CONTENT ANALYSIS

CONTENT ANALYSIS

1. WHAT IS CONTENT ANALYSIS?

   • This is essentially what content analysis does – it applies significance or meaning to information you have collected and helps to identify patterns in the text. Content analysis is an extremely broad area of research – its coverage includes both quantitative and qualitative approaches to analysis. As a research technique, content analysis has been used in a variety of ways and within a number of contexts. It has been successfully used to analyse text and solve issues of disputed authorship of academic papers: the techniques used have included an examination of prior writings and a frequency count of nouns or commonly occurring words to help determine the probability of authorship.
   • In the early 1990s content analysis was used to establish the identity of the ‘anonymous’ author of the fictional text Primary Colors (Foster 1996), and it has also been used, controversially, to help determine how many people actually wrote the plays attributed to Shakespeare (Mostyn 1985). Content analysis can be used as a powerful research tool to determine, from the content of a message, sound inferences concerning the attitude of the speaker or writer. It has been usefully employed as a descriptor of diverse research techniques used for systematically collecting, analysing and making inferences from messages (North et al. 1963). More recently, Krippendorff has stated content analysis to be simply ‘a research technique for making replicable and valid inferences from data to their context’ (1980: 21). Other examples of texts suitable for content analysis include essays, journal articles, books and chapters in books, discussions, newspaper articles or stories, speeches, conversations
   • In essence content analysis is based on the assumption that an analysis of language in use can reveal meanings, priorities and understandings, and ways of organising and seeing the world. Content analysis, which has been used as a research instrument for many years, was initially a time-consuming process. While we can now analyse varying types of datum using computer software, in the early days of content analysis much of the work was done with little automation. This often restricted analyses to simple frequency counts of identified words and terms. The introduction of evermore sophisticated processing equipment (in the form of mainframe computers and the advancing development of personal computers) has enabled content analysis to move beyond word counts to consider such issues as the relationship between words and phrases. It is now becoming a powerful agent in the development of artificial intelligence – being used to draw conclusions and make inferences from messages.

2. BASIC PRINCIPLES

   • Whilst the tools used for analysis may have changed and developed, the essential principles of content analysis have remained constant. To conduct a content analysis, the data (the text of an interview, speech or focus-group discussion) are coded or grouped into categories which are tested for their reliability and validity (whether or not they accurately represent what is being said, in a transcript for example). These categories or codes will include words or themes, word senses, phrases or whole sentences. Once coded, the textual data are interpreted and the results of the analysis provided.
   • CONTENT ANALYSIS: Identify the topic of your research, establish content categories, test the categories generated, collect data and analyse content of data and provide results. Either of two methods is usually used: conceptual analysis and relational analysis.

3. CONCEPTUAL ANALYSIS

   • Of the two main approaches to content analysis, conceptual analysis is by far the more popular. This approach examines either the incidence or the frequency of concepts (themes/issues, words, phrases, etc.) in a text. It quantifies the occurrences of the concepts you have chosen for examination. Conceptual analysis is also referred to as thematic analysis because it is the themes or issues in the text that the researcher intends to analyse.
     • DECIDING ON THE UNIT(S)/THE LEVEL OF ANALYSIS
       • Before textual data are analysed, a decision has to be made concerning the unit(s)/the level of analysis.
     • IDENTIFYING THE CONCEPTS
       • Concepts are the issues, themes or concerns you will search for in your analysis. At the concepts stage you must develop a list of codes or categories to be used in your subsequent coding. You must decide how many concepts you will code for, which ones are the most relevant, and how much flexibility you will allow in your coding. Indicating at the outset of your research the number of categories or codes to be included will allow you to quickly analyse text and code for very specific things.
     • DEFINING THE CONCEPTS
       • In this stage it is important to be specific about the concepts that are to be coded as the same issue. This stage is where you clearly define your codes (in writing) to allow you (and possibly other researchers) to consistently apply the right codes to the analysed text. Your concept definitions may provide for words with closely similar meanings to be included.
     • CODING FOR INCIDENCE OR FREQUENCY?
       • Once the researcher has selected a number or set of concepts for coding, a decision has to be taken on whether to code for their incidence or for their frequency of occurrence – an important decision because it has implications for the coding process itself.
   • At this level, conceptual analysis is relatively straightforward and simple. However, the themes for conceptual analysis may be implicit rather than explicitly stated. Stages in the conceptual analysis process: decide on the unit(s)/level of analysis, identify the concepts, define the concepts, decide whether to code for incidence or frequency of concepts, establish coding rules, trawl through the information, code the information and analyse the results.
     • ESTABLISHING CODING RULES
       • Clear coding rules will allow you to ‘collapse’ codes into broader coding categories. Establishing, and communicating, clear coding rules will help to prevent words being incorrectly classified. Coding rules are sometimes referred to as ‘transition rules’ – he or she provide, for example, that ‘TV’ should be classified as ‘home entertainment’ rather than ‘watching movies’, and that any subsequent occurrence of the term should be interpreted as ‘home entertainment’. Applying and consistently observing such coding rules allows for comparability and maintains the validity of your data
     • TRAWLING THROUGH THE INFORMATION
       • When you begin to trawl through or examine your data you will discover that not all of the information lends itself to coding within the codes established. You must decide what to do with this information. If it is beyond your coding frame then you may seek to amend your codes to include it. However, if the additional information adds little to your study then it may be worth considering excluding it in order to focus more on the codable content. Often words such as ‘and’, ‘the’ and ‘that’ can be disregarded as he or she have no impact upon the analysis.
     • CODING THE INFORMATION
       • Following development of your codes or categories, you can begin to code your text. In many instances, this can be done by hand when reading through each interview transcript, applying the codes to the information as you progress.
     • ANALYSING THE RESULTS
       • The last stage in any conceptual content analysis is to analyse the results of your coding. Here you would provide information as to, say, how many times the word ‘health’ appears compared to the occurrence of the word ‘education’ in a group of interview transcripts. Whilst conceptual analysis is a very useful and much-used tool, it is a quantitative exercise. Some argue that it fails to take account of true meanings behind information and data. To better understand the information or the text before us, we must consider the relationships between words and phrases and explore their emphases. This is what relational analysis attempts to do.

