The Big Three types of research misconduct—fabrication, falsification, and plagiarism—are clear failures to meet science’s epistemic responsibility for results published in peer-reviewed journals. They are condemned by virtually all national policies on research misconduct (Altman, 2006; Resnik et al., 2015; U.S. Federal Policy on Research Misconduct) and are easy to comprehend, criticize, and convey to the lay public. However, they are peripheral to the current epistemic problem.

The sources of current failures are subtle, hard for non-scientists to grasp, and morally ambiguous—hence the general label “questionable research practices,” or QRPs (Gardner et al., 2005; Martinson et al., 2005; John et al., 2011; Fang et al., 2012; Ioannidis et al., 2014).⁴ Although the “replication crisis” in psychology may be most widely known, it is a tip of the QRP iceberg.⁵ In some cases, non-replicability (or non-reproducibility) can be traced to QRPs (Begley, 2013). But results that are not replicable may not have involved QRPs; there may be no attempted replications of results in which QRPs may or may not have occurred; and studies that are QRP-free may not be submitted for publication at all if the results are negative or mixed.

The general context is one of the failures in mechanisms for producing and publishing research results in a reliable way. Failures in research include various ways of nudging results of a study in a desired direction. Failures in publication include practices that result in epistemically damaging distortions in a field’s publication record. Some QRPs are common enough to have labels. “p-value fishing” or “p-value hacking” includes analyzing data in various ways in order to obtain a result that falls within the p ≤ 0.05 threshold for statistical significance, the conventional level at which one can reject the null hypothesis. “Adaptive sampling” involves stopping data collection prematurely once a wanted result is obtained, or continuing to collect data if results are nearly, but not quite, statistically significant; sometimes data are excluded after looking at its impact on statistical significance. Sometimes dependent measures that do not reach statistical significance are not reported in order to eliminate “imperfect” results, which are difficult to publish. The “file drawer effect” is when studies that have negative or non-confirmatory results are not submitted for review, given that peer-review journals prefer high-impact results as well as new studies (non-replications). This yields systematic bias in the pool of papers submitted for review and in the publication record (Easterbrook et al., 1991). There is also lack of thorough peer review of submissions (Begley, 2013; Alberts et al., 2014).

These and other technical—to outsiders, mysterious—misbehaviors “present greater threats to the scientific enterprise than those caused by high-profile misconduct cases such as fraud” (Martinson et al., 2005). QRPs may or may not be a slippery slope to fraud in individual cases (Crocker, 2011), but they can make a field’s entire publication record unreliable. John et al. (2011) estimated that in psychology the actual prevalence of various adaptive sampling techniques was 100% based on high rates (62–72%) of self-admitted cases. Kühlberger et al. (2014) found evidence of a cumulative effect of QRPs: in their sample of 1,000 published papers across psychology, about three times as many studies just reached the p ≤ 0.05 threshold than just failed to reach it.

Despite the widespread epistemic damage they can cause, QRPs are much harder to eliminate than the Big Three. A field’s publication record cannot be quarantined via retraction the way selective publications by a single researcher can be. QRPs are also morally, legally, and professionally ambiguous. The Committee on Publication Ethics Code of Conduct (Committee on Publication Ethics, 2011) for best peer-reviewed journal editor practices lists “ensuring the integrity of the academic record” in its formal code; the associated recommendation states that “errors, inaccurate or misleading statements must be corrected promptly and with due prominence.” QRPs are not clear cases of any of these. At best, they “lie somewhere on a continuum between scientific fraud, bias, and simple carelessness,” and while many scientists admit to “alteration” or “modification” of data, they also do not think this amounts to falsification (Fanelli, 2009). To date, QRPs are treated as “inconsequential” rather than “career-ending” sins (Simmons et al., 2011); “there is no real consequence for investigators or journals” for engaging in them (Begley, 2013).

Note that QRPs are not limited to psychology. Martinson et al.’s (Martinson et al., 2005) misconduct survey was of researchers who had received funding from the U.S. National Institutes of Health; Ioannidis et al. (2014) report evidence of QRPs in neuroimaging and preclinical studies, including animal studies; and Fanelli (2009) report that misconduct was more frequently reported in their survey by medical/pharmacological researchers than others. However, even if the current crisis were limited to medical research and psychology, these areas are major sources of science news.

