Confronting Uncertainty: The Contours of an Inferential Community

This monograph addresses the question of how journalistic knowledge work, and in particular inferential reasoning, as a process of uncertainty reduction is manifested in news texts. We argue this takes place both in and in-between news media within a community of practice. The main premise is that journalistic texts reveal communal processes of knowledge creation and it is within these texts that we see the contours of what we term an “inferential community.” The backdrop to this, is that the digital (news) landscape, political developments, and global issues produce an environment rife with uncertainty. We focus on three contemporary cases around the current U.S. presidency. We are, however, not arguing that the processes we study are altogether new; journalists have always, alone or together, grappled with uncertainty. Rather, we present here a conceptualization based on the premise that current circumstances offer a window into the more fundamental processes of journalistic knowledge work based on inference.


Introduction
In March 2017, BuzzFeed News's Politics Editor Katherine Miller (2017) adroitly commented, "Given the current level of uncertainty (does Trump really mean X?) and the sheer volume of incoming information (what will Trump do tomorrow?), each day demands your judgment. Is this normal? Is this normal? Is this normal?" (emphasis in the original). To what extent we are witnessing a new state of normalcy is an open question, but what seems certain is that in recent years, journalists have experienced a growing sense of uncertainty both in terms of what and how they cover news. While uncertainty has always been present to some degree, what has arguably changed is the pace and manner in which it is introduced into our lives. As news ricochets rapidly online, it has created, as Miller opines, an unending stream of alerts that something-an uncertain something-has happened. What also remains constant in this historical trajectory of confronting uncertainty is the role of journalists and news in working to reduce such uncertainty. Lippmann (1922) described this as providing a "true picture" of the world which the rest of us cannot see, recounting news events which (as he writes) come at us "now fast, now slowly"; they nevertheless come at us (p. 1). Whether by steamer or tweet, news consumers have looked to journalism to resolve uncertainty, to get a "true picture" of reality. Yet, in a news landscape characterized by a multiplicity of online platforms with divergent origins and epistemologies, it has become increasingly difficult-for both citizens and journalists-to construct and maintain such "true pictures" while navigating what may seem a complex maze of facts. As Waisbord (2018) writes, in the current moment, the nature of certainty, not to mention facticity, has become contested, as " [r]elatively esoteric questions about truth-telling, philosophical realism, and mass deception suddenly gained currency in public debates" (p. 1866).
It was within this context that we addressed uncertainty in "Negotiating Uncertain Claims" (Eldridge & Bødker, 2018). This monograph builds upon and expands that article, which identified a series of textual markers of newswork and journalistic practice; and posited that journalists, confronted with news stories that were rife with uncertainty but nevertheless warranted coverage, engaged inferential practices to arrive at best-formed conclusions. That study offered our first take on inference as a dynamic of uncertainty resolution, that is, "an initial probe into the public performance of certain journalistic practices that (may) be increasingly prevalent in an age of uncertainty" (Eldridge & Bødker, 2018, p. 1921. This monograph adds a more solid theoretical foundation based on literature on journalism as knowledge, on inference, as well as on more specific ways of thinking about how journalists make probable claims. All this forms the basis for revisiting the two cases at the center of the pilot study, a new case, and more extended reflections on our conceptual and analytical approach. This monograph presents an investigation of how communal processes of knowledge construction through inference leave textual traces in the products of journalism. Such textual traces can provide insights into how journalists attempt to construct feasible accounts of events within what Ananny (2018) calls a "networked press" and a complex "infrastructure," where journalists, institutions, audiences, and technologies interact and co-constitute each other. We argue that processes of communal knowledge creation have become both more necessary because of a growing level of uncertainty and more visible through the ways in which digital news become more processualthat is, published incrementally, before the whole narrative has been fully researched and then updated as new aspects become available-and more intertextual, with greater interlinking including through hyperlinks and embedded media artifacts, such as tweets. We see this as amplifying dynamics previously seen in journalism, as news has to some degree always been intertextual as journalists monitor their competitors, citing and following other outlets' reporting, and engaging in an explicit or implicit dialogue between other texts and other journalists.
Some aspects of today's media landscape and changing news ecology, however, make these processes of intertextuality via linking, embedding, and of course quoting increasingly significant. First, a seemingly greater number of highly complex and global issues, from climate change to tax evasion, necessitate international collaboration. Second, because journalists can now produce "more differentiated forms of knowledge" (Nielsen, 2017, p. 104) online, both news users and journalists are able to piece together what they know, or can know, from a range of publicly available types of knowledge. Some of these have developed instrumentally, as solutions to particular reporting challenges. Others seem to reflect an increasingly networked environment. This has resulted in a growing variety of types of content and of media forms-from brief reports offering "news-as-impressions" to complex, distributed, and datainformed "news-as-relations" that "go beyond the individual article" (Nielsen, 2017, pp. 101-102, p. 108). This signals, at least partly, how journalism has expanded to include a wider range of digital outlets which locate themselves at different points on the scale of verification and newswork, and where news takes a variety of forms (Eldridge, 2018).
Our research here considers not only how both traditional and emergent digital outlets demonstrate awareness of each other, operating with different allegiances toward journalistic verification, but also how they use each other's output to construct a foundation for news claims. In developing the notion of journalism as an inferential community, we build an empirically grounded theoretical conceptualization of the process of knowledge construction as collaborative within and between different institutions operating in a complex news landscape. Such a perspective will not only help elucidate certain contemporary traits of journalism but also help focus on a relatively neglected aspect of journalism studies, namely, how most journalistic texts contain intertextual elements that implicitly or explicitly reveal a dialogue with other journalistic texts across both time and space.
We explore three cases where an evident lack of information challenges conventional practices of journalistic knowledge production. Specifically, we attempt to identify how journalists come together in confronting such situations through textual manifestations of what we term an "inferential community." To consider how communities of journalists coalesce in such circumstances, our pilot study showed how journalists build on scarce facts by demonstrating their own authority and that of their peers so that their news stories seem definitive to the public. Borrowing from British philosopher Elizabeth Anscombe's (1957) work on intention and inference, we see inferences performed within and between news texts. Lasorsa and Lewis (2010) use the concept of inference to distinguish between "deceptive" news and "legitimate" news. Inference is a "statement about the unknown based on the known," in contrast to a report, which is "a statement capable of verification" (Lasorsa & Lewis, 2010, p. 379). We, however, apply the notion of inference to news where journalists are forced-for various reasons-to deal with information that is very difficult, even impossible, to verify, yet carries potential societal significance that cannot be ignored. This results in journalistic narratives pointing to "a theoretical conclusion about the means to the end you pursue/are pursuing" (Gjelsvik, 2014) where the "end" is a clarification of facts through the deflation of uncertainty. Inference is thus seen as (textual) means to avoid the uncertainty endemic to journalistic cultures (Hanusch, 2009). When viewed as intentional activities expressed through discourse, these convey an authoritative journalistic performance around the expected role of providing information to a public in their interest (Eldridge, 2018).
Following the election of Donald Trump as president in 2016, two events exemplify the dynamics that we explore here. The first was the publication of a dossier of alleged but potentially incriminating activities linked to, among other things, possible collusion between (then) President-elect Trump and Russia, published in full by BuzzFeed News in January 2017. The second was President Trump's unsubstantiated tweet in March 2017 accusing President Barack Obama of wiretapping Trump Tower. Without clear factual details, these events posed specific challenges for journalists, yet they similarly demanded attention. The dossier was a set of largely unverifiable facts. While the rumors in the dossier had circulated during the campaign, BuzzFeed News's publication made its claims public. While this, on one hand, produced a rift between news outlets, it also revealed how knowledge was created across these outlets and how-by agreeing this was an important issue to discuss-journalists came together to negotiate their positioning as, using Zelizer's (1993) terms, "subcultures of interpretation within the larger community" (p. 226). Such communal negotiations are not predicated on agreement, Zelizer notes; rather, they emphasize matters of common concern. The tweet, on the other hand, was presented as fact, but this was unverified. It shook political circles when it was sent early one morning, without evidence, and in response journalists were left searching for possible explanations. Both cases draw attention to how facts are negotiated when authoritative voices are absent (as with the scarce support for Trump's tweeted claims), and when digital journalism changes the ways that information reaches the public (as with BuzzFeed's publishing the dossier).
These cases also pose further questions as to what these dynamics mean for a journalistic community that now interacts and coalesces through hybrid practices of community-building and value-adding, and as different actors support and at times supplant traditional roles (Benkler, 2011;Chadwick, 2013). To further investigate such practices, we also look at a third case: the coverage of the summit between President Trump and Russian President Vladimir Putin in Helsinki in July 2018. We chose this summit as a third case to see if the dynamics found in the first two continue to emerge in a very different set of circumstances. Because it was a planned event (as opposed to the first two cases) and took place 18 months into Trump's presidency, this case allows us to reflect on processes of inference at some distance from the novelty of the new administration. We are aware that working on three cases all related to the coverage of President Trump may undermine the wider applicability of the ideas. Yet, our reasoning is that the peculiarity of Trump and coverage of his presidency has opened a window to more fundamental aspects of journalistic practice. Put differently, the ways that journalists react to Trump may reveal aspects of reassuring "normalcy" precisely because Trump breaches expectations (Berkowitz, 1992). We have, therefore, focused our attention on the structural aspects of news coverage without being drawn (too much) into discussions of the peculiarity of Trump.
Regarding our overall aims, we are primarily proposing and developing a conceptual approach rather than presenting a definitive empirical study. While the study builds on both inductive and deductive processes, the amount of empirical data and the types of news institutions used here point toward further studies. Our hope is nevertheless to conceptualize communal practices of journalism, exploring how engagement between journalists found in texts reflect a field of journalists coalescing as a community of practice (Lave & Wenger, 1991). A community of practice resembles an "interpretive community" (Zelizer, 1993). While the latter mainly is concerned with how journalists collaboratively develop perceptions of pivotal incidents, the former is more focused on journalistic production practices, which are, however, related to broad perceptions of communality. What we propose with the notion of an inferential community is a lens on how practices link up with communal aspects of interpretation, reasoning, and inference.
Below, we first turn to our two main and interrelated theoretical building blocks, namely, journalism as knowledge and as inference. In setting the stage, we are not providing conventional literature reviews but rather attempting to outline our theoretical foundation. Having said that, the section on inference eventually does discuss literature focused on journalism and uncertainty reduction. As inferential reasoning is an important element in the creation of knowledge, we first discuss journalism as knowledge, which we conceptualize as interactions between contexts of production and consumption, interactions that revolve around journalistic texts. Based on that, we progress toward inference and how this is publicly performed in news texts. Next, we turn to our methodology and analytical framework where we present our approach to the analysis of the textual traces of inferences both within individual news texts and between texts. The aim is thus to identify knowledge as interactions and inference as performances in texts, well aware that newsroom practices and audience practices constitute further vital contexts of public knowledge construction in and around journalism. Taking the notion of an inferential community (or communities) beyond texts would be a highly relevant next step. Finally, we turn to our three cases, in chronological order and we end by summarizing our main findings, offering concluding reflections, and suggesting possible paths for more research related to issues raised in this monograph.

