4 " Unamerican Views " : Why Us-developed Models of Press-state Relations Do Not Apply to the Rest of the World

The article shows the limitations of the " indexing " hypothesis, an influential conceptualization of state-press relations based on the notion that the media tend to reproduce the range of debate within political elites. The hypothesis, as confirmed by an international comparative investigation of the elite press coverage of 9/11 in the US, Italy, France, and Pakistan, cannot be applied outside the American context. The analysis finds that the variation in the levels of correlation between elite press coverage and governmental discourse are explained by previously neglected variables: national interest, national journalistic culture, and editorial policy within each media organization. The article argues that more international comparative research and multidisciplinary approaches are needed in order to renew old paradigms, especially at a time when the distinction between foreign and domestic politics is disappearing. Introduction An important conceptualization about the media-foreign policy relationship is the indexing hypothesis. The idea of " indexing " was originally developed by Bennett in 1990 in a study about the New York Times coverage of U.S. funding for the Nicaraguan contras. The study supported the elaboration of the hypothesis that 'mass media news professionals, from the boardroom to the beat, tend to " index " the range of voices and viewpoints in both news and editorials according to the range of views expressed in mainstream government debate about a given topic' (Bennett 1990, 106).


Introduction
An important conceptualization about the media-foreign policy relationship is the indexing hypothesis. The idea of "indexing" was originally developed by Bennett in 1990 in a study about the New York Times coverage of U.S. funding for the Nicaraguan contras. The study supported the elaboration of the hypothesis that 'mass media news professionals, from the boardroom to the beat, tend to "index" the range of voices and viewpoints in both news and editorials according to the range of views expressed in mainstream government debate about a given topic' (Bennett 1990, 106). Subsequent applications of the hypothesis confirmed Bennett' findings and contributed to refining it by detailing the extent of the indexing of media coverage to domestic political discourse in different circumstances. Bennett and Manheim (1993), for example, analyzed New York Times coverage of the Gulf War and found a correlation between the range of coverage at each stage of the crisis and the range of debate among officials. Among other studies (Althaus at al., 1996;Livingston ______________________________ Westminster Papers in Communication and Culture © 2008 (University of Westminster, London), Vol. 5(3): 4-26. ISSN 1744-6708 (Print); 1744-6716 (Online) and Eachus, 1996;Mermin, 1999) Zaller and Chiu (2000) observe that media coverage during forty-two foreign policy crises, from the Soviet takeover of Poland in 1945 to the war in Kosovo in 1999, followed the cues of officials, even if after the end of the Cold War media tend to become more independent of the US congress and the president. Billeaudeaux et al. (2003) examined New York Times and Washington Post coverage between September 11 and the beginning of the War in Afghanistan and concluded that the language of the US administration produced effects ('discernible patterns') in the editorial contents of the two newspapers. Hutcheson at al. (2004, 44) observed that 'the discourse of U.S. journalists about U.S. national identity in the aftermath of September 11 closely paralleled that of government and military officials'.
Indexing has been confirmed over and over again for at least the past 15 years, to the point that the idea that the media tend to reproduce political messages is almost taken for granted. Some studies point at the empowerment new technologies are giving the media to evade the official framing of events (Livingston and Bennett, 2003;Livingston and Van Belle, 2005), but they represent exceptions rather than the rule within the literature. In fact, particularly after 9/11, a general belief has consolidated that we have 'ushered in a new era of media subservience to the state when it comes to matters of international crisis and conflict' (Robinson et al., 2005, 952). The implication is that the media are somehow "failing" to fulfill their function of democratic "watchdog" (Bennett, Lawrence and Livingston, 2007). The tendency of the media not to question the official line is regarded as detrimental to public debate. Through biased reporting journalists fail to inform citizens (Philo and Berry 2004). By being deferential to authorities they are accused of contributing to the government's propaganda machine. Wolfsfeld (1997), for example, refers to the media as 'faithful servants ' (p. 69). Guardian columnist George Monbiot (2004a), writing about the coverage of the 2003 Iraq War, suggested that 'journalists help governments to lie to the people'. He went as far as arguing that 'the falsehoods reproduced by the media before the invasion of Iraq were massive and consequential: it is hard to see how Britain could have gone to war if the press had done its job' (Monbiot, 2004b).
This article questions such widespread ideas about the role of the media in society by showing that the extent to which they reproduce the official line varies across countries and media organizations and that the final shape of the coverage is not the journalists' responsibility only. While it is media professionals who physically construct the news, the choices they make in their newsgathering and newswriting routines are actually guided by a much broader social and political context than their contingent interaction with political actors.
