104 Interviewing for socia/ scientists
Background knawledge
Competent interviewing and listening are closely associated with background
knowledge (see Chapter 8). Where is this sort of information available? The
literature review is an excellent source of background reading. If your study is
topical, newspaper and magazine articles can provide insights into popular
viewpoints. Alternatively, if it is to do with policy making, find out whether
minutes of relevant meetings are public documents, and if so, read them.
Having some prior understanding with regard to individual interviewees, and
their own set of circumstances, is also important. For example, when I (Hilary)
was interviewing people with pl~ysicaldisabilities, I did not always understand
the seriousness and full implications of the disease or condition they were
suffering from. Looking back, some of the questions I asked must have seemed
inane and meaningless. In a subsequent project involving disabled children, I
prepared for each interview by reading about the child's particular medical
condition and was then able to ask informed questions.
Demonstrating that you are knowledgeable about the area in which you are
interviewing is valuable in two ways. First, you will have more credibility with
the interviewee if you can demonstrate in your questions that you are familiar
with the context of the study. This is an especially important factor when
'interviewing up', that is interviewing people higher in status than yourself.
Secondly,there are implications in terms of the trustworthiness of the study. It is
less likely that interviewees will try to be misleading or deceitful because they
will fear being detected.
Personal appearance
. Dress and personal appearance may affect an interview, in the sense that the
interviewee may be assessing and making judgements about the (ability of the)
interviewer on the basis of what they can see. The literature (Warren, 1988)is full
of examples of research projects where investigators have adopted differentkinds
of dress an$hair style in an attempt to establish rapport and gain acceptance.
Collinson (1992) describes how he was concerned to look 'professional' and
'competent' when bterviewing managers. Looking 'well dressed' was particu-
larly important, and involved wearing a suit, polished shoes and carrying a leather
briefcase. In marked contrast, for his interviews with members of the shopfloor
workforckhe %re 'relaxed' and 'informal' clothes,joked and swore as the men
did, and generally tried to lessen any class or status differentials. Collinson's attire
was acceptableto the respective groups, which in turn encouraged people to talk.
The strategy worked for him, and there is every reason to think it will work for
you. This may well mean you will have to invest in outfits that include the casual
and informal,the fashionable and trendy, and also conservative-looking suitsthat
signal you are a professional.
Achieving a successful interview 105
Recording the data
We have given you some advice about asking questions,and persuadingpeople to
answer them. Now it is time to think about how to record responses in an accurate
and retrievable form. As the first section indicated, with survey work this is
generally a case of ticking boxes, or circling numbers in @re-coded) answer
blocks on the interview schedule. Alternatively, answers can be entered directly
on a computer. The options available when qualitative data are involved include
hand-written notes, audio- and videotaping.
Note-taking is cheap; you need only paper and a couple of pens or pencils. On the
other hand, it can be slow, is open to charges of selective recording and requires
practice and skill. It might be that you take notes as a 'fall-back' measure, as
Hilary did when she was interviewing someone whose voice box had been
removed because of cancer of the throat. However, if you envisage taking notes
on a regular basis then it might be worthwhile devising your own shorthand or
other form of customized speedwriting. Another useful aid is a simple form
containing sections or headings that reflect the main topic areas to be covered
during the interview. Key words, significant terms and the occasional verbatim
comment can be written down in the relevant space. As soon as possible after the
interview use these 'triggers' to help you expand on what was said. Bear in mind,
though, that the longer you wait the more detail you are likely to forget.
The use of hand-written notes, in conjunctionwith tape recordings,is described
in Box 7.7.
Audiotaping
Audiotaping is probably the most popular method of recording qualitative
interviews. There are a number of advantages. The interviewercan concentrateon
what is said. There is a permanent record that captures the whole of the conver-
sation verbatim, as well as tone of voice, emphases,pauses and the like (but note
that when agreeing to a study taking place, ethics committees sometimes make
it a condition that the tapes are destroyed afterwards; see Chapter 9 for further
discussion about obtaining ethical approval). Using a tape recorder demonstrates
to informants that their responses are being treated seriously. Finally, the costs
involved in purchasing a good quality tape recorder, microphone and cassette
tapes are not too prohibitive.
