Transcript extract: focus group interview with Y9 pupils

Analysing Qualitative Data

381

wished

36. 

you could speak French?

37. 

P4: They had like lots of (.) kind of sayings which when they’re translated into English

38. 

they don’t like mean the same thing in English. So even if you’ve like translated with 

a

39. 

dictionary, it still doesn’t mean anything to you.

40. 

P2: I would say the English. You can’t explain to them like when they want to do

41. 

something and you don’t want to do it, you can’t say it politely! Because you can’t say

42. 

‘I’m sorry, I really just don’t want to do this’. You have to say ‘No’.

43. 

P5: I had loads of arguments with mine actually.

44. 

R: Really? With your partner?

45. 

P5: Yes.

46. 

R: What sort of thing?

47. 

P5: I got told to stop being rude!

48. 

R: Rude? Who by? By him?

49. 

P5: The mum.

50. 

R: By his mum? Right! What were the arguments about?

51. 

P5: About going places, because mine didn’t want to go anywhere and it was my 

birthday

52. 

party and they wouldn’t let me go.

53. 

R: Oh, right.

54. 

P6: I had to translate for her! Hers got really angry at me for saying what she was 

saying! (Laughter)

As this extract shows, the interviewer is constantly making decisions about what to

follow up from the interviewee’s responses; and these decisions are made on the basis

of what one might call ‘online analysis’, made however fleetingly and instinctively by the

interviewer during the flow of the interview. For instance, Pupil 3’s comments (lines

23–24) that her partner’s parents spoke in English to her all the time interested me

because this could potentially reveal something about how the pupil saw the relationship

between language learning and personal communication. My first hypothesissuggesting

question (that the parents’ use of English was ‘annoying’) was rejected by

the pupil. But my second hypothesis-question (that communicating in English had a

comforting effect) proved more acceptable and seemed to have the agreement of at

least one other pupil in the group. One could therefore argue that for a brief moment

what was happening here was a form of ‘collaborative analysis’.

One might object at this point that we are here precariously treading the delicate line

between informants’ and researcher’s constructions of the meanings surrounding

events and experiences being researched. However, as noted earlier, qualitative research

paradigms to a certain extent legitimize researcher input in the process of data elicitation,

provided this input is explicitly acknowledged and critiqued. In the case of young

pupil informants, the strategy is arguably particularly needed in order to draw out their

analytical thinking.

Analysing Qualitative Data

382

DEDUCTIVE AND INDUCTIVE ORIENTATIONS IN QUALITATIVE ANALYSIS

There  are  two  competing  tensions  when  embarking  on  qualitative  analysis  which  are 

best seen as two ends of a continuum. On the one hand, qualitative researchers have an 

overarching idea of what it is they want to investigate in the data and some notion at least 

of the different areas in which this sought-after information can be grouped. This ‘deductive’ 

approach  seeks  to  generate  and  examine  findings  in  relation  to  pre-established  themes 

which may not be exhaustive or totally defined at the outset, but which nevertheless provide 

overall direction to the development of the analysis. At the other end of the spectrum, the 

inductive approach, adopted, for instance, by grounded theorists, takes an entirely open 

minded approach to the data and uses themes which emerge from the data themselves 

as tools for the analysis. In reality, educational researchers dealing with qualitative material 

usually employ a mixture of the two approaches, but, as Huberman and Miles point out, the 

different emphases appeal to different types of studies:

There is merit in both ‘loose’, inductively oriented designs, and ‘tight’, more deductively 

approached ones. The former work well when the terrain is unfamiliar and/or excessively 

complex, a single case is involved, and the intent is exploratory and descriptive. Tighter 

designs are indicated when the researcher has good prior acquaintance with the setting, has 

a good bank of applicable, well delineated concepts, and takes a more explanatory and/or 

confirmatory stance involving multiple, comparable cases. (1998: 185)

In either of the two orientations, the systematic character of the analysis is largely dependent 

on the use of a coding system to organize and structure the examination of the material. 