The concept of a conflict of interest, in both research and media ethics, is usually narrowly defined in terms of possible personal gain or loss avoidance. In today’s scientific climate, the primary conflict of interest is between the pursuit of publishability and the pursuit of truth (Nosek et al., 2012). There is an analogous conflict of interest in journalism whenever the ethics of inquiry take a back seat to the imperative of public appeal. This conflict is exacerbated when journalists are unable to satisfy the epistemic norms of their profession, as discussed in the next section.

Many journalists take, or frequently have no choice but to take, a stance toward science characteristic of a member of a lay community, which lacks the expertise to assess results and is not involved in the research yielding those results (Grasswick, 2010). Laypersons judge what to believe by judging whom to believe (Anderson, 2011:145). In this case, scientists have specialized knowledge, and the lay stance in science journalism is a response to this knowledge deficit. It is a perspective from which science happens in a separate box (Lief, 2015). It induces a sharp asymmetry in perceived epistemic responsibility between scientists and journalists regarding the production of reliable science news.

The lay stance is institutionalized and reinforced by the practices of assigning reporters without scientific knowledge to science stories or assigning those with background in one science to report on another. In the U.S., if not in all countries, budget and staff cuts in mass media (Brainard, 2009; Murcott, 2009; Nature, 2009; Russell, 2009) entail that big-Journalism science reporting from the lay stance is the rule, not the exception. Institutions that do not have the public interest at heart can step into the breach (e.g., Göpfort, 2007, in Germany). Although general reporting experience is more highly correlated with reliable science stories than training in a science (Wilson, 2000), it is unlikely that only experienced reporters are being assigned science stories.

There is ample pragmatic justification for these assignment practices. As one British science news editor put it:

If a guy’s got a paper in Nature that’s been subject to peer review I have absolutely no qualms about quoting everything he says in full and being unquestioning. That sounds awful in a way but we’re a high speed operation, you know. (Hansen, 1994)

But the lay stance is not essentially a pragmatic response to time pressure. It is rooted in the historical fact that science reporting initiated (in the U.S.) as a cheerleading enterprise that involved translating science for lay audiences in order to win public appreciation for the benefits science provides to society (Lewenstein, 1992; Brainard, 2008; Rensberger, 2009). However, while science’s high socioepistemic status has been tarnished by greater recognition and reporting of the bad consequences of science, there has been little change in journalists’ inability to recognize bad science. While science has become more quantitative—adding mathematical and computational modeling to statistics and other traditional mathematical tools and reasoning methods (e.g., Castellano et al., 2009; Baronchelli et al., 2013)—many journalists still cannot do the math (see below). As long as the lay stance persists, it is misleading to characterize this journalistic shift as one from “science lapdog to public watchdog” (Rensberger, quoted in Brainard, 2008).

The epistemic helplessness of the lay stance becomes obviously problematic when we have good reason to be skeptical about general research and publication practices in a field. Engber (2016) notes “We’d like to think that a published study has more than even odds of being true.” From the lay stance, the reporter is unable to improve upon these odds. The actual reliability of a particular result is not affected by quoting other scientists. The public still may as well toss a coin to determine its credibility. Compounding this with the journalistic bias for novelty, it is more likely than ever that the public will be exposed to unreliable science news (Dunwoody, 2008/2014; American Association for the Advancement of Science and Center for Public Engagement with Science and Technology, 2015). Even in a field with a reliable publication record, an individual study flawed by QRPs is a problem for science news.

The obvious response is that since the scientists possess the knowledge, they are responsible for the epistemic integrity of science news. This position is unjustified. Suppose the science journalist sees her epistemic responsibility as limited to not inserting noise into the communication channel between scientists and the public. Without someone to do epistemic triage in an affected field—a function that properly conducted science makes moot—even a noiseless transmission channel will spew reliable and unreliable research alike into the public sphere. Epistemic norms would counsel not covering the affected fields until researchers get their act together. Suppose the journalist sees her professional role as more than being a pipeline between scientists and the public. She must still avoid epistemically misleading framing choices and other mediations. The lay stance makes it a matter of luck or intuition whether her choices are epistemically virtuous ones.

