The impact of free access to the scientific literature: a review of recent research

Studies based on surveys and interviews

Over the past thirty years, periodic studies of the information-seeking and reading behavior of scientists in the United States have revealed three trends: (1) scientists are reading more articles while spending less time searching and retrieving them; (2) scientists now read from a broader group of journals and extend their reading into the older literature; and (3) scientists have become more reliant on institutional (library) access to journal collections, although they also make use of informal sources such as preprint servers and colleagues [13–21].

Overall, most surveyed researchers indicate that they have adequate access to the scientific literature and that access conditions are generally improving. An international survey of authors conducted in 2004 revealed that 61% of respondents indicated that they had “good” or “excellent” access to the journal literature. Three-quarters (76%) of respondents also felt that access was easier than it was 5 years ago [22]. A similar study of immunologists and microbiologists conducted in 2004 revealed that two-thirds of respondents (67%) reported they had “good” or “excellent” access to the literature [23]. Nearly 84% claimed that access was much better now than it had been 5 years before, and almost all (97%) reported that they were “very up-to-date with the current literature in their area,” with less than 35% reporting any need for greater access to the research literature. In a list of 16 potential barriers to higher research productivity, journal access ranked 12th, below everything but greater conference/networking opportunities, better research leadership, more general management training, and clearer legal and ethical frameworks [23].

Surveying international scientific authors in 2007, Ware found that 69% of respondents reported having “good” or “excellent” access to the literature, a figure that varied by region [24]. Authors in the United States and Canada were most satisfied (85% reported “good” or “excellent” access; 3% reported “poor” access), while those in other countries were significantly less satisfied (53% reported “good” or “excellent”; 15% reported “poor”). In Ware's latest study, conducted in 2010 [25], 93% of all respondents reported having “very easy” or “fairly easy” access to the journal literature, a finding which varied from 97% for North American respondents to as low as 78% for African respondents.

Ware also surveyed small and medium-sized commercial enterprises in the United Kingdom, including companies involved in pharmaceuticals, chemicals, computing, and medical and precision instruments [26]. More than 70% of respondents claimed that they had “very easy” or “fairly easy” access to the journal literature, and 60% reported that their level of access had gotten better over the past 5 years. Among enterprises with more than 250 employees, 82% reported having easy access to the journal literature. Ware's study used a sample of convenience, however, and only 4% of potential respondents completed the survey.

While greater access to the scientific literature has undoubtedly altered the behavior of scientists as readers, it is not among their main priorities as authors. In deciding where to submit their work, an international survey of scientists revealed that authors were most concerned with each journal's reputation, readership, impact factor, and speed of publication [27]. In contrast, access status and copyright-related issues were ranked last.

Although these large-scale studies are prone to several forms of bias, such as sampling bias and nonresponse bias, they are confirmed by a number of smaller but more rigorous studies of authors' preferences. For example, authors submitting manuscripts to the British Medical Journal reported that impact factor, reputation, readership, speed of publication, and quality of peer review were all major factors in deciding where to submit their work. In contrast, only 13% rated free access as important [28].

In-depth interviews with authors and other stakeholders reveal similar values and priorities. A series of interviews with faculty at the University of California, Berkeley, revealed that authors were most concerned about the mechanisms that ensure the quality of journal articles and the integrity of the peer-review process. Other considerations, such as the costs of the scholarly publication system, did not have much impact on their work [29]. Although many librarians “strongly perceive a crisis in scholarly communication,” most faculty do not, underscoring a fundamental perceptual difference between these two stakeholder groups.

A recent and more extensive study, by Harley and colleagues, explored the values, motivations, and behavior of 160 interviewees at 45 research universities across the United States [30]. The recurring theme in this report is that the scholarly community is conservative, with a firm dedication to disciplinary norms and an unyielding commitment to external peer-review and assessment. There is little room for experimentation with new forms of publication, especially for new academics. In particular, many faculty view author-pays open access publishing with suspicion due to the potential conflict between publication fees and rigorous peer-review. Overall, Harley and colleagues “heard little about a crisis in scholarly communication” [30]. The only faculty who perceived a crisis were some of those in the humanities, who noted that the diminishing output of university presses made it more difficult to publish their work as academic books.