The article does not deny the validity of the original formulation of indexing, but it does challenge its use as a paradigm to explain the media-foreign policy nexus outside the US by raising five important points. Firstly it demonstrates the inapplicability of the hypothesis to non-American contexts showing that there are additional and broader variables explaining why journalists choose to reproduce the official line in their reports. Secondly it questions the empirical meaning of "indexing" by showing that the correlation between political debate and media coverage it postulates can be the outcome of methodological choices rather than an objective reality. An in-depth qualitative comparison between political debate and media coverage raises doubts about the applicability of the hypothesis, even to the US context. In fact the decision concerning the level of detail at which the tightness of fit between political debate and media coverage should be approached has a direct impact on the assessment of whether indexing exists or not. Thirdly, the article points out that, despite the apparent trend towards confirming the original formulation of the initial hypothesis, each single study which followed Bennett's (1990), operationalized the detection of indexing in its own way, making rigorous testing and accumulation over time problematic. As a fourth point indexing cannot apply to foreign countries because the dynamics it postulates are based on a specific idea of the role of the journalist in society and the nature of press-government relations. This is the outcome of the particular way in which the American political and media system- Hallin and Mancini (2004, Chapter 4), for example, call it the North Atlantic or Liberal Model to distinguish it from the Northern European or Democratic Corporatist Model, and the Mediterranean or Polarized Pluralist Model-have developed over history. Finally the hypothesis might have more accurately reflected the US media-foreign policy relation in the past rather than in the current information age.
The article raises these points by drawing on the evidence provided by an international comparative study of the elite press framing of 9/11 in the US, Italy, France, and Pakistan. The study quantitatively and qualitatively analyzed both governmental statements and elite press coverage over two months following 11 September 2001 in the four countries. The governmental statements and the coverage were then compared within each case to measure the extent to which the coverage reproduced the official framing of the events. The objective of the study was identifying the variables which shaped the coverage of 9/11 and its aftermath. Beside the hypothesis that coverage would reflect domestic political statements (indexing) the study also explored the extent to which coverage was shaped by international media flows, by the globalization of media networks, and by norms at the level of the single media organization. The discussion of the results presented here is therefore illuminated by such broader insights.

The Problems with Indexing
Despite being apparently straightforward, indexing reveals a much greater complexity when empirically unpacked. Althaus at al. (1996) directly address a clarification of what 'indexing to political elites' debate' should mean in practice. As the authors put it, 'Indexing seems to predict that the media hold a mirror to elite discourse and reflect that discourse to the public' (Ibid,417). The empirical objective of the indexing concept is explaining 'the closeness of fit between media discourse and debate among political elites' (Ibid,408). This could lead, in their opinion, to three predictions: 'the breadth of media policy debate is less than or equal to, but never greater than, the official debate;' 'the news closely reflects the distribution of expressed views among officials'; or 'some combination of the first two' (Ibid).
Even this clarification, however, raises some questions. What does it mean that the media 'breadth' of debate is smaller or equal to the official one? In what does the 'reflecting' the 'distribution' of 'views' exactly consist? The difficulty in understanding what media content being "indexed" to the political debate actually means relates to several aspects. The first is defining what the nature of "indexing" is: it could either be a correlation, a compatibility, or a reproduction of political debate's contents within the news. Depending on the definition of what indexing is, the data analysis could lead to different assessments of the levels of indexing within media coverage: "reproduction" of political messages-i.e. contents detected within the coverage should be exactly the same as those in the political discourse-could be a stricter criterion to meet than "compatibility". In the latter case contents would not have to be the same. They could be just "similar," although, in turn, a new definition would be needed about how similar contents should be in order to be coded as "compatible". And even when such clarification is made, what would the nature of such "contents" be?
On this last point the literature does not provide clear answers. Livingston and Eachus (1995, 425) talk about 'thematic shifts in news frames'. Billeaudeaux et al. (2003, n.p.) use a wide range of terms: 'messages', 'language', themes', 'positions', 'voices'. Althaus et al. (1996, 409) refer to 'views', as do Entman and Rojecki (1993, 170). Bennett (1990, 106) refers to 'views', 'voices and viewpoints' within the coverage. The same terms are also used by Livingston and Eachus (1995, 423). White (1994, n.p.) refers to the presence of 'frames' in the news. Indexing is operationalized as the expression of support vs. opposition to the government agenda in both Bennett (1990) and Mermin (1999). It is the ratio of positive vs. negative coverage in Zaller and Chiu (2000).