Current norms of objective journalism or the “journalism of verification” (Kovach and Rosensteil, 2001) sidestep journalistic epistemic responsibility. The objective journalist strives for verifiable facts, accurate reporting of events, impartial reporting and writing, and a detached, impersonal point of view. Practices associated with this norm—modeled on the scientific method (Nelkin, 1987; Schudson, 2001)—include using neutral language (detachment), fairly representing positions and actors (impartiality), and getting both sides of the story (balance) (Schudson, 1978, 2001; Mindich, 1998; Schudson and Anderson, 2009; Vos, 2011). But these practices guide descriptions of people and their words and actions. They presuppose the ability to verify facts and ensure accuracy, which the science journalist reporting from the lay stance does not have. Trust is essential to journalism, but not blind trust. In this case, “trust but verify” becomes “trust, and trust some more.”

When divorced from the ability to verify, these practices can wreak havoc on the epistemic status of science news. When reporters cannot ascertain the actual level of credibility of research, they may be more inclined to frame a science story in ways that raise or lower its appearance of credibility instead (Dixon and Clarke, 2013). They may be more likely to use epistemically damaging metaphors, such as by drawing unjustified analogies to human stereotypes (e.g., Hoffman, 2016). They may use balance (or competing voices) without considering that this practice transforms any degree of uncertainty about the facts into a difference of opinion about the facts. This is epistemically catastrophic on the science beat. We expect scientific opinions to be based on the best available evidence, but uncertainty framed in terms of differences of opinion suggests all scientific opinions are equally valid and evidence is secondary or irrelevant. The practice also lowers the credibility of science journalists (Jensen, 2008 for cancer news) and empowers industry to manipulate public opinion and sow doubt (Antilla, 2005 and Wilson, 2000 for climate science).

Problems in climate science communication illustrate how standard journalistic practices can transform reliable science into unreliable science news. Nearly unanimous scientific consensus on anthropogenic causes of climate change (if not on specific consequences) enabled environmental reporters to do their jobs much as Boffey did decades ago. But journalistic norms of balance have contributed to public confusion (Boykoff and Boykoff, 2004; Antilla, 2005), while social psychological research has shown the inadequacy of the “deficit model” of science communication, according to which science illiteracy is the reason behind public opposition to scientific findings, new technology, and environmental action (Kahan et al., 2011 in the U.S., Besley and Nisbet, 2011 for a more general view in the U.S. and U.K.). If climate science turned out to be riddled with QRPs, we would be back to the start of this paper.

As with QRPs in science, these practices also have pragmatic justifications. Usual pressures of time and audience appeal have been ramped up into a hypercompetitive reporting atmosphere that incentivizes science “infotainment” (Rehmann, 2013). The quantity of science available on the Internet promotes reliance on press releases from science journals or research institutions and homogeneity in coverage [a loss of “infodiversity” (Granado, 2011: 810)]. Nevertheless, as Nature editors (Nature, 2009) remark, “society needs to see science scrutinized as well as regurgitated if it is to give science its trust, and journalists are an important part of that process.” Scientists are asking for effective watchdogs. But short of becoming scientists, how can journalists shoulder their epistemic responsibility?

In complex epistemic communities, no one individual has the expertise to verify a result and epistemic virtue depends on the efforts of the group. Interdisciplinary scientific collaborations are a primary example of such communities. Science journalists and scientists are another, with science news the interdisciplinary product. Each group is highly motivated by practical pressures: the scientific equivalent of “What does this mean for the everyday person?” is “What does this mean for a follow-up study and another grant?” Each group individually can undermine the reliability of science news by engaging in profession-specific QRPs. They need to work together to ensure an epistemically virtuous product. Specific steps might include the following.

The author confirms being the sole contributor of this work and approved it for publication.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.