Journalism as Knowledge
Issues related to uncertainty are connected to journalism's role in modern societies, which is linked to the development of journalism away from mainly happening in relatively local domains toward a more professionalized and complex institution. Giddens (1991a) refers to such processes as "disembedding," through which social practices are less and less embedded in our immediate social and temporal contexts. Rather, they happen through larger and centralized institutions at some distance from our social contexts. This movement, Giddens argues, necessitates a significant level of trust in distant (expert) systems. Something similar has happened to the way knowledge is constructed within journalism, as news (in a very broad sense), has moved from being a more integral part of local knowledge in the 19th century to also being the domain of increasingly large, centralized, and bureaucratic institutions "disembedded" from local contexts in the 20th. In parallel with Giddens's argument about trust, such a "distancing" of journalism arguably necessitated a more visible textual demonstration of uncertainty reduction; and this was, in a broader context, needed for journalism to be accepted as the main platform for the exchange of information and knowledge for society (as argued by Anderson, 1983;Ekström, 2002;Park, 1940;Tuchman, 1978;and others). This has at times been an uneasy development; Deuze (2014) describes as paradox whereby our ability to know our wider societies confronts the way, through media, we alienate ourselves from people and local contexts. Nevertheless, drawing on Finnemann's (2001) argument that all societies are information societies, modern societies and modern news media depend on these developments. Indeed, Giddens (1991b) argues that "the global extension of the institutions of modernity would be impossible were it not for the pooling of knowledge which is represented by the 'news'" (pp. 77-78).
Following these broad arguments and the subsequent needs for the institution of journalism to maintain its authority, we will elaborate below on journalistic knowledge as constructed through interactions between modes of production and consumption, interactions that turn on journalistic products. Journalism, writes Ekström (2002), is "clearly among the most influential knowledge-producing institutions of our time," responsible for producing the "necessary" knowledge for well-functioning modern democracies (p. 259). This function is embodied in a more or less implicit social contract through which rights and obligations of democracies are balanced between states, their institutions, their citizenries, and journalists (Sjøvaag, 2010), where the journalistic side of the bargain consists of its "claim of being able to offer the citizenry important and reliable knowledge" (Ekström, 2002, p. 260). Given this implied social contract, as the processes of knowledge production became more centralized, more industrial, and less transparent in the 20th century, scholars began to question how journalistic knowledge was being produced. This was linked to the perception that the news "industry" was increasingly defining our "reality" for us. By applying (various mixtures of) sociological and ethnographic methodologies and perspectives to explore newswork, pivotal studies such as Tuchman's (1978) Making News: A Study in the Construction of Reality connected newswork practices with the mediated presentation of reality. The resulting insights into the practices through which news comes into being remain important for understanding the conditions for producing journalistic knowledge (see, for example, Thomsen, 2018).
Journalism as a form of knowledge work is, however, more complex than what can be understood by looking at the practices and interpretations that lead to the making of news, as a journalistic product. In fact, one could say that the knowledge produced through journalism merely begins with the published product. Standing apart from other types of knowledge, a key attribute of journalism as a form of knowledge is precisely that it is public. In his article "News as a Form of Knowledge," Park (1940) expands on this: In order that a report of events current may have the quality of news, it should not merely circulate-possibly in circuitous underground channels-but should be published, if need be by the town crier or the public press. Such publication tends to give news something of the character of a public document. News is more or less authenticated by the fact that it has been exposed to the critical examination of the public to which it is addressed and with whose interests it is concerned. (p. 679) As journalism is a specific form of public knowledge that can be evaluated by audiences as well as other journalists, it also (ideally) addresses and calls into being a public. This occurs through the "circulation" (Warner, 2002, p. 50) of news texts. News as a form of knowledge-if we follow Park-is also therefore found in the social life or the "talk" related to the public documents of journalism. Moreover, the circulation of public documents undergirds the formation and maintenance of publics as interpretive communities that "determine lines of interpretation, found institutions, and set boundaries based principally on their own internal dynamics" (Lee & Li Puma, 2002, p. 192). The "interpretive communities" referred to here are, however, not made up of journalists, who-in Zelizer's (1993) application of this concept-are drawn together through the interpretation of events pivotal to the profession. Rather, these interpretive communities are drawn together by the circulation of news. The knowledge produced through journalism involves the coming together of two types of interpretive communities, namely those of production and those of consumption. While the former manifests itself in journalistic texts (in the widest sense of that term), the latter grows out of the circulation and interpretation of those texts and is manifested in various forms of mediated and unmediated "news talk" (Bird, 2011) and circulation among publics (Warner, 2002).
This stark and dichotomous rendering of two types of interpretive communities, however, ignores the many ways in which journalists and non-journalists interact on various participatory media platforms-for instance, those of social media. Such platforms have in recent years allowed journalists and non-journalists to engage in collaborative processes of knowledge production that somehow breach internal dynamics of somewhat disparate realms. Despite this change, journalism remains a specific type of public knowledge in the sense that it is constructed at the intersection of journalistic methodologies, their textual manifestations, and the material and interpretive processes through which it becomes part of public life. From the side of production, this is largely an issue of epistemology, that is, the specific ways in which journalists and publishers apply "rules, routines and institutional procedures" (Ekström, 2002, p. 260) to produce authoritative public documents. From the reception side, it is largely a hermeneutics question, namely, the specific ways in which texts get meaning through processes of interpretation.
The realms of epistemology and hermeneutics are, however, intertwined in different ways: First, because the production of the journalistic text occurs through interpretive processes related to the various sources and processes involved; second, because the text is also the product of an ongoing monitoring and interpretation of how the specific event is covered, including by competitors; and, third, because the construction of truth claims presumably is at some level linked to journalists' perceptions of what interpreters are likely to accept. Ekström (2002) alludes to this in arguing that epistemology in a communicative perspective is "partly a question of what is considered acceptable and sufficiently true knowledge" by those who produce news and by those who consume it (p. 261). As Hall (1980) might have paraphrased it, public knowledge of journalism is created through articulated processes of encoding and decoding.
News products continue to connect knowledge produced by journalists and knowledge acquisition by audiences, particularly those news products that (still) largely emanate from journalistic institutions. In arguing that news products present a public form of knowledge, Park (1940) distinguishes (via William James) between two types of knowledge: A broader "acquaintance with" and a deeper "knowledge about." The former refers to lived and accumulated everyday knowledge people use to navigate in the world. The latter reflects "formal, rational, and systematic" (p. 672) knowledge. An important difference here is that "knowledge about" is "communicable" in a way that "acquaintance with" is not (p. 673). Yet, Park argues, these two forms should be seen as on a continuum: "In such a continuum news has a location of its own" (p. 675) as it reports specific events that "orient [as much as inform] the public, giving each and all notice as to what is going on" (p. 677).
This type of knowledge is first related to the "disclosure of the fact itself, in all its singularity" (Meditsch, 2005, p. 130). Yet the knowledge constituted through journalism is also drawn out over time in the sense that news gains further meaning (or becomes knowledge) as it accumulates as social or public memory (Bødker, 2016(Bødker, , 2019. Put differently, because news involves a somewhat systematic reporting of events, it (ideally) offers continuous relevance for understanding and navigating the everyday. As such, news is the soundtrack to our lives, a continuous and accumulating stream where the most startling, emotional, and significant become part of our collective memory. Journalism thus sits between the systematic and formal knowledge of science and the accumulated, informal, commonsense knowledge of everday life. Meditsch (2005) observes, "While Science develops by rewriting common-sense knowledge in formal and esoteric languages, Journalism toils in the opposite direction" (p. 130).
The fact that journalism operates "in the logical field of the dominant reality" (Meditsch, 2005, p. 129) gives journalistic knowledge both its weakness and its strength. "As an analytical and demonstrative method, it is frail, since it cannot be detached from pre-theoretical notions in order to present reality. And it is strong, as those same pre-theoretical notions guide the principle of reality of its public" (Meditsch, 2005, p. 129). Such a conceptualization of news as knowledge has, in stronger and weaker forms, undergirded much of post-World War II research on news as knowledge; it was especially strong in critical theory emanating from the Frankfurt School and parts of the later (and partly related) cultural studies that leaned toward materialism and structural readings. Important aspects of such approaches are still prevalent in critical discourse studies of news. But, while such studies of news may be detached from considerations of its modes of dissemination, this is not necessarily the case. Meditsch's conceptualization above is linked to the position that journalism ideally seeks a "universal audience" and it is through this that it finds "its main social justification" (Meditsch, 2005, p. 130). While this is not about dissemination as such, it is linked to a view that (or a time when) what may be called mainstream journalism and mass media were relatively unchallenged. While a universal audience as an ambition of legacy media was never fully realized, it is now being supplemented by different modes of address in the contemporary heterogeneous and digital media landscape.
To make sense of news as knowledge in this heterogeneous space, and explicitly drawing on Park (1940), Nielsen (2017) suggests three "ideal-typical forms of mediated, public knowledge" for digital news: "news as impressions, decontextualized snippets of information"; "news-as-items, [. . .] self-contained, discrete articles" (the kind of knowledge conceptualized by Park); and "news-as-relations," a new explanatory form enhanced by data (p. 93). He argues this is critical to understand "the properties of digital news as different forms of knowledge-rather than a form of knowledge" (p. 93, emphasis in the original). While the kind of news that Park (1940) conceptualized remains a central part of news output, what we see now, Nielsen (2017) argues, is an expansion of mediated news toward both ends of Park's continuum, that is both more analytical and science-like news and more everyday impressions. News occupies a greater span of the continuum between "acquaintance with" and "knowledge about." Arguably this diversification of journalism as knowledge means that various forms of journalistic knowledge must be positioned-implicitly or explicitly-in relation to other forms of journalistic knowledge. This positioning takes place, now, in reconfigurations of the collective and hierarchical nature of news production. To begin with the collective, while the production of news to some degree has always been collective and cumulative in the sense that the production of news relied on what others already had verified, in a digital environment this is something that is happening to greater degrees among interconnected platforms and producers (Bruns, 2018), including on news aggregators and similar platforms (Weber & Monge, 2011). Regarding the hierarchical nature, while the production of news has also always been relational within a hierarchy of forms of knowledge, for example, broadsheet and tabloid, the increasingly heterogeneous media landscape has partly reconfigured these relationships as news content are also mediated through a broader range of platforms, for example, social media, which are structured along very different hierarchies; and such modes of dissemination leave traces in the form of news, for example, embedding tweets or linking to other outlets (Eldridge, 2019). The intertextuality of journalism related to both its collective and hierarchical relationships is thus increasingly complex and increasingly visible.
This means that some of the ways in which journalists embed textual elements meant to legitimize journalistic productions as knowledge have changed, including by becoming more explicit. In the 20th century, a great deal of authority rested with the institution, the journalist, and the profession; therefore, texts arguably did not need to work so hard to establish their authority. Yet, journalists to some extent always have relied on texts to demonstrate their status as authorities, in what Broersma (2010) calls "journalism as performative discourse." These days, this performance happens in a more complex and interlinked context, because news presents knowledge of a specific kind. This complexity is partly visible in news pieces themselves, including in the different ways that texts showcase the underlying newswork and journalistic authority that went into their production (Eldridge, 2017). This includes narrating various reporting practices, and articulating decisions of news judgment, observation, and interviewing to distinguish themselves from other media actors-including aggregators, leakers, and social media (see Coddington & Holton, 2014;Eldridge, 2017). As practices that legitimize newswork as knowledge work, these are visible in news analyses that attempt to raise interpretive implications of the news and in meta-journalistic discourses linked to news events in which the institutions try to legitimize their specific approaches to handling news events (Berkowitz, 1992;Carlson, 2015).
Yet, given the diversity of forms of knowledge, different attempts to legitimize journalistic authority have also emerged, including those which speak to the basic level of a decision to publish something that others might not. These legitimizing practices are grounded in different bases of justification, ranging between what is legally permissible and what is ethically sanctioned. Particularly with new digital actors claiming journalistic authority (cf. Carlson, 2017;Eldridge, 2018), Foley (2018) argues digital actors "exploit the gap between legality and professional ethics" (p. 5). As we shall see, these justifications also surround BuzzFeed's decision to publish the Russian dossier, and other journalists' reactions to this decision. At the center of this debate was the question of verification, processes of legitimizing the journalistic production of knowledge, and transparency.
In the context of knowledge work, the transparency of journalistic practices within news products, which enable readers or audiences to understand how a journalistic product came into being, bestow greater levels of confidence in the knowledge produced. This returns us to the social contract mentioned above, and the relationship between journalism and publics, where the journalistic side of the bargain is the pledge to produce the necessary knowledge for the functioning of democracy, and society can assume news will provide them, to a degree, with this knowledge. The reciprocity of the contract is further embedded in the structures of modern democracies where states guarantee free speech and access, and journalism guarantees the knowledge that society needs. This relationship has become embedded in the functioning of modern democracies. As Ryfe (2009) argues, "by the early 20th century, government had created an entire publicity infrastructure to support the newsgathering efforts of reporterseffectively subsidizing the news business with free information" (p. 202). Yet, this was never a straightforward barter. First, "the stories with the most potential public value . . . [often] are sensitive and confidential" (Ananny, 2018, p. 35). Second, as political communication has been increasingly professionalized, public utterances from state actors (and here, we primarily mean politicians) reflect strategically crafted pieces of knowledge, which often leave room for interpretation. In the wake of such professionalization, news analyses and commentary have emerged as a prominent form of interpretive knowledge work-including, in the digital context, through microblogging as a form of interpretative knowledge work. Social media streams, for example, "blend news facts with the drama of interpersonal conversation, and combine news reports with emotionally filled and opinionated reactions to the news in a manner that makes it difficult to discern news from conversation about the news" (Papacharissi, 2015, p. 32). Despite social media allowing for new modes of conveying knowledge (through news) and interpretation (through commentary) to publics, however, challenges remain to fulfilling the social contract between journalists and their publics. This is embedded in practices of investigative journalism, when knowledge production is aimed specifically at overcoming journalists' lack of access. This can, for instance, include not only the request of access to specific records but also the publication of unverified documents or the citing of anonymous sources. Such practices are reliant on the relative autonomy of journalists, who are not acting as informers to the state but rather as intermediaries between the state and the public (Ananny, 2018). When journalists have very limited access to people and documents, they have, broadly speaking, a choice between not going forward with a story and trying to fill the gaps between the dots in different ways. Without sufficient sources, the question is how can "investigative reporters accomplish the fundamental and very practical task of knowing what they know" (Ettema & Glasser, 1987, p. 4). This leans heavily on journalists "weighing the evidence" (Ettema & Glasser, 1987, p. 19). Ekström's (2002) assertion that "journalists must bear the responsibility for their assertions" is demonstrated in the ways that journalists present their work with varying degrees of probability, thereby guarding against certainty where it does not exist (p. 271). It can also be found when news reports rely on "accumulated" collective knowledge, as building blocks in a broader patchwork of information. This brings us to the dynamics of reporting examined in this monograph, during situations in which journalists feel obliged to publish new information, even if they cannot produce verification due to a lack of information from authorities and/or a lack of access to those who could provide such verification. In the section below, we explore reasoning practices to understand how journalists present performances of their newswork that reflect the navigation of this uncertainty; from this perspective, processes of inference are a fundamental part of the construction of journalism as a knowledge.

Inference
In positing that an inferential community of journalists is emerging in times of uncertainty, we draw attention to the ways in which journalists as social actors are particular knowledge workers who collectively see themselves as committed to understanding and then explaining the activities of other social actors. In this section, we first unpack our understanding of inference as a reasoning process, highlighting how inference aims to reduce uncertainty. We aim here to connect work on uncertainty reduction to ways of thinking about knowledge and inference that are particularly relevant in the contemporary setting.
While practices of inference and the formation of inferential communities may be more visible today, such practices have been constituent elements of how journalists evaluate claims posed by authorities who, due to their status, warrant journalistic consideration (Harman, 1965;Schudson, 1995). Thus, the natural skepticism that may come with confronting the claims of political actors, for instance, is met with the ability or obligation to see whether their statements warrant journalists' consideration. To understand how journalists rise to the task of explaining the unverified or unverifiable, we look to inferential reasoning to describe newswork in these circumstances as a process of drawing reasonable conclusions from the best available resources.
As a type of logic, inference is examined within several disciplines concerned with determining outcomes, from philosophy to probability (Anscombe, 1957(Anscombe, , 1974(Anscombe, /2005Gjelsvik, 2014;Menges, 1974). We outline practices of inference here in the context of journalists attempting to remove as much uncertainty from the pronouncements of political actors in particular, as both a response to misinformation (Waisbord, 2018) and as a shoring up of their own journalistic authority (Carlson, 2017). Inference helps to explain public-facing reasoning expressed in newswork, so this discussion will identify where inference fits within knowledge practices of journalists who aim to resolve uncertainty to the maximum possible extent.

What Is Inference?
Inference refers to the way we think about the "if/then" conditionals that people encounter and how, through processes of reasoning, we can try to determine the most likely outcome or outcomes. As a part of knowledge work, inference is primarily a process of conclusion-reaching, where reasonable assumptions allow a person to move from knowing a few things toward understanding larger truths. In short, if we know (or assume) something to be true, then we can also assume certain outcomes based on that knowledge (Anscombe, 1957;Gjelsvik, 2014). Inference, of course, also has a more everyday meaning related to how we infer meaning from what we think is implied in someone's speech, for instance, or infer from someone's actions to make sense of what we think best explains their motives (Von Wright, 1963). In this way, inference describes not only how people reason through the likely implications and explanations of people's actions but also how we can reasonably guess at their intentions.
We see journalists' work as devoted to this type of conclusion-making. Inference can be explored in the way journalists move from the information and facts they have gathered in reporting, toward explaining their meaning to the public. Inference also emerges when, trying to understand the actions of others, journalists assess the reasons for and the likely outcomes of decisions being made. This is a core aim of inference: to locate the most probable outcome of a series of behaviors (Menges, 1974). An emphasis on probable, or likely, explanations draws from Charles Peirce's foundational work on inference (cf. Kloesel et al., 1993). Peirce articulates processes of resolving uncertainty as processes of reasoning through what is known to be true to then determine what is likely to come next.
Peirce describes inferential reasoning as a logical process, reaching conclusions based on taking things that are known to be true (or generally assumed to be true) about a specific circumstance and linking these logically, to move forward toward conclusions (Kloesel et al., 1993). Inference has been applied extensively in mathematical probability and statistics, where it is specifically not concerned with subjective or personal inference (Menges, 1974). Peirce and those who have built upon his work, however, regard inference not merely as a mental exercise of logical thinking but also as reflecting more general practices of individuals. Practical inference takes into consideration the actions of individuals to understand the judgments behind such actions and their outcomes (Gjelsvik, 2014). Journalists examine the actions of others in society in terms of their intentionality; they assess both the practices (the actions) of individuals and their likely outcomes (conclusions), as well as the judgments necessary in shaping those practices. It also reflects the ways journalists work backward, moving from newsworthy statements or events (as outcomes) to try to ascertain the most likely facts which preceded those outcomes (Anscombe, 1974(Anscombe, /2005. Inference, therefore, offers us a lens through which to understand news texts as reflecting reasoning practices, allowing us to operationalize inference as a specific type of journalistic knowledge work. Because inference often occurs unconsciously (Von Wright, 1963), philosophers find it difficult to categorize such reasoning (Gjelsvik, 2014). Yet, precisely because the inference process is so natural, it has unique purchase within our discussion of sense-making around the work of social actors like journalists, who work to produce knowledge about the everyday social world (Baird & Baldwin, 2001). In doing so, the casting of likely implications, explanations, and intentions into a news narrative which seeks to explain a set of activities is central to the production of news. In opening a discussion such as this, which extends philosophies of reasoning into the messier confines of the social world within which journalism sits, the first demand is to make practical the considerations of these discussions. We are aided in this attempt by understanding inference as a part of human reasoning which tries to produce knowledge and reduce uncertainty by filling in gaps in explanations of social actors' practices and judgments.

Practical Inference and Journalism Practice
In terms of its contribution to understanding the social world, journalistic newswork aims in various ways to associate actions of members of society with their intents. Journalists search for the likely explanation as to why specific actions were taken (particularly when this is not explicit) to place such activities in a larger context. The overall goal is to relay the goings-on in the world, while explaining to the degree possible why those events occurred, and any implications they have for the public (Carey, 2007). This is at the core of journalistic work (Hanitzsch & Vos, 2018), as presenting newsworthy occurrences within "socially shared frameworks" of understanding to explain what happened, why it happened, and for whom it is relevant, and so on (Conboy, 2013, p. 2). At the most basic level, this can be seen as the sense-making that is at the core of journalistic practice (Goode, 2009). More often than not, when it comes to sense-making around the actions of individualsparticularly public officials and political figures-this is not merely a description of incidental occurrences, but of the motivations and implications of their intentional acts.
Such practices of sense-making are also at the heart of practical inference, as the area of logic which explores the relationship between premises, judgments, and resulting actions. These are what Von Wright (1963) describes as "(necessary) means to an end" (p. 160). Gjelsvik summarizes this as the process of developing "a theoretical conclusion about the means to the end you pursue/are pursuing" (Gjelsvik, 2014). Weinberger (1998) describes practical inference as the consideration of premises and outcomes (or "operation results"; p. 125): Premises: Comprise the ascertainment of the existence of an intention of an agent, and the claim that a certain action (or: only this action) will lead to the fulfillment of this intention; Operation result: The performance of the action (or the thesis that it is necessary to perform this action).
Practical inference thus means paying attention to reasoning around the actions of social agents to determine what are the necessary and likely steps (premises) in order for an agent to produce a particular result (conclusions). Put differently, this is the thinking through either the steps taken, or those one would need to take (which Gjelsvik labels "means") to reach an eventual outcome (which he labels "ends"). Gjelsvik (2014) examines ways to look at both judgments and actions involved in an occurrence. In most of these descriptions, we are talking about inferring forwardfrom premises toward conclusions-as a series of steps involved in reaching an eventual outcome.
The work of Elizabeth Anscombe on practical inference (1974/2005) and intention (1957) points to a more concise way to think about this type of reasoning, albeit reasoning backward through the practices of external actors. To reach this point, Anscombe (1974Anscombe ( /2005 first specifies that practical inference requires us to consider actions which are intended-rather than incidental-to then examine hypothetical means and ends. Following Anscombe's guidance on intention, we consider communicative actions or speech acts as intentional acts (Habermas, 1984)-seeing publications and tweets as more than accidental utterances-whose premises and possible implications can likewise be subject to inferential reasoning. Following this line of thinking, and given that the President is "the single most symbolically potent and legitimate source of authority" (Schudson, 1995, p. 1), journalists are-at least initially-almost duty bound to take up processes of inference focused on Trump's utterances. Harman (1965), as outlined further below, makes a more general point in relation to this in arguing that even when the statements of politicians prove erroneous in the long-run, their pronouncements in text or speech should be considered as intended in the moment they are made, based on the gravity of their status, and the possible implications of even minor statements they make.