The two challenging issues here are: firstly how to establish the relationship between political debate and media coverage in empirical terms; and secondly how to draw conclusions from the data. In fact political actors are a newsworthy source of information for journalists (Cohen, 1963;Sigal, 1973;Gans, 1979). To an extent a certain level of "indexing" is always to be expected. The indexing hypothesis claims to expose a lack of independence by the press. While some would regard reporting official views as part of the press' democratic duty of voicing the views of elected officials (Bennett 1990, 103), others would object that the press, by always remaining within the boundaries of official debate, is not fulfilling its role of democratic "watchdog" (Hallin 1989, 5). What is indexing supposed to reveal to the researcher? At which point indexing starts telling the investigator something at all?-and something interesting rather than obvious? In fact, how close should the content of the political and media discourses be to allow the researcher to conclude that the media is not being critical enough rather than it is simply doing its job?
The key to understanding, let alone measuring, any relationship between media coverage and political debate is in the definition of the nature of the contents in question, their detection in both political and media discourses, in the assessment of their comparison as well as quantification. These tasks present challenges and need careful operationalization through an approach that records small contents differences and allows the quantification of the extent to which political messages are reproduced within the news discourse. These aspects are described in detail in the following methodology section.

Method
The validity of the indexing hypothesis was tested on the case study of 9/11. The case was chosen for having been the focus of substantial media coverage worldwide. The context allowed exploring simultaneously, on the same event, both global dynamics of news reporting as well as local coverage features. The research design involved international comparisons among four embedded cases: the US, Italy, France, and Pakistan. These countries were selected to control for three aspects which were initially thought to be able to affect the closeness of fit between press coverage and political debate. One was a country's involvement in the 9/11 events: the cases spanned from a country directly affected by the attacks, the US, to European countries with no direct involvement (France and Italy), to a case very close to the theatre of the subsequent military operations in Afghanistan (Pakistan). A greater involvement was expected to produce, in comparative terms, a higher volume of coverage and a greater attention by a country's press to the respective national official statements. A second aspect affecting the extent to which news in different newspapers could be "indexed" to authorities' statements was governmental ideological orientation. The US at the time of the event was run by a Republican administration; France had a centre-left PM and centre-right President; Italy was led by a centre-right government. The choice of two newspapers per country, one liberal, the other conservative, as it will be explained later, served the purpose of testing whether a newspaper would report more of the statements by officials sharing a similar political leaning. A third variable was a country's form of government. While the US and France are presidential systems, and Italy a parliamentary Republic, Pakistan does not fit comfortably with the Western definition of a democracy: the country is run by General Musharraf who came to power in October 1999 following a military coup (Talbot, 2005, 375-78). The inclusion of a non-democratic country would have allowed exploring the possible effects of press censorship, restrictions for journalists in accessing official sources, governmental pressures on media professionals.
The time span of the analysis, 11 September to 14 November 2001, covers: the 9/11 events; the first reactions to the terrorist attacks; the developments leading to the intervention in Afghanistan (starting 7 October 2001); as well as the military operation Enduring Freedom until the conclusion of its first phase (fall of Kabul, 13 November 2001). The scope, sixty-four days, is wide enough to capture both immediate "spontaneous" reactions to the events and more reasoned interpretations of the situation, within political discourse and media coverage alike.
The political discourse was approached as all political statements (interviews, speeches, press conferences) by governmental actors (presidents, heads of state, prime ministers, foreign and defense ministers) in each of the countries under study. The media coverage sample included daily first-page and editorial articles from two elite newspapers per country. As Benson and Hallin (2005, 5) argue, elite newspapers are not representative of the entire media coverage of a country, but they 'occupy similar positions in prestige and influence in each society, making them suitable for a controlled comparative analysis'. In each country the newspapers selected for the study had respectively a centre-left (or liberal) and centre-right (or relatively more conservative) political orientation. They were, respectively, the New York Times (NYT) and the Wall Street Journal (WSJ) in the US; Libération (LIB) and Le Monde (LM) in France; La Repubblica (REP) and Il Corriere della Sera (CDS) in Italy; the Dawn (DAWN) and the Nation (NAT) in Pakistan.
The content analysis focused on the process through which political actors, journalists, editors, as well as sources within the news text constructed (framed) the war on terrorism, more specifically the issue of 9/11. The study faced the methodological challenge of approaching news stories in a way that made their contents comparable across the embedded cases. The problem was tackled by approaching contents in terms of a framing process (De Vreese, 2003). The study used as a reference Entman's basic definition of frames.
A frame, according to Entman (1993), fundamentally involves selection and salience: To frame is to select some aspects of a perceived reality and make them more salient in a communicating text, in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation for the item described (Ibid, 52, his emphasis).
