Theories of the circulation of information in the media.

Many theories attempt to establish a model of the flow of information in, and its circulation via, the media. Here, we only describe those that are the most relevant for our study. For instance, Katz and Lazarsfeld’s two-step flow theory [15], developed in the 1970s, questioned the effects of media on the public and the role of opinion leaders (e.g. how those who relay the information can indirectly influence individuals). This approach remains a reference in many areas, such as marketing, and is also very relevant in studies on the diffusion of innovations [16]. Undeniably, the notion of "opinion leader" is effective in describing the logic of influence between the various participants. Moreover, the “Uses and gratifications” model questions users’ motivation and reveals the communicational value of the information exchanged. This theory is also mobilized in knowledge management research [17] in order to identify the motivations behind the sharing of information on academic forums (e.g. information sharing, problem solving, debates, shared interests, etc.). In the present study, this concept allowed us to understand the motivations of Internet users to cite a research article. However, the analysis of the social meaning of the motivational actions to share information have often been sidelined in this model, which has led research in cultural studies to further contextualize media usage by taking the variety of profiles and life experiences into account, and considering the general public as participants in their own history.

More recently, research on mediacultures or on the convergence culture [18] has fostered a new understanding of the links among media, the general public and culture by showing how participants can build a specific form of reality through media mediation.

Theories of the ‘dissemination’ of knowledge.

Scientific studies have theorized ideas of knowledge circulation mechanisms in the media based on three main paradigms [19] [20] [21] which are directly related to the media theories previously discussed. The oldest model is the "lack of scientific knowledge by the public". This top-down linear model, which prevailed in the 1970s and 80s, considers the general public as a relatively passive homogeneous whole. Later, in the 1980s, in the prevalent context of questioning scientific authority and of public scientific controversies, a model "of dialogue" emerged, urging its users to take the various participants in the sphere of sciences into account. Since then, the participatory model developed in 2000 has considered science as an area of coproduction of knowledge, where knowledge amateurs intermingle with knowledge experts [20]. This third model is based on the renewed vision of knowledge production modalities developed by Nowotny & al. [22]. According to their descriptions, the two modes of knowledge production are in opposition: Mode 1 describes science as governed by the academic interests of a specific community and Mode 2 is characterized by collaboration not only between different scientific disciplines but also between scientific communities and the socio-economic world. In fact, in the example analysed here, it is important to point out that it was the fishermen who alerted researchers to the change in catfish feeding habits. This third model, the so-called "participatory culture", is a very similar to the one developed by Jenkins [18] on the culture of convergence, which unites multiple participants and media in a collaborative approach. It is based on Web 2.0 features, which draw on the media to broaden the audience for specialized content. Today, there can be no doubt that the general public has "access to information and resources previously accessible only for journalists, for enlightened amateurs, or for researchers” [23].

Nonetheless, it is essential to emphasize (c.f. Bensaude-Vincent [19]) that these knowledge dissemination regimes now coexist and are not mutually exclusive. In this paper, we show that their coexistence reflects an emerging situation of knowledge circulation, appropriation and creation based on the analysis of the circulation of a scientific article in the media.

Ranking by media type.

We ranked the sites relaying the information by media category and pooled the data in an analytical table.

Six categories were defined:

• Online version of traditional media (VL): La Dépêche (Fr), France Inter (Fr), The Week (US), The Blaze (US), The connexion (GB), Birdwatch (GB), etc.
• Sites Native to the Web (NW, information sites existing only online): Actuzz (Fr), Carré d’info, Fun-buzz (Fr), Business insider (US), Grist (US), Mail Online (GB), Descrier (GB), etc.
• Content Aggregators: NewsYahoo (Fr), Google actualités (Fr), Geekosystem (US), Yahoo News (US), Travel AOL (GB), Noxxed (GB), etc.
• Forums: SilureGlanis (Fr), Carpe Alsace (Fr), io9 (US), Carp Forum (GB), Fishkeeping (GB), etc.
• Blogs: 2tout2rien (Fr), Littlefishing, America Blog (US), ZME Science (US), Pigeon Mania (GB), Following the nerd (GB), etc.
• Sites (administrations, corporations, associations): Terra Femina (Fr), Amis des Amis (Fr), 10000 Birds (US), GeekyDump (US), Aqcenter (GB), Grindtv (GB), etc.

S1 Fig shows the proportion of categories of each site in each country.

Boundaries between the chosen categories are complex and porous. The choice of these categories nevertheless proved to be useful for analysing the circulation of the PLOS ONE article in the Internet media sphere. Each type of site has, by its nature, a specific editorial line, a different editorial status engaged (the expert-researcher, the journalist or the amateur), distinct reader agreements [24] and various ways to appropriate and process information.

Ranking by theme.