Formulations of Practical Inference
We use Gjelsvik and Anscombe to structure inference as a process of reasoning where the conclusion is dependent on each premise as a building block toward that conclusion. This requires each premise, first, being agreed to as legitimate, and second being conditional; that is, premises are part of the necessary processes for working toward a specific goal (Von Wright, 1963). In practical terms, we can reimagine this conditionality when evaluating any act as the outcome of a series of if/then conditionals, whereby the conclusion only holds true if each premise is treated as also holding true. Gjelsvik (2014) proposes this in his formulation of practical knowledge (P) and judgments (J) necessary, here, for a journey to Stockholm (p. 164): Starting Proposition: I am driving to Stockholm (P); Proposition 2: If I am driving to Stockholm, I am turning left at crossing X (J); Proposition 3: This is Crossing X (J); therefore, Concluding Proposition: I am turning left (P).
In this formulation, two specific actions are spelled out-driving to Stockholm and turning left-where we link an individual's practices to awareness of these practices as intentional actions, because neither heading toward Stockholm nor turning the steering wheel are incidental occurrences. These actions are then linked to the judgments made in Proposition 2 and Proposition 3. A chain of premises and actions lead to an outcome. This offers a way to think about intentional actions as both conditions for and conditioned by other decision-making processes, within a series of practices and judgments. We can relate this, for instance, to the ways that utterances by political figures are both prompts for news coverage (conditions for), and how inference may be applied within news coverage to explain an utterance or speech as responding to likely explanations (conditioned by).
Transposing this route-mapping exercise onto other occurrences shows how we can infer, from propositions and premises, the rationale behind political or state activity, for instance, how a declaration of war is based on premises of treaty-violations, or how a series of policy positions link political propositions to legislative outcomes. However, we risk applying an overly rigid formulation to the ways in which one set of social actors (e.g., journalists) infer the intentionality of any other given actor (e.g., politicians). One way of breaking from this rigidity comes by reformulating inferential reasoning in terms of "if/then" statements. The example from above would thus look like this: This offers a formulation closer to "everyday inference," from which we can look at the intentional actions of social actors in terms of inferred expectations (If I am doing X, then I can expect to do Y next) and inferred outcomes (If I have done X, then it is an outcome of Y). In doing so, as Gjelsvik (2014) notes, and for the overall reasoning to have explanatory heft, we still must either assume or agree to the legitimacy of the premises posed. Anscombe's (1974Anscombe's ( /2005 essay Practical Inference addresses this conundrum; we too wrestle with this, as inference insists on a certain acknowledgment that some premises, and some conclusions, appear to be more legitimate than others. This comes into focus when applying inferential reasoning to the actions of external actors whose judgments and practices may be (partly) obscured. Anscombe (1974Anscombe ( /2005 argues practical inference can nevertheless apply to making sense not only of those premises asserted but also those premises that are "merely supposed" (n.p.). This is particularly useful for grappling with uncertainty around the intentions of external social actors when we are availed only of some of their judgments or practices. In this sense, inference is applied to make the link between propositions-making sense of outcomes (Thens) as presumably based on conditions (Ifs), or vice versa assuming that when conditions have been met (Ifs), there are natural outcomes (Thens). This distinguishes inference toward a conclusion [If A occurs, then infer B will occur] from inference of the "facts to what would be premises of such a conclusion" [If B occurs, then we can infer that A did occur].
Work on practical inference offers important caveats as to the degree to which we can be conclusive in reasoning through premises and conclusions. These caveats force thinking about the ways sense-making does not occur in a vacuum. While applying reasoning to understand the judgments of premises and actions around interactions between different actors with different roles in society, we also need leeway to acknowledge that inferring does not insist on all premises being true. Instead, it operates on assumptions that allow reasoning to proceed as if they are true. Harman (1965) describes this as "inference to the best explanation" (p. 92). He argues that particularly when authorities speak, we are "warranted" in making such assumptions-at least to a degree-and treating these as true premises which contribute toward conclusions. Harman goes on to describe how even if statements turn out not to be true, there is a certain way in which inference naturally builds on accepting those premises (Harman terms these "lemmas") posed by authorities because of their authority. What consideration is warranted, therefore, is tied to questions of journalists' obligation to report on prominent figures due to their status, even those who are unreliable narrators. Harman argues, in line with Schudson (1995), that even if an authority's claim later turns out to be false, journalists need first to engage in the process of reasoning through these premises toward likely explanations, which may show the initial statement as false.
Inference to the best explanation also applies when either actions or premises are obscured from the observer (e.g., the journalist). Anscombe (1974Anscombe ( /2005 applies this to cases "when the conclusion is something to aim at bringing about, and the premises are possibly effectible truth-conditions of this, or means of effecting them" (n.p.). Fact claims can be considered as premises that have likely conclusions. We can also look at claimed conclusions as actions with certain likely premises. These "truth-conditions," which are also supported in Harman's (1965) work, allow inference to proceed on a path where each resulting premise and conclusion is woven together on the basis of each being true. As a reasoning process, this allows knowledge work within newswork to continue while recognizing that both premises and conclusions may later be determined to be untrue. Within journalism as a type of knowledge work this is especially important as, notably, the premises and conclusions journalists engage with can be precarious due to an array of factors that may have an effect on them. We will return to this idea below, where we connect inferential reasoning more directly to journalistic practice. It is also worth noting we are intentionally treating our cases without the benefit of hindsight, and considering practices of inference as precarious irrespective the eventual (long-term) conclusions of the stories being reported (Harman, 1965;Hawkins & Hastie, 1990).
The discussion above suggests a few features of inference. First, this reasoning process considers social actors' judgments and their practices. Second, philosophies of practical inference raise questions about how such judgments and practices are related in terms of their intentionality. Third, inference offers a way of reasoning through likely premises and likely outcomes, or at least those that are most likely under the condition that a conclusion can only be considered true (or likely true) if premises are also considered true (or likely true). Fourth, this process can proceed forward from premises toward conclusions (reasoning from means toward determining ends), or backward from conclusions toward premises (determining means from ends). Fifth, working in either direction, inferential reasoning is at its best when considered in relation to the contexts within which practices and judgments occur, where reasoning rests on the assumption that premises are true and where the actions examined are intentional; together, these conditions allow a movement toward the "best possible explanation" of these actions.

Inference, Uncertainty, and Journalistic Performance
We situate inference within the specific sense-making routines of journalists when confronted with uncertainty; this practice is reflected in the performative discourses of journalism and is aimed at uncertainty reduction. Dimmick (1974) helps us by discussing inference in his theoretical work around gatekeeping. Citing Coombs's (1964) A Theory of Data, Dimmick describes gatekeeping in terms of inference as a set of reasoning processes aimed at uncertainty reduction. Notably, however, the institutional structures at the core of Dimmick's conceptualization of gatekeeping and inferential reasoning in the 20th century (Cohen, 1963;Dimmick, 1974) are now diffuse in community "gatewatching" practices that may make such reasoning a more networked activity online (Bruns, 2005).

Uncertainty and Journalism
The goal of resolving uncertainty is not unique to journalism. It is an end goal for all types of knowledge work aimed at offering an accounting of reality on which others can build, whether as a foundation for further discovery (as with science) or for navigating the decision making of everyday life (such as journalism). Nevertheless, the status of uncertainty and its counterpart certainty differ across knowledge fields. Scientific fields are more at ease with the impossibility of certainty and strive to emphasize uncertainty to demonstrate their awareness of the limits of scientific discoveries (Anderson, 2018). But journalism as a field of knowledge production has been less enthusiastic in highlighting the limits of its conclusions. Instead, journalists express devotion to being as certain as possible, incorporating more and more measures of surety into their practices (Carlson, 2017).
Indeed, working against uncertainty remains a core concern of journalistic practice, as both journalism's commitment to knowledge work and its expressions of journalistic authority have been geared toward demonstrating certainty-wherever possible (Barnhurst, 2005)-and offering the best explanation distilled from disparate facts (Ekström, 2002;Park, 1940). This is embedded in journalistic routines geared to presenting news-as-knowledge and in journalism's normative claims of truth-seeking (Hearns-Branaman, 2016;Hermida, 2015); this reinforces journalists' authority as knowledge workers and status in society. However, considerations of certainty and uncertainty do not merely reflect how knowledge workers treat the confidence surrounding their conclusions. They are also embedded within the public perception of different types of knowledge practice, and the interactive constellation of self-perceptions (how journalists see themselves) and external perceptions (how they are seen) which shape the practices of social actors. This is at the core of the tension between internal control and external evaluation, which defines the dominant vision of the journalistic field (Bourdieu, 2005).
Having uncertainty reduction as an end goal introduces a particular strain on journalistic practice as it depends on other social actors' contributions to convey news stories with authority. This is reflected in journalists' reliance on sources and accounts of reality external to their own to paint a picture of reality (Tuchman, 1976). The field thus would benefit from emphasizing uncertainty, should those external actors' accounts or motivations later be exposed as contradictory to any conclusions which journalists are first able to present (cf. Eldridge & Bødker, 2018;Lipari, 1996). Yet when it comes to the field's claims of presenting a definitive vision of the social world (Bourdieu, 2005), and its practices as geared toward presenting an authoritative picture of reality, journalism overall demonstrates an aversion to uncertainty. Too much uncertainty could diminish the field's ability to claim it is engaged in truth-telling (Ettema & Glasser, 1984Waisbord, 2018), and thereby lessen its authority (Carlson, 2017). As a result, journalists often emphasize producing as definitive accounts of events as possible, with the most definitive explanations possible. Journalists accomplish this by placing emphasis on the evidentiary basis for the conclusions drawn (e.g., emphasizing sources consulted, data examined, and expertise on subject matter) rather than the uncertainty of possible outcomes. As a result, "[n]ews stories tend toward the proclamation of certainties while masking uncertainties" (Carlson, 2017, p. 58).
This is not to say uncertainty is altogether absent in news coverage. Science news in particular plays with ideas of uncertainty more prominently than other news genres (Corbett & Durfee, 2004;Dunwoody, 2010;Miliauskas & Anderson, 2016;Olausson, 2009). It is merely to highlight that when it comes to journalists' reinforcing their contributions to society and presenting accounts of otherwise-opaque events in the world, journalists emphasize what can be concluded, rather than what cannot, in order to convey certainty about the political world (Cook, 1998;Matthes, Morrison, & Schemer, 2010).
Notably, the differences in accounting for uncertainty seem at least partly based on the dual symbolic and commercial incentives for journalists in being seen as expert knowledge workers. Journalistic authority rests on a public recognizing that news and newswork contribute something of value in knowing the world (Carlson, 2017;Nielsen, 2017;Park, 1940). The public then supports such work through financial contribution (Conboy & Eldridge, 2018). So, both symbolic and economic capital are potentially targeted in the uncertainty-averse way journalists present their work. Thus, to reinforce its public perception and value, journalism is committed to resolving uncertainty to the degree it can. The journalistic field's boundaries are patrolled by practices that reinforce certainty and yield as little ground to uncertainty as possible.

Resolving Uncertainty as Journalistic Practice
For all journalism's attention to emphasizing certainty, we begin with the premise that certainty may be, ultimately, unobtainable. Uncertainty is also an uncomfortable reminder that we cannot predict or control our social worlds. Highlighting uncertainty can lead to what Anderson (2018) describes as the "uncertainty twitch." So, while uncertainty may be unavoidable, it is understandable that journalists start from the position that trying to orient one's daily life through news is easier for readers when they can do so with a sense of certainty. Journalistic practices accordingly respond by drawing on an array of resources to bring certainty about, or at least what seems like certainty.
Further exploring uncertainty, Dimmick (1974) theorized journalistic gatekeeping routines specifically as "processes of uncertainty reduction" (p. 3) processes which simplify journalists' and editors' practices of deciding what items seem newsworthy, and those items that seem most likely to be true and worth taking seriously. He treats gatekeeping routines as "input/output" processes that take place in two stages. Journalists and editors first determine the universe of possible news items (N) as inputs into a gatekeeping routine, to then determine which news stories or outputs will make up the daily news content (n) to be printed or broadcast (Dimmick, 1974). They involve "a sensing or input identification process and a valuating or output defining process" (Dimmick, 1974, p. 2, emphasis in original). Sensing processes include those routines which allow journalists and editors to reduce the uncertainty about the potential universe of items to cover, to a smaller portion of possible news items they will cover, while valuating moves further from identifying possible stories toward shaping how those stories will be developed (Dimmick, 1974).
In determining "the potential universe of news items" (Dimmick, 1974, p. 6), journalists and editors consult a variety of sources to identify which news items they should pursue into news stories, and prominent among these resources are other news institutions and journalists. Dimmick specified as "allied" resources groups and institutions both internal and external to the newsroom. Internally, these include peer journalists and editors with experience on a specific beat and those with awareness of institutional policy (and therefore, aware of what is most likely to make it all the way through editorial review). Externally, Dimmick mapped how editors at smaller newspapers have been shown to consult larger wire services in shaping their daily news budgets, how regional afternoon papers have looked to metropolitan morning dailies, how journalists grappling with specialist stories have looked to prominent specialist reporters, and how network television looked to large national dailies.
Through these routines, journalists relied in part on the work of their more prominent peers to simplify decision making in their own institutions, quoting Breed (1955), "[t]he influence goes 'down' from larger papers to smaller ones, as if the editor is employing in absentia, the editors of the larger paper to help 'make up' his front page" (p. 278). 1 By consulting the array of media resources-or "reference institutions"editors could more quickly reduce the range of stories that could be covered to those that should be covered because of their newsworthiness, because news institutions wanted to avoid being "scooped" by their competitors, or because they knew such stories would resonate positively with their superiors or institutional policies (Dimmick, 1974, p. 14). This process also helped accelerate individual journalists' own reporting, including by relying on the reporting of others "as a news source (i.e., using his facts or adopting his interpretation of an event)" (Dimmick, 1974, p. 11). By consulting other journalistic professionals and other news media, individual journalists and editors "use[d] the social groups and institutions to which he is allied to reduce uncertainty concerning the propriety of believing and acting in certain way" (Dimmick, 1974, p. 9). It further allowed journalists and editors alike to feel confident that their decisions aligned with what a fellow journalist agreed was newsworthy; what Dimmick (1974) described as a "validation process" (p. 9). Such "allied" resources were also utilized by journalists seeking a different angle on news items (i.e., opting not to cover something which has already been reported on or using an alternative news wire to their competitors).
Thus, Dimmick argued that gatekeeping was not only a solitary decision-making process by one editor or newsroom, but one which drew on shared processes of uncertainty reduction where the work of a variety of actors and institutions contributed to the decisions of individual journalists. While many of the conditions Dimmick's work refers to have changed over the years, including the makeup of the news media environment, writing several decades later, Cook (1998) continued to find journalists relying on a panoply of resources when assessing what items to cover, and what stories were likely to gain traction. He also highlighted the ways journalists consulted within peer groups when resolving uncertainty, identifying among examples of these "allied" resources, the Washington breakfasts for reporters and officials [. . .] or the daily internet compendium of campaign coverage, the Hotline, that provide opportunities for a lot of journalists to find out instantly what a wide range of colleagues are thinking and saying-not to mention a chance for the reporters who get diffused to have their work validated and reinforced. (Cook, 1998, p. 80) These accounts paint a picture of journalists and editors working within the larger community of practice toward reducing uncertainty around specific news items warrant further investigation. However, these routines have also been criticized, as they can contribute to journalists' pack mentality covering the same stories through largely similar lenses across media, something also noted by Tuchman (1972). Nevertheless, considering the practices of journalists in terms of how they engage with those resources to which he is allied, we can see from Dimmick a longer history of journalists and editors engaging in reasoning processes within inferential communities, allowing journalists to draw conclusions about the newsworthiness and definitive account of an event/item by relying on the editorial work already pursued by other institutions and journalists. Uncertainty reduction, in these routines, is linked to the ways that individual journalists engage in the constellation of practices to reinforce their own reasoning around news practices. The goal of these practices, while on one hand steeped in an acknowledgment of uncertainty, is nonetheless using communities of practitioners, their work, and their "allied" resources to present an authoritative case of reality.

Inferential Communities: Journalistic Evaluation Within Performances
To move toward seeing journalists operating within an inferential community, the evaluation of fact claims needs to be further situated within journalistic practices. These practices are linked to epistemologies of journalism and journalists' knowledge work, which are further evident in the textual traces of newswork based on these dynamics. In their work on epistemologies of investigative journalism and in later work on the "moral voice" of investigative journalists, Glasser (1984, 1998) explore these as two discrete processes of justification and verification. These two activities come into play when journalists engage in examining premises and outcomes of events. Ettema and Glasser's (1998) approach to considering what one is justified in believing and what can be verified can be linked to the attention to "virtue" in work on practical inference, which assesses the legitimacy of believing a premise or conclusion (Gjelsvik, 2014, p. 164). For Gjelsvik, virtue refers broadly to whether one is justified in believing a premise when engaging in inference. Linking this practice to newswork, journalists engaging in inference can strive toward the larger aim of relaying "true accounts" through news, as long as their practices are carried out truthfully (virtuously), even if they are fundamentally limited in an ability to lay out "the truth." Ettema and Glasser reinforce this, noting that while "the truth" may be an unrealistic goal for journalists, some journalistic practices are undertaken in pursuit of truth. As they argue, for knowledge work, truth is not merely applied to aspects of a story (as true premises toward true conclusions), but is an outcome of an overall process of truth-telling: To understand truth telling as a practical endeavor, the truth must be conceptualized and analyzed not merely as a condition that some statements happen to fulfill but rather as the outcome of a process for generating and defending those statements. (Ettema & Glasser, 1998, p. 132) The nature of truth in this dynamic is in the representation of things that are true (or considered to be true), and how the representation of such things contributes to truthful accounts of reality. Put differently, journalists engage in "a particular process for the production of social knowledge functions" and through that process work "to understand what it means in the context of that process to tell the truth" (Ettema & Glasser, 1998, p. 14).
Examining journalism's epistemology-how journalists know what they know- Ettema and Glasser (1998) distinguish between journalists' knowledge work around the verification and the justification of propositions. Verification relates to empirical validation; justification focuses on the grounds upon which the proposition rests. "[T]he practical tasks of collecting, evaluating, and assembling evidence to establish grounds for acceptance of a belief are all included in our conception of verification" (Ettema & Glasser, 1998, p. 157). Verification requires the empirical demonstration that facts are set out to be true. Justification, on the contrary, is more complex, and its connection to truth more conditional: "Justification and truth are conceptually distinct, however, because a true belief may not be justified (it may, for example, have been a lucky guess); conversely, a justified belief may not be true (the evidence may have been deceptive)" (p. 157). In a 1984 work, they point out that there is nothing peculiar about journalists acting only on how credible they judge their beliefs to be: [V]eracity is often not a very useful criterion for judging the quality of our beliefs, because veracity requires verification and verification requires experience; and experience, all too often, is either impractical or undesirable. (Ettema & Glasser, 1984, p. 27, ff. 2) Often, therefore, we have little choice but to act on what we have good reason to believe will be true, for what we know is true is too often too limited to be the sole guide for our conduct. Rarely, it follows, is the justification of a belief its verification. More often than not the justification of a belief is grounded in its credibility. Thus, the important issue is not whether a belief has been verified but whether its justification can be defended as rationally credible.
The contribution made by justification becomes clearer when looking to the ways in which journalistic actors make sense of other social actors. When considering hard-toverify or unverifiable truth claims made by political actors, as in this monograph, justification enters in force. This not only illustrates news practices in times of pronounced uncertainty, it is part and parcel with journalistic practices that are by their nature involved in confronting the sometimes-messy interactions of the social world (Ettema & Glasser, 1998). In particular, this connects to inference when justification is involved in evaluating the "truth-conditions" (Anscombe, 1974(Anscombe, /2005 under which utterances are made. In the same way Ettema and Glasser are interested in "conditions of justification"-and point to "a discursive process through which beliefs become justified" (Ettema & Glasser, 1984, p. 4)-practical inference provides a means of reasoning that is in service to this larger goal of offering reasonable, truthful, accounts. Reasoning becomes further embedded within journalistic practices for examining things known or assumed to be true, even if they do not purport to stand in for "the truth." In this case, true accounts are based on things that are warranted to be believed as true (Harman, 1965). Knowledge-seekers proceed from that basis to assess the premises and determine outcomes which follow as something that can be reasonably arrived at.

