Chapter 1: Introduction

• 
  Complete participant: The researcher is fully immersed in the group, perhaps through covert research or taking on full time job in the area of study.
  

In this study the researcher identified the role of observer as participant as the most effective method. As a former journalist in local British newspapers the researcher was already familiar with the sub-culture under study and it was therefore not necessary to become fully or partially embedded in the group, as there was already a level of understanding and knowledge. However it was necessary to understand the case studies in depth, with the researcher having no prior knowledge of these two particular newspapers and this could only be achieved through some level of interaction. It was also felt that the subjects would be put at ease and act more naturally and openly if the researcher was not a distant clinical complete observer, but engaged in conversation with subjects to convey a willingness to learn and understand the subject’s perspectives. It must be noted however that on one occasion the researcher transitioned from the role of observer as participant to the next stage of participant as observer. This was during a breaking news story when the researcher took video footage of a live news event due to a lack of resources and equipment within the Leicester Mercury office. This footage was then used on the newspaper website and became part of the reporting which the researcher was observing. This is discussed in more detail in Chapter 5.

The researcher spent three weeks observing at the Leicester Mercury (see Photo 4.1) during October 2010 and two weeks observing at the Bournemouth Daily Echo (see Photo 4.2) – one week during November 2010 and one week during January 2011. The first week of observation at both case studies was observation alone but the subsequent weeks were intersected with conducting journalist interviews. During this time the researcher observed different factions of the news room at each case study which were initially sampled purposively and then took a snowball structure, similar to the interview selection process described in section 4.3. At both case studies the researcher observed the news desk, web desk, reporters and attended daily conferences and editorial planning meetings. All areas were observed at varying times of day including early and late shifts, from 7am through to 10pm, Monday to Saturday. The majority of observation centred around the news desk and webs desk due to their strategic position in co-ordinating the bulk of journalists and online content.

Photo 4.1: Leicester Mercury news room

Photo 4.2: Bournemouth Daily Echo news room

During the observation periods the researcher recorded a variety of factual, subjective and reflective information as it happened on an electronic netbook. Less-structured observation has no fixed design and researchers are encouraged to record any data that seems “relevant or interesting” depending on the “opportunities that arise” (Sapsford and Jupp, 1996, p.81). The researcher therefore devised an observation guide and observation theme list during their initial week at the Leicester Mercury to record verbal and non-verbal information as illustrated in Appendix 3. The guide recorded the location, time, subject being observed, actions, conversations and comments, general environment (physical and social), subject feedback and reflection on researcher influence and observer thoughts. Validity was built into the observation process through the measures of putting subjects at ease, as discussed above, to diminish reactivity. Secondly the observations were recorded as they happened so there was no delay and no reliance on memory. Observer misinterpretation was avoided by the observer-as-participate approach which allowed the researcher to clarify details with subjects and discuss observations with them allowing for respondent validation. Thirdly the researcher continually reflected on their observations and their influence upon them, making a note in real time. The fourth validation was triangulation of data with interviews. The study therefore had the build in validity checks of triangulation, reflexivity and respondent validation recommended by Sapsford and Jupp (1996).

The reliability of the observation was tested through a pilot study which was carried out at the Leicester Mercury over a two day period in October 2010, prior to the official study beginning. This enabled the researcher to develop an appropriate, effective observation guide and consider what was valuable information to record and eliminate the recording of information not relevant to the research questions. It was also helpful in deciding the best way of recording observations as both handwritten notes in a notebook in longhand and shorthand were trialled and notes typed direct to a netbook. The researcher concluded that a netbook was the quickest and most accurate way of recording observations.

4.5 Online content analysis

Content analysis is a systematic, objective and quantitative (McMillan, 2000) method often used to analyse complex static data. In the field of journalism studies it is most traditionally used to study the content of newspaper pages and/or television news programmes (Lewis et al, 2008; Semetko and Valkenburg, 2006; White, 1950) but in recent years it also has emerged as an effective tool to research the content of web pages and websites, including user comments (Anstead and O’Loughlin, 2010; Trice, 2010), personalisation (Thurman, 2011) and user generated content (Jonsson and Örnebring, 2010). It has also been successfully used in triangulation studies combined with interviewing journalists (Hermida and Thurman, 2007) as this study will replicate. It has therefore been identified as an appropriate method to address RQ2a: What is the nature of Web 2.0 audience participation in British local newspapers? This method will also explore the interactivity of participation and how frequently readers are interacting with one another, and how often journalists interact with readers. This will help to inform the first and final research questions: RQ1a: How does Web 2.0 change the nature of audience participation in British local newspapers? RQ3: How is Web 2.0 impacting on the role of journalists in local British newspapers as traditional gatekeepers?

The content analysis takes three forms. An analysis of reader comments on newspaper website stories, an analysis of journalists' use of Twitter and an analysis of journalists' use of Facebook. These are described as the units of analysis.