There are disadvantages, though. In particular, transcribing the tapes can be a
lengthyprocess; as we have noted before, a one-hourtape can take up to ten hours
to transcribe hlly. Further, the idea of taping the interview might increase
nervousness or dissuade frankness.When I (Hilary)asked aprominentergonomist
whose work involved acting as an expertwitness in personal injury litigationcases
whether I could tape our interview, he agreed but indicated that his responses
would then be 'public' rather than 'private', elaborates the distinction between

106 Interviewingforsocial scientists
Box 7.7 Note-taking while interviewing
One of us did a series of semi-structured interviews with well over 100
young children. The childrenwere told stories or given a problem and then
asked questions. The responses they could give were quite limited in
number and could be simply summarized, there and then. They were also
asked to explain their thinking, and it was simple to paraphrase the reasons
they gave. Analysis of the notes gave a pattern and reference back to the
tape recordings of the interviews supplied apposite quotations.
Interviews with 178 schoolteachers were routinely summarized in note
form.For example,teacherswere askedwhether they thought teaching was
a profession. Their answers werepredictably 'yes', 'yes with reservations',
'not sure', 'no, but ...' and 'no'. They were asked to explain their answers
and notes were made of the ideas they used in those explanations: for
example, one set of notes reads 'Yes: (a) training, (b) level of knowledge,
(c) status, (d) ethics.' Again, these points could be fleshed out by reference
to the subsequent transcripts and to the tape recordings, as the need arose.
For most purposes, the notes were sufficient for data analysis and
considerably speeded up the process.
'publ:c' a d 'private' accounts.(Cornwell, 1984.)In other words, withoutthe tape
recorderrunningt e was preparedto disclose 'insider' information about theworld
of occupational injury claims. My on-the-spot decision was to put the machine
away, produce a couple of sheets of paper and write furiously!
We \otl&yw of researcherswho have 'lost' interviews in the sense that they
ended up with either no recording or one which was inaudible. To try to ensure
this does not happen to you, Box 7.8 gives hints on audiotaping.
Videotaping
Recent developments in video technology mean that it is now more widely
available.None the less, videotaping is not commonly used to record interviews
apart from, perhaps, focus groups. Its current limited role reflects factors suchas
the expense involved, the specialist training needed to operate the technology
effectively,the vast amount of material to be anal~sedand the possible intrusion
it may create in the actual interview setting. Certainly, it is questionable how
comforzab/ejnterviewees are about video recordings. In one focus group Hilary
knows of, the participants refused to take part in the event if i f was going to be
videoed.
Despite their disadvantages, video recordings produce a wealth ofinformation,
both verbal and non-verbal. And if facial and bodily expressions, gestures and the
like are as important to the study as is the content of what is said, then this is the
medium for You - assuming that you also have the time required to undertake
Achievinga successful interview I 07
Box 7.8 Advice on audiotaping
Read the operating instructions for the tape recorder.
Practise using the equipment, trying out features like the pause and
record buttons.
Check the batteries, and change them frequently.
Always carry sparebatteries and spare cassettetapes. We find C90 tapes
1 work best.
1 Before each interview, test out the optimum setting for the recording
level. A poor quality recording causes problems later on when it comes
to transcribing.
At the start of the interview,position the tape recorder and microphone
close to those who will be speaking; there is more scope for this with
battery operated machines.
Try to eliminate background noise, for example ask for windows to be
closed if there is heavy traffic outside.
Turn the microphone on!
After each interview, write down the informant's name, and the date of
the interview, on each tape used. If any tapes are likely to be seen by
people not directly involved in the research, use an identifier code in
order to preserve anonymity. Other information might be useful, for
example the length of the interview.
* If finances allow, make a second copy of the tape, especiallyif they are
being sent away to outside transcribers. The duplicates can always be
wiped and reused at a later date.
the subsequent analysis. (See Jordan and Henderson [I9951 for a more detailed
considerationof the use of video technology.)