Taking Entman's problem definition, causal interpretation, moral evaluation and treatment recommendation as a basic structure for approaching the way an issue is framed, the 9/11 events could be seen as terrorist act (problem definition) by evil people (moral evaluation) motivated by hate for freedom and for America's way of life (causal interpretation). In order to bring to justice the perpetrators the world should declare a war on terrorism (treatment recommendation). This, however, is the framing by a specific actor, in this case US President Bush, at a specific time, the evening of 11 September 2001 (Bush, 2001a). At a later stage, on 19 September, in occasion of a meeting with the Indonesian President Megawati, the same 9/11 events are framed slightly differently. The events are a 'crime' (problem definition) committed by 'evildoers' (moral evaluation) motivated again by hate for freedom (causal interpretation) against whom a coalition of freedom-loving countries should fight a global war on terrorism, which involves not only eradicating terrorist organizations and punishing those who support them, but also freezing terrorists' funds (treatment recommendation) (Bush, 2001b). As seen the contents of a frame, depending on the level of detail at which they are analyzed and on the time span under analysis, can change.
In order to capture small content changes over time within the framing process the analysis focused on smaller units of analysis than a whole frame: IEs elements (IEs).(i) They belong, depending on their relationship to the issue being framed, to different aspects of framing (Problem Definition, Causal Interpretation, Moral Evaluation and Treatment Recommendation). IEs are simple ideas within the news text. For example 9/11 can be defined as either 'war', 'attack', 'tragedy', or 'event'. In relation to the causal interpretation, terrorists could be motivated by either 'hatred for human life', 'hatred for America', 'religious fanaticism', or 'ideology'. The idea of 'war', depending on the context in which it is used, can be coded also as a Treatment Recommendation. President Bush says, for example, that we should wage a 'war against terrorism'.(ii) IEs are used both for quantitative and qualitative analysis within the study. They are processed quantitatively to establish, for example, to what extent news and political discourse present similarities, but they are also regarded as the building blocks through which social actors-in this case sources "speaking" within the news text-construct meaning (i.e. frame the 9/11 issue). In the following "Results" section I will refer to the framing of 9/11 within the news stories by providing examples of IEs.
The content analysis also recorded sources: they are all actors (politicians, commentators, international organizations spokespersons, media organizations like newsagencies) who, within news stories, contribute to framing the issue of 9/11 with at least one IE. They do so either through a statement ('President Bush said: "[quote]"') or through an attribution ('BBC reported that …'). It is important to note that sources are here understood as newsworthy actors making a statement in the news text, not as sources of information approached by journalists in the newsgathering process.
Testing the closeness of fit between media coverage and political discourse involved a comparison between the IEs detected within the governmental discourse of each country and the respective first page news and editorials. The comparisons are conducted among batches of IEs: "framing samples". Framing samples are a selection of the most frequently appearing IEs within either the political statements or the media coverage (either at national level, individual newspaper level, within first page news only, or editorials only), and which constitute at least 50% of all recorded IEs. Framing samples are therefore constituted by those IEs which would appear more salient to a hypothetical reader. The important point is that the IEs of the framing samples are selected as a proportion of all IEs coded either in the political statements or the media coverage. The number of IEs within each sample can therefore change and the sample can be bigger or smaller depending on the characteristics of the political statements and media coverage themselves: how many IEs they present and how diverse the IEs within them are.(iii) The comparison is based on the quantitative assessment of the extent to which political messages are reproduced in the news. A distinction is made, however, between the proportion of the political discourse that is reproduced in the news and the proportion of the news the same contents represent. In other words it could be that all a political actor says is reported (100%), but this just constitutes a fraction of the overall coverage (10%). The study investigates the content of the remaining 90%. It asks where it originates from by analyzing its sources. It also questions, in a qualitative perspective that takes into consideration the way IEs combine with each other within the framing process, to which extent alternative IEs within the coverage challenge the official framing.