The advantage of the content analysis data is that it is unobtrusive, accepts unstructured material, is context sensitive and can cope with a large volume of data (Krippendorff, 2004). Holsti (1996) explains the primary purpose of content analysis as describing the characteristics of communication, to make inferences as to the antecedents of communication, and to make inferences as to the effects of communication. Through RQ2a this study aims to describe online participation and secondly it aims to make inferences to the effects of Web 2.0 on journalists' communication with readers. McMillan further advocates the use of content analysis in online research explaining that “both descriptive and inferential research focused on web-based content could add value to our understanding of this evolving communication environment,” (2000, p.81). Even a decade later this is a relatively understudied field in journalism studies and Trice (2010) acknowledges that “little work has occurred that examines how people use these comment fields on new sites,” (p.3).

In order to create a valid content analysis, this study followed the recommendations of McMillan (2000) by following five steps: draw up research questions in the context of theory (see Chapter 1 and 2), select a sample (case study selection discussed in Chapter 3), define categories for coding units and context units, train coders, collect data and analyse. Selecting a sample was a simple process as the two case studies were already pre-existing samples and it was natural for the content analysis to isolate itself to these two newspapers. However drawing up the boundaries of the context units in this study was particularly challenging as it seeks to explore participation not only on each of the case study websites but also through their use of external social media websites such as Twitter and Facebook. McMillan identifies that “as analysis of the web matures entirely new context units may need to be developed to address phenomenon that are completely nonexistent in traditional media” (2000, p.93). This is discussed in further detail below.

To meet the requirements of step four, training the coder, the researcher undertook a Research Training Programme at The University of Sheffield in research methods. During this module the researcher completed content analysis exercises and gained a greater understanding of the methodology. Prior to implementing the content analysis at the two case studies, the researcher carried out a pilot on the website and social media network of the Northampton Chronicle & Echo, a newspaper and website the researcher was familiar with. Once coding and context units were designed the researcher tested the code on the two actual case studies and brought in a second researcher for cross coding checks and development. The resulted in several changes to the categories for all three units of analysis (comments, Twitter, Facebook). The categories for each set of coding units went through at least four versions before they were finalised. A reliability test was then carried out by a second researcher on a 10 per cent sample of each of the units of analysis which resulted in a 70 per cent reliability score. With increased knowledge about who the journalists were communicating with, the reliability increased to 90 per cent.

Due to the emerging prevalence of reader comments on news stories online (Robinson, 2010; Domingo et al, 2008, Hermida and Thurman, 2007) this study deemed it necessarily to gain a greater understanding into the nature of this type of participation. Trice (2010) maintains that comment fields are ubiquitous in social media networks and “these web applications are then mimicked by established media sites to capitalize on the phenomenon known as the social web” (p.3). This research therefore explores both newspapers’ use of social media comments (see section 4.5.2 and 4.5.3) and the more traditional approach of comments on the newspapers’ associated websites. In particular the researcher wanted to explore the richness of reader comments and how they interact with one another. Building on the work of Trice (2010) a content analysis was devised to understand how people engage with news stories, journalists and other readers via comments.

The following definitions were used:

• 
  Comment: A post by an individual with a username. This is directly below a specific article on the website.
  

• 
  Thread: A series of comments by one or more posters which appear below one specific article. This can appear in chronological or reverse-chronological order depending on the website.
  

• 
  Poster: Someone who leaves a comment beneath a specific article on a website.
  

Figure 4.3: bournemouthecho.co.uk home page with Most Commented list

The context unit provided a total of 439 comments (coding units) from thisisleicestershire.co.uk and 730 comments from bournemouthecho.co.uk.

It must be noted that each individual comment was a separate coding unit and therefore if one individual poster left several comments, each which would have been counted. The content analysis did not calculate how many individual posters there were or how many comments they left on one thread.

There were five coding categories in total which reflected the richness of the comment, the interactivity level and who was interacting with who. This is important to understand the nature of participation and subsequent interaction on British local newspaper websites between different actors. The categories were developed from the four categories identified by Trice in his 2010 content analysis of comment fields on six news websites. Trice's (2010) study explored the richness of comments from simple opinions, new content to complex arguments and analysed the interactivity of comments to a more limited degree, in particular whether the comments referred directly to the story or referred to another username. This PhD study includes the added dimension of who is interacting with who, and distinguishes between reader to reader interaction and reader to journalist interaction.

The categories were defined as:

• 
  No relevance to article (Post)
  

• 
  Refers to article (Content Interaction)
  

• 
  Refers to another user (Poster Interaction)
  

• 
  Website host interaction (Newspaper Interaction)
  

• 
  New content (Advanced Content Interaction)
  

A dominant category approach was taken when it was felt a comment fell into two categories, rather than counting comments twice and placing them in two separate categories. For example if a comment included a poster interacting with another poster about a topic of no relevance to the article the stream was attached to this was counted as a Post rather than a Poster Interaction, or both. It was felt the irrelevance of the comment was more significant than the interaction, since the interaction was not related to the news story in question. This is illustrated in the following comment:

In this comment Mr LFE is interacting with Supporter Not Customer, England, but it is not in response to the article. Instead it is a remark/insult about the boring nature of the comments of Supporter Not Customer, England. By contrast in this second example below from the same thread, two posters are interacting with one another about the article in question Leicester City’s ‘Fosse Boys’ fear permanent band from the Walkers Stadium. This was therefore counted as a Poster Interaction.

