42 Interviewing for social scientists
Box 3.4 Feelings about an interview
(This is an extract from a researcher's fieldnotes)
I'm a little phased by having been away from home for two days, not sure
that this interview will rgally happen and dogged by low-level anxiety
about some personal difficulties.But I've done hundreds of interviews and
think I'm skilled at relating to and being easy with people. Spend some
time on breathing, relaxation and a h a t i o n s ...
'Like all colleagues in this study, he [the informant] is a prolific and
engagingtalkzr, with nice, open and engaging body language and a speech
style that leaves little room for me to talk. To begin with, I'm happy to put
aside my three page interview guide and let him shape the conversation.It
is levant interesting and, above all, it tells me what he thinks is
1+'
important. ...I become increasingly worried about what we're not
covering. ..I'm enjoying listening to my colleague and, as with other
interviews in this series, I feel an emerging sense of who he is, as well as
warmingto this person. But I can feel tension -I'm doing it wrong, I'm not
getting through the questions on the guide. I start looking for points to
interrupt him. I manage to slip in a rather complicated question and, when
he simplifiesit, I settle for his interpretation at the cost of losing data that
would have been interesting for comparative purposes. Eventually, well
aware of the time, I rather derail my informant, cover some of the closed
questions in the protocol, scan it to see what important areas we've not
covered (very few, thank God!) and turn him towards the remaining big
area.
Relief.It more-or-less worked.Confidencehigh and self-esteemgroomed.
4 Designing an Interview-Based Study
This chapter is based on the assumption that you will be doing a project in which
interview approaches offer a good prospect of getting answers to your research
questions. Most of the points we discuss have to be confronted regardless of
whether an enquiry is interview-based or led by other methods. It also sets the
scene for Chapters 5 and 6, which also focus on design issues, although the dis-
cussion becomes increasingly concerned with design issues that are particularly
important for interview research.
We begin this chapter with some thoughts on how you might identifya topic to
investigate. Then, you have to choose the enquirymethods,which we assumewill
include interviews. Here, the guidingprinciple isthat the methods shouldbe fit for
the purpose - they should have the power to provide data that you can use to
answer your particular research questions: different questions, different research
designs. While you are engaged in the twin processes of choosing the topic and
the methods, you should also be doing a search of the literature. We suggest that
this is the right time for you to begin drafting your report and we make some
suggestions for drafting the literature review section of it. Lastly, we address
some of the more detailed design considerations that you will have to consider in
an interview-based study- sampling issues, the trustworthinessof the design,and
the claims that might be made about its generalizability.
Staking out the territory
What is the topic that you will investigate? This decision causes beginners
problems and is not unknown to experts either. Thekind of rationalthinkingabout
research that pervades much of the writing about social research points to the
following strategy for identifying a research topic:
Your reading on a topic might make you think that there's an aspect, an
area, or an application of an idea that hasn't been pursued. You might wonder
whether findings would apply to different groups, or to different situations.
You might suppose that the conclusions were a result of the research methods
used, and that differentmethods (substitutinginterviewsfor questionnaires,for
example) or differentuses of a method (substitutingunstructuredfor structured
interviews) would lead to different conclusions. Or you might reasonablywant
to do something that is commonplace in natural science, which is to replicate
(copy as closely as possible) a study in order to check out the strength of the
conclusions.

44 Interviewing for social scientists
Our experienceis that people find topics by other means as well -sometimesit
seems tkat more topics come from the less publicized sources of inspiration than
from those that are commended in the research literature. For example:
.Interest. You've got a feeling for a topic, such as housing issues, single old
people, children andthe internet,discipline inhigh schools.Alternatively,atopic
may be&fr&ng a lot of media attentionand seems to be ripe for a small-scale
but systematic enquiry that is sensitive to the relevant academic literature.
You sensibly believe that an important part of doing research is the relation-
ship that can be created with the supervising tutor. Consequently, you pick a
congenial academic and say that you want to work in the area in which she or
he is expert.