Results
The notion that the range of elite press coverage contents, here approached as number of IEs, is contained within the range of debate within the political elite is confirmed only in the US. By comparing the most recurrent batches of IEs (framing samples) in the governmental discourse with those of the press coverage it can be observed that the news framing of 9/11 in American newspapers contained less IEs than in the political statements. Both the NYT and the WSJ, for example, presented 24 IEs about 9/11 against 34 by the Administration (Table 1, next page). In both Italy and France, instead, 9/11 news coverage contained more ideas than in the political statements-REP's coverage, for instance, contained 54 IEs against the official 39 (Table 2, see also Table 3, both below). In Pakistan first page news framing of 9/11 also contained more ideas. DAWN had 35 IEs against the official 20 (    The most interesting aspects of the testing of the indexing hypothesis involve the qualitative analysis of the IEs which were or were shared/not-shared between the political debate and the press coverage. It is true that the project focuses on the correlation of the coverage to governmental statements rather than to the whole debate among a country's political elites. It could be argued that the contents that are not matching the governmental statements originate from other political, but non-governmental, political actors. The analysis of the sources confirms that this is not the case. From a qualitative perspective it is not possible to detect a particularly tight fit between political statements and press coverage. Indeed, even where there appears to be a higher proportion of governmental messages shaping the coverage, as in the US, there are still some key IEs which challenge the official interpretation of the events. In fact 66.67% of NYT's first page news about 9/11 is constituted by IEs also appearing in the political statements. Editorials even present a 72.22% commonality with the political framing (Table 5).  Table 5. Comparisons between political statements, first page news, and editorials in the US

US 9/11
The first page news framing of 9/11 in the coverage of both newspapers, however, does not present several of the core contents of the official framing: it neither reproduces the idea of 'freedom' (IE461) as target of the perpetrators; nor the notion that terrorism has to be fought through a variety of measures (IE1068); nor that the perpetrators should be found and brought to justice (IE745). The NYT's editorials emphasize the need of building a world coalition (IE843) rather than a generalized 'cooperation' (IE842), as suggested by the political statements. Editorial pieces advance new ideas: they recommend improving domestic security (IE1003) and reject any restriction of civil liberties (IE1016).
The French case confirms the importance of considering both the proportion of political statements being reproduced in the coverage and the proportion of coverage in common with the political statements in the analysis: a higher proportion of political statements is reproduced within the coverage of both news and editorials of LM than in the US newspapers. 54.84% of the French political statements is reproduced in LM against 47.06% of the American political statements reproduced in the NYT, for example, which would lead to draw the inference that there is "more indexing" in France than in the US. The common contents, however, constitute just a fraction of the overall media coverage: this time 29.82% in LM's first page news against 66.67% in NYT's news (compare Tables 5, previous page, and 6, below).

Table 6. Comparisons between political statements, first page news, and editorials in France
The first observation is that the framing by the French elite press is broader and more varied than in American newspapers. The remark begs a first question: where does the framing that is not in common with the political statements come from? Looking at the sources contributing to the framing leads to establishing a relationship between number of sources and variety of ideas in the coverage. The number of sources is greater than in the US: 32 sources in LM and 22 in LIB against 17 in the NYT and 13 in the WSJ. While sources on American elite newspapers are mainly domestic, French sources cover US, French, Pakistani, Afghani, EU, Middle Eastern sources as well as international organizations and foreign media such as the American and EU press, and Al-Jazeera. The overall French media framing appears to be broader than the governmental one due to such variety of sources. The first important observation is that the number of IEs within the press coverage depends on number and country of origin of sources quoted: the wider the spectrum of sources, the more varied the framing (in terms of diversity of IEs) of an issue within the coverage.
The second observation is that the relationship between the number of IEs in the political statements and the number of IEs in the coverage directly affects the prominence of governmental framing: the lower the number of IEs in the press coverage, the more prominent the political framing within it. This is supported, in the previous comparison between the American and French case by the fact that the whole coverage in the US, where the official framing is more prominent, presents 272 IEs, against 311 in France. The trend is even more strongly confirmed by the Italian and Pakistani cases (respectively presenting 288 and 362 IEs in their coverage). Both REP's and CDS's first page news reproduce 38.46% of the political statements, which constitutes a fraction of overall media framing: 27.78% of REP first page news, 30% of CDS coverage ( gives voice to an apparently smaller range of sources (12 rather than 25 as in REP) and involving US, Italian, Pakistani, Afghani, and Italian media, but the contribution by Italian academics, almost every single day of coverage, outnumbers the framing any other source and introduces a wealth of alternative interpretations of the 9/11 events.  Tables 5, page 13 and 8, below). The relatively high reproduction of the political statements within the first page news of both newspapers is explained by the growing media interest in Musharraf's statements in the context of the developing international situation. The reason for the low proportion such statements constitute within of the overall press media coverage, instead, is the framing competition by voices other than the national government, as confirmed by the analysis of the sources contributing to the framing. The DAWN's sources include Pakistani religious leaders, American officials, Afghani sources, the EU, international organizations, CNN, and international newsagencies. The NAT presents similar sources, with the exception of international organizations.

Table 8. Comparisons between political statements, first page news, and editorials in Pakistan
This last points leads to a third observation: the number and origin of sources quoted varies depending on the newspaper. The differences are related to the country, but also to the agenda of each media organization. The analysis shows that some newspapers, such the WSJ, LIB, CDS, and the NAT appear more focused on domestic issues and rely on a narrower range of sources that their more internationally-oriented counterparts-respectively the NYT, LM, REP and the DAWN. For example, while both WSJ and NYT tend to reproduce more domestic sources than Italian, French or Pakistani newspapers, the only non-American source out of a range of thirteen (representing 5% of all sources) in the WSJ is represented by the Afghani government. The NYT also gives voice to the Afghani government, the Pakistani government, the Northern Alliance, British PM Blair and Russian President Putin.