Coding is the process by which a text is examined thematically according to certain categories 

(codes) which are either predetermined or emergent from the data. The categories serve 

the purpose of reducing the total mass of data elicited in order to focus on what they tell us 

about the particular themes we are interested in. The process also allows the researcher to 

identify in the data evidence links to the different categories, to closely organize and inspect 

that evidence, and to use the process to inform and develop the themes indexed by the 

categories.

Coding framework

One of the first steps in this process is the development of a coding framework which 

can be applied systematically to the analysis of qualitative data. The framework should in 

the first instance remain fluid and develop on the basis of preliminary readings of some of 

the transcripts or at least of the pilot data. The following example is of a coding framework 

which I drew up in a study of pupils’ code-switching between English and French in an 

online bulletin board project between pupils in England and in francophone countries. The 

framework was applied to the analysis of interviews I carried out with some of the English 

pupils involved in the project (Evans, 2009).

The codes, listed in the left-hand column (Table 10.1), refer to topics which I wanted

to identify and for which I wanted to collate evidence from the pupil interviews. The

right-hand column provides a brief description of what each code represents. This

Analysing Qualitative Data

383

description is useful for supporting the validity claims of your findings and therefore it

is advisable to include it in your methodology chapters in your dissertation or report.

The descriptions are also useful to you as a researcher in that they help you to crystallize

in your mind exactly what it is that you are targeting with each code. In the above example,

I grouped the codes according to different dimensions of language use identified in

the sociolinguistics literature. The types of concepts referred to by my codes in this

framework are fairly descriptive or factual, in that they sought to locate references to

these topics either directly or indirectly in the pupils’ interview responses, with minimal

levels of researcher interpretation at this stage.

SCHOOL-BASED RESEARCH

164

The codes, listed in the left-hand column (Table 10.1), refer to topics which I wanted 

to identify and for which I wanted to collate evidence from the pupil interviews. The 

right-hand column provides a brief description of what each code represents. This 

description is useful for supporting the validity claims of your findings and therefore it 

is advisable to include it in your methodology chapters in your dissertation or report. 

The descriptions are also useful to you as a researcher in that they help you to crystallize 

in your mind exactly what it is that you are targeting with each code. In the above exam-

ple, I grouped the codes according to different dimensions of language use identified in 

the sociolinguistics literature. The types of concepts referred to by my codes in this 

framework are fairly descriptive or factual, in that they sought to locate references to 

these topics either directly or indirectly in the pupils’ interview responses, with minimal 

levels of researcher interpretation at this stage.

Table 10.1   A coding framework for the analysis of interview data

Ideational Dimension 

Code choice

Reasons for using French or English in posts

Codeswitch

Explanation of code switching and code mixing

Content

Choice of topic in post

Word choice

Meaning-related explanations for the choice of a particular word

Opinion

Opinion-related influence on the content of post

Interpersonal Dimension

Borrow

Copying text from native speaker posts

Interact

Perceptions of interactions with other members of the group

Read

Comments on the experience of reading other people’s posts

Tu/vous

Rationalization of choice between the two forms of address in posts

Metalinguistic Dimension

Know

Reference to what they know

Learn

Reference to learning goals, outcomes and experience

Schoollang

Reference to the nature of their school language learning

SayWrite

Describing the online communicative process 

Different types of coding

The second point to bear in mind is that researchers have identified the need for differ-

ent kinds of coding and decisions. Although different writers define several distinct 

types of categories in the coding of qualitative data, it is best to focus initially on two 

main types: open coding and thematic coding. I shall from now on, in line with the literature, 

refer to categories as ‘codes’, defined as the names or labels that refer to concepts 

(synonymous with ‘theme’ in common parlance). The denotation of ‘concept’ is not 

trivial, however, since even at its most basic, descriptive level, coding is a process of 

translating raw data into conceptual references. Let us look at the example provided by 