To complete this analytical table, we designed a thematic classification to analyse the editorial treatment of the PLOS ONE article citations. Depending on the site, we observed two distinct editorial choices: in the first case, the change in the catfish feeding behaviour was highlighted while, in the second case, the oddness of the information or even problematic fishing policies were underlined.

Seven themes were identified, some of which were broken down into the following sub-themes:

• General information (News): 20min (FR), Canal plus (FR), Fark.com (US), Grist (US), The connexion (GB), French News Online (GB), etc.
• Media and Technology (M&T): Actuzz (FR), Divertissonsnous (FR), Reviewer (US), Geekosystem (US), Business Insider (US), Gismodo (GB), Digital Spy (GB), etc.
• Fauna and Flora (F&F): La pêche et les poissons (FR), Actualités news environnement (FR), Animal Fact guide (US), LiveLeak (US), Fishkeeping (GB), Pigeon Mania (GB), etc.
• Science and Technology (S&T): Sciencesetavenir (FR), Futura-Sciences (FR), Discover (US), SciTech Daily (US), etc.
• Odd and Funny (H& I): Mortderire (FR), vidéo-buzz (FR), Blog's avenue (US), Tobefun (US), Boreme (GB), Anorak (US), etc.
• Academic: Paul Sabatier University (FR), CNRS (FR), Smithsonian (US), etc.
• Diverse: (category that includes disconnected themes such as rural life, the paranormal, health and individual initiatives): Nous ne sommes pas seuls (FR), Fédération des acteurs ruraux (FR), EarthTouch TV (US), Grindtv (GB), etc.

Some themes or types of sites maintained rather vague borders (S2 Fig). For example, some media were not present in the whole corpus. This was the case for academic sites such as the CNRS or University Paul Sabatier, for which there are no equivalents in the United States or Great Britain. Likewise for the “Science and Technology” theme, which was not represented in Great Britain.

This element led us to develop thematic categories not from the way the contributors themselves defined the themes of their websites, but rather through a content analysis of our corpus websites.

Key role of leading sites, true diffusion nodes.

In analysing the circulation modalities of the article on the European catfish, we quickly identified the leading sites, the “real nodes of diffusion” in the three countries of our study. Similarly to what the two-step flow theory of communication identifies as "opinion leaders," these "source" documents - true "opinion formers" - seemed to emerge as unavoidable diffusion nodes on the Web. These leading sites were more than just relays: they played an advisory role in advertising content worthy of being webcast by others.

The main US source was the blog, Not Exactly Rocket Science, (webcast of 5^th December) by Ed Yong, a British science journalist for the American promotional magazine, Discover, (26.5% of the US articles analysed cited this source).

In Great Britain, 15% of the analysed records cited the Daily Mail (webcast of 7^th December), an online version of the British newspaper, while only 9% mentioned Ed Yong’s blog.

In France, the main opinion leader’s website was the Passeur de Sciences (PDS) blog (webcast of 6^th December), run by Pierre Barthélémy, a science journalist for Le Monde, a daily newspaper (34.5% of French documents referred to it). Just like "pure scholar readers" [10], "journalist bloggers" turn away from their usual practices to write promotional articles and, as a result, participate in the circulation of scientific information. These blogs, run by scientific journalists, play a major role in the dissemination of scientific knowledge.

On 7^th December, two other sources played a key role in France: an article from La Dépêche du Midi (LD), a regional daily newspaper (20.5% of the documents referred to it) and the Agence France Presse (AFP), a world news agency, (19.2% of documents referred to it).

This first analysis shows that the principal information relays, whether in France, the United States or Britain, are journalistic sources (such as articles in the online media and journalists’ blogs). Today, journalistic sources remain references, even on the Web. Gradually, the information moved further into the amateur sphere [19] [20], through other participants and other types of media support (blogs, association sites, etc.).

Furthermore, in France, citation analysis of the documents identified showed that most of them repeated the information webcasted by the three nodes mentioned above but without making explicit reference to them. It was on the basis of "key words", numbers, geographical indicators and extracts of interviews of F. Santoul that we identified the intertextual relationships between these different documents.

Likewise, only 20 French sites cited Passeur de Sciences as a source when, in fact, 26 of them used the same wording. For the AFP, the gap was even larger as only 3 out of the 15 identified articles explicitly referred to the source. What was more important was the dissemination of the content and not the secondary source that attracted attention to the same content.

S4 Fig below gives a clear account of the complexity of the circulation network of the PLOS ONE article in the Web-based media sphere and highlights one of the relay nodes analysed in France: the scientific blog Passeur de Sciences.