4. QUALITATIVE CONTENT ANALYSIS

   • In essence, the stages of the content analysis process, just outlined, focus on the number of times a word, theme or issue emerges, and meaning is applied to that numerical data. Qualitative researchers argue that because some form of quantification occurs, the real meanings of the data are overlooked in favour of frequencies (Berelson 1971). Qualitative research is concerned with capturing the richness, and describing the unique complexities, of data. Some argue that counting numbers so dilutes the quality of the information collected as to make it of little use.
   • Additionally, there are many researchers who argue that themes and issues can be determined more readily through a qualitative, more holistic, approach rather than by employing advanced statistical and analytical techniques. In work championing the usefulness of qualitative content analysis, it has been stated (e.g. by Berelson 1971) that quantitative content analysis should be used only when three factors are present: that is to say, when: you are interested in very precise results, there is the possibility of the collected data being biased and the data collected will be statistically related to numerical data.

5. RELATIONAL ANALYSIS

   • As with other content analysis models, relational analysis begins by identifying themes or issues to explore. However, unlike the majority of other analyses, this approach attempts to explore and identify relationships between the themes or issues. Using this more qualitative model of analysis allows researchers to establish significant relationships between words or phrases.
     • DECIDING ON THE QUESTION
       • Stating your question explicitly at the beginning of your analysis gives focus to your work. Carefully crafted research questions can limit the number of themes and issues, and their types, to be explored, making the whole process more manageable. A question, or theme, suitable for relational analysis might be what is being said in a number of newspaper articles about the cost of health care.
     • FRAMING THE ANALYSIS
       • When the question has been established, it is necessary to frame your analysis.
     • STAGES IN THE RELATIONAL ANALYSIS PROCESS
       • Decide on the question, frame the analysis, decide which types of relationship to examine, code and categorise the text(s), explore the relationships, code the relationships, analyse the relationships, map the relationships.
     • DECIDING WHICH TYPES OF RELATIONSHIP TO EXAMINE
       • When examining relationships between words, phrases or other units of analysis, a number of approaches are possible. These include:
         • AFFECT EXTRACTION
           • A relationship analysis based on affect extraction develops an emotional evaluation of themes and issues. While useful, this can be a difficult analysis to perform as emotional assessment of themes and issues can change with time and between researchers. Nevertheless, this analysis tool can usefully be employed to help determine the emotional/psychological state of a writer by exploring his or her language use.
         • PROXIMITY ANALYSIS
           • Proximity analysis explores text for the presence of words or themes and determines their proximity to other (pre-defined) words or themes. Using this approach, the researcher would begin by defining a string of words (such as a sentence, line or paragraph of text) within which the analysis will take place. This ‘window’ is then scanned for the presence of the pre-defined words. Meaning is inferred from the proximity of words. Often, when proximity analysis is used, matrices and clusters of relationships are established which focus on the distance between words, and then apply weightings and meaning to those distances.
         • COGNITIVE MAPPING
           • Cognitive mapping provides a further level of analysis once affect extraction or proximity analysis has been performed. This relational analysis technique allows researchers to generate a model of the overall meaning of the text rather than of specific elements within it. Often cognitive mapping techniques result in a graphical representation of the linkages between words or concepts to provide a network of the relationships that exist (Carley 1990).
     • CODING AND CATEGORISE THE TEXT
       • The decision about the type of analysis to perform will inform decisions about which words or concepts to code and categorise.
     • EXPLORING RELATIONSHIPS
       • A number of tools are available in the relational analysis model that allow researchers to explore linkages and relationships between words or concepts. The strength of a relationship can be displayed by examining words or concepts and establishing (given their location in the text, number of times used, association with the central theme or themes of the research, etc.) the clarity of the connections that may exist.
     • CODING THE RELATIONSHIPS
       • This stage of analysis provides the main difference between conceptual analysis and relational analysis. While conceptual analysis focuses on the words or concepts, in relational analysis the focus is on the relationships between words or concepts. Therefore the relationships themselves are coded under this analysis. Many codes and relationship permutations are possible – such as codes indicating weighting, codes indicating positive relationships and codes indicating negative relationships.
     • ANALYSING THE RELATIONSHIPS
       • Analysis of your coded information using a relational approach can involve extremely complex statistical techniques. For example, you can examine all your heavily weighted codes and seek to establish linkages between positive and negative words or concepts.
     • MAPPING THE RELATIONSHIPS
       • Relational analysis allows a graphical representation of relationships. This is often useful when analysing your data and presenting it to others. You can quickly establish the strengths and weaknesses of relationships, including positive and negative attributes, visually.

6. DIFFERENT APPROACHES TO CONTENT ANALYSIS

   • The first approach, using a conceptual analysis model, is essentially quantitative – categories are developed and coded, and the number of occurrences of themes or issues is recorded. This is a very popular method with researchers new to content analysis and who are accustomed to analysing data in the more traditional, numerical way. A number of researchers (Berelson 1971; Krippendorff 1980) believe an approach based on relationships is a more accurate way to interpet the content of textual material. We believe that both the conceptual and the relational approaches are valid tools for analysing what is being said in any communication, whether in newspaper articles, journals, speeches, etc.