Furthermore if a poster referred to the author of an article or letter published on the website this was counted as a Content Interaction rather than Poster Interaction or Newspaper Interaction. This was viewed as a poster interacting with the content provided by the author rather than engaging with another ‘live’ user.

The nature of reader participation via commenting on newspaper website articles has been addressed in this study via the content analysis described in section 4.5.1. However this is only one of the ways in which Web 2.0 is changing the way in which readers participate and interact with newspaper journalists. The use of social media networks has exploded in the past five years and due to its limited history there is little literature or empirical evidence on how this is impacting on journalists and readers. This study therefore aims to fill this gap in knowledge by exploring how Twitter and Facebook are being used by local print journalists. Due to the global interconnected nature of Twitter it is not possible to create a typology of how audiences participate in British local newspapers via this social media platform. Firstly this could only be achieved by obtaining the individual usernames of thousands of readers and secondly it would be an extremely problematical task to identify which of their tweets related to the two case studies in particular. However it is possible to explore the journalist-to-reader relationship and identify how journalists interact with their readers via Twitter, since they are much smaller in number and easily identifiable. Although this does not directly address the nature of Web 2.0 audience participation in British local newspapers, it does conversely inform our understanding of the nature of journalist interaction with their readers. It can therefore help to explain whether this interaction is impacting on the role of journalists as traditional gatekeepers as discussed in RQ3.

Due to the different approach needed to analyse Twitter a different coding system was needed which shifted the focus from reader interaction to journalist interaction. Unlike the content analysis of comments, the researcher was unable to identify similar studies as work in this field is limited and has focused on mass individual Twitter users rather than on an individual user or organisation. Kwak et al (2010) for example coded 41.7 million user profiles to understand trending topics. Meanwhile other research to date has investigated the location of users around an event (Yardi and boyd, 2010), Twitter user influence (Cha et al, 2010), uses of Twitter (Java et al, 2007) and Twitter as a conversational tool (Honeycutt and Herring, 2009). This research however seeks to discover how individual staff members within specific organisations use Twitter in their professional role.

Current Twitter research tends to use searching software which is able to capture retrospective trends or key words, such as Twitter Search. It is possible to develop bespoke search software since Twitter has an open Application Programming Interface, enabling programmers to create their own applications, widgets or websites that interact with Twitter. Meanwhile some researchers (Anstead and O’Loughlin, 2010) choose to use existing Twitter search tools such as Twitter Search where they can enter specific search terms and specific time periods to capture all data that matches those criteria. However external applications like Twitter Search were not appropriate, or indeed any use, for this study as it was not analysing trends or key words, but looking at individual users and their entire output over a period of time. Therefore the only way to capture the data was to cut and paste tweets over a given time frame into a Word document. This included all information given in each tweet including usernames, hyperlinks, retweets and hashtags. The information captured was only public content and excluded direct messages. This was viewed as appropriate because direct messages have a similar function to emails and are a private interaction, therefore were discounted in this data set. Capturing the data in this cut and paste way proved problematic because due to the large volume of data, Twitter could not go back further than an unascertained number of tweets. A two week sample from each user appeared to be the maximum data set that could be collected using this method at the first attempt. A two week sample was therefore taken at each case study coinciding with the observation period (October 2010 and January 2011). However a second attempt was made seven months later in June 2011 after a new version of Twitter was launched. This version could handle higher volumes of data and a one month sample was taken from each user. These two samples enabled the researcher to compare how the journalists' use of Twitter had developed over the interim period.

Data was captured from the most prolific Twitter users at the two case studies, defining them as the context units. The two week and one month data sets included every single tweet within those time frames. More journalists used Twitter at the Leicester Mercury therefore data was collected from four users (see Table 4.1). During the research period only two journalists at the Daily Echo consistently used Twitter so data was only collected from them. In total 2,588 individual tweets were coded, made up of 1,134 from four Leicester Mercury users and 1,454 from two Daily Echo users.

Table 4.1: Twitter context units

The coding categories (see below) expanded during the cross checking and development phase with the second researcher. It became apparent that the levels of interactivity were quite complex and it was necessarily to divide them into journalist-to-reader and journalist-to-colleague although these two categories could be collapsed into one to show overall interactivity levels. The same system was carried over to the sharing categories. This coding system meant it was easier to identify how much of the social media interaction or sharing was with external actors such as readers, and how much was internal within a news room. However it should be noted that external actors could include journalists at other news organisations and indeed people working in public relations.

The categories were defined as:

• 
  Traditional
  

Promotional link to own website competition

• 
  Informal
  

Informal news (including live updates)

Comment (inc. comment on current affairs)

Personal message to readers

• 
  Personal
  

Non-current affair related comment

Interactive/sharing with friend (no direct work relevance)

• 
  Sharing: colleagues
  

Share colleague’s content (inc. lists)

• 
  Sharing: external
  

Share external user’s content (inc. lists)

• 
  Interactive: colleagues
  

Asking a question of colleague

• 
  Interactive: external
  

Asking a question of readers

Asking readers to do something (inc. send in photos)

Setting up vote