Journalistic Performances
In line with practical inference and the overarching insistence on knowledge workers representing reality truthfully, one way that newsworkers can demonstrate their evaluations is by emphasizing that they have evaluated premises and outcomes of news stories. This includes explicitly identifying those claims that journalists have determined warrant further explanation (Harman, 1965), as well as explicitly noting certainty around fact claims within these discussions wherever possible (Anderson, 2018), while also openly acknowledging the conditional nature of premises and conclusions in any process of reasoning (Anscombe, 1974(Anscombe, /2005. Seeing news practices as reasoning practices of inference emphasizes the importance of journalists providing credible accounts of society. Such practices have certainly been studied, for example, in ethnographies of journalists. Gans (1979), Schudson (1995), Tunstall (1970), Usher (2014), and Zelizer (1993) are among those who have tried to explain journalists' interactivity and inter-reliance on peer journalists.
This now takes place, we argue, not only across news institutions, which Dimmick (1974) examined, but is now also evident publicly, including through specific online practices of inter-reliance (e.g., hyperlinking), and within the narratives of journalistic routines within news texts. Because texts are performative spaces where credibility must be earned, they bear traces of this effort. Texts partly reveal journalists' reasoning practices, which they narrate explicitly as newswork routines that allow the news to be seen as credible, meaningful, and "true." Berkowitz (1992) has described how making reporting practices evident in texts allows journalists to convey to audiences that their routines can be trusted, and that journalists following established routines "can be effective even in times of significant pressure" (Berkowitz, 2000, p. 130). Similarly, Broersma (2010) treating this as a matter of form and style, argues that over time news genres and forms have developed such that the discourse conveys traces of these practices. While texts make apparent the practices of journalism, this is often only partly transparent; texts reveal only those practices that journalists want readers to see-the practices that reinforce journalists' credibility. Broersma (2010) notes two interlinked connotations of performativity: First, that of (re-)staging; retelling events and by doing so putting meaning on events.
[. . .] On the other hand, to ensure the effect of authenticity and truthfulness, journalistic texts rely on a set of professional practices, routines and textual conventions that were developed during the 20th century to guarantee that this process of construction or representation is as accurate-or mimetic-as possible. (p. 17) At the core of this performance is the marshaling of familiar resources, including recognizable narrative styles and the use of familiar experts and public voices. Such resources are "the basis of a shared social code between journalists and their public" (Broersma, 2010, p. 16).
As does Broersma, Eldridge (2017) emphasizes the performance of journalism: Not only are the processes and practices made evident but also the effort itself, the work. This is particularly apparent in newswork around stories whose digital nature (their novelty) risks overshadowing their journalistic contribution. Here, journalists are interested in both providing a truthful account of reality and demonstrating their role as providers.
Particularly with news stories that may reach audiences through other channels outside traditional news media control (Ekström & Westlund, 2019)-independently on websites or intermediated through social media-performing newswork helps make explicit the authoritative contribution of journalistic actors (Eldridge, 2018).
Taken together, performances of newswork in texts become narrations of the service which journalists provide as navigators of facts. Such "narratives of newswork" (cf. Eldridge, 2017;Ettema & Glasser, 1984) demonstrate journalistic role performances in various news formats. Texts are not merely conveyances for facts but places for performing the anticipated practices of sense-making, including inference when necessary. They exhibit the "shared social code" (Broersma, 2010, p. 16) and "shared social frameworks" (Conboy, 2013, p. 2) of interpreting reality. In other words, texts are where the basis for conclusions are explicated, reasoned through, and examined in public view for members of society, as well as other journalists, to better understand their reasonable conclusions.

Theoretical Conclusion
In the cases examined in this monograph, inference occurs when journalists confront unverifiable or hard-to-verify claims. These claims nevertheless warrant coverage because of the status of the speaker or whomever is being spoken about. When a president exclaims that something definitively occurred, or when an allegation about a president is made, this pushes journalists to act. Yet as we first explored (Eldridge & Bødker, 2018), when tweeted statements and undeterminable dossiers came to the foreground in 2017, they did so without premises or conclusions spelled out; so inference emerged in the work of journalists trying to piece together the best possible explanations. That they did so while drawing on the work of colleagues, across a networked array of media, offered an early indication that these practices were moving from the isolated practices of newsworkers in 20th-century newsrooms (Dimmick, 1974) toward the networked spaces of the 21st (Benkler, 2011;Heinrich, 2011).
In the sections above, we proposed seeing journalistic knowledge through the interactions in and around texts, that is, that the knowledge work of journalists partly is visible in news texts. It is in relation to these that journalistic knowledge becomes part of broader social contexts. Inference, as a vital element in knowledge construction through uncertainty reduction, is situated within the same contexts of interaction. Relatedly, we see the textual performativity of journalists as attempting to "win" legitimacy for the knowledge presented/constructed. The theoretical foundation is thus seeing journalistic knowledge and inference as interactions linked through textual performativity. In a digital setting, this is arguably more apparent but nevertheless tied to journalists' familiar routines and practices. In this monograph, we examine such practices in the seemingly unprecedented context of uncertainty which surrounds the Trump administration and its neglect for norms of validation (Zelizer, 2018).
Our monograph offers an interpretative framework for textually visible routines and practices dealing with uncertainty. As we show below, these practices are particularly evident in newswork performances when the verifying statements is either difficult or impossible for journalists to accomplish. Instead, journalists offer a reasoned best explanation of propositions and conclusions. In making this case, we are concerned with the efforts of journalists to speak with a degree of certainty and authority. We also consider the ways that the negotiating of propositions toward demonstrating reasoned conclusions-journalistic judgments and evaluations-can be observed in news texts where inferential reasoning is evident. We argue that in journalism's increasingly networked environment, the range of "allied" resources which journalists can consult and draw upon has expanded beyond established news institutions to include, for example, social media accounts, websites, and other digital media outside the journalistic field. Thus, the possible universe of items on which journalists can report has also widened, and a wider array of actors can contribute to developing news stories. Within this environment, journalistic practices of reducing uncertainty continue to be "made visible" through discursive performances of inference within news media: Texts reveal reasoning practices such as inferring from premises to conclusions.
Particularly online, this representation of newswork is also intertextual: The inclusion of links and hyperlinks to others within the "networked Fourth Estate" (Benkler, 2011) gives rise to a new imagination of "the social groups and institutions to which [one] is allied to reduce [their] uncertainty concerning the propriety of believing and acting in certain way" (Dimmick, 1974, p. 9). The embedding of media artifacts, including tweets, screencaps, and documents reveal online newswork that builds on premises posed by one media, toward conclusions by another (cf. Eldridge, 2019).
Online performances reveal even more clearly than print and broadcast journalism the practices of inferential reasoning, making even more apparent the continuing inferential community of journalists collectively engaging in such practices, and particularly in response to the challenge posed by unverified or unverifiable fact claims in times of uncertainty.

Methodology and Analytical Framework
The theoretical framework outlined above forms the basis for examining how journalists engage in knowledge work. This section highlights the specific ways we operationalize our theoretical foundation. Guiding our analysis are two overarching research questions: Research Question 1 (RQ1): How does knowledge work, and in particular inferential reasoning, as a process of uncertainty reduction manifest in news texts? Research Question 2 (RQ2): How does inferential reasoning take place within and between media as a community of practice?
To address these, we look for journalistic knowledge work via the textual traces of that work in texts, which are the public-facing performances of knowledge work. We outline below the categories of markers we expected to find in news texts, through which we can locate inferential reasoning. This involves taking our initial (2018) study as a first iteration of research toward understanding inference within contemporary journalistic practice, one that was intriguing but nevertheless under-theorized and thus fell short of fully explaining the dynamics at play. The empirical analysis in that pilot study prompted us to expand on what we saw as inferential practices in three main directions: First, we offer a stronger theoretical foundation for making sense of the traces of journalistic practice found within texts. Second, this foundation enables an enhanced, more nuanced analysis of the textual markers in news coverage. Third, we added a case to test our initial findings and explore further the dynamics of knowledge work. That is, this monograph offers a more theoretically informed analysis of two cases we introduced before and analysis of the new case, which-as will be described below-was chosen to address new issues.
Analysis was carried out both individually and together. Each researcher, first, performed a close reading and analysis of the corpus of news articles for each of the three cases, which were then analyzed inductively during open-ended, interactive research sessions in 2017 and 2018, each taking place over several days. Maxwell (2013) advocates an inductive open-ended approach to qualitative research, where interaction between researchers serves as not only a tool to build categories of interpretation and understanding but also a way to address validity concerns, with participating researchers openly challenging the individual interpretations of the other researchers within a reflexive qualitative approach. During these sessions, we adopted Maxwell's approach to revisit the categorizations and connections developed individually and to further refine these collectively. Recursive assessments of the news texts in our corpus enabled us to improve the analysis and findings (Srivastava & Hopwood, 2009). This was done interactively not to repeat analysis, but as Srivastava and Hopwood (2009) argue, to incorporate critical reflexivity at three points: Selfreflexivity (as researchers), reflexivity about what was studied (our cases), and reflexivity about the audience (the wider academic community). We returned to these forms of reflexive engagement to avoid deterministic interpretations of texts, and to contextualize our analysis within the larger body of scholarship to which we aim to contribute. We embrace the richness that follows from qualitative approaches based on a reflexive design.
To trace dynamics of inference within and across media, we initially explored news coverage of two events (described above and then again below) where information was presented publicly as either unverified or hard-to-verify (Eldridge & Bødker, 2018). We examined coverage in 2017 in The New York Times, a legacy newspaper; Politico, an online political niche media; and BuzzFeed News, a "clean-sheet" (digitalnative) news site (Küng, 2015, p. 2). This offers a most-different set of media, each with distinct types of news coverage. For each of the first two cases, we collected 144 (Dossier) and 135 (Wiretapping) articles to create a corpus that is suitable for exploratory analysis. Articles were obtained through LexisNexis with the search strings "dossier AND Russia AND intelligence" and "Obama AND wiretap," from January 10 to 16 (New York Times) and January 10 to April 12 (Politico) for the "dossier," and March 5 to 21 (New York Times) and March 5 to 22 (Politico) for "wiretap." From BuzzFeed, posts were gathered manually from its archives at Buzzfeed.com/Archive, from January 10 to 19 and March 4 to 26. A handful of duplicates per case turned up within the Times and Politico samples, consisting of re-published articles particularly on weekends. After removing duplicates, we sampled 25 to 35 articles per outlet per case for closer reading (200 articles total).
The additional case was news coverage from the same three outlets of the July 16, 2018, summit between Trump and Russian President Vladimir Putin. This time we worked with a wider set of parameters to capture reporting in anticipation of this event, as it was planned (in contrast to the previous two cases), beginning just prior to the summit and continuing until coverage waned. As a result, based on an initial scoping of texts, we gathered 50 articles from The New York Times, using the search string "Trump" AND "Helsinki" from July 15 to 17; 80 articles from Politico from July 15 to 19, with the same string; and 27 articles from BuzzFeed News from July 16 to September 12, which we manually gathered using its archive page. This netted an overall set of 157 articles, which we subjected to close reading for overall trends and then closer analysis for specific dynamics, examined below. Analysis was done on the web version of each article, rather than the LexisNexis and downloaded (BuzzFeed) archive we created so as to allow for intermedia analysis (including of hyperlinks and embedded media). In total, then, each researcher first closely read the full corpus of 357 articles before sharing findings regarding overall trends and dynamics. Then each researcher individually assessed a subset of the texts to produce the analysis (closely analyzing approximately 30-40 articles per case). This second round of analysis was completed for each case, based on the exchange of overall trends and dynamics. After each round of individual analysis, we came together to analyze coverage collectively over several days first in 2017 and then in 2018.

Analytical Framework
Our preliminary close reading revealed the repeated use of factual disclaimers and assertions of journalistic authority including demonstrations of "proper" journalistic performances. It also showed journalists both writing about the content of the claims (in the dossier, or tweet) and providing meta-commentary on the nature of journalistic practices, including how these relate to the current political landscape. Further iteration of analysis at each stage allowed us to outline two partly overlapping sets of discursive markers (Figure 1). The first contains markers of four aspects of journalistic practices which we would anticipate, based on the literature, to appear in news stories that try to address uncertain claims; these elements include disclaimers/doubt and caution, explicit newswork (e.g., explicit textual references to the activities of reporting, such as sourcing, investigating, verification, etc.), disclosure, and media-to-media references. The contextual marker "media-to-media references" can be part of both ordinary news practice and placing the event in broader contexts of journalism. The second of categories of language emerged out of the analysis, situating aspects of inference in the broader contexts of journalism: Subject/object switching, meta-commentary, and ethics/moral statements-all operationalized below. These markers allowed us to study performances of journalistic practice and practices of inference within news texts. They also offer a way to analyze journalists' progress in resolving uncertainty over time when they move from unsuccessful attempts to infer at one level toward attempts at inference at another. However, these practices did not occur uniformly, and markers often emerged together within the same article (or even paragraph).
We initially found that these discursive processes of inference occurred when journalists could not independently confirm or refute fact claims; yet, by linking and referring to the work of peers and drawing on both individual and institutional expertise, journalists could show they were working to verify fact claims. Through these processes, within a community of newsworkers, they could offer some semblance of uncertainty reduction around these stories. Our more fully developed analysis here both corroborates and expands analysis of such processes and news texts are analyzed as public-facing performances of journalistic belonging among an in-group of journalistic peers (Eldridge, 2017). Having expanded from the pilot study by adding new cases (including with the Trump/Putin Helsinki Summit in July 2018) as well as a wider theoretical framework, we explain these eight ways that, in the interests of authoritative knowledge work, journalists perform and offer explicit markers of their reasoning practices.
Disclaimer/doubt/caution. Following Harman (1965), and Anscombe (1974Anscombe ( /2005, we continue to focus on doubt and caution as important elements of inferential reasoning, particularly in instances of uncertainty around specific premises or conclusions. Harman's (1965) discussion of "inference to the best explanation" helps theorize the role of doubt and caution in presenting conclusions. Doubt reflects the limits of any conclusion, so, for example, journalists will highlight aspects of doubt within news texts by emphasizing the "if true" conditionality of reasoning. In particular, we expect caution to be present in texts that note that while the premises may not be fully confirmed, they are nevertheless necessary components to explain a news event (i.e., the premises necessary for reaching particular conclusions). To take a different example, not from our corpus here, reporting on Trump's June 2019 claims that Russia had pulled out of Venezuela, Reuters (2019) said, It was unclear exactly who Trump was referring to or how this was communicated to him, and the White House did not provide any clarity. There was also no immediate word from Moscow.
However, if true, it could mark a significant setback for Maduro. This example illustrated an expressed caution around the as-yet-unverified claim ("unclear exactly"), while still presenting a conclusion ("significant setback") if the claims end up being correct ("if true"; Reuters, 2019). These are presented as "necessary" conditions, even where they are "merely supposed" (Anscombe, 1974(Anscombe, /2005. Explicit newswork. Following Broersma (2010), the writing of all new reports contains traces of newswork within their form and style, and when it comes to practices of inference, we see such traces as explicit descriptions of how conclusions are reached. The presence of explicit textual markers of newswork, such as explanations of the investigative and verification steps taken in assessing certain fact claims, reinforce not only where conclusions may be drawn but also how the journalist draws such conclusions and the limits on any conclusions presented. These include demonstrations of how journalists have engaged with source material (explaining investigation steps, questioning sources, etc.) and clear explanations of reasoning processes (e.g., building if/then statements into such narratives, identifying gaps where inference is applied). This specific type of newswork is more explicit and makes clear to the reader how conclusions about a specific set of facts have been reached.
Particularly in cases when conclusive explanations are lacking-those scenarios which prompt the types of inferential reasoning discussed here-such explicit narratives of newswork strengthen the authoritative nature of the news story. When it comes to practices of inference, journalists "perform" journalism by articulating the necessary premises for reasonable conclusions (Gjelsvik, 2014;Von Wright, 1963). These performances also include how journalists formulate the intermediate conclusions necessary for presenting a truthful account of reality by evaluating evidence (Ettema & Glasser, 1998). Reporting on the National Rifle Association (NRA), to take one example, Washington Post journalists stipulated five times in one story that NRA officials "did not respond" to requests for comment, and so the reporting is based on "tax filings, state charitable reports and NRA correspondence reviewed by The Washington Post" (Reinhard, Zezima, Hamburger, & Leonnig, 2019).
Disclosure/transparency/normative claims (journalistic authority). Journalism remains indebted to public evaluation of journalists' work and public recognition of its value, which is closely linked to journalistic authority. Journalism is deemed valuable because its institutionalized methodologies lend it authority. Textual traces of journalistic authority are seen in phrases such as "according to sources close to Mr. X," which underscore a journalist's access and ability to evaluate information from various sources. When presented publicly, textual markers of reassurance and certainty (e.g., "While initial reports suggested X, multiple sources confirmed Y") can reinforce authoritative accounts (Carlson, 2017). Within these presentations are assurances of the roles of journalists in informing society. We see texts as not only emphasizing normative expectations but also reinforcing the idea that reporting on salacious, speculative, or sensational stories is necessary when they concern prominent public figures. Therefore, these discourses become closely linked to the evaluation of what type of coverage is "warranted" (Harman, 1965), how this is evaluated against the work of other colleagues within the journalistic community (Dimmick, 1974), and the role of judgment within the larger moral imperative which journalists adhere to (Ettema & Glasser, 1998).