Several months after the Harley report, Schonfeld and Housewright released another report on the perceptions and behavior of faculty with regard to scholarly communication [31]. Their findings, based on a series of longitudinal surveys conducted every three years since 2000, were remarkably consistent with those of the Harley report. With regard to publishing, faculty attitudes are fundamentally conservative and are guided almost entirely by career advancement. Faculty expressed little interest in transforming the scholarly communication system, and across all disciplines, free access to journal content was consistently ranked last among the factors considered by authors when selecting a journal in which to publish. In contrast, the absence of publication fees was ranked as the second most important factor, suggesting, as Schonfeld and Housewright [31] pointed out, that the author-pays model of open access publishing might be at odds with the attitudes of many faculty. As numerous studies have shown, the primary goal of most faculty is to publish in journals that are widely read by scholars in their field [23, 27, 28, 32]. If transforming the scholarly publishing system is a goal of faculty, that goal is nonetheless eclipsed by issues of career advancement.

Table 1 (online only) presents a summary of the major studies described in this section.

Studies based on unobtrusive methods

Distinguishing primary literature from secondary and user-generated sources

The Pew telephone surveys list many sources of medical information including websites, blogs, commentary, and podcasts, but they do not ask respondents to distinguish between scholarly and non-scholarly resources [74, 82]. In particular, they make no specific mention of journals or scientific articles as sources of medical information, although they do distinguish “Internet” from “printed reference material” [82], as well as identifying cell phones and mobile applications [83]. Likewise, the Health Information National Trends Survey (HINTS) of the National Cancer Institute asks several questions about sources of health information, but the available responses represent a potentially confusing set of information providers, media, and delivery mechanisms [84]. For example, question HC02 asks, “The most recent time you looked for information about heath or medical topics, where did you go first?” and offers response categories such as “Internet” (61.0%); “Doctor or health care provider” (13.9%); “Books” (8.4%); “Brochures, pamphlets, etc.” (3.8%); and “Magazines” (3.4%). It is difficult to discern what the researcher is implying—or, indeed, what the survey respondent might be thinking—when asked about Internet use. Magazines, books, brochures, and health care providers can all be accessed via the Internet, just as doctors and other health care providers communicate in person, via the Internet, and through books, brochures, and magazines. A more recent (2010) survey of the health-seeking behavior of adults separated “Internet” from “Media,” the latter consisting of newspapers, magazines, and television [85]. Neither of these studies distinguished between the primary medical literature (e.g., scholarly journals) and popular health magazines.

Most medical and health-related websites suffer from significant problems in terms of accuracy, bias, and completeness [86, 87]. Nongovernmental websites are especially prone to these problems [88]. While laypersons claim that they use a number of criteria in evaluating the credibility of medical websites, few of them actually check the authority of web resources or are able to later recall the sources of their information [79]. Indeed, just 15% of telephone survey respondents claimed that they “always” checked the source and date of the information they found online, and 10% claimed that they did so “most of the time” [89].

Evaluating websites retrieved through a series of online searches for medical terms, Laurent and colleagues found that the user-generated online encyclopedia, Wikipedia, ranked higher in the search results than professional sites such as MedlinePlus, which is maintained by the National Library of Medicine and the National Institutes of Health [90]. Indeed, Wikipedia appeared on the first page, among the first ten results, for the overwhelming majority of medical keyword searches conducted in Google, Google UK, Yahoo, and MSN. Despite significant errors of omission and the absence of source attribution [87], Wikipedia is a prominent source of online health information.

No study has systematically evaluated the extent to which the general public makes use of the primary medical literature rather than secondary and user-generated sources such as Wikipedia and WebMD. Nonetheless, the available evidence shows that “the Internet” (however survey respondents might define it) is the primary source of health-related information for the American public and that typical Internet users are far more likely to encounter secondary sources of health information than the primary health sciences literature. Research has shown that the quality of consumer health information can be improved through better integration of the primary literature into online health resources and through public involvement in the development of such resources. However, even these efforts involve the use of tools and interfaces that repackage, summarize, and simplify the results of medical research—not on the unmediated reading of the scientific literature by the general public [91–93]. Overall, the published evidence does not indicate how (or whether) free access to the scientific literature influences consumers' reading or behavior.