It is suggested to you that you might like to work in a certain area.
You set no real choice. You have to do a piece of action research while on- -
placement, or in your place of work. So, you might be expected to do some-
thing about the education of hospitalized children, about post-natal supportfor
opiate-addicted mothers and neonates, or about prison health services. And
vou have to do it within the situationin which you are working, which may not
be a very promising research setting.
Who cares? From topic to topicauty
Social research may be valuable because:
It gives the researcher a chance to practise and master research techniques -
apprenticeship.
The findings add to knowledge about the topic, or call into question existing
concepts, assumptions or findings-substantive value.
Researchtechniques are appliedto new areas, or new techniques are developed
-methodological value.
Research reports should generally have substantive or methodological value,
even in the case of reports that comefrom apprenticeshipresearch. In otherwords,
they will be more than collections of information. It is not enough to have a topic
to investigate, nor to collect a lot of information about it. Fair responses to that
information would be 'So what?' 'What does the information mean?' 'Why is
it significantor topical?' 'Who cares?' The true story in Box 4.1 is an example of
a very poor choice of topic while Box 4.2 indicates the difference between
information collection and useful research, a distinction that is developed in Box
4.3. The heart of the difference lies in purposefully connecting the information to
a conceptual or theoretical base.
Once a topic is identified,it needs to be connected with relevant other work and
thinking so as to make the value of the study clear. This is illustrated with
reference to 'action research', which couldbe (uncharitably)describedas all action
and no research. It is done with the aim of trying to make a difference to a
Designing an interview-based study 45
Box 4.1 A proposed study with no obvious usefulness
A mature, overseas student, on.a course concerned with the assessment of
students' learning, insisted that he wanted to study the operation of the
educational bureaucracy in his country. He agreed that there was nothing
unexpected likely to come out of this study; that the Minister of Education
already knew that his ministry was a featherbed for his family and friends
(after all, the minister himself had lifted them gently into that bed); and that
a research report would make no difference (exceptthat the student,had he
completed the study, would have probably been wise to make himself very
scarce).
The story, which still puzzles Peter, illustrates the 'so what?' question,
describing a study with no topicality and no usefulness.Nor was there any
indicationthat new conceptswould have been developed, or old ones applied
and extended. It seemed quite pointless, an example of interest without
thought.
Box 4.2 A way of answering the 'so what?' question
A research study, for example, would not be very interestingif the write-up
said, 'I used x, y and z methods to gather information and the findings
appear below.' Rather, it is the theoretical or conceptual base in which
particular methods are couched that makes them interesting. Consider the
following statement: 'Because I was interested in exploring the conceptual
position taken by authors in this area, methodsx, y and z helped me answer
the following questions. The findings are presented below and are com-
patible with theories and ideas derived from the literature. The methods
employed were adequate to answer the questions I had raised' (Grasha,
1996: 95).
situation.It typically involves examiningpractices and trying out ways of improv-
ing them. What this means is that the researchneeds to be purposefuland focused,
not just information collecting. In fact, the need to be aware of the literature and
other thinking is all the greater here, since other people's research and writing are
a fruitful source of ideas about the most promising approaches to making a
difference. Where work is purposeful, uses theory and researchfindings-when it
is connected - then it is better, not least because it contains an answer to the 'so
what?' question. And the more that answer is emphasized, the more attention is
likely to be paid to the findings.
It might be objected that progressive focusingwas recommended in Chapter 1,
whereas here we are commending clarity and focus from the start. There are two

46 InteM~ilrgforsocial scientists
Box 4.3 Commentaryon a selection of research topics
Research topic
A study of a day care
centre
An action research pro-
ject in Ward M1
How effective is Social
Work Team PR5?
How effectivedo
clientsof Social Work
team PR5 believe it to
be?
A study of teenagers'
attitudes to pollution
How do police deal
with rowdy youths?