Media-to-media references (with/without hyperlinks).
Media-to-media references involve two kinds of shared practices. First, Dimmick (1974) suggests comparing fact claims against a journalist's "allied" (p. 9) resources (consulting journalistic peers' reporting, news media, and social groups and institutions to which the reporter is allied) and seeing these resources as drawn upon through the input/output function of gatekeeping. This is particularly evident in texts online, where this process is more public via gatewatching and monitoring of the public flow of information alongside internal reporting and newsroom practices (Bruns, 2005). Second, our pilot study (Eldridge & Bødker, 2018) highlighted the ways in which journalists convey this consultation by including the array of accounts they consult within their own news stories via quoting, embedding of peer media (including via social media), and hyperlinking. This is all the more visible in an online context. Subject/object switching. Subject/object switching refers to differences in what stories focus on. With subject attention, the content of a speech act (or of a publication, such as of a tweet or dossier) garners the bulk of journalistic attention; with object attention, not what is said but how it is said is emphasized, including what medium is used (tweeting, for example). We draw this distinction based on the way journalists refer to the act of speaking as an object that warrants attention due to the status of the speaker or the way something goes viral. This is in contrast to instances when what they saythe speaker's subject claims-is the basis on which further news attention builds (Eldridge, 2019). When the subject matter gets the primary attention, this often reflects at least an initial acknowledgment that that material itself has some likely importance that warrants further consideration. When the object gets the primary attention, this often reflects the journalists' inability or unlikelihood to verify the claims any further, or alternatively that they see little value in trying to do so and do not take the subject seriously. Particularly in the digital environment, reporting contains mixtures of subject and object coverage (as many things are communicated via tweet, online video, etc.) so we focus on where the dominant attention is placed. An example could be a politician announcing their candidacy for public office via Twitter. Some attention would be paid to the object (the tweet), although reporters would likely pay most attention to the subject of the tweet-the politician's candidacy. It is also worth noting, as we explain below, that even when subject claims are not taken seriously as fact claims warranting further reporting, they can still warrant journalistic attention because of the position of the person making the claims.
Meta-commentary/meta-journalistic commentary. The dynamics presented above also play out within evaluations of how well journalists confront speculation by their sources and how they deal with information that is unverified/unverifiable. This is simultaneously embedded in the specific news content being reported and in a dialogue that transcends the subject matter being discussed. This meta-journalistic commentary (Carlson, 2011(Carlson, , 2015 reflects journalists' engagement with verifying fact claims at the community level, where reflections on processes and practices of journalism around inference emerge, including through debates over journalists' approaches. This contributes to a larger societal commentary, within which journalists respond to the necessity of covering, for example, the speech acts of the president-it is what journalism must do; the coverage is warranted. In the current climate, however, such coverage poses challenges to the normative function of journalism as offering truthful accounts of reality. Ethics/morals statements. As an extension of journalists' meta-commentary, specific discussions of both journalistic and societal ethics and morals reinforced the bases for certain types of reporting. While ethics issues are complex and potentially open up important professional and social discussions, we focus on ethics more narrowly in relation to inference as an expectation of journalism's "moral voice" (Ettema & Glasser, 1998). We examine the ways in which moral voices are performed, including in work that evaluates claims to present reasonable conclusions or judgments made in that process. This includes what is "warranted" for consideration by journalists, who are duty bound to examine the claims of and about prominent public figures.

Further Reflections on Methodology
It is worth noting that before deciding to keep the set of markers identified in the pilot study, outlined in Figure 1 and described above, we discussed whether the more elaborate theoretical framework developed for the monograph should be operationalized in a more discrete and detailed list of textual markers that could be assessed deductively. We decided against doing this for two reasons. First, in the second close reading of the material, we found the initial markers remained relevant to the broader arguments around explicit narratives of newswork and journalistic practices of reasoning. Second, we found that the more nuanced theoretical foundation could be applied within the structure of the original markers while also supplying us with a number of new perspectives, concepts, and distinctions that we found of interest and which we had not considered initially. These nuances needed to be brought under the auspices of the initial study, to demonstrate whether these still resonate with efforts to understand journalistic work and the ways these emerge in texts. Therefore, we decided a recursive and iterative approach that re-engages with the initial findings offered the greatest opportunity for theoretical development.
We also explore here an increasingly apparent inter-reliance on "allied" resources, in particular other news media (including by not only referring to these specifically in text and/or via hyperlink but also more generally in mentioning "media" or "news media"). We now specifically assess intertextuality, to capture the way journalists consult other media through in-text, hypertextual, and embedded media-to-media references in the initial stories reported by each outlet in each case. For each case we map all intertextual references to other media, including textual (quoted/referred to), hyperlinked, and embedded artifacts-such as tweets-in initial coverage. This offers insights into the ways in which media draw on a variety of resources to assess fact claims, and the ways in which this differs among the three types of media we focused on-traditional, political niche, and digital-native. We should also emphasize by way of context our aim is toward developing a conceptual understanding of the role of inference in news stories addressing unverified or unverifiable information.

Case 1: The Russian Dossier
Our first case is what Politico's media columnist cheekily dubbed "Goldengate" (Shafer, 2017). This news story involved the publication by BuzzFeed News on January 10, 2017, of an unverified dossier containing potentially damaging information on Trump as President-elect, that is, on improper sexual behavior (that accounts for the "Goldengate" nickname) and, more importantly, possible collusion with or interference from Russia during the U.S. presidential election. Publication by BuzzFeed gave rise to coverage by The New York Times and Politico. The dossier comprises multiple memos written by Christopher Steele, a former British intelligence officer who authored the dossier for a private Washington, D.C.-based investigative company, Fusion GPS. Over time, reporting confirmed that conservative interests first paid Fusion GPS to gather this material, and later attorneys linked to the Clinton campaign did so; but at the outset of this case and in the reporting we analyzed, the dossier's provenance was somewhat unclear. The broader coverage of its content addressed not only the political implications of the claims (whether true or not) but also issues of journalism ethics related to BuzzFeed News's decision to publish the full dossier (Bensinger, Elder, & Shoofs, 2017). Further coverage expanded this discussion, including in a follow-up BuzzFeed article in which BuzzFeed Editor-in-Chief Ben Smith defended the publication of the dossier (Nashrulla, 2017); the article included a link to a CNN interview with Smith. Both New York Times and Politico articles explicitly referred to these BuzzFeed articles and to a tweet of a staff memo from Smith. The New York Times and Politico articles focused on the claims within the dossier (Shane, Goldman, & Rosenberg, 2017;Wright, Bade, & Bresnahan, 2017) as well as the wider circumstances around the dossier's circulation and publication (Ember & Grynbaum, 2017;Shafer, 2017).
These initial articles provide a sense of how inferences factor into journalistic practice within a journalistic community. Such an inferential community seems to be coalescing around practices that focus on evaluating the implications for society of the claims within the dossier, if proven true, while engaging in processes aimed at ascertaining the truth value of the dossier. When such attempts of verification ultimately proved unsuccessful, the publication of the dossier itself invited audiences to make their own inferences. This raises important questions about journalistic practices of verification, inference, and knowledge work and with these, journalism's societal role in the contemporary digital media landscape. Overall, we saw textual traces of • • Shared journalistic endeavors of verification, including publicly demonstrating practices of inference in relation to truth value of claims that are very difficult to verify within the dossier; • • Debate over how journalism should engage with unverifiable documents circulating within the communities of politics and journalism; this in turn raises important questions about journalism as the arbiter of public knowledge.

Overview: Findings
Given the character of the dossier, the inferential work demonstrated in this case primarily relates to what Anscombe (1974Anscombe ( /2005 and Von Wright (1963) describe as inferential reasoning that moves from premises toward conclusions. Regarding premises, the claims made in the dossier are described as warranting consideration because they concern the President-elect (and possible collusion). Practices of inference, therefore, follow an if true, then what? reasoning. That is, these texts require readers to engage with them, specifically requiring readers to speculate about what conclusions would be reached if what is stated within the dossier is true (Harman, 1965).
Other news outlets, such as Mother Jones magazine, had initially, before the dossier was made public, alluded to the contents of the dossier in rather general, non-descriptive terms. Once the dossier was public, however, journalists attempted to gather information that could demonstrate journalists' ability to ascertain to what extent the dossier's allegations could plausibly be argued to be truthful. These attempts involved consulting what Dimmick (1974) refers to as "allied" resources, drawing on past reporting by journalists within their own institutions and references to the previous reporting in Mother Jones. One aspect of coverage that emerged reflects the dossier's circulation among journalists and political figures prior to it being published by BuzzFeed, and stories noted its existence as a somewhat open-secret in Washington, but otherwise unknown among the larger public. This aspect of the story highlights important questions about the role of journalists as co-creators of public knowledge, in the sense that they interact with other knowledge brokers, including sources but also citizens who look to journalists and news outlets to develop their awareness of the world around them. In the ways different news outlets took different approaches to engaging with this material, journalists reflected on the choice to report on something that had already been discussed in certain circles (and as BuzzFeed Editor Ben Smith argues, such discussions about the incoming President should be shared with the wider public). Below, we will discuss how such processes developed in three stages: First, as what if? newswork, which is performed at both the subject and object level (discussing the content of the dossier, and the dossier itself); then through intermedia reporting practices, including references and links to other news media; and finally as relatively extensive meta-commentary about journalism ethics related to disclosure and the potential effects of a new political and journalistic landscape on such considerations.

What If?
Many journalists mentioned that the dossier had been circulating within elite circles, including among journalists. To take one of several examples, Politico's media columnist Jack Shafer wrote, "The dossier, which many top newsrooms saw late in the campaign and passed on, makes for wonderfully pervy reading" (Shafer, 2017). While this "fact" in itself reflects certain types of community formation-an in-group community of political journalists-it also raises questions of how journalists negotiate semi-public allegations. Differences in journalists' responses to this question opened a rift among news outlets that centered on the limits of journalistic inference, namely the ability, or rather inability, to eliminate uncertainty to present news that comes across as authoritative. On January 4, Politico highlighted this uncertainty, noting, "intelligence officials delivered a report to the President-elect last week outlining allegations that Russia could have compromising information about him. Although the details of these revelations remain murky and unverified" . The article went on about the potential problems these allegations would have for Trump-if true. These dynamics are also apparent with BuzzFeed, whose coverage emphasized the existence of the dossier, attempts at verifying its contents, and that the President-elect had been made aware of it by the "intelligence community" (Watkins, 2017a). Further efforts to verify the dossier's contents are intermediated: That is, BuzzFeed referred to the Guardian's reporting, saying it was based on interviews with "intelligence sources" (Whittaker, 2017). While such attribution is vague, it positions journalists as authoritative, with access to sources and knowledge that cannot be revealed in detail but can nevertheless signal that what they are talking about may have wider implications, particularly given the subject of the information-the President-elect. The articles presented their findings as following from reasonable assumptions that allow the journalist to report conclusively as if the premises are true, reinforced by the knowledge that "the intelligence community" had inferred similarly and thus found this important enough to inform the president. This represents practices of explicit newswork; that is, they broadly portray the processes journalists engaged with, processes aimed toward uncertainty reduction.
Another aspect of this is the inclusion of doubt. Doubt runs parallel to what if? coverage. The journalists behind the stories we analyzed openly expressed skepticism about the dossier by suggesting that its contents were probably not true-or, as Shafer in Politico expressed it: "Almost too good to be true" (Shafer, 2017). Doubts about the truthfulness of the allegations in the dossier are, in part, linked to journalists' inability to access credible, named sources (to which we will return); this is further linked to the complexity of the relationships behind those involved in funding (both conservative and Democratic political interests), compiling (a former British spy), and commissioning (Fusion GPS) the dossier. On this latter point, Politico, much later (in April 2017) quoted Senate Judiciary Chairman Chuck Grassley (R-IA) reacting to Fusion GPS's relationship with other clients, including with links to Russia: "The relationship casts further doubt on an already highly dubious dossier" (Arnsdorf, 2017). Doubt was, therefore, a core element of early coverage and continued to be so. Even BuzzFeed News Editor-in-Chief Ben Smith said in a tweet linking to his staff memo explaining the full disclosure of the dossier, that "there is serious reason to doubt the allegations. We have been chasing specific claims in the document for weeks, and will continue to do so" (Mack, 2017).

Explicit Newswork
The primary journalistic response to uncertainty is to seek information that can corroborate or falsify fact claims. Journalists made various attempts to do precisely this, as is evident in coverage that shifted between subject and object-that is, between attempts to verify the claims within the dossier and/or the authenticity of the dossier itself (these are portrayed as closely linked concerns). Even the initiating BuzzFeed article (Bensinger, Elder, & Shoofs, 2017), which included links to the full dossier, reported on (a few) factual errors that, according to suspicious journalists, undermine its authenticity, for example, "It is not just unconfirmed: It includes some clear errors. The report misspells the name of one company, 'Alpha Group', throughout. It is Alfa Group" (Bensinger, Elder, & Shoofs, 2017). Most of the articles reflected similar attempts of verification, and these are often presented as more reliable attempts at verifying or falsifying the dossier's claims than leaning on Trump's pronouncements. Trump tweeted: "FAKE NEWS-A TOTAL POLITICAL WITCH HUNT!" This tweet is embedded in coverage presented as providing a contrast to Trump's reaction through substantive narratives of newswork, even as journalists also admit an inability "to confirm the claims" (Shane, Goldman, & Rosenberg, 2017). This is further hampered by an inability and/unwillingness of the authorities to provide information that can help journalists in their reporting, leading to necessary practices of inferring likely outcomes or conclusions. This is highlighted in coverage, for instance, by BuzzFeed News that cites Obama's press secretary saying, "[T]he president-elect has fallen short of meeting the 'bare standard of transparency' in proving that accusations lodged against Trump are false" (Georgantopoulos, 2017). In coverage the next day, BuzzFeed continued to underline this: Director of National Intelligence James Clapper told President elect Donald Trump Wednesday night that he is dismayed by leaks to the press, and that intelligence agencies have not made any judgment as to whether the documents published by BuzzFeed News are "reliable." (Hernandez, 2017a) What followed was a steady slide toward covering the dossier at the object level rather than its contents, including attempts to determine its origins and how it reached political actors and journalists. This is repeatedly based on officials "who spoke on the condition of anonymity because they were not authorized to speak about it publicly" (Shane, Goldman, & Rosenberg, 2017). The reporting emphasized uncertainty, highlighted unsubstantiated claims that "[d]etails of the reports began circulating in the fall and were widely known among journalists and politicians in Washington" (Shane, Goldman, & Rosenberg, 2017). This recurrent theme served as a community-building discourse, binding journalists together on two points. First, journalists had a shared inability to significantly reduce uncertainty about the dossier-due partly to the murky background of the dossier, and partly the unwillingness of reliable and on-the-record sources to confirm the contents. It was a failure of conventional newswork as well as related practices of inference that led BuzzFeed to disclose the existence of the dossier and its contents, which both made visible a community of inference as well as important rifts within it. This was evident in a New York Times report: "Raise your hand if you too were approached with this story," Julia Ioffe, a journalist who has written extensively on Russia, wrote on Twitter, adding that she had not reported on the information in the document "because it was impossible to verify." (Ember & Grynbaum, 2017) Second, a community coalesced around journalists' ability to proceed, nevertheless, with reporting that infers from the premises upon which more solid conclusions can be based. They went on to produce reporting as if the premises are true, and injecting doubt when premises seemed not to be true. An important aspect of this coverage presented journalists' explicit newswork to address factual uncertainty in contrast to political reactions-including the President-elect's claims (e.g., "FAKE NEWS!"). Rather than push back against journalists' pursuit of facts, such responses led journalists to infer further, including speculating about what the lack of any official response might mean. This proceeds along the lines of, What does it mean that no substantial, official rebuttal is issued? However, it also highlights how the lack of access and transparency (Lasorsa & Lewis, 2010) forces journalists to work with whatever they have. BuzzFeed references this lack of access explicitly, noting they only have Trump's tweet to rely on as an official rebuttal as it had been 167 days since Trump had held a press conference (Cormier, 2017).
The immediacy of Trump's response, combined with this lack of access, aligns with broader doubt and uncertainty surrounding the President-elect, highlighted in The New York Times: Trump strongly denies all this, the dossier has zero public evidence behind it, and it should be treated with skepticism. But it reflects an unprecedented uncertainty: There is a disorienting kernel of doubt about whether we can fully trust the man who will occupy the Oval Office. (Kristof, 2017)

Intermedia References
At the center of intermedia references and journalistic meta-commentary in this case is the publication by BuzzFeed of the entire dossier. Its publication produced various, often conflicting, perspectives as to how journalists should act on information circulated within policy and journalistic circles; this further calls into question the status of such information as knowledge, when it cannot be verified by conventional newswork. Most of the articles we analyzed refer to and link to BuzzFeed's "disclosure" article and to a previous CNN report stating that President-elect Trump and President Obama had been briefed on the dossier's existence as well as on some of its overall claims. This demonstrates how the news built up across media and how-according to The New York Times-" [t]he reports by CNN and Buzzfeed [sic] sent other news organizations, including The New York Times and The Washington Post, scrambling to publish their own articles" (Ember & Grynbaum, 2017). Thus, while many of the subsequent articles include disclaimers about the dossier and largely condemn its publication (especially the Times), they felt obliged to refer to it. These discussions make visible a journalistic community of inference focused on news at the object level. This is, arguably, an indirect acknowledgment that traditional newswork around the subject material and within one institution cannot produce informed conclusions, further demonstrating why inferences must be drawn (partly by journalists and partly by readers) from the information dispersed across the news landscape. Such coverage indicates that the practices of leaking-which have been dripping "nonstop since Trump was elected" (Wright, Bade, & Bresnahan, 2017)-and moreover the practice of circulating leaked information, including this dossier, within a journalist community takes place through an informal division of labor between outlets where some are more willing to push boundaries in publishing such information. This allows news media to build coverage between media, which adopt different positions toward leaks and different publishing tactics, ranging from full disclosure to highly qualified paraphrasing. This case also shows these practices now extend to a wider online environment, where the likelihood that aspects of these stories will be made public seems greater. This has changed the array of allied resources (Dimmick, 1974) on which journalists can draw in their reasoning practices. We see this reflected materially in the intertextual mapping; Figure 2 shows the pronounced inter-reliance on other media to examine the fact bases for this story. While all media engaged with a range of other media, Politico stands out for the variety of allied resources it drew on from traditional media. This is likely due to the nature of early coverage by media columnist Jack Shafer, although it is worth highlighting that every reference by Shafer to other media includes hyperlinking. In contrast, the online version of The New York Times referred to BuzzFeed and CNN but without hyperlinking. This may reflect an effort to mention something that has been made public, without amplifying its publication; the Times explicitly objected to giving any additional publicity to the dossier. The Times described their own reporting and The Washington Post's coverage as responding to the release and "scrambling" to follow up on its content, and referred to other media that had alluded to the dossier in earlier reporting prior to BuzzFeed publishing the dossier, for example, the progressive magazine Mother Jones. BuzzFeed is the only one that did not refer to any other mainstream media save CNN, which also published details of the dossier. Instead, they linked to the dossier itself, and as with the Times, referred to and linked to Mother Jones's reporting. While not exactly an island, this intertextual mapping does show BuzzFeed apart from other analyzed media in its intermedia dynamics.