Strauss and Corbin (1998: 106–9) illustrating their coding of part of an interview with a 

young adult talking about teenage drug use. The following extract is taken from the 

beginning of the transcript presented by the authors, and the labels in square brackets 

are the codes the researchers applied to the text:

11-Wilson-Ch-10.indd   164

8/31/2012   5:41:25 PM

Table 8.1 

A coding framework for the analysis of interview data

Different types of coding

The second point to bear in mind is that researchers have identified the need for different

kinds of coding and decisions. Although different writers define several distinct

types of categories in the coding of qualitative data, it is best to focus initially on two

main types: open coding and thematic coding. I shall from now on, in line with the litera-

ture, refer to categories as ‘codes’, defined as the names or labels that refer to concepts

(synonymous with ‘theme’ in common parlance). The denotation of ‘concept’ is not

trivial, however, since even at its most basic, descriptive level, coding is a process of

translating raw data into conceptual references. Let us look at the example provided by

Strauss and Corbin (1998: 106–9) illustrating their coding of part of an interview with a

young adult talking about teenage drug use. The following extract is taken from the

beginning of the transcript presented by the authors, and the labels in square brackets

are the codes the researchers applied to the text:

Interviewer: Tell me about teens and drug use.

Respondent: I think teens use drugs as a release from their parents [“rebellious act”]. Well, 

I don’t know. I can only talk for myself. For me, it was an experience [experience] [in-vivo 

code]. You hear a lot about drugs [“drug talk”]. You hear they are bad for you [“negative 

connotation” to the “drug talk”]. There is a lot of them around [“available supply”]. You just 

Analysing Qualitative Data

384

get into them because they’re accessible [“easy access”] and because it’s kind of a new 

thing [“novel experience”]. It’s cool! You know, it’s something that is bad for you, taboo, a 

“no” [“negative connotation”]. Everyone is against it [“adult negative stance”]. If you are 

a teenager, the first thing you are going to do is try them [“challenge the adult negative 

stance”].

One can see that the codes attached by the researchers to this passage are doing more 

than just labelling: they are lifting the specific points made by the interviewee to a more 

generalized, conceptual plane. The code ‘rebellious act’ is an idea which has certain ‘prop-

erties’ and ‘dimensions’ which differentiate it from other related concepts, such as ‘mindless 

act’ or ‘self-destructive act’, and is applicable in other contexts and in relation to other 

people. The ‘property’ in this case might be that the act is directed against authority fig-

ures, and the ‘dimensions’ of this property might be ‘parents’, ‘teachers’, ‘society in general’. 

Even at this initial stage of the analysis where the researchers are essentially attempting to 

identify and itemize the content of the data, the process involves a degree of conceptual 

interpretation. The above is an example of ‘open coding’ which Strauss and Corbin define 

as ‘the analytic process through which concepts are identified and their properties and 

dimensions are discovered in data’ (1998: 101).

The act of open coding has the effect of fracturing the transcript into different fragments 

which are labelled according to extensive lists of codes. Sometimes the same fragment can 

have several different codes attached to it. This process therefore results in a reduction of 

the material, which in ethnographic studies especially can be dauntingly voluminous. These 

fragments can then be examined in groups, thus enabling the researcher to focus on the 

concepts and the related evidence from the data. The danger of such an analytical strategy 

is that a textual fragment (and therefore the evidence) becomes detached from its original 

context and might lead ultimately to a distorted or inaccurate reading. To guard against 

this, the analyst must attempt to preserve as far as possible a balance between the aims of 

categorization and contextualization.

Another feature of the process of open coding, which is illustrated by the Strauss and 

Corbin quote above, is the use of ‘in-vivo codes’. These are words or phrases which are 

borrowed from the data and used as an open code. In this example, ‘experience’ is an 

in-vivo code since its use as a code is prompted by the interviewee’s use of the word as 

part of their own explanation of their drug consumption. In-vivo codes are particularly 

useful in inductive analysis orientations as they help to ensure that analysis and resulting in-

terpretations remain close to the original material and reduce the risk of extraneous ideas 

influencing the interpretation of the data.