An interview study of
beat officersin two
contrasting areas
What is the best way of
teaching the concept of
place value to six-year-
olds?
Under what circum-
stancesdo 'empty
nesters' give to
charity?
Comments
Topic too big. Focus unclear.No indicationof
what the value of the study might be, nor of
who might be interested by it
TOO general -what aspects of Ward M1 are
going to be studied? Why? Who will be the
audience and who will benefit? In what ways?
And with action research, it is important to
know that you can make changes, see them
through and evaluate them
More focused than the other questions but the
notion of 'effectiveness' is a minefield- it is
really too big to tackle in one undergraduate
study. The concept of effectiveness also
appears in Box 7.4
Much better. With the right literature review
this could be good, because work on effective-
ness tends to be producer-centred, not client-
centred. This could question the validity of
that tendency
'Pollution' is too broad a topic. More seriously,
this research is unlikely to tell us anything
new -it is unlikelyto have anyvalue beyond
apprenticeship
Good to see a focus on contrasting areas and on
'amanageable sample. However,defining
'rowdy' is going to be tricky (but not impossi-
ble). The serious mistake is to forget that
interviews do not give information about
practice: they tell us whatpeople believe they
do
A well-formed question that will not yield to
interview methods. A long-term, experimental
design will give the most secure results
There is value in knowing why people think
they give to charity but it is well known what
people actually do is often quite different from
what they believe they do: it is an areawhere the
public voice tends to overwhelm the private
voice. Unsurprisingly, then, research into
contirzued
Designing an interview-based study 47
altruism, of which this is an example, is
dominated by experimental studies
A survey of the needs In many ways a well-defined study. However,
of recently bereaved is a survey the right way to proceed? Yes, if
unemployed men there has been a good reading of other research
below retirement age and some unstructured or semi-structuredpilot
work. Otherwise. ..
Your tentative research Your comment
question
responses to this apparent contradiction. The first is that progressive focusing
should be seen as progressively focusing on the best ways to understand and
investigate the research question, which will lead to some changes in the initial
question and which might lead new questions to be brought intothe research.This
response is essentially that progressive focusing takes place from the start of
research into a question of value. The second response sees progressive focusing
coming from the topic. As you read more about the topic, talk to people and,
perhaps, do some pilot research, questions arise, which are refined and developed
as the project progresses. In this case progressive focusing leads you to the
research question, which is then sharpened by further focusing. Often, it is only
through progressive focusing that researchers identify the value and significance
of their work. The difficulty is that can take up time: and time is always in short
supply in research projects.
If research is to be valued, it needs to have a purpose, to be shapedby important
substantive or methodological concerns. Making the case that these concerns are
important involves relating them to the literature, as well as listening to stake-
holders' claims that this is an important issue for them. However, the research
community is unlikely to be very interested in a story of significanceto one social
work team unless it is related to the prevailing research discourse-to the existing
literature. The following advice on doing a literature review should be read in
conjunction with that in Chapter 12 on writing up the results.
Reviewing the literature
To illustrate the importance of reviewing the literature, we introduce the action
research approach to enquiry. It is usually small-scale research that is intended to
help the researcher or practitioners to take actions that are planned to make a
difference and whose impact will be evaluated. That evaluation will often set off
a further round of research, leading to the analogy that action research is a spiral
of activity, a process. Because the problem and situation are immediate,there is a
temptation to jump straight in to the research, sometimes without thinking too
carefully about the methods, let alone about the instrumentation in any detail.
Understandable though that is, there are three major disadvantages:

48 Interviewingforsocial scientists
Rediscovering the wheel: the investigator has to invent instruments when
perfectly good ones already exist. Furthermore, if they were used, the investi-
gator would get data that could be directly compared with the study from
which they were drawn, broadening the interest of the study's findings.
Missing significantfindings: the investigator does not realize that some good
answers to this problem already exist; or that there are good conceptual frame-
works that can profitably guide the research; or that this is an impossible
enquirybecause it demandslarge-scaleand sophisticatedresearch and analysis.