Self-evaluation
erviews, especially in the early days of a research study, can leave a lot to be
ired. And regardless of how experienced you are, there is always a need to
standback and assess how the interview went. Accordingly, it is importantthat at
regular intervals throughout the study you take stock to see what worked well,
what did not work and where there is scope for improvement.Having diagnosed
a particular problem area, see if you can think of how best to overcome or
circumvent it. If it helps to talk with someone outside the study, then ask friends,
your tutor or supervisor, or an experienced interviewer for assistance.
Some general advice about appraisingyour interviewingtechniqueispresented
in Box 7.9.

140 Interviwing for sociaI scientists
that you do not have to rely on inconvenient public telephones can be very
reassuring. Before you set off for the interview, give a friend or colleague written
- details of your movements, including who you have arranged to interview (whilst
emphasizing the need for confidentiality), where you are going, your route and
when you expect to be back. If you are staying away overnight, leave the address
and telephone number of your accommodation. You might choose to arrange to
check inwonyour return from the interview. And if your plans change, then let
someone know.
We have summarizedthe main points to bear in mind regarding personal safety
in Box 9.3.
L.. / A,
Conclusion
In this chapter,we have consideredthe risks of harm to both research participants
and researchers, as well as presenting techniques aimed at reducing any negative
effects. It could be argued that we accentuated the negative aspects of inter-
viewing, and downplayed the positive. This was deliberate, in the sense that we
wanted to concentrate minds on the issues that we think are important, and in this
way encourage best practice. However, we now want to redress the balance by
making one last key point: keep things in perspective. Do not dramatize the issues
so that they assume a level of risk far higher than interviews - even those con-
cerned with sensitive topics - typically entail (Hammersley and Atkinson, 1995).
10 Transcribing the Data
...committing verbal exchanges to paper seems to result in their immediate
deterioration: context, empathy, and other emotional dynamics are often lost or
diminished, and the language seems impoverished, incoherent, and ultimately
embarrassing for those who have cause to read back over their contributions
(including the interviewerlresearcher!). (Poland, 1995: 299)
Transcription is a part of the organization and management of the data. It is the
production of a written record of the interview.However,recalling the discussions
in Chapter 1,we observe that many social scientists would deny that there is one,
real version of reality to be captured. Accordingly, a transcript is one interpre-
tation of the interview, and no more than one interpretation. Furthermore, most
transcripts only capture the spoken aspects of the interview, missing the setting,
context, body language and 'feel'. Ln many, but not in all transcripts pauses and
hesitations are edited out. Decisions are made about the ways in which speech is
represented, there are invariably guesses about what was said, and there is the
issue of how to turn speech into written prose, all of which extend the distance
between a transcript and the interview event. Mishler (1991) draws a helpful
parallel between atranscript and a photograph. Just as a photograph is one, frozen,
contexted, printed and edited version of reality, so too with transcripts. So, in a
sense, the question is not so much whether a transcript is accurate as whether it
constitutes one, careful attempt to represent some aspects of the interview.
The transcriptionof data can be done at many levelsof detail.The level of trans-
cription will depend on the research purposes. With surveyresearch,transcription
is scarcely needed, and it is better to talk of data capture. With more qualitative
interviews, notes summarizing the key points are sufficient for some purposes,
although in other cases it is important to have transcriptions that include grunts,
'er', 'well . . . ' and 'mmm', as well as timed estimates of the length of pauses.
Mercifully, this level of transcription is not normally needed in many studies,
where the researcher is interested only in the meanings and not in the hesitations,
false starts and throat clearing that accompany them. However, these featuresare
important to linguistics researchers and those interested in discourse analysis.
Consequently, transcription is neither neutral nor value-free. What passes from
tape to paper is the result of decisions about what ought to go on to paper.
Sometimes, 'bad' language gets edited out. Sometimes, a typist decides to type
onlywords, not pauses and 'er', 'mmm' and 'huh!'. Similarly,there is the notorious
problem of how to punctuate speech: where should full stops, semi-colons and
commas go? What about paragraph marks? These decisions are the more acute
the further one moves from closed questions to open questions, from concise

142 lnterviewih for socialscientists 1 Transcribing the data 143
respondents to voluble informants. Transcriptions are, quite unequivocally, inter-
pretations. For that reason, if for no other, it is wise to keep interview tapes as an
archive to which Merence can be made if transcriptions prove to be inadequate
for the level of analysis that becomes necessary.