Discussion
The analysis of the framing of 9/11 in eight elite newspapers across four different countries suggests that coverage is doubly constructed. It is constructed by media professionals, who physically assemble it by gathering information. It is also constructed in meaning by sources. They might well include editors or journalists themselves, "speaking" within the news text and competing among each other to communicate to the public their interpretation of events. The aspects which were initially thought to affect the levels of correlation between governmental discourse and elite press coverage-national level of involvement with the 9/11 events, governmental ideological orientation and a country's form of government-did not produce significant variations across the countries and newspapers under study (for a detailed discussion see Archetti, 2007, especially Chapter 6, pp. 280-284). The framing of 9/11 in the elite newspapers under analysis appears, instead, to be effectively explained by the selection of newsworthy sources within the news. It is the range of sources, their variety of origin (foreign rather than national) and identity (politicians/intellectuals/social actors/religious leaders etc.) that determines the scope and variety of the coverage.
The variables which were found to have the greatest impact on the selection of sources were national interest, national journalistic culture and editorial policy of each news organization, which I am now going to illustrate in greater detail.
National interest, as the analysis of the governmental discourses in the four countries suggests, can be defined as the combination of national identity, existing foreign policy agendas and international relations, particularly alliances with foreign countries and membership of international organizations. National interests appears to drive the journalists' selection of national vs. international sources. The foreign sources reported in the French coverage (81.17% in LM and 85.84% in LIB), for example, including the Arab world and Africa, mirror both the historical relations with those regions and the 'diplomatic activism' in which the country was involved in the aftermath of 11 September (Davison, 2004, 76). The number and country of origin of French sources sharply contrasts with the limited number of sources, mainly American, in the coverage of the NYT (80.13%) and WSJ (95%). This appears to reflect the more unilateral approach of the American hyperpower to international relations.
It could be argued that the focus on domestic sources in American elite newspapers is dictated by the fact that 9/11 occurred on American soil. While this line of argument should certainly be acknowledged the comparison with the sources in French, Italian, and Pakistani coverage still offers strong support to the role of national interest in shaping journalists' selection of sources. Although France and Italy could be regarded as positioned virtually at the same distance from New York, their range of sources are markedly different. The French sources reflect the French self-perception of a world power; the Italian sources mirror the national identity as a regional player in the Mediterranean and Central European area (apart from US officials, Talibans and general Musharraf, they also included the UK and the EU press). Pakistani sources clearly show the country's interest in an international community whose reactions deeply affect Pakistan's economic situation and the country's Muslim identity. They include respectively the UN, the Afghan ambassador to Pakistan, the UN representative for Afghanistan, as well as Pakistani religious leaders, ulema (Islamic scholars), the Afghan Islamic Press [AIP] among the newsagencies. In this view the limited range of American sources is a result of the way the country has constructed, over the long-term, its role in relation to the rest of the world.
National journalistic culture shapes the ratio of official vs. non-official sources. It is the set of moral ideals, as well as reporting and editing practices, which characterize journalists in a country and that lead to their different perceptions of their own role within that country's society. It affects the way they gather news, the way they approach sources, and write their stories. Within the study the main distinction between journalistic cultures is between objective and interpretative journalism. The former aspires to reporting "only the facts" and requires journalists to be detached and to abstain from expressing their own opinions. As Adolf Simon Ochs, founder of the NYT put it in 1896, it means 'to give the news impartially, without fear or favour, regardless of any party, sect or interest involved' (Berger, 1951, 107-8). This kind of journalism is mostly practiced in the Anglo-Saxon world (Chalaby, 1996) but within the study was found to represent a model for the Pakistani elite press, too.
Interpretative journalism, instead, is precisely about producing opinion. The founder of Le Monde Beuve-Méry, for example, wanted the newspaper to have opinions that were 'independent and unpredictable, opinions that would make a difference in the public life of France because readers would read, value, and respect them, even if they did not agree with them' (Thogmartin, 1998, 219). Ermini (2003, 27-28, his emphasis) an Italian journalist writing for Corriere della Sera, similarly says that a newspaper should not only 'inform, but also give its own views so that the reader can build his/her own opinion, even a different one from that of the newspaper he/she is reading'. In this perspective absolute objectivity does not exist ('everybody knows that'), but this does not mean that the author of an article should just express his/her own opinions (Ibid, 27). As also Zincone (2003, 76) confirms, as a columnist you know that 'readers want to be told what they already know'. However, unless one is writing for a 'party-newspaper', for people who just see the world in a certain way, a journalist's function should be similar to that of a 'boutique', a 'luxury supermarket' providing customers (readers) with a wide range of products (opinions) (Ibid,75). This different understanding of the journalist function affects the selection of sources in the coverage. 31.9% of the sources in the coverage of REP and 46.33% in the reporting of CDS are academics, writers, columnists, journalists or members of the public rather than domestic political actors (10.43% for REP and 20.9% in CDS). Political actors totally disappear from the most prominent part of the coverage when a provocative article by writer and journalist Oriana Fallaci (2001) on 29 September starts a debate that will rage on the pages of Corriere for the following three weeks and will spread to Repubblica, too (Eco, 2001).
Editorial policy of each newspaper has a two-fold effect. Firstly it shapes the news agenda, i.e. the very amount of space devoted to a foreign policy issue. While all editors in France attribute to opinion greater newsworthiness than in the US, where just fact becomes news, there are further differences at the level of each newspaper. Some newspapers do not deal with certain issues but very marginally because they are not newsworthy to the organization's specific agenda. The study finds, for example, that Libération does not cover the issue of 9/11 in as much detail as Le Monde. This difference in priorities concerning the coverage reflects the fact that Libération was established in 1973 as the voice of 'all the leftist political ferment' of French society and was therefore concerned with domestic issues (Thogmartin, 1998, 248). The very title of Le Monde, "the world", reveals instead a strong international vocation. Having been allocated more space in the news agenda by the editor means for journalists both the possibility of reporting a greater number of sources and reporting more of what they say (their framing of an issue), with the effect that the number and variety of ideas within the coverage will increase. Secondly editorial policy affects the level of objectivity/commentary against the national "norm". Within the same American objective journalism the Wall Street Journal shows a higher presence of evaluative statements (interpretation) in the news than The New York Times. This reflects the editorial philosophy of the business news organization, which openly states: 'We often take sides on the major issues of politics and society, with a goal of moving policies or events in what we think is the best direction for the country and the world' (DowJones.com, n.d.).
More commentary by journalists, if they disagree with the official framing, can offer contrasting ideas with the official interpretation of a foreign policy event within the coverage.
The analysis therefore suggests that 1) the "media" should be understood as organizations of journalists and editors operating according to their own norms and those of the broader national social and political context; 2) understanding the relationship between political actors and the media involves the consideration of a wider set of relationships than the pure authorities-journalists interaction; 3) at the same time explaining the media-foreign policy nexus means also taking into account the effect of variables at a higher level than the national dimension.

Conclusions
The analysis of the 9/11 foreign policy crisis and its aftermath suggests that the indexing hypothesis does not apply outside the US. At a closer qualitative scrutiny, depending on methodological choices, it might not apply to the American statepress relations either.
The point is not that a correlation between media coverage and political debate does not exist. The study of the international elite press framing of 9/11 presented here does detect some correlation between coverage and governmental discourse, higher in the US than elsewhere. The fact that the media reports governmental authorities' interpretations of international crises, however, is not at all surprising.
When it comes to issues of national security and terrorism, as in the case of 9/11, it is reasonable to assume that political actors are newsworthy sources in virtually every country. What is interesting, instead, is the variation across countries and newspapers, which suggests that domestic power relations between authorities and journalists are not enough to explain such patterns of correlations.
The reliance on national political actors by American newspapers, however, is made more extreme in the US firstly by the country's relatively lower interest in foreign countries than in France, Italy or Pakistan. Secondly by the country's objective journalistic culture: paraphrasing a well known quote by Sigal (1973, 25), news stories are mostly reliant on what a small range of newsworthy sources, mainly national politicians, say happens. Commentators are not allowed to introduce a variety of views within the news discourse. All these elements contribute to keeping the scope of the US elite press discourse narrow and generally reliant on political elite debate. The relative impact of political actors is lower where the scope of the media discourse is wider-as in Italy, France, or Pakistan-because the framing by political actors in these countries faces competition by alternative sources.
There are therefore additional and broader variables than domestic dynamics to consider whose effects, when one analyzes the coverage from inside one's country, just cannot be seen. A country's involvement with the rest of the world (national interest), the self-perception of the role of a journalist in a society-shaping the way reporters relate to sources and write their stories-(national journalistic culture) do shape the way journalists report about foreign policy, but cannot be detected but in an international comparative dimension. Even if the role played by editorial policies of each individual media organization can be detected within a national context, the full extent of their effects only becomes visible when journalistic practices in different parts of the world are compared.