If the study is never connectedto general thinking, it stands alone. Readers can
still draw their own conclusions,but they may remain unaware of some of the
possible significancesof the results.
The case has been that it is important, even in action research, to examine the
relevant literature. How might this be done? Suppose this is to be a study of
femalewashroom attendants. This is manual labour, performed by women, in cir-
cumstances that are full of taboos and symbols. It is a service occupation, poorly
paid, held in low esteem and sometimes quite unpleasant. The way that female
washroom attendants regard this work is related to their sense of self and self-
esteem, which means that there is a greater or lesser interplay between their
identity and cultureas a washroomattendantand their other identitiesand cultures.
At first sight, it is rather hard to see why washroom attendants would be taken
as a subject of intrinsic interest. However, there could be a lot of value if these
people were studied in order to test out or develop theories about how women
balance work and home lives; self-esteem; occupational motivation; taboos; or
women's perceptions of their bodies. Different theoretical concerns imply
different research questions, different methods and different ways of making the
claim that the research is of value and significance. One aim of the literature
review will be to see which methods and questions are normally applied when
investigatingany one theoretical line.
Another aim will be to see what the big issues and questions are in any
theoretical stance. Suppose that the attendants' work is changing. They are now
expected to cover two washrooms, not just the one they tended in the past; their
pay is not increased annually, overtime is no longer available andjob security is
threatened as budget pressures grow. There is talk of introducing charges for the
use of the facilities and also of having women cover both male and female
washrooms, and of male attendants doing the same. These changes could be
interpreted within the literature on the intensification of work, as well as within
the often-rqated literatureon women's work. Both are quite diverse and extensive
fields of study, as a literature review will show. One question for the prospective
researcher is whi'ch interpretations within the literature appear to be more
convincing and more relevant, which implies a critique of the others on grounds
such as they are incomplete, their conceptualization is inadequate, or there are
problemk'6ifli'the empirical research. In turn, this will help to clarify the
directionsfor the study and set the tone for the report.
This rather assumes that there is little problem in identifying the relevant
literature. Nothing could be further from the case. Searches of CD-ROM or
Designing an interview-basedstudy 49
on-line data bases can be infuriating. On the one hand, they can throw up too
many references (it is a good idea to limit the first search to the past fiveyears, or
less). On the other, many references will be impossibleto access and many will be
trivial or of little interest. And there are all the valuablereferencesthat are missed
because no human imagination would guess the right keywords to find them. A
good technique is, first, to locate a couple of recent and quite comprehensive
references and follow up some of their citations. Finally, keep in mind a sense of
proportion. Within the time available, there are limits to how much can be read.
Usually, the aim is not to produce an exhaustive literature review but one that
links an enquiry to wider concerns and issues. In other words, the review is there
for a purpose, to provide one answer to the 'so what?' question.It establishesthe
substantive significance of the investigation and, where the enquiry is
methodologically innovative, it shows the reader what there is that is innovative
about the methods - how they are an advance on the methods used by earlier
researchers.
At somepoint, the reading will need to be written up and presented as a section
of the research report. Table 4.1 is a guide to one way of doing that, but because
the way you organize the literature review will have to depend on the research
question, it has to be appreciated that other structures are feasible.
Three further points deserve mention:
Be sparing on historical background (and one of us has had a career as a
historian). It is legitimate to survey the history of an issue, but only when it is
clear how this affects the proposed research. Where the purpose of a historical
overview is unclear, it can damage a review by making the reader suspicious
of a writer who appears to be inflating the report with irrelevantmaterial.
The literature review may be concise.
Reading the literature on a topic generates research questions (see Box 4.4 for
an example). However, research questions develop and new ones can emerge
during the research. The implication for the literature review is that whatever
is read and written early in the study will be re-written later in the study in the
light of new emphases. Research manuals often imply that the review is done,
then the fieldwork, and then the findings are found. In fact, the review, and the
reading that goes with it, are usually revisited and rewritten after the findings
are in.