Meta-Commentary, Disclosure, and Ethics
Two types of newswork and journalistic performances were closely linked to the presence of meta-commentary debate. The first is journalists shifting between subject-and object-focused reporting, and the second is the sustained interlinking between news outlets. The first Times article was all about meta-commentary. Published the same day as BuzzFeed's disclosure, its headline read "BuzzFeed Posts Unverified Claims on Trump, Igniting a Debate" (Ember & Grynbaum, 2017). This story, describing a "swirling debate over journalistic ethics," discussed whether publishing an unverified report can ever be justified. The implication is that this is never justified. The Times quoted Dean Baquet, its own executive editor, saying "we're not in the business of publishing things we can't stand by." The most sustained aspects of meta-commentary on journalistic practice are BuzzFeed's own responses to its work. BuzzFeed Editor-in-Chief Ben Smith argued for the disclosure in a staff memo, in an article in BuzzFeed News (Nashrulla, 2017), and in an interview on the CNN program appropriately called "Reliable Sources," hosted by Brian Stelter. "When you have a document that is circulated so widely within the elite," Smith told Stelter, "the argument for keeping it away from the public has to be really, really strong." In relation to knowledge, he explained to BuzzFeed after the interview: "'let them know what you know', Smith said, adding that the audience should be treated with respect" (Nashrulla, 2017). This debate points to transparency, inference, and the public expectations of journalism, even when journalists disagree on how these should be best practiced.
In effect, Smith argued the American people should be able to make their own inferences. But, CNN's Stelter responded, "how can the Americans or anyone else make up their own minds without providing reporting to them?" Stelter describes the role of reporting as resolving uncertainty, that is, public processes of inference that weigh premises, judgments, and outcomes. Alternatively, drawing on Ettema and Glasser (1998) and Gjelsvik (2014), serious investigative reporting involves processes that assess reality. From such a perspective, premises that cannot be justified as credible should not be disclosed. In the CNN program, a distinction between "publishing and reporting" was made, with CNN putting itself alongside legacy print media on the side of the latter. Having tried to figure out whether BuzzFeed was The Washington Post or, conversely, WikiLeaks, Stelter concludes that that BuzzFeed was "trying to be both." He stressed a "profound difference between legacy media and digital media," where the latter shied away from proper reporting and "annotation" (processes of inference), leaving the bulk of the inferential work to other outlets and even the audience. This highlighted both shared interests of reporting facts and substantial rifts within journalism in relation to the construction of public knowledge not least between heritage news media and digital natives known for publishing leaked information (cf. Eldridge, 2018;Lynch, 2012).
Politico's media columnist Jack Shafer sided with Smith. While Smith linked his decision to the contemporary media landscape-this is "how we see the job of reporters in 2017"-Shafer went further: In this "new regime, if something exists, somebody is going to publish it." He continued, "The odd thing isn't that CNN and BuzzFeed went with the story, but that it took this long for a news outlet to pull the trigger and finally snuff the old journalistic order" (Shafer, 2017). While what Shafer called the "old order" did not adjust journalistic ethics to this new situation, he argued for a realignment in relation to a new media "regime." Moreover, he complained of a nontransparent Trump camp, which hinders journalists' ability to infer in more systematic and traditional fashion. This will be expanded on in the discussion of the second case.

Case 2: The Wiretapping of Trump Tower
The second case addresses the coverage of four of Trump's tweets, made early on a Saturday morning, March 4, 2017, asserting President Barack Obama had "wire tapped" Trump Tower. In a string of tweets, 2 Trump wrote, We initially analyzed BuzzFeed News's initial reportage, which linked the claim made in the tweets to a rumor circulated on conservative talk radio (Melville-Smith & Mack, 2017); BuzzFeed's follow-up coverage (Melville-Smith & O'Connor, 2017); initial reporting in The New York Times and Politico (Shear & Schmidt, 2017;Stokols, 2017); and follow-up stories in which journalists also traced conservative radio links (Baker & Haberman, 2017;Bennett, 2017). Our initial study showed two clear dynamics: Immediate attention paid to tweets from President Trump, and immediate (BuzzFeed) and near-immediate (Politico and The New York Times) efforts to locate a possible explanation for Trump's claim, particularly in rumors from talk radio. This offered an initial indication that journalists engage in inferential reasoning and, in particular, sought likely premises that would justify the conclusion presented in Trump's tweet (Gjelsvik, 2014;Von Wright, 1963).
Turning toward further developing the conceptualization of an inferential community, coalescing in response to uncertainty, we see engagement with what Anscombe (1974Anscombe ( /2005 describes as an unpacking of the "facts to what would be premises of such a conclusion." This is reflected in the way journalists address Trump's conclusive claims and then describe their practices of inferring possible explanations (premises) for such a conclusion. Analysis of the public-facing performances of journalistic practice found within the texts suggests three key points, which we see as a set of shared journalistic ambitions: • • By examining the veracity of the truth claims in the tweet, journalists develop and present a fact-based account of society; • • By presenting their work to a public, through authoritative performances of newswork and inference, journalists introduce transparency around the opaque explanations for the tweet; • • In their coverage, journalists present the most likely explanations for the tweet, so as to decrease uncertainty through demonstrations of reasoning and justification.

Overview: Findings
Analysis of the wider coverage of this tweet showed clear examples of inferential reasoning within and across news outlets, driving toward what Anscombe (1974Anscombe ( /2005 identifies as inferential reasoning from a conclusion toward premises. Critically, for journalistic practice during times of uncertainty, this usually appears as "inference to the best explanation" (Harman, 1965). Journalists not only outlined the premises for any possible conclusion but also noted these as speculative, not going any further than reason allowed-toward suggesting certainty-based on the few details available. Notably, we also found news texts almost universally emphasized doubt about Trump's claim. From this basis of doubt, subsequent reporting interwove skepticism into the presentation of possible premises for Trump's conclusion, building toward a reasonable (if unlikely) explanation. The practices of inference and negotiation of doubt presents a challenge for journalists in this case. As a result, journalists articulated two parallel reasoning practices: The first, that Trump had drawn a conclusion that was unlikely to be true, yet due to the nature of the speaker it warranted attention, and the public was owed an explanation of the necessary premises for such a conclusion to be true (Schudson, 1995). The second, because the necessary premises for such a conclusion can be shown to be untrue, or likely untrue, even the most plausible explanations for the conclusion can be ruled out, and the conclusions posed by Trump in the tweet rejected. Described in terms of inferential reasoning, this would look like this: If: Tweet (conclusion) is true (or, treated as true) Then: X (premises) must also be true (or, x premises must also be treated as true).

And
If: X (premises) are shown to be untrue (or likely to be untrue) Then: Tweet (conclusion) must also be untrue (or must also be treated as untrue).
Within these stories, further reasoning practices are found in performances of journalistic authority, with reporters emphasizing with some surety the unlikelihood of the tweet's underlying claims. They do so by drawing on their own and their peers' subject expertise. We thus found journalists engaging in community-building discourses as they drew together premises from their own and peer journalists' work to expand on the scarce evidence base within Trump's tweet. Finally, we saw a particular discussion of media and journalism in meta-commentary that addressed the subject claims made in Trump's tweet as based on conservative media, and the object of the tweet itself, highlighting the implications for journalism as they contend with the unique communicative practices of the time-that is, Trump tweeting.
This was also apparent in the intertextual mapping in Figure 3. This figure shows that all news media heavily relied on tweets from Trump as well as others (former Obama staffer Ben Rhodes, for instance), also linking to Breitbart and Mark Levin's conservative radio show as the source of the rumor. BuzzFeed and The New York Times even embedded tweets, including Trump's, although Politico did not. Finally, while all news media refer to the Times (including the Times itself, both in text through quoting and through hyperlinking), neither Politico nor BuzzFeed referred to one another. The Times did not refer to other news media except for The Washington Post, which they referred to both in text and via hyperlink.
This mapping signals a complex and visible intertextuality of journalism, embedded within the equally complex and collective intermediated processes of knowledge production. This intertextuality is further evident in the subject/object switching: Reporters addressed the subject claim-alleged wiretapping-while emphasizing the information object-the tweet-both as an event and as the prompt for further newswork. This was done both individually within discrete articles and in the marshaling of "allied" resources across news media (Dimmick, 1974;cf. Cook, 1998, p. 80). Extending the analysis to the wider news reporting which followed, inference, knowledge work, and uncertainty reduction became more nuanced, and particular dynamics more prominent.

Doubt/Disclaimer/Caution
In contrast to the claims within the dossier, treated by journalists as "unverified, but potentially true," here coverage presents Trump's tweet as a case of "unverified, and probably not true." Politico, for example, described the subject matter of Trump's tweet as "Trump's evidence-free accusation" (Bennett, 2017). The Times described a claim made with "no evidence to support the notion that such an order exists" (Shear & Schmidt, 2017). BuzzFeed presented it as a claim made "without offering a scintilla Key is the use of language associated with prompting inferential practices. We found prompting language in stories where journalists first indicated uncertainty over a premise or claim before presenting their assessments of the claim or premise. In this case, journalists repeatedly framed their reasoning around Trump's claims in terms of what Trump "appears" to say, and what is "alleged." Journalists then presented their own efforts at explaining, if these claims were true, what would be the necessary basis for any such wiretapping. Indeed, nearly every report was couched in uncertainty, and across the content we analyzed, his tweet was described as a "claim." While this uncertainty is hardly unexpected, it is clearly indicated as a prompt for the reasoning practices that assessed the strength of the claim which followed, and we found language of everyday inference as well as inference to the best explanation outlined in the discussion of inference, above.
Articles quickly and clearly demonstrated journalists' attempts to expand on scarce evidence through explicit newswork which incorporated inference; newswork that showed that, despite little likelihood of confirming Trump's claim, journalists diligently tried to assess its premises. This is particularly evident in BuzzFeed's reporting, which incorporated and quoted many sources, including current and former political actors and elected officials dismissing Trump's claim (Villa, 2017a(Villa, , 2017b. This reporting also included a few conservative supporters in Congress (Cheney, 2017;Melville-Smith & O'Connor, 2017;Shear, Goldman, & Huetteman, 2017), although supportive Republicans were quoted as suggesting that while unlikely, if Trump's claims were true, they would be troubling. Therefore, these sources said Trump needed to provide a firmer evidentiary basis for his claim (Loop, 2017;Villa & Georgantopoulos, 2017;Wright, 2017).
Journalists set out possible theories (means) from which readers and journalists can infer possible conclusions of what Trump might be referring to (ends). Doubt became prominent in stories, also reflecting journalists engaging with reasoning based on premises and conclusions which they nevertheless must assume are true. It is as if to say, the reporters themselves are skeptical but they were nevertheless demonstrating the bases for their skepticism by walking the reader through likely explanations. BuzzFeed did this by posing questions and possible answers of what Trump "may have been referring to" (Melville-Smith & Mack, 2017). The New York Times presented possible premises in the form of conditionals (e.g., If: Wiretapping is a result of ongoing investigation, Then: Trump's claim is possible, although unlikely): During the 2016 campaign, the federal authorities began an investigation into links between Trump associates and the Russian government, an issue that continues to dog Mr. Trump. His aides declined to clarify on Saturday whether the president's allegations were based on briefings from intelligence or law enforcement officials-which could mean that Mr. Trump was revealing previously unknown details about the investigationor on something else, like a news report. (Shear & Schmidt, 2017) This example from Politico offered a similar construction (If: True, Then: Authorities had obtained a warrant; If: They had obtained a warrant, Then: Trump's might have accessed this classified information): Trump's allegations raised questions about whether federal authorities had obtained a legal warrant to tap Trump's or his associates' calls, perhaps through the Foreign Intelligence Surveillance Court, which can authorize wiretaps against potential foreign spies. Those orders are typically classified, and Trump's comments sent Washington scrambling to understand his allegations. (Cheney, 2017)

Performances of Journalistic Authority
While indebted to the contributions of sources and peers who engaged in inferential reasoning and evaluation of the conclusions and premises in Trump's tweets, journalists reporting on this claim also utilized their own authoritative statements. This emerged in line with what Carlson (2017, p. 73) refers to as the "isolated and impersonal" authoritative voice of journalists. Such a voice is especially apparent in The New York Times and Politico. Drawing further on the subject expertise of its reporters, the Times, for example, noted that such wiretapping would be "a highly unusual breach of the Justice Department's traditional independence" (Shear & Schmidt, 2017). Politico used its authoritative voice to present a hypothetical but "plausible" explanation that ongoing investigations may have included campaign communications (Bennett, 2017). This interweaves authoritative accounts with elements of doubt, couching explanations as premises which could explain the "supposed tapping" (Shear & Schmidt, 2017), including that Trump is "under scrutiny for possible ties between his campaign and Russia and increasingly fixated with rooting out leaks" (Stokols, 2017). In neither outlet did authoritative statements rely on external sources beyond the journalists' own authoritative voices.
The media we studied differed in how they communicated authority in other ways as well. The Times and Politico used authoritative statements that reflect what their reporters already knew, and therefore could restate as true. Politico offered what could be read as a particularly assertive retort: "No evidence has emerged to suggest that Obama ordered surveillance of Trump or his associates, and FBI Director James Comey said last month that there was no evidence to support Trump's claim" (Conway, 2017). BuzzFeed, however, relied more heavily on quoting, linking, and citing other expert voices to demonstrate certainty or authority, and less often on stating something as "true" without such sourcing. For instance, BuzzFeed suggested that the wiretapping claims were unlikely to be true by referencing legal statutes in a summary: Under US law, President Obama could not have simply ordered the surveillance of an individual without first obtaining a warrant from the Foreign Intelligence Surveillance Court, or FISA court. If such a warrant were to exist with regards to Trump, then the president could order it be declassified. (Melville-Smith & Mack, 2017) All the same, each outlet made clear that the possible explanations they outlined, based on inference, were unlikely. This came through in the ways they pushed back against those supportive Republican voices, highlighted above, which suggested that such wiretapping would be troubling if it occurred. In their approaches toward evaluating the claims Trump has made, journalists were engaged in the "input/output" practices of evaluating information to determine what was newsworthy as Dimmick (1974) described. These examples further reflect how news practice involves journalists resolving "decision problems" to identify what is newsworthy, including by calling on a mix of journalists' own information authority and information from external sources.