Current conceptualizations of the media-foreign policy relation are not able to account for differences at the cross-national level. This is largely the result of most of Political Communications research having been conducted at a national level (Gurevitch and Blumler, 2004). The indexing hypothesis, however, cannot apply to foreign cases not only because it is a description of the interaction between political elites and the media at a domestic level. It cannot apply because it comes with a normative baggage, which is rooted into the Anglo Saxon journalistic model. This, in turn, implies a specific idea about the way democracy should work and the role of the media within its framework. The notion is that 'a strong, adversarial press must be ready to raise its own and other grass-roots voices against government officials who would exclude these voices from deliberations about the national interest' (Bennett, 1990, 104). It is this ideal which actually leads researchers in drawing conclusions from the data about the correlation between coverage and political debate. A high correlation (although it is not clear how high) means that the press is not doing its job. The level of correlation is used as a litmus test of the health of a democracy. The fact is that journalists' role is different in other parts of the world. In France, for example, journalists do not appear to think that their duty is to give a voice to the people. They see themselves as 'high literary creators and cosmopolitan thinkers' (Schudson, 2001, 166). What they tend to express in their articles, to a much greater extent than in Anglo-Saxon countries, is their own voice. In developing countries journalists might want to reproduce official views. This does not mean they are not fulfilling their democratic duties as this is precisely part of their perceived role as supporters of fragile institutions (Nasser 1983, 48). The relationship between political actors and the media, and with it the media-foreign policy nexus, needs to be evaluated according to cultural, political, and social country-specific criteria.
Besides, the indexing hypothesis might have more accurately reflected American state-press relations in the past than in an era in which information travels instantaneously across global media networks, bypassing national boundaries and official gatekeepers. The development of communication technologies leads on the one hand to changes in the policy processes. Growing interconnections on a global scale are the root not only of a blurring between the domestic and international politics. The increasing use of the term "intermestics" (see Brenner at al., 2002, 204, for an example) suggests that, to an extent, there is no longer a distinction between the two dimensions. On the other hand the way in which journalists work is also changing. More specifically the very fact that each reporter has access to a whole virtual network of information sources across different countries allows reporters to compare different, potentially contradicting, pieces of information and develop a critical attitude towards official briefings (Archetti and Brown, 2008).
This article, to conclude, argues that some of the current paradigms in political communications are not only getting old. When they are applied to a different social, political, and cultural context, they are also inadequate. The media-foreign policy relation can be truly explained only in a multidisciplinary perspective. The 9/11 study suggests that a fruitful avenue is the combination between Political Communications, International Communications, and News Sociology. This needs further development but must not merely involve adding up the variables from each field. It should rather be based on reworking current paradigms completely, by starting from the elaboration of an appropriate ontological perspective, which brings together the micro-interactions of journalists within news organizations and between journalists and officials (which the indexing hypothesis in its original formulation was attempting to establish), with long-term macro-processes of international politics and the global changes in information production and distribution.

Notes
This approach is different from Ferree et al. (2002), from which the study borrows the label of "idea elements." Here the purpose is gathering IEs as a dynamic flow (the "framing process") developing over time. Ferree et al. (2002), instead, group the IEs into eight "frames" derived from the legal, political and social movements' context of the countries (US and Germany) involved in their study. 2 A codebook was specifically designed for guiding the detailed content analysis of the news coverage. The codebook provided a "9/11 datalist" and an "Afghanistan datalist". These are lists of ideas (IEs) actually identified within the news discourse during a pilot study preceding the creation of the codebook. The Access database used for the analysis also allowed adding new IEs to the list while the analysis was underway. The "9/11 datalist" eventually contained 599 IEs, the "Afghanistan datalist" 286. The coding of political statements by four different governments and media coverage of 8 different newspapers over sixty-four days generated 1058 entries into the database. They involved a total of 12,922 IEs recorded in relation to the issue of 9/11 and 5602 IEs in relation to the intervention in Afghanistan. The codebook, as well as coding examples of both political statements and news texts can be obtained from the author. 3 The selection process starts by running a query on the Access database where all IEs are stored to gather all relevant IEs at the required level (for example either US political statements, or US media coverage, or New York Times's first page news coverage) including dates. The data is then exported to SPSS and entered into a table. A frequency analysis is run on the IEs by selecting from the Tool menu the option "Analyze" "Descriptive Statistics" "Frequencies". The analysis format is set so that the frequency is going to be ordered by descending count. The Output Sheet presents a list of IEs with relative frequencies. The list also contains the cumulative percent in relation to the whole coverage. Several trials have supported the decision to select 50% as proportion of the coverage to be taken as the framing sample. 50% is a balanced option in so far as it represents both a substantial selection of the overall coverage, as well as a large enough sample to be representative of the contents of the overall framing process.