Data capture
With survey interviews, data are normally transcribedby the interviewercompleting
a schedule as the respondent answers. This is only possible because the schedule
and interview have bothbeen designed so that this transcription involves the inter-
viewer in making few inferences about the meaning of what the respondent says.
Interviewer training will also have provided training in interpreting responses
and, where computer-assistedtelephoneinterviewingisused (CATI), the reliability
of the interviewer is monitored. So, transcription is relatively simple, although
Dijkstra, van der Veen and van der Zouwen (1985) have shown that survey inter-
viewers do vary in the way they ask questions and code (transcribe) responses.
There is another possible source of error in transcribing survey data. It has to be
moved from the interviewer's schedule to a computer database. It is common to
pay clerks to type the information on the schedules into a computer program,
which is a potential source of error, the extent of which will be greater or lesser
depending on the motivation and accuracy of the typist. There is a faster, cheaper
and more accurate method available. It involves designing the schedule so that
respondents' answerscanbe read by an opticalmark reader (OMR). Thesemachines
are fast, as accurate as is the marking up of the schedule, put the data straight into
a database and cost less to operate than a team of typists, although their capital
cost needs to be taken into account.
Transcribing qualitative data
The commonest procedure is to make audio tapes of interviews and then produce
a typed version of the words on tape. Video tapes are an alternative method of
data capture and have their own additional transcription problems, discussed
below.
The transcriber's work
Undergraduate and postgraduate students may not have the funds to pay for trans-
criptionand have to do it themselves.Transcribingthe tapes yourselfhas undoubted
advantages: you become familiar with the data; you are reminded of the tone of
the interview; and you should get a transcript that is acceptable to you, whereas
when someone else does the transcription it is always necessary to hunt out the
mistakes and mistranscriptions. But, if you try to transcribe the tapes yourself,
you need to be a skilled typist. 'Peck and hunt' typing, applied to transcription
work, is a recipe for misery. Box 10.1 presents a further option that is becoming
more attractive with the development of technology (see also Anderson, 1998).
Box 10.1 Computer-assisted transcription
For some years Peter has been transcribing tapes by means of a voice
dictation programme. He uses IBM's Sinply Speaking Gold which has
been superceded (and super-superceded) by more robust and powerful soft-
ware. Affordable, entry-level programmes can be efficiently handled by
all new multimedia PCs, although software affordability is at the price of
sophistication. With these cheaper, older programmes each word has to be
sounded out separately and distinctly whereas more sophisticated software
does a reasonable job of recognising continuous speech. Then again, older
PCs may falter with top-end software.
The process is to listen to a tape through the headphones and then, phrase
by phrase, to dictate a 'clean' version directly into the word processor. This
is cheaper than using audio typists, keeps the researcher in control of the
transcription process and brings about greater familiarity with the data.
However, some people find it hard to persuade the software to recognize
their words accurately and feel that typing is faster. That has not, however,
been Peter's experience.
The main snags are that the program makes mistakes, which have to be
correctedmanually, and that it stilltakesabout four or fivehours to transcribe
an hour of tape. On the other hand, the researcher becomes more familiar
with the data, which speedsanalysis, and theprogrammeproducesaproperly
laid-out, correctly spelled transcript.As the sophisticationof theseprograms
increases, this is likely to become a preferred way of transcribing interview
data.
Traditional means of transcription are mechanical, repetitive, fatiguing and can
be stressful, especially when the recording quality is poor. As we said in Chapter
7, it is advisable to use the best recording equipment that funds can be stretched
to. Even so, there are many reasons why parts of some tapes will be barely
audible. Sooner or later, the interviewer will forget to switch a microphone on, the
interviewing site will have poor acoustic qualities, or low battery power will
produce a faint recording. Transcribers seldom have machines sufficiently
sophisticated to overcome these problems, and not everyone can use a computer
in conjunction with a hi-fi system so that the graphic equalizers lift the
conversation from the background noise. Some parts will not be transcribable,
and this should be noted in the transcript. Elsewhere, guesses will be made, and a
good transcriber will be trained to alert the researcher to the fact that a section is
a best guess.