Thematic (also labelled ‘pattern coding’ by Miles and Huberman (1994: 57–8)) coding is a 

form of analytical coding which involves the search for thematic patterns in the coded data, 

often at a higher level of abstraction than open codes. Grounded theorists

like Strauss and Corbin, for whom the ultimate purpose of qualitative analysis is the

construction of theory, refer to the process as ‘axial coding’. As Robson has noted (2002:

494), the process is ‘about linking together the categories developed through the process

of open coding’. As such, it can be described as a process of analysis which is at a

level which is further removed from the textual origination of the data. The focus now

Analysing Qualitative Data

385

is on the codes themselves and their theoretical connotations. In this way, a broader

theoretical argument surrounding the emergent themes begins to develop. For

instance, in the example of the study referred to above, once I had gathered the relevant

quotations from my data in relation to ‘code choice’, ‘code switching’, and ‘word

choice’, I was able to examine them in relation to the three concepts and to compare,

for instance, the pupils’ explanations of their decisions in relation to the three phenomena.

This allowed me to see whether different types of considerations were at play in the

pupils’ minds and therefore to develop an argument about the pupils’ communicative

priorities and the constraints on these.

The practice of coding: by hand or by computer

The first step in open coding is to go through the transcript and, line by line, label those

bits of it that correspond to the codes which you have previously identified or new ones

which emerge as you proceed. If you are doing this by hand, then the codes should be

written in the margin with some form of highlighting of the selected text. Working on a

word processor, the codes can be entered in brackets within the text, usually immediately

after the relevant segment.

One of my former Masters students gives the following vivid and honest account of how, 

having started off by entering her data onto a computer software analysis program

(namely, NVivo) and begun the process of coding the data, she eventually abandoned

the medium in favour of the less technologically sophisticated but time-honoured tradition

of pen and paper:

Confident that the programme would help me formulate an assertion about gender and 

motivation, I started to form trees or families of nodes with different attribute values. It did 

not take me long to realize that my data was being transformed into a computer version 

that had very little in common with the original group interviews. So I stopped the pro-

cess, printed all the units of analysis from the open coding and displayed them on a wall.

The benefit of doing the analysis manually is, as this student found, that you maintain

a sense of overall control and viewing of the data globally. For some researchers, the

physical ability to cut up the different quotations and to sort them according to the differ-

ent code headings and to view the groupings simultaneously can provide a valuable

perspective to stimulate emergent interpretations and findings. This retention of an

overall view of the data is part of what helps the researcher to hold on to the original ver-

sion of the events rather than, as the student above observed, transforming it into a com-

puter version. On the other hand, there are drawbacks to the manual approach. Firstly, it is 

only feasible to do this with relatively small amounts of data. For studies involving several 

interviews with several informants or large amounts of transcripts of classroom data, the 

scissors-and-glue approach could quickly break down. Even with a small quantity of data, 

the ability to manage the data manually is very restricted: for instance, you will often want 

to mark the same piece of text with two or more codes. This is not easily done on paper.

Programs such as NVivo or Atlas.ti are quite costly and it may not be economic for a 

Analysing Qualitative Data

386

teacher to invest in one for personal use. However, if you are enrolled at a local university 

for your study, then it would normally be possible for you to make use of that facility.

It is important to recognize from the outset that computer-aided qualitative data analy-

sis software (or CAQDAS) does not analyse the data for you: the programs are simply 

support platforms which facilitate the storage and management of the qualitative data and 

phases of analysis. You still have to do the analysis and interpretation yourself. So what are 

the main ways in which a program such as NVivo can be of use to a hard-pressed practi-

tioner researcher?

жүктеу/скачать 32,4 Mb.

Достарыңызбен бөлісу: