal name) is rarely mentioned in news stories (e.g., 12%-13% see: Lewison et al., 2008; De Semir, Ribas & Revuelta, 1998). Hence, this article introduces an automatic method for comprehensive journal-level analysis of press uptake of scientific articles. This method can also be used to aid the evaluation of the wider impacts of academic research.

This paper introduces and assesses an automatic method to estimate the coverage of published academic research in digitised newspapers on a large scale and uses it to investigate recent UK press coverage of research. Major newspapers were investigated since these are a coherent set and most likely to cover academic research. The UK was chosen as a clear unit of analysis and to ensure familiarity with reporting idioms. Relating to the two broad goals, the following specific questions are addressed, updating previous large-scale studies and broadening knowledge about press treatment of academic research using the new automatic method.

1. Can citations to scientific journals be automatically identified in news stories with an acceptable degree of accuracy?

2. How likely are different academic subjects and journals to be cited by major UK newspapers? This updates previous manual research and, for the first time, includes an analysis of journals.

3. Has newspaper coverage of academic research changed in the last decade?

4. Are recent newspaper stories citing scientific journals in-depth, positive and/or critical?

The research design was to apply the new method to identify mentions of academic journals in news stories for a large-scale systematic analysis involving many scientific journals. A follow-up content analysis was conducted to assess the accuracy of method and to give detailed information about the news stories found.

The method identifies mentions of journals in newspaper stories by searching for journal names in the ProQuest Newspapers database with a small set of curated queries, then using a program to automatically extract and filter citations from the news stories identified. For this method WoS was used as a convenient large-scale source of important journals, although any other source of non-abbreviated journal names could be used, and ProQuest was chosen for its press coverage, full text sources, and query syntax.

To apply and evaluate the new method, UK subset of newspapers in ProQuest database (The ProQuest UK Newsstand) was selected to have a large enough but manageable number of English language news stories. Eight major UK daily newspapers were selected: Daily Mail, Times, Daily Telegraph, Daily Mirror, Guardian, Independent, Sun and Herald (Scotland).

On a stylistic note, most newspapers have “The” within their titles (The Sun, The Guardian, The Times, The Daily Telegraph, The Independent, The Herald) but these are omitted here, and academic journals are referred to without definite articles.

In answer to the main research question, 100% of the news stories extracted by the method in the manually checked random sample were technically correct (i.e., mentioned an academic journal) and 94% cited the journal to discuss its research. Thus, the method has a high precision (few false matches), although its recall (percentage of matches found) is unknown. More details are given in the content analysis section, after a discussion of the overall results.

The newspaper citation search method in this study has a several limitations. Heuristics were used to identify mentions of journals in news stories. Although the manual checks of a sample of the results found no false matches (i.e., an estimated 100% technical precision), there might be a few errors in capturing journal names in the whole data set. For instance, the software may identify journal names incorrectly in some complex cases, such as in “published in Scientific Reports, a Nature journal”, where Scientific Reports is a correct match and Nature is not. For practical reasons, it was not possible to search generic and short journal names (e.g., Science, Nature or Cell) without adding the term journal to the ProQuest query to avoid downloading a huge number of news stories using these above terms in other contexts (“Science fiction stories are”, “Nature conservation issues are”). Nevertheless, there could also be relevant results not mentioning the term journal, such as “a study published in Science” or “The findings, which have been published in Science magazine”, where these are indirectly flagged as academic journals.

Many news stories do not explicitly name their sources. For instance, internationally, newspapers (The New York Times, Le Figaro and Le Monde, El País and La Vanguardia, La Repubblica, and the International Herald Tribune) seem to cite journal articles in about 13% of their scientific news stories (De Semir, Ribas, & Revuelta, 1998) and only 12% of US newspaper stories in 2003 about cancer (n=3,600) identified journal articles (Lewison, Tootell, Roe, & Sullivan, 2008). A content analysis of 640 BBC News stories with science themes in 2010 also showed that a third named a publication. Online news stories were more likely to mention journal names, however. Two thirds of online BBC news stories about research explicitly mentioned journal names. Only a fifth of the stories naming journal articles had hyperlinks to the main articles online (Mellor, Webster, & Bell, 2011). The recall of the method (i.e., the percentage of news stories about research that the method found) is therefore probably low and almost certainly lower than studies in which complete sets of newspaper articles were individually read to identify mentions of academic research (e.g., Bartlett, Sterne, & Egger, 2002). The level of recall also probably varies between journals and may be lower for those with long names that newspapers might abbreviate or with short names that heuristics were used to identify (e.g., adding “journal” before). Moreover, it was not practical to estimate the overall recall of searches for individual journals with generic names, such as Science, Nature, Cancer or Cell. This is because searching above journal names in the ProQuest Newspapers database gives many results. For instance, searching the term “Science” in the full-text of the eight UK newspaper during 2006-2015 retrieves over 162,000 news stories which requires extensive manual work to download and to identify relevant results.

Editorial policies of newspapers about mentioning academic sources might have influenced the results and may change over time. For instance, newspapers may interview authors or researchers about their published research without mentioning journal names or article titles. Hence, newspapers with more emphasis on mentioning research sources would presumably have a larger presence in the overall results. The recall of the method may therefore differ between newspapers, making comparisons between them potentially misleading.

There are some discrepancies between the ESI classification for journals and their subject categories within SCI or SSCI. The subject category Public, Environmental & Occupational Health is in both WoS SCI and WoS SSCI. Some journals have been indexed in Public, Environmental & Occupational Health in WoS SCI rather than WoS SSCI. However, ESI has classified them as Social Sciences, General, presumably because there is no public health category in the 22 ESI research fields. Such issues are a limitation with using ESI for subject analyses of journals.

Finally, the ProQuest UK Newsstand database seems to be incomplete for some of the newspapers and years, so comparisons over time are not reliable.

In answer to the first research question, the content analysis showed that the method used here captures mentions of academic journals in news stories with high levels of precision (94% technically correct, meaningful matches) but probably low recall (a majority of relevant news stories were probably not found). Since the method is general, it could also be used to extract citations to journals from other news sources in text format such as US newspapers indexed by ProQuest or other online news sources (e.g., BBC, CNN or Reuters).

In answer to the second research question, few WoS Science (5.2%) and Social Science (5.7%) journals were cited by any of the eight UK daily newspapers 2006-2015. Whilst it was already known that some journals are more likely to be covered by the press, this is the first large scale evidence that most are completely ignored. The results also show differences between newspapers. For instance, the Daily Mail had the most and the Sun had the fewest citations to different academic journals. About half of the cited Science (42%) and Social Science (43%) journals in the eight UK newspapers were medical (including Clinical Medicine, Immunology and Pharmacology) or about Psychiatry/Psychology respectively, indicating that UK newspapers tend to report medical and mental health care research more commonly than other subjects. The British Medical Journal (BMJ) was the most frequently cited journal (from 12% in both the Daily Mail and Daily Telegraph to 22% in the Sun). New England Journal of Medicine and other specialist British journals, such as Lancet, British Journal of Cancer, British Journal of Sports Medicine, and The British Journal of Psychiatry were also frequently reported. Within Social Sciences, a range of Psychiatry/Psychology journals were frequently reported in news stories, such as British Journal of Psychiatry and Journal of Epidemiology & Community Health (a BMJ journal). This suggests that UK newspapers tend to report a significant proportion of research from British journals (see also Entwistle, 1995) in addition to prestigious multi-disciplinary journals like Nature and Science. One reason might be that UK newspapers prefer to report research findings that have more general interest or benefit for public in the UK such as public or mental health issues and British journals may publish more research of this kind. Another reason could be that science journalists or editors have more access to British-based information from journals, such as publishers’ press releases, authors or experts in the field for writing up their news stories. For instance, there is evidence that UK science writers more frequently look at popular science journals such as Nature and Science in addition to the UK’s leading medical journals, such as BMJ and Lancet for ideas for their reports (Weitkamp, 2003). Discoveries by UK academics may also be intrinsically more newsworthy for UK newspapers and UK university press offices may have more contacts with UK-based science journalists.

In answer to the third research question, overall the Daily Mail and Daily Telegraph had the most coverage of scientific research over time, eclipsing the broadsheets that may be thought of as the more logical source of research news, whereas the Sun had the lowest. During the period analysed there was no general trend for an increase or decrease in the coverage of science, although there was an increase for one newspaper and a decrease for another. In the current era of increasing competitiveness of newspapers and cost cutting, it is encouraging that coverage of research has not suffered, at least in major UK daily newspapers.

In answer to the fourth research question, almost all academic journals mentioned in the sampled news stories were cited to report research and only 6% were used in a non-research context. Most news stories reported research extensively (about 60%) and beyond one source (53%), typically other published research or interviews with experts. There were differences in the reporting good or bad implications from the research between newspapers, with the Daily Mail for reporting more good news (48%) than bad news (24%) compared with the Sun reporting more bad news (47%) than good news (24%). Nevertheless, the keyword-based topic analysis of news stories citing Science journals suggests that risk factors for cancer, heart diseases and medical issues about children and women were particularly common topics
in all newspapers (Tables E and F in the appendix). This is the first reported comprehensive analyses of these issues.

In terms of practical implications for academics, the results emphasise the difficulty in getting major press coverage of published research, since most journals were ignored. Health continues to be an exception, as are prestigious journals, and so work fitting these criteria should continue to be promoted to the media. The most encouraging news is for the social sciences. Although the study confirms the ongoing low level of coverage of social research, as a proportion of SSCI/SCI journals the difference between the sciences and social sciences is small. Hence, social scientists that publish health related research or in prestigious journals should not be discouraged from promoting their work to the press. Articles are likely to be discussed in reasonable depth using additional sources but without subjective evaluations. Thus, press releases that point to full text versions of papers and make available other sources of information might be particularly helpful for journalists.

Finally, the automatic method introduced here could help researchers, universities, news organisations, governments and scientific publishers to assess the press uptake of published research. With the new method, this can be achieved more quickly and on a larger scale than before.