Unsurprisingly, there are problems when there are hundreds of tapes to be
transcribed. Very few people can work full time at audio transcription, and a
5050 ratio of transcription to other activities is necessary if the audio typist is not
to quit and if productivity and accuracy are not to plummet. Clearly, this sort of
transcription can produce a serious bottleneck in the research process, something

144 Interviewing for social scientists
of which both authorshave experience.Transcriptionagencies offer an alternative
solution for research teams with sufficientfunds.
There are a number of transcription conventions, each designed to capture
greater or lesser detail about different features of the interview - Mishler (1991),
Silverman (1993) and Poland (1995) give examples of different conventions. TO
illustrate that transcription can be more than simply typing out the words, we
reproducea sliverof a transcriptproducedfordiscourseanalysis(Potter, 1996: 137).
Counsellor: Wha- 0.what bppened at point.
Woman: At point, (0.6) Jimmy ha- 0.
My- Jimmy is extremelyjealous.
Ex- e x m l yjealous =:son.
Has dways-been, fromthe da:ywe &.
Transcribersneed to be trained in using the convention that is most appropriateto
your researchpurpose. They must be stoppedt o m tidying up the tape in unhelpLl
ways. Patton (1990) tells of a studentwhose transcripts had all been put into good,
transactional prose by a 'helpful' transcriber. The result was data that were unfit
for the researchpurpose.
Box 10.2provides somerules ofpracticethat canmitigate some of the problems
mentioned in this section -but nothing mitigates them like money does.
Partial transcripts
A partial transcript is where the researcherkeeps full interview notes and has only
key sections of the tape transcribed. If the purpose is to use the interviews to get
understandingof the range of ideas used, then this may be acceptable, particularly
if there were two researchers in the room. One of us has used this technique with
focus groups, where the second researcher made notes and identified portions of
the tape recording that were especially worth transcribing. However, in order to
do this, the focus group's discussionshad to be quite tightly managed: fortunately,
informants' commentstended to be quite short and free of the complex,wandering
and looping structure that is commonwith one-to-one, in-depth interviews. Since
the purpose was to get a sense of the things that mattered to people, not to probe
those things in depth, notes-and-partial-transcriptswere fit for the purpose. It was
also a relatively cheap approach and quick to do, both of which were importantin
this project.
The underlying assumption, in the context of the research purpose, is that the
data are fairly unproblematic.
Full transcripts the director's cut?
Full transcription is expensive. Estimates of the time it takes to transcribe an hour
of tape vary between seven and ten hours, and poor sound quality will mean that
even longer may be n'eeded. Someone working on piecework rates is unlikely to
chargeless than 50 per hour of tape for this. Operating on tightbudgets,researchers
often have full transcripts made of the first ten or so interviews and of a sample
thereafter.kif6f?td partial transcriptions are used for the rest. While this canbe
Transcribing the data 145
Box 10.2 Some hints on transcription practices
I If you are to transcribe your own tapes
I
I Use C60 or C90 tapes - C120 tapes are more likely to break.
Knock out the tabs on the cassette so that you cannot inadvertently
record over your interview (you can always use sticky tape to cover
them later).
When typing, don't be mean with space. Leave at least a line between
each paragraph. If you will be using software for the data analysis,
check the format your transcript needs to be in. For example, with the
NUD*IST software package (see Chapter 1I), it is a good idea to begin
each sentence on a new line (or type it normally and save as 'Text with
linebreaks'). If you will be using software for analysis, don't waste time
with Bold, Underline, Italic, or any other effects.
When the interviewer speaks, begin the paragraph in capitals, in a
standard and consistent way, with letters that show it is the interviewer
speaking: 'INT: How did you feel about that, then?' for example.Do the
same for the interviewee(s).
I If someone else is doing the transcription for you, also
1 Make it clear that they are not to tidy up the prose. Tell them how you
want pauses, repetitions, laughs, hesitations and the like handled.