Sources of bias in the research literature

Most studies on the use of the primary scientific literature reflect the opinions and behaviors of those who are well integrated into the system of scholarly communication. This might be regarded as a form of bias because the true population of interest includes not just those who currently rely on the research literature, but those who might make good use of it if they were more fully aware of its utility. Arguably, the greatest value of open access journals and archives is their potential to make scholarly information available to those who consume, but do not contribute to, the scientific literature. At present, however, there is little evidence that this potential has been realized. Admittedly, the existence of a resource (such as a library collection or a body of open access literature) has value independent of the number of downloads, citations, or other indicators of use: option value (the value of the personal right to use the resource in the future), existence value (the value derived from the knowledge that the resource is available for others' use), and bequest value (the value of making the resource available to future generations of scholars) [94]. However, these kinds of utility are difficult to measure; are not as readily apparent to authors, readers, and publishers; and are of less immediate value than the indicators of use that drive the present-day scholarly communication system.

Impediments to broader access

The studies presented here suggest that publication fees are perhaps the greatest impediment to broader participation in open access initiatives. While fewer than 30% of open access journals charge publication fees, those journals represent half of all open access articles [95]. Moreover, the proportion is much higher for particular disciplines and publisher types. For example, publication fees were levied for approximately 69% of the open access biology articles published in 2008/09 (all publisher types combined) and for 76% of the commercially published open access articles in 6 subject areas [95]. At the same time, there is clear evidence that faculty generally do not like publication fees [31], and many fear that the pay-to-publish model may compromise the integrity of peer review [30]. It is important to realize that individuals' opinions of general publishing practices may be contrary to their attitudes about particular journals. For instance, PLoS Biology and PLoS Medicine both charge relatively high publication fees, but neither seems to have trouble attracting submissions. The business model adopted by these 2 journals may or may not be transferable to other journals or publishers, but it does demonstrate that at least some free-access journals have the potential to transcend the limitations noted here.

Alternative routes of access to the scientific literature

Most research on access to the scientific literature assumes a traditional and hierarchical flow of information from the publisher to the reader, with the library often serving as an intermediary between the two. Very little has been done to investigate alternative routes of access to the scientific literature. If consumers of the scientific literature operate in the same way as consumers of cultural media (music and video, for example), studies that focus on publisher-initiated communication might overlook important components of the scholarly communication system. While interlibrary lending is one such component, the direct transfer of documents among colleagues may represent an even larger flow of information. Gaulé's [59] study of access to scientific information in India suggests that informal peer-to-peer sharing is very common in countries with a history of poor access to the scientific literature. For authors, the practice of fulfilling reprint requests by physical post has largely been replaced by transmtting articles by email or by directing readers to publicly accessible online archives or the public Internet [96].

A 2009 mandate of the US National Institutes of Health requires researchers in receipt of federal funding to deposit their final, peer-reviewed manuscripts in PubMed Central [97]. Several research universities have also set forth their own self-archiving requirements [98]. Open access archives such as PubMed Central may attenuate existing access inequalities, but only if authors' manuscripts are adequate substitutes for the published record. This requires, among other things, that manuscripts are posted online in a timely fashion, made readily discoverable by readers (through indexing or other mechanisms), and preserved for future use.

Further research

As noted earlier, there is a clear need for empirical research dealing with the impact of free access on the use of the scientific literature by the general public and by communities that consume, but do not contribute to, the scholarly literature. It is also important to gain a better understanding of the constructs and measures that are commonly used in studies of scholarly publishing. For instance, to what extent do downloads and citations correspond to the careful reading of articles?

Research on free access might also be improved through more careful attention to the various forms of bias that persist in both surveys and observational studies. For instance, studies of the impact of free access on citation rates have been hindered by a number of methodological problems including selection bias and incomplete model specification (e.g., failure to control for all relevant confounding variables). Studies of free access are likely to benefit from a greater understanding of these problems and a stronger consensus on the most effective ways of dealing with them.

Finally, further research should investigate the extent to which scholars rely on informal sharing of the scientific literature. This kind of sharing is often poorly documented, since individuals are not required to track and report document transactions to the publisher, nor are institutional repositories, subject repositories, public websites, or peer-to-peer file sharing programs. By relying entirely on publisher-provided usage data, researchers may under estimate the extent to which scholarly works are disseminated through such secondary and tertiary pathways. It is possible that these alternate methods of access help to attenuate the current inequalities in formal access to the scholarly literature.