The identification of this singular content in the various relays allowed us to reveal the intertextual phenomenon occurring on the Internet in a broader way. References and citations did circulate from one site to another; Internet users modified and appropriated the content that they, in turn, published. The Passeur de Sciences case was not isolated since it appeared for each of our three main opinion leaders in France. This phenomenon corroborated the role of hub-sites as opinion leaders playing an implicit role as advisors of mediatized contents. The fact that the leading sites took time to mention a certain piece of information eventually led many "trend-setting" users, usually members of diverse communities of interest, to take over this information. By doing so, these “trend-setting” users seemed to aim more at publicly displaying their membership of the digital sphere and social networks than at contributing to the visibility and recognition of leader sites (through references or citations).

Academic references.

This dissemination was duly sourced from an academic point of view and, from this perspective, we are coming closer to understanding the circulation modalities of scientific information in academic blogs [13]. References to the original article in PLOS ONE were present in almost half of the articles analysed in the three countries. Consequently, the article from PLOS ONE was cited or mentioned through a link in 46.9% of the websites in the US, 35.1% of the websites in France and 30.3% of the websites in Great Britain (S5 Fig).

Institutions (laboratories or universities) were slightly less referenced but in similar proportions: 40.8% of the US sites cited institutions versus 27.3% of the French sites and 24% of the British sites.

This phenomenon may seem surprising, since, on the one hand, the relaying sites were rarely mentioned but, on the other hand, the analysed sites maintained no explicit link to the academic world. Therefore, the citation and referencing codes in this realm were not supposed to be known by the Internet users. However, the authors of the identified documents had obviously made the effort to refer to the original article in their own published article. But what were the intentions of these authors? Was it an ethical concern or a rhetorical strategy to support their views and demonstrate their knowledge of scientific codes, and hence their membership of academia? Luzon [13] has shown, in the case of scientific blogs, how academic references and citations were identified as increasing not only the value of the blog but also the credibility of the author vis-à-vis a community of interests (a community of individuals sharing common interests) [26]. A similar phenomenon was observed in our study: the reference to the original article, clearly identified as coming from the established academic world, increased the credibility of the websites in question.

Nonetheless, there was a fundamental difference between the three countries: even though the PLOS ONE article was co-written by six authors, they were not cited in the same way in the documents analysed. Their rank in the list of authors was decisive, as in the bibliometric analysis of scientific publications. However, R.K Merton stated, "As we examine the experiences reported by eminent scientists we find that this pattern of recognition, skewed in favour of the established scientist, appears principally (i) in case if collaboration and (ii) in cases of independent multiple discoveries made by scientists of distinctly different rank." [27]. Only two authors out of six were cited here: the first on the list of authors, Julien Cucherousset, who did the experimental work, data analysis and co-editing, and Frédéric Santoul, last author on the list, who was the scientific director of the research project, identified as having participated in all phases of the research from experimentation to writing the article. Depending on the country from which the articles came, the editors did not refer to the same researchers. In the United States and in Britain, Julien Cucherousset was the representative scientist whereas, in France, the most often cited researcher was Frédéric Santoul. These two different treatments can be partly explained by the origin of the sources that the contributors used or, as demonstrated by Merton [27], by the necessity to highlight Julien Cucherousset, since his status was directly linked to the financing of the catfish study (Julien Cucherousset received an “ERG Marie Curie” grant (PERG08-GA-2010- 276969) in the EDB laboratory, part of the “Laboratoire d”Excellence” (LABEX) named TULIP (ANR -10-LABX-41)).

In the American and British network, Ed Yong and the Daily Mail cited Julien Cucherousset as the first co-author, reflecting the specific citation modalities of the academic world. In contrast, in France, Frédéric Santoul was the only author cited (23.4% of the sites analysed). As a team leader, he was interviewed by two of the three main diffusion nodes identified above (Passeur de Sciences and La Dépêche). From then on, his name was quoted almost systematically by the other websites, providing this source with greater legitimacy each time. This researcher quickly became emblematic of the research on European Catfish in France. The concepts of source and copyright experienced significant changes in terms of the relays identified. While circulating, information became loaded with new embedded values put forward by the different participants who contributed to its circulation.

Citing photos and videos.

The audiovisual material also played a decisive role in the intertextual relations between the original article in PLOS ONE and the subsequent Internet postings. The role of images and videos, used as eye-catchers in the dissemination of scientific blogs, has notably been highlighted by Ranger & Bultitude [28]. In our case, the media scrutinized were especially relevant to this phenomenon: the video posted by PLOS ONE, followed by YouTube, flourished internationally on the Web. On S6 Fig, we can see that this video was present in 89.7% of the American websites studied, in 69.5% of the French websites and 66.6% of the British websites. The Internet postings did not necessarily refer to the original video link on PLOS Media but rather cited a link to a video posted several times on YouTube by Internet users. The spectacular nature of this video was an essential element in explaining this extensive diffusion and circulation [29].