Get them to use a standard,non-usual symbol to showwhere they've not
been able to hear what was said, or they're unsure that they transcribed
correctly. Peter uses ** or zx, which can rapidly be located using the
'Find' function of a word processor.
Try to give them one interview to do at a time and set a deadline for its
return (on disk and in hard copy).
Make sure you check the first couple of transcripts quickly, so that you
can act if the transcriber is unsatisfactory.
Insist that the transcriber makes at least two disk copies of the files.
Impress upon the transcriber that the transcripts are confidential.
One way of getting transcriptions done is to send the tapes to an agency.As
long as the tape doesn't contain an identification of the informant,
confidentialityis secured.It can be more difficultto be as confidential when
the transcriber is a person who lives or works more locally. It is your
responsibility to see that transcribers are not given anything that would
identify informants or those connected with informants and to insist that
transcribers do not discuss their work with other people.

146 Interviewing for social scientists
pragmaticallyjustified, the price is thatmany meanings are reduced to the neatness
of note summaries, disguising the complexity and subtlety of the interview.
The transcripts then need to be checked for accuracy and suitability. Both
have to be done by the person who did the interview, and this can be time-
consuming. There are two main approaches to checking for accuracy. One, the
faster, is to read the transcript andif it makes sense, leave it at that, correctingonly
the obvious errors. Up to an hour of checking for an hour of taping is needed for
this method, depending on the quality of the tape and of the work. Alternatively,
read the transcript while playing the tape. This is more scrupulous and more time
consuming. Probably, the decision about which checking method to use rests on
a decision about how much it matters if the faster method misses points that
the slower method would pick up. In other words, is 95 per cent accurate trans-
cription acceptable, or does it have to be 99.5 per cent? Notice that 'accuracy' is
necessarily a relative concept. Even where the aim is faithfully to put words into
print, decisionsaboutpunctuationmeanthat more than one 'accurate' transcription
could be produced.
However, a bigger question is whether the transcription is a suitable represent-
ation of the interview,in terms of the researchpurposes. Conversationhas features
that seldom make it into print. For example:
Abbreviations (isn't, aren't, weren't) -sometimes transcribed as 'is not' etc.
Verbal tics, like 'er' and 'um' -usually ignored.
Pauses -either cut or shown simply by three dots ( . . .).
Repetitions (for example, 'What I mean. ..I mean. . .what I want to say is . ..
1me& that it is a real problem') -this might simply be rendered as 'It is a real
problem'.
There is a good case for weeding these features out during transcription. For
example, t s ~ n i o reducation official gave an excellent and stimulating interview.
When transcrib2 literally, it was full of 'er'. This tic broke the flow of the prose
and made the person seem dumb. He was embarrassed when he saw the transcript.
What did we gain by retaining the 'er'? In this case, nothing. Where other people
silentlypause fora moment's thought,he kept his voice going with the 'er' sound.
We understood it as a sign of thought, not of uncertainty. Unless the level of
intendedanalysisneeds transcriptionsthat retain tics, pauses and the like, they can
be excised. Does that matter?
For most social sciencepurposes, where it is the ideas, logic, beliefs and under-
standingsthat are wanted, this editing is acceptable. It may be that a straightprose
transcription is unsuitable, and that pauses, hesitations and tones of uncertainty,
weeded out of the transcript,need to be restored. There are some areas of research,
notably linguistics,where it isvital that transcripts are literal records of the sounds
on the tape, or as nearly as possible, and that pauses are exactly timed and
recorded. Unfortunately, the tone of voice - enthusiastic, bored, confrontational,
mocking - easily and routinely does not make it into the transcript. So too with
body language. There is no reliable way of conveying either in transcripts,
although it is usual to put, for example, [he/she/interviewer laughs], if that can be
Transcribing the data 147
heard. The question is whether it is more suitable to try for richer descriptionsof
the interview, or whether it is acceptable to settle for an 'accurate' rendering of
the spoken words.