Meta-Commentary: Assessing a Community
Discussions of media by media-meta-commentary-emerged most often in reference to the polarized media environment. Media polarization is hardly new but is certainly pronounced in the contemporary era (Faris et al., 2017). BuzzFeed explicitly highlights this: "Like much of what Trump tweets, the Obama wiretap claim appears to have followed a path through the pro-Trump media" (Melville-Smith & Mack, 2017). The New York Times is also explicit, calling Trump's accusation "remarkable, even for a leader who has repeatedly shown himself willing to make assertions that are false or based on dubious sources" (Shear & Schmidt, 2017). Adding that Breitbart had promoted the rumor, the Times called it a conspiracy theory, and "[l]ess than 24 hours later, the president embraced the conspiracy in a series of Twitter posts" (Baker & Haberman, 2017). The Times article underlined this polarization by contrasting Trump to past administrations: Previous presidents usually measured their words to avoid a media feeding frenzy, but Mr. Trump showed again over the weekend that he feeds off the frenzy. Uninhibited by the traditional protocols of his office, he makes the most incendiary assertions based on shreds of suspicion. He does so without consulting some of his most senior aides, or even agencies of his own government that might have contrary information. After setting off a public firestorm with no proof, he then calls for an investigation to find the missing evidence.
Politico also emphasized political polarization by noting that Republican members of Congress were appearing on conservative talk radio shows and Fox News to support Trump's claims (Wright, 2017).
As noted in our outline of markers of inference, meta-commentary can also include specific discussions of newswork and the bases for certain types of reporting. In both accounts, the status of the speaker goes some distance toward justifying journalistic attention. All three outlets made clear that they were obligated to examine the claims of prominent public figures and the likely impact of even their smallest statements (Harman, 1965;Schudson, 1995). Politico described Trump's tweet as "explosive," highlighting impact, and then quoted Sen. Marco Rubio to underline the link to Trump's status: "The president put that out there, and now the White House will have to answer as to exactly what he was referring to" (Cheney, 2017). The Times also described the allegations as "explosive," quoting former Obama officials describing the accusation as "unprecedented" (Baker, 2017), even though Trump had established a reputation for questionable claims on Twitter: "Of all of Mr. Trump's Twitter eruptions, his March 4 outburst on surveillance may have been the most disturbing, both for those who believed it and for those who dismissed it as outrageous nonsense" (Shane, 2017). Indeed, Times reporters made this argument from their first story on this tweet: The president's decision on Saturday to lend the power of his office to accusations against his predecessor of politically motivated wiretapping-without offering any proof-was remarkable, even for a leader who has repeatedly shown himself willing to make assertions that are false or based on dubious sources. (Shear & Schmidt, 2017) Such meta-commentary has dual intentions. First, it provides a public explanation for journalists devoting attention to an unverified claim (the source-the president-is afforded such attention). Second, it provides a bulwark against accusations of bias or sensationalism because of his status. By situating their engagement with unverified claims within a larger discussion of media and Twitter use by the president, journalists reinforce publicly the "input/output" decision making at the core of inferential newswork processes used by journalists when evaluating what information to convey to audiences as true, where to insert doubt, and how this is arrived at. Journalists made the case that this is what journalism must do, in the current political climate, particularly in the face of uncertainty. BuzzFeed referred to this directly: There's so much discordant noise that just making out each individual thing and tracking its journey through the news cycle requires enormous effort [. . .] People's reactionstheir theories about what's really going on here-end up occupying as much space as the actual action. (Watkins, 2017) Examining this case, meta-commentary extended to the at times pronounced performativity of journalism, where reporters articulated the ways in which they were addressing uncertainty around Trump's claims by interrogating its possible links to other investigations (e.g., an inquiry into Russian contacts made by the former National Security director Michael Flynn, who had resigned on February 13, 2017, after misleading FBI investigators about these discussions), presenting numerous interviews, sources, and overall investigative reporting (Griffin, 2017). Such explicit performances of newswork also anticipated the need for a public explanation of journalists' truth-seeking practices (Harman, 1965).

Subject and Object Coverage
As expected, we found significant attention to the object of the tweet itself, aka the "tweetstorm" (Stokols, 2017). This became something of a back-and-forth between justification for coverage of the object (a set of tweets) and assessment of its subject content (the claims made). Over time, the object of the tweet itself became a point of rebuttal and discussion for reasoning around the claims within it, as "presented, without evidence, in a tweetstorm Saturday morning" (Griffin, 2017). On the occasion of Trump's first 100 days in office-a traditional milestone for evaluating the performance of a new president-The New York Times presented his tweets in an online article "Fact-Checking President Trump Through His First 100 Days"; after embedding the tweet, the Times added, "The next day, Mr. Trump repeated his claim, again without offering evidence" (Qiu, 2017).
BuzzFeed's reporting dealt with the tweet similarly, revisiting it repeatedly as a point of spectacle within coverage of the new administration, also using it as an artifact and illustration for related news (Melville-Smith & Mack, 2017;Miller, 2017). BuzzFeed did engage with its content and Trump's tweets in general, emphasizing the ways in which reporters and officials reacted to them: "Interpretation of what Trump meant by his tweets has been a matter of debate, to the extent that Schiff on Wednesday said the White House has 'been all over the map' on the matter" (Villa, 2017b).

Preliminary Discussion
We see inferential reasoning in journalists' engagement with premises and conclusions, where texts demonstrate reasoning through the likely and unlikely premises of the conclusions presented in our first two cases. We see this supporting our thesis about an inferential community: Reporters drew connections among the larger community, explicitly through links and textual references to other content both within a news organization's own work (this occurs repeatedly with The New York Times) and across the journalistic field (this regularly featured in BuzzFeed content). On one level, this practice is more common in digital-native work (Eldridge, 2019), which points to the emergence of a shared practice of resolving uncertainty. We draw this in the comparison of fact claims gathered by one outlet, set against allied resources from others, marshaled in the input/output processes of gatekeeping (Dimmick, 1974). This is simply a more public demonstration of the piecing together of possible conclusions in a post-gatekeeping era, with its proliferation of "gates." Bruns (2005) argues that conventional gatekeeping as a means of ensuring complete, balanced coverage is no longer necessary and has been replaced by "gatewatching," defined as a process of harnessing the collective intelligence and knowledge of communities of bloggers, citizen journalists, commentators, and activists to filter the news flow. Nonetheless, the manifest inferential reasoning demonstrates a need to resolve the premises and conclusions around which reasonable accounts can be formed.
These first two cases offer insights into how reporters respond when confronted with unverified and unverifiable claims, and when the White House is not forthcoming with factual assistance. Over time, we saw markers of journalists negotiating this challenge, incorporating a growing skepticism as to the facticity of Trump's statements on Twitter. This negotiation has continued to be on the agenda of journalists' reflection and meta-commentary. A pertinent example at the time of writing was a media debate over whether airing a presidential address was appropriate, considering the likelihood of lies and false statements (Chotiner, 2019). In any case, a community of journalists committed to knowledge work formed, doing so with an aim of reducing uncertainty in presenting a fact-based account of reality. This community formation helped reinforce journalistic authority through professional performances of their social standing and authority as knowledge workers. This included projecting their work as necessary for understanding news events and as serving a public equally committed to understanding the goings-on of society.
As our thinking around inference has deepened, including through theoretical engagement as well as the iterative analyses reflected in this monograph, one point that continues to emerge is that inferential practices can work in several directions: Inferring forward from facts as premises could lead to various possible outcomes. Inferring backward from conclusive claims can lead journalists to infer possible premises which would be necessary conditions for such conclusions to be true. Inference can thus be seen as a complex practice within journalistic knowledge work that seeks to resolve uncertainty at multiple levels. We now examine this complexity in further detail in our third case.

Case 3: The Trump/Putin Helsinki Summit
We have thus far highlighted the role of inference in public-facing performances of journalism that strive to provide likely explanations of news events marked by uncertainty and unexpectedness. As a field committed to reducing uncertainty regarding the goings-on of society, journalists presented their best effort at verifying the unverified and, when obstacles to certainty emerged, accounting for the unverifiable.
Our third case analyzes coverage of the July 16, 2018, closed-door summit meeting between Trump and Russian President Vladimir Putin held in Helsinki, Finland. We again examine coverage by The New York Times, Politico, and BuzzFeed News. However, this third case differs in several key aspects from the first two. In the "dossier" case, journalists embarked on efforts to verify a compilation of damaging, albeit alleged, information on Donald Trump, and then to debate journalism's role in the face of such uncertainty. In the "wiretap" case, journalists confronted a tweet claiming something highly unlikely, albeit warranting coverage due to the nature of the speaker, Trump. Both cases involved coverage in reaction to unexpected events. The summit case examines the practices of the journalistic community regarding a planned event; it highlights the role of inferential reasoning in journalists' attempts to make sense of what turned out to be a rather opaque event.
This case builds on the first two both temporally, as it occurred 18 months into the Trump presidency, and conceptually, in terms of the development of the community of journalists covering the White House. The predicate knowledge that journalists bring into covering any presidency will inevitably be greater 18 months into the administration; journalists will have more information about the president and the administration than they began with. In this specific case, however, this knowledge grew to include several reasons for caution, and our analysis shows that uncertainty had become an ongoing feature of such coverage. Journalists' explanations of their newswork clearly showed that the nature of resolving "unverifiable" claims had not been made any easier by a White House that reinforced dubious remarks. Statements by Trump had at this point regularly been shown to be untrue, and his approach toward working with foreign leaders had departed from past practices of U.S. presidents. Nevertheless, this meeting triggered a particularly strong reaction (and sense of surprise). Political actors and journalists alike reasoned out loud in news texts in an effort to explain what might have occurred behind closed doors, including by inferring from the statements made at a press conference at the end of the summit.
Turning to the conceptualization of the inferential community in this analysis, we found specific types of inference emerge in coverage both before and following the summit. These are as follows: ing by quoting and linking to other news media, linking to other official and political sources to reinforce their conclusions (including social media posts), and doing so in ways which allowed them to make sense of their own explanations, as inferred and presented.
Inferred expectations. Most of the early coverage was guided by inferred expectations that move forward, from premises toward conclusions. This is evident in the way that reporters outlined expectations for the summit and the particular issues it should deal with. That is, a range of pending/unsolved issues with Russia, including possible collusion in the election, had accumulated since Trump took office. Thus, earlier coverage outlining what was expected from the summit formed a basis for later coverage, which compared what was expected against what actually occurred (or seemed to occur). Here is an example of how expectations, based on the ongoing investigations into Russian activities around the 2016 U.S. Presidential Election, guided the coverage of the summit and its outcomes: Politico's coverage noted that charges had been filed against Maria Butina, a Russian gun-rights activist, for trying to infiltrate Republican groups to promote Russian interests in the run-up to the U.S. election. These charges came "on the same day that President Donald Trump met privately with Putin in Helsinki, Finland." Politico's account added, "Hours after their meeting, Trump told reporters Putin made an 'extremely strong and powerful' denial that Russia interfered in the 2016 presidential election" (Loop & Leopold, 2018). Contextualizing the summit and the press conference with important issues regarding United States-Russia relations exposed the discrepancy between what journalists had outlined as expectations and what they could determine actually happened. Journalists underlined this by pointing to the criminal charges against Butina and indictments against 12 Russian military officers for election interference as a context for the summit. These would later form the basis for reporting on Trump's response at the press conference concluding the summit, discussed below. Coverage ahead of the summit emphasized both the lowered expectations of serious diplomatic work among journalists and those they quoted in their coverage as well as the unpredictability of the Trump presidency. For instance, in outlining what the summit would likely entail, The New York Times incorporated a range of experts' viewpoints to establish parameters for the summit, while also referencing the positive relationship between Trump and Russia and the risk of the summit being a "Helsinki lovefest" (as one source described it; Higgins & MacFarquhar, 2018). Journalists focused on what policy experts said should have happened and what Trump should have said in the summit's closing press conference.
When it became apparent following the summit that these expectations were not met, journalists did not merely infer conclusions. Instead, journalists established the parameters of interpretation; that is, the larger context within which the readers should make sense of this event. The focus was on context, partly based on expectations as to what should have been the outcome and partly based on evaluations of what was actually said at the press conference. Indeed, most of the articles the three news outlets carried did not quote at length from the press conference, which BuzzFeed News called "bizarre" (Tamkin, Lewis, & Dalrymple, 2018). Instead, journalists for the three outlets quickly recounted specific aspects of the press conference after which they emphasize its abnormality: BuzzFeed reported, sarcastically, "Here are 10 other totally normal things that happened during the joint press conference" (Abdelmahmoud, 2018). The press conference as an event was thus largely ridiculed, and Trump's and Putin's statements cast in doubt. Press conferences must always be treated with caution as managed news events; therefore, any information emanating from them is carefully selected (Ananny, 2018). Still, the coverage conveyed the sense that what was actually said at the press conference was only valuable as spectacle. As New York Times reporter Julie Hirschfeld Davis (2018) reported, The 45-minute news conference offered the spectacle of the American and Russian presidents both pushing back on the notion of Moscow's election interference, with Mr. Putin demanding evidence of something he said had never been proved, and Mr. Trump appearing to agree.
More substantively, the lack of clarity in the statements made at the press conference allowed journalists for all three outlets to display doubt regarding what actually happened when Trump and Putin were alone. But, lacking access, conventional newswork can get no closer. As the Times commented, "We don't know everything the two men spoke about-only translators were present-but their 45-minute news conference afterward was a remarkable spectacle" (McDermid, 2018). Readers and journalists alike were thus left with what was said and either skepticism or acceptance that mirrored what actually happened.

Inferred explanations. Once it became clear that journalists' expectations of what
should happen at the Helsinki summit were not going to be met, coverage shifted to trying to explain what might have gone on behind the closed doors, and to make sense of its public outcome. Anchored around the surprising press conference as an event, the statements establish hooks for journalists to infer from, relating inferred expectations with outcomes. Saying that the end had left observers "head-scratching," the Times's live updates made this clear by noting "although international affairs was expected to dominate the session, Mr. Trump turned again and again to a defense of his own political legitimacy" (New York Times, 2018). This example shows public-facing performances of inference from conclusions (i.e., the press conference) toward premises, in this case, identifying the possibility of what might not have been said behind closed doors. At the same time, any explanations are cautious, concluded only to the extent possible, only with pronounced uncertainty.
News analysis and opinion articles first emphasized how, using the wording of New York Times columnist Maureen Dowd (2018), it was "befuddling and alarming to watch him [Trump] kowtow to Putin." Similar commentary appeared in all three media. As Politico's Nussbaum (2018) wrote, The news conference left observers gobsmacked, as Trump-who delights in flaying opponents and perceived opponents, and has made a habit of attacking American alliesrefused to say a single negative word about Russia and used the international stage to praise the country's strongman leader and attack American institutions.
News articles shifted from covering the "remarkable" and "extraordinary" event (Hirschfeld Davis, 2018) to outlining possible reasons for Trump taking this approach. Inference is made apparent in the use of qualifiers, such as "appeared." For instance, describing how a comment made by Trump "appeared to absolve Moscow of many irritants in the relationship with Russia, including the election hacking" (Hirschfeld Davis, 2018), one journalist used the word "appeared" to emphasize the journalist's inability to conclude this as the explanation for what occurred. In a Politico article, explanations are inferred from "suggestions": "The Russian government appeared to be taking advantage of the secretive meeting, suggesting that Trump and Putin had reached agreements on key issues despite denials from U.S. officials" (Nelson, 2018). Similar phrasing returns in another Politico article as "appearing": "Putin was asked whether he directed any of his officials to help Trump's presidential campaign, but Putin appeared to sidestep that part of the question" (Murray, 2018).
In trying to explain their approach, journalists typically began by recounting the press conference and then the story shifted to various reactions to what Trump said there. As is notable in the intertextual mapping of initial follow-up coverage (Figure 4), The New York Times relied on its own body of work to source its coverage, repeatedly quoting or linking to its previous and parallel reporting on US-Russia matters. They quote, but do not hyperlink, Trump's interview on Fox News immediately following the summit. Otherwise they, as do Politico and BuzzFeed, link to the shock and backlash which played out primarily on social media. While the Times did not hyperlink to Senator John McCain's (R-AZ) comments (detailed below), BuzzFeed and Politico do. BuzzFeed's follow-up coverage differs in one key way: It embedded several tweets, thereby making more prominent the reactions of other U.S. officials, while hyperlinking CNN among other media sites in text. Similarly, Politico hyperlinked tweets by Trump and other officials in text, but embedded more prominently a tongue-in-cheek (presumably) comment in a tweet from the Russian foreign minister. Notably, none of the media we studied cites each other's work, although they do cite Fox News and CNN. These particular intermedia references were used to highlight what observers regarded as the bizarre nature of the summit.
Subsequently, uncertainty as to what was said during the meeting was a key thread in news articles. In resolving this uncertainty, journalists posed "if/then" conditionals in their explanations. These infer backward from comments made publicly by Trump and Putin, during the press conference; journalists treated these as conclusions, to then identify possible topics discussed during the meeting as likely premises. For example, BuzzFeed, the Times, and Politico writers all tried to determine what was behind Trump's statement that Putin had made an "incredible offer" to facilitate interviews with Russians accused of interfering with the election, and to determine what Trump meant when he said "President Putin may very well want to address it and very strongly, because he feels strongly about it and he has an interesting idea." Journalists inferred that this "interesting idea" was Putin's offer to exchange interviews with Russian officials for an interview by Putin of long-time Putin adversary Bill Browder, but they emphasized that Trump did not make this clear himself and Politico said explicitly that "it was unclear what the 'interesting idea' was that Trump mentioned" (Morin, 2018). Politico outlined Browder's campaigns for sanctions against Russian oligarchs, connecting Browder's concerns to a 2017 Trump Tower meeting between campaign officials and a Russian lawyer connected with the Kremlin-who were under federal investigation at the time of the summit (Morin, 2018). As with many possible explanations, Politico tried to piece together disparate statements made at the conference, and ultimately let readers infer from these connections, rather than spelling out causality.
Inference aimed at divining some explanations for Trump's performance was presented in ways that clearly marked the nature of any conclusions as speculative; readers got only possible explanations for the outcome of the Summit (e.g., that Putin had compromising information on Trump [Prakash, 2018], or that Trump was referring to an exchange of interviews including with Bill Browder [Morin, 2018]). To shore up the tenuousness of any inferred explanations, reporters quoted possible explanations from other political actors, including Sen. Chuck Schumer (D-NY), who "suggested that the president's performance on Monday would fuel speculation that the Kremlin is in possession of compromising information," and Trump allies, who argue the "best explanation is that Trump has a deep-seated fear of being seen as an illegitimate president" (Nussbaum, 2018 Or, as Times columnist Maureen Dowd (2018) suggested, with explanations both perverse and profane, "Perhaps it's an Oedipal thing, that Putin reminds Trump of his authoritarian father. Possibly it's blackmail or his fear of people suspecting that Russia saved his businesses." Inferred evaluations. The inferred expectations and explanations above were intimately tied to evaluations: The unmet expectations more or less implicitly defined the summit as a failure. Beyond this, however, several articles conclusively categorized the meeting as a failure even before it begun. Simply agreeing to meet "is viewed by Moscow as a victory," said the Times (Higgins & MacFarquhar, 2018). A BuzzFeed News headline similarly claimed, "This Is How Vladimir Putin Owned Donald Trump-The Russian president didn't need concrete deliverables to come out of the meeting looking like a success" (Tamkin, 2018). Such statements, alongside the failed expectations to talk about difficult issues, allowed journalists to evaluate the summit as a failure; in the absence of U.S. sources with key insights from the talk, reporting focused on other reactions to frame their post-summit evaluations. BuzzFeed noted the "Russian delegation appeared positively giddy" (Tamkin, 2018). Politico's Jack Shafer similarly reported, "Trump attacked the Mueller investigation of Russian election meddling, calling it a 'disaster for our country' and said he held 'both countries responsible' for the Russian cyberattacks," to underscore "Trump's genuflection to Russian President Vladimir Putin at the joint press conference" (Shafer, 2017).
Once the summit concluded, journalists drew on a range of experts and politicians to substantiate their evaluation that the summit was a failure for the United States (Ansari, 2018). The most vocal source was Trump critic Senator John McCain (R-AZ); all three news outlets quoted the Republican as saying about the summit: "No prior president has ever abased himself more abjectly before a tyrant. . . . Today's press conference in Helsinki was one of the most disgraceful performances by an American president in memory" (Mimms, 2018). Indeed, many articles relied on the same sources; they all listed, at length, the tweets, statements, and posts from a range of political and government officials to corroborate their own evaluations.
Efforts at evaluation and explanation are fairly common journalistic responses to new information. They became pronounced in this case, however, when the limits on journalists' ability to develop plausible conclusions through their usual routines are exposed. For example, journalists seeking to clarify Trump's comment on Putin's "incredible offer" might, using conventional approaches, speak with the White House press office. Yet, as Hounshell recounted, the press spokeswoman declined to clarify. This left journalists with little other option than to infer. Efforts to resolve uncertainty also evidently led to inferential reasoning practices. The summit case to some degree follows the mechanisms identified in the first two cases. Yet, while some measure of newswork was possible in the first two cases, this was impossible here. As such, the actual meeting remained a black box. Furthermore, although a lack of source transparency marked the other cases, it was a dominant obstacle here. According to BuzzFeed, "Information ahead of the talks-on what form the summit would take and what topics would be covered-came not from the US side, but from Kremlin aide Yuri Ushakov, who was present at the summit" (Tamkin, 2018). Processes of inference are, to some degree, dependent on a level of transparency into what has occurred, because journalists need something to work with; journalists need material on which to build. In the case of the Helsinki summit, there was not much to follow, leading to coverage built overwhelmingly on inferring expectations of what should occur; then, once the summit was over, inferring from these expectations to offer explanations of what might have occurred during the summit; and finally building toward journalists' own evaluations of the summit.
Our prior cases showed a journalistic community reacting to unanticipated publications (dossier) and unexpected claims ("wire tapping"). This third case involved a scheduled news event, one that was not only planned but that would presumably follow a relatively familiar set of routines. Meanwhile, the journalistic community had moved some distance in developing practices for dealing with uncertainty. Nevertheless, here, when confronted with few resources to confirm or refute possible explanations of what occurred, an inferential community emerges.
One aspect that we repeatedly considered was the transparency with which journalists present their reasoning processes. When confronted with uncertainty and pushed toward inference, as was the case before and during the summit, journalists publicly narrated their own negotiations around unverified and unverifiable facts. Moreover, where they infer they also made clear the limits posed by uncertainty. At the same time, within a community of practice they drew on "allied" resources to make reasonable judgments around (best) possible explanations. In that sense, even when they could not entirely resolve uncertainty, merely by reporting on such events and the larger contexts of those events journalists work to reduce uncertainty. However, here this uncertainty is limited to the explanations for why the summit was a failure, and not the evaluation that it was a failure. This point seems largely agreed upon across the media we analyzed. Some journalists-especially commentators and opinion writers-were vehement in their conclusions that this summit was unsuccessful and even a victory for Russia, but even within non-opinion news articles this evaluation was presented as the result of inferred expectations. At that point, journalists let the public make the final step in inferring an explanation for this conclusion.