On the other hand, the photos picturing the significant difference in size between the pigeons and the catfish and showcasing competition between two natural environments were much less posted: 38.7% of the US websites used them versus 29.6% of the French websites and 24.2% of the British websites. The use of these photos and videos played a central role in the circulation of the reference of the original article, even if they served to show the spectacular nature of the phenomenon rather than the change in the European catfish feeding habits. Yet, beyond their eye-catching nature (or maybe because of it?) these videos and images played the content accreditation role in very much the same way as the references to the original article did.

Furthermore, some native websites, blogs and forums highlighted different photos from the ones provided by PLOS ONE or by the key site–notes identified above. In these articles, this difference from the original article was due to the authors’ background - often curious individuals, pigeon fanciers or fishermen. Although posting personal fishing photos (often of the fisherman’s own catch) is common in France (7.8% of the websites analysed), we also observed pigeons pictured as doves in Great Britain, which could have had a particular effect on the recipients of this information. The utilization and the staging of these photos and videos also reflected a particular appropriation of the content by the users. Depending on their target audience, contributors may divert visual symbols to raise awareness to their cause, e.g. pigeon preservation.

Diverse communities of practices and interests.

The "Fauna and Flora" theme appeared repeatedly in France and Britain, followed by the fishing community, who seized upon the information. The "Fauna and Flora" section was fairly representative of the link between types of sites and themes. In fact, whether in France or Britain, the percentage of forums was relatively high compared to other types of media (22% and 27.2% of the web sites analysed). Most of these forums were devoted to fishing (57% of the forums in France and 66.6% in Britain) [30]. The fishing community, understood here as a singular community of interests and practices, was largely mobilized in both countries. The information was of interest to them not only through its ecological nature but also for its spectacular elements.

Conversely, this section was under represented and even marginal in the United States and consequently represented a small proportion of forum documents in our corpus (only 2% of the American websites analysed). In this case, fishermen were not particularly interested in the information since catfish are quite unusual in American waters.

Nevertheless, fishermen were not the only community to be interested in the science of catfish or to appropriate the original PLOS ONE article through information poaching and alteration processes [31]. Communities of interests as diverse as technology amateurs or online entertainment fans seized this scientific information and appropriated it through discursive regimes often very close to information spectacularization.

All of these communities played a specific role in the diffusion of this information on the web and participated in the media hype that this first series of documents generated. We specifically studied this feature in France. As shown in the chart below (S7 Fig), 21% of the French articles did not mention the behavioural change in catfish feeding but rather discussed the hype generated by this information on the Web. The buzz dimension appeared very early in the circulation of the PLOS ONE article. Importantly, it was not in the "Odd and funny" nor "Media and technology" sections but in two rather journalistic themes: "News" and "Fauna and Flora". In fact, it was mainly the journalists and the fishing community that contributed to the circulation and the media boost of the spectacular dimension of this information.

Geographical anchorage of communities of practices and communities of interests.

The communities of practices and interests were also often communities anchored within a specific geographical territory. The research was conducted by a team of French researchers from the Midi-Pyrenees Region, located in the South West of France. Moreover, the catfish story was also rooted in that specific area and the scientific observations of its occurrence took place in that same region. Finally, many participants in the media, the scientific and the associative spheres were also from the same geographical area.

This phenomenon was rapidly identified in the leading article published by the regional daily newspaper, La Depeche du Midi. It stated the characteristics linked to the geographical anchorage of the research and its particular knowledge of that area. Hence, the characteristics of the place where the research was conducted were specifically mentioned: primarily the river Tarn but also Albi or its Cathedral (the small town of Albi, listed as a UNESCO World Heritage site, is locally renowned for its Cathedral) and it included finer idiosyncratic details like "the teacher and his students are photographed on the parapet of the Old Bridge, their eyes fixed on the island located below the dam". Finally, local specificities, such as the name of the university in Toulouse, also reinforced the local dimensions of the information.

More generally, this territorial anchorage reflected the proximity and the regular exchanges between the different people and social categories behind the circulation of this information: researchers, journalists and fishermen, were all guarantors of the disseminated information. This was demonstrated not only in the interview of the iconic scientist, Frédéric Santoul, who leads the research on catfish in France (key interview sentences were repeated on numerous occasions by many Internet users who, de facto, appropriated the researcher’s words without referring to the original article) but also in an interview with Stéphane Cabanes, the president of the local angling association.

The presence of visible local references, such as geographical statements or interviews of locally influential personalities, was rarely used in the American and British corpus we analysed; obviously these references to local peculiarities were not relevant for the audience abroad. This shows that a theme is more developed when it can be anchored not only in communities from a common area but also in local communities of interests. These two factors explain why the PLOS ONE article mostly circulated in France.