Transcribing video tapes Video tapes are used because interviewers recognize
that an enormous amount of interaction takes place in non-verbal ways, and this
channel of information is completely lost on audio tape. Here, video recordings
give a better record of the interview. This is especiallytrue of focus groupswhere
the researcher wants to know who said what, and cannot infer it from an audio
recording.-
Needless to say, there areproblems, the most frequentlymentioned of which is
the obtrusivenessof the camera.It can alsobe hard to sitea camerain a focusgroup
so that everyoneis in frame. But, if there areproblemswith making a video record,
the problems of transcribing it are greater.
Since video recording has been done to capture information that gets missed by
audio recording, it follows that the transcript must be designed to capture that
information. It might then be in two columns, one for the words and the other for
non-verbal material taken from the video. But how is that non-verbal material to
be reliably captured? Should there simply be a description of behavioursthat need
little inference,such as sitting upright,looking at the interviewer,or gesticulating?
Or should there be more subtle readings of body language and inferences about
mood and attitude? The problem is that it is hard to make these inferences and
they are disputable. How are they to be made reliably? The classic answer is that
two researchers watch the tape and code the behaviour, discussingdisagreements
until agreement is reached. Leaving aside the point that all this shows is that two
people can come to an agreement (but not that a third, fourth or ntlz person would
reach the same conclusion), the main objection is the cost of doingthis.It can only
be justified, we suggest, under three conditions:
Where the techniques exist to describe non-verbal information accurately and
then to analyse it in ways that offer purchase on the research problem.
Where the research demands that such extensive data are collected (that is
rarely the case).
Where the aim is to produce excerptsthat can be used for training or illustrative
purposes, or that look as though they will have exceptional significancein the
findings.
Here, we are in essence repeating the message that the decision to use video tape
and thinking about how to transcribe it depend upon what would count as the best
way of discharging the research purpose.
Conclusion
The way that transcription is done reflects the research purpose and design and is
also effectively a part of data analysis. The form of transcription affects the ways

148 Interviewingforsocial scientists
in which the data can be analysed. The case of transcribing video tapes has been
used to highlight the way in which transcription is an act of interpretation. This
can even hold good for surveys, where the interviewer may need to usejudgement
in order to decide which pre-specified category best captures a respondent's
words.
It is to full-blown data analysis that we now turn.
1 1 Meanings and Data Analysis
Literarystructuralistshaveconcentratedon theformalstylisticfiltersin communication
between people, to the extent that some seem trapped in an impossibilism. . .
too often these theories are phrased in deliberate obscurity, self-referring in their
complexity. ..it is easy to forget the important messages that do get across. ..
to forget that the informant had something to say; in short, to stop listening.
(Thompson, 1988:24617)
If I were to try and put my finger on the single most serious shortcomingrelating to
the use of interviews in the social sciences, itwould certainly be the commonsensical,
unreflexive manner in which most analyses of interview data are conducted.(Briggs,
1986: 102)
In qualitative research, little is ever usually written about the process of analysisat all
. . .little is said about who the analysts are, . ..which particular perspectives they
adopt. ..how are disagreements resolved. ..whether full transcripts are used, how
much is reported, what level of uncodable or unsortable data is tolerable, what basis
is used for filtering data. . .(Powney and Watts, 1987: 174)
The analysis of data is perhaps the most demanding and least examined aspect of the
qualitative research process. (McCracken, 1988:41)
The analysis of the data gathered in a naturalistic inquiry begins the first day the
researcher arrives at the setting. The collection and analysis of the data obtamed go
hand-in-hand as theories and themes emerge during the study. (Erlandson et al.,
1993: 109)
The message of the quotations with which this chapter starts is one that most
qualitative researchers have learned the hard way: data analysis is difficult and
can take the novice - and the more experienced researcher as well - longer than
expected. That is an unwelcome discovery when analysis has been planned to fit
a short space of time before the report is finalized and presented. To a lesser
extent, it is also true of the analysis of survey data as well.
We begin this chapter from the position that one reason why data analysis can
be so complex is because there are embedded difficulties with knowing what it is
that the data could plausibly be said to mean. From there, we proceed to look at
the analysis of data from closed questions, including the use of statistics, to
indicate meanings that we see in the data. This is followed by an extended
discussion of procedures involved in analysing data from open-ended questions,
focusing on the analysis of interview transcripts. We review the use of computer
software to support this process.