Concluding Discussion
We have studied how journalists, working within a field of knowledge production, endeavor to explain prominent events. In particular, we explored how journalists do their work when faced with unverified and unverifiable, albeit newsworthy, information. Below we bring together the dynamics we explored to offer a forward-looking conceptualization of journalists working as an inferential community to resolve uncertainty. Our conceptualization originated in the context of particular understandings of conventional ways of doing journalism. But new realities compel us to revisit those understandings, so we highlight what our findings offer for new ways of making sense of public-facing demonstrations of newswork incorporating inferential reasoning practices in increasingly uncertain times.
First, we are fully aware that in choosing to focus narrowly on news surrounding the person of Donald Trump, both before and after he was inaugurated as U.S. president, we analyzed a unique set of cases. Nonetheless, this choice exposes practices with wider implications. We chose cases that address three discrete types of uncertainty, and each case engages with reasoning around uncertainty in different ways: A particularly unwieldy subject matter-politically damaging, if true, content gathered together in a dossier; politically charged accusations, if true, in a tweet; and, politically uncertain agreements, if made, in a closed-door dialogue between political leaders. Wrestling with these high-stakes cases illuminates a community of journalism practice at its most active. In that sense, journalists were engaged in, as Berkowitz (2000) puts it, "the process of taking extraordinary occurrences and reporting on them in a way that makes journalistic work appear competent to news media audiences" (p. 129).
In one sense, the circumstances of these cases are specific to the current political climate and the U.S. political situation in particular. At the same time, they draw on established thinking around journalism practices that have informed the shaping of the field thus far. They also make evident how these are changing in a digital age. For all the attention to Trump's "unique" nature, journalists' efforts to resolve uncertainty, to reason through explanations, and to deliver news-as-knowledge to a public are obviously not unique to this context. Similarly, for all the attention paid to Trump and a dossier, a tweet, and a "bizarre" press conference, salacious information about public figures, unfounded statements by politicians, and awkward public performances are far from restricted to him or now. The Profumo affair in the United Kingdom was a parallel scandal to the dossier case; then, too, newspapers and particularly the News of the World highlighted possible "kompromat" linked to a politician's alleged sexual proclivities (Conboy, 2016, pp. 126-127). We could point to any number of cases of political surveillance for political gain, Watergate most infamously. In a case of striking similarity to the Helsinki summit, then-president George W. Bush famously said he trusted Putin, having "looked into the man's eyes" during a summit in Slovenia in 2001 (Goldgeier & McFaul, 2002).
Beyond their historic parallels, these cases offer three notable elements with wider resonance for journalistic practice. The first involves the deliberations that investigative and political journalists go through when choosing to make information publicwhich now occur from within a more interconnected journalistic field. The second addresses the communications by public figures who sidestep journalists through social and online media and in doing so make public pronouncements that journalists must then react to. The third deals with new types of politics and political actors who defy expectations, even while they are working within otherwise familiar news structures and are reporting within the highly familiar structures of news events. Locating inferential reasoning and shared sense-making practices in covering each of these aspects opens up wider understandings of these reasoning practices.
While we argue that an inferential community has become pronounced in response to these circumstances, Dimmick (1974), among others, long ago examined newsroom reasoning as part of the input/output function of gatekeeping. Furthermore, while we argue knowledge work in public is both more pronounced and intensified in the current environment, we also draw a thread of constancy from Park's (1940) and Ekström's (2002) engagement with knowledge work in the relationship between journalists and their publics. Finally, while certainty and uncertainty are more apparent when journalists confront potentially dubious information and unreliable public narrators, this has long been a tension for news workers (Anderson, 2018). Yet, our research also suggests new challenges for journalism and journalism scholarship. Bourdieu (2005) describes journalism as a field endeavoring to present a vision of society to the public, which the public then agrees to. In their efforts to do so, journalists make judgments (Ettema & Glasser, 1998), which allow them to express with some certainty (Anderson, 2018) and authority (Carlson, 2017) news that can be regarded as a truthful and knowledgeable account of reality (Nielsen, 2017;Park, 1940). These ongoing practices are playing out in ways that foreground shifts in journalism practices that are unfolding more publicly than they had in the past.

Knowledge Work, Publicly Performed, Toward Reducing Uncertainty
The knowledge work of journalism occurs both before and after the publication of a news item (Ekström, 2002). We found this knowledge work to be apparent not only between journalists and their traditional public (of audiences, or citizens) but also among peers of journalists who engage within a community of practice. Journalists both observe and respond to the practices of their colleagues and peers in a kind of monitoring that now takes place online. This monitoring enables journalists to evaluate their own reporting and validate their own work; it simultaneously prompts journalistic meta-commentary. This meta-commentary signals how members of the field engage in introspection, examining the boundaries of journalistic practice, such as when one actor steps out ahead of her journalistic peers, leaving others to either follow or explain their reasons for not doing so. Thus, boundary work (Carlson & Lewis, 2015) around uncertainty reduction entails a distinct, shared practice of responding and reacting to both established and emerging peers. This practice, particularly evident in response to BuzzFeed's publishing the dossier, is not straightforward. As shown here, alongside an attempt to castigate BuzzFeed for its decision, news outlets, including the Times, nearly inevitably built upon the dossier's claims. This balance between dismissing and engaging with leaked information has become more common in a diverse digital media environment (Eldridge, 2018).
In these complex interactive spaces, inference is important for navigating the sharing of information on the peripheries of the norms and routines of journalistic work, whether defying norms of publicity (with the dossier), disintermediating (with the tweet), or playing with the familiar practices of foreign relations and foreign news events (with the summit). Inference comes through in these cases in both "everyday" and more sophisticated understandings of reasoning. Moreover, journalists show their efforts to make sense of the building blocks which explain social actors' actions. In particular, the conceptualization of inferential reasoning as journalistic practice helps us understand outcomes and the premises for such outcomes, particularly those which otherwise are obscured. This is evident when journalists present their best possible explanations for something having occurred, even when these are not definitive (Harman, 1965). While it can be distilled to the simple construction of "if/then" conditionals, inference within the public performances of newswork reveals a highly complex dynamic of knowledge work.
One of our findings is that inferential reasoning expressed in narratives of newswork is bidirectional (Anscombe, 1974(Anscombe, /2005. When journalists are presented with premises upon which conclusions might be based, journalists build forward from these bases toward probable conclusive outcomes. Forward inferential reasoning was apparent in the efforts to resolve uncertainty in the claims made in the dossier case. When conclusive statements are presented-as in the president's tweet about a "wire tap"journalists engage in backward inferential reasoning, moving from the conclusion posed to unpack what would be necessary premises for such a conclusion. They can then reject these premises as improbable based on their own expertise and authoritative knowledge work. As a result, we see inferential reasoning as a necessary tool in journalists' toolkit. It allows them to confront and thereby reduce uncertainty in an increasingly uncertain political arena. Finally, we can see the evolution of this practice in the ways particular "abnormal" actions are made sense of through inferential reasoning-for example, in the Helsinki summit case, when applied to the "abnormal" nature of Trump's press conference statements. Here, inference is a reasoning practice for setting the parameters of understanding in anticipation of a news event, establishing premises for what is anticipated as a conclusion, then shifting quickly toward explaining the unanticipated result. Inference as a type of knowledge work has a long history. What differs in the cases explored here is the degree to which these practices take place publicly, not only in terms of responding to publicly (rather than privately) sourced information about or from prominent figures, but also in the ways in which the public narration of knowledge work takes place publicly, and is now made apparent through shared meaningmaking practices including through hyperlinking. Journalists' work across the news ecosystem show how they all address what both journalists and their publics consider acceptable and sufficiently true knowledge. This is particularly evident in intermedia linking and textual sign-posting. In part, this may be a reaction to the pace of news unfolding; our second and third cases offer particular demonstrations of this. But certain revelatory practices of journalistic reinforcement also allow the public to see and perhaps appreciate how newswork is accomplished. This potentially bolsters audience perceptions of journalists at a time when their work is regularly derided, including by public figures. Relatedly, by doing their work so transparently, journalists also allow audiences to perform their own inferential work if they so choose.
Importantly, news performances and texts studied here show that explicit narratives of newswork involve "inference to the best explanation" (Harman, 1965, p. 92). Reinforcing inferential reasoning can indicate, albeit not always resolve with certainty, potential explanations for a particular occurrence or event. While uncertainty cannot be eliminated, journalists often have no choice but to first proceed by "believing that what we are told or what we read is correct" (Harman,p. 92). This is important for journalists, as our analysis shows, because knowledge workers are expected to make sense of what prominent sources say or about a prominent claim. By the time of these cases, Trump's authority with fact claims was already shaky. Yet, his prominence meant his statements were afforded the opportunity of being believed. Moreover, once journalists made this decision, consistent with Harman (1965), they engaged in reasoning processes necessary to reach the best possible conclusion. Our analysis finds that this process can end in various ways: With uncertainty still being salient, as with the dossier; with uncertainty being reduced and the initial claim being ruled unlikely, as with the tweet; or with uncertainty being explained and evaluated, as with the summit.
Texts continuously bear markers of reassurance, of certainty, to establish authoritative accounts (Carlson, 2017). Our cases show how texts emphasize normative expectations, including why engaging with unproven fact claims, including alleged corruption alongside salacious, speculative, or sensational stories, is necessary when they revolve around highly prominent public figures-here, the U.S. President-elect (Von Wright, 1963). News discourses relay the evaluation of "warranted" coverage (Harman, 1965), which entails weighing decisions made by other colleagues within the community of practice (Dimmick, 1974).
Clearly, the space(s) in which journalists perform these practices have opened up. Bruns (2005) calls this "gatewatching" to highlight new practices of monitoring online news, responding and reacting to and building on news texts, including by referring to previous reporting, often through hyperlinks. At one level, we also saw this. However, focusing primarily on the monitoring, linking, and interlinking of content belies the unfolding of a more complex dynamic. Here, journalists were reacting not only to online news content but how digital media more widely inject themselves into public view. The complexity is not merely a technological feature, although this is clearly a concern as information reaches the public arena through new and newly digital avenues. Indeed, the pace and prominence of social media require journalists, whether with legacy media or political niche outlets, to be able to respond to and to provide context to what might otherwise be overlooked-that is, were it not also so public. At the same time, dealing with such uncertainty does not happen in a vacuum. The complexity of making sense of unverified and unverifiable information highlights a challenge for all journalists, who remain committed to reducing uncertainty in news stories. While this has become more dramatic in the digital age, what is certain and uncertain is up for negotiation (Waisbord, 2018). And journalists who are regarded (and who see themselves) as responsible for clarifying events, especially political events, are even more cognizant that their own authority depends on doing so reliably.
During the writing of this monograph, we continued to see examples demonstrating the importance of inference in newswork. BuzzFeed Editor-in-Chief Ben Smith, who already was the one to defend publication of the dossier, was back in the news defending BuzzFeed's reporting of another political scoop (CNN, 2019)-he went to pains to defend as an important process reporting which inferred from pieces of information the larger truths about Trump's business dealings in Russia. An op-ed writer for The New York Times applied inferential explanations to the activities of former Trump campaign manager Paul Manafort, activities that are secret and therefore push those writing about them to make the best explanation possible: "All of the available explanations for Mr. Manafort's self-destructive path seem highly implausible, at best. So which hypothesis is the least implausible?" (Litman, 2018). Discussing the same legal travails on National Public Radio's (NPR) Politics podcast, reporter Phil Ewing drew his conclusions based on what he could: "we don't know what we don't know at this time." Ewing acknowledged that his reporting was based on snippets of public information in legal briefs, noting that his conclusions were based on "inference . . . [and] further inference" (NPR, 2019).

Knowledge, Communities, and Beyond
We largely, and somewhat schematically, argued above that journalistic knowledge is constructed through the interactions of two types of interpretive communities, namely those of production and those of consumption. Our conceptualization of an inferential community focuses on how knowledge is produced within communities of practice, particularly how this is done and/or becomes visible in textual form. It is, in other words, a community that partly forms through the production of textual knowledge that displays processes of communal inference. As such, it is a public-facing window into how knowledge is produced within and across texts and within and across journalistic outlets. At one level, therefore, this is simply a community of practice. At another, it is also an interpretive community-beyond the historically facing community described by Zelizer (1993)-in the sense that it offers a public weighing of material, arguments, and practices that are interpreted across the community as the coverage unfolds.
What we have not studied are the roles of audiences in the production of knowledge. Audiences increasingly produce texts that, in various ways, constitute public knowledge in their own rights and/or play into the journalistic production of texts where authority is constructed somewhat differently. An appeal to publics is, however, visible in the texts. This is mainly the case when conventional newswork fails to deliver enough material for processes of inference deemed strong enough for plausible conclusions, leaving journalists to either give up or to make available what they have unearthed, then leaving it up to audiences to make their own inferences. This was clear in the dossier case and in reporting on the Helsinki summit, where articles linked verifiable information on criminal cases linked to Russia and collusion as a context of the summit. In these cases, no explicit inference or conclusion was drawn. Readers were left to connect the dots and infer on their own. A next step would be to design a study to investigate how knowledge is constructed across boundaries of production and consumption. This is no easy methodological task: It necessitates the ability to follow the interaction of various processes over time. While digital methods tracing digital texts may be able to unearth how this takes place across different contexts (Günther, Buhl, & Quandt, 2019;Zamith, 2019), another entire layer of uncharted offline processes cannot be accessed through semi-public texts. This would necessitate more anthropologically oriented methods. Nevertheless, more could be done with texts to enhance the case studies presented here. A first step would be to expand the range of media investigated in terms of types (e.g., TV and radio news) and geography. Then sites not traditionally thought of as news media (e.g., social media) could be included. With a broader range of media types and thus (journalistic) texts, researchers could work to operationalize the textual markers proposed in this monograph to analyze conduct. We hope that the concepts, analytical approaches, and issues raised here may indeed inspire further studies of how knowledge production take place through processes of inference across a complex media landscape where journalists' texts no longer necessarily hold the authority that they once enjoyed.

Authors' Note
This project is a joint intellectual endeavor by Henrik Bødker and Scott Eldridge; this article represents an equally shared product of this partnership and a 2-year period of robust debate and dialogue.