Much more extensive advice about reading research reports is to be found in
Locke, Silverman and Spirduso (1998). Cooper (1998) goes into considerable
detail about the processes by which a set of research reports is drawn togetherinto
a synthesis, such as the literature review described here.
Making your research credible
Readers want to know that the research has been carefully done, so that findings
can be trusted. Within the positivist tradition, this involves taking care to
maximize validity and reliability, while taking steps to ensure that the findings

50 Interviewing for socia1scientists
TABLE 4.1 A literature review template
Focus
Preview of the issue
Main literature findings - empirical
Are there any notable exceptions to the general
trend?
Are there any explanations that need to be
considered? [Beruthless here: your study may
not be much concerned with why so much as
with what, or with the effects]
Are there any problems with the ways in which
existing research has been conceptualized?
Are there any problems with the sampling?
So, explain w k t issues you are going'to
examine: wha't are your research questions?
Who cares about your fmdings?
What methods have been used? Are they
suitable? Would other methods enrich
understanupf.tQ phenomenon? Justify this
claim
Example
Throughout the Western world, those infull-
time employment havefound that they are
frequently being expected to work longer, to do
more. ..This has ofien been called
intensfication
Studies have been conducted in a number of
countries, involving both low- and high-status
occupations. Thefindings have generally been
that. ..
However, .. .[You may not need this section]
[There is general agreement that economic
factors are the major cause of intensification,
although ideologicalfactors also need to be
considered]
A majorproblem with the notion of
intensification is that it assumes that objective
changes, such as requiring workers to take more
responsibility, areperceived injust one way
(as an imposition rather than as,say,
empowerment)
Attention has been concentrated onprofessio~zal
groups, and not on low status groups, such as
women who do manual work. It is not clear,
then, whether intensification is a widespread
pherzomenon, or a concern of the 'chattering
classes ' alone
Thisstudy will therefore . .. Thejindings will
help to decide whether. ..
Much of the research has takenplace on the
assumption that demonstrating increases in the
demands made on employees is the same as
demonstrating the existence of intensification,
which has largely been done on the basis of
documentary sources and survey research.
However, whether changes are seen as
intensification is a matter of individual
perception. Exploring theseperceptions
requires the use ofinterview methods. ..
from the sample can be readily generalized to the population. First, we discuss
validity and reliability within that tradition, since they are important prompts to
thinking well about research design and they are directly applicable to survey
research.However, they cannotbe straightforwardlyappliedto qualitativeresearch,
Designing an interview-based study 51
BOX4.4 From research topic to research questions
(This is based on a study in which Peter was involved)
The researchers became interested in the occupational socializationof new
higher education lecturers. The topic is important, since patterns estab-
lished in the first years in the job persist. They approached the topic with
three principles in mind: that it was necessary to consider the 'whole
person' when studying academic staff; that socializationshouldnot be seen
as a one-way process; and that informal socialization and professional
development were likely to be more significantthan formalprocesses (this
point came from a major study of schoolteachers in which Peter had been
involved). These principles were to be tested against the evidencecollected
in the research.
The literature review threw up far more research questions than the
duo could pursue. The research guide included those that could be fairly
pursued through long, loosely structured interviews; which would shed
light on the three principles; and which seemed to have the potential
to advance thinking about the topic. As the research progressed, the
researchers continued to find new literature, which influenced their inter-
viewing, although it was not necessary to revise the interviewguide.
Typical interview questions were: Which of four views of academic
work (taken from the literature) is closest to yours? Which of fourviews of
the goals of undergraduate teaching (taken from the literature) is closest to
yours? How would you describe the relationship between your work and
your out-of-work life? (Prompts taken from the literature.) Were there any
formal induction arrangements? If so, please describe them and say how
good you found them. Have you got a mentor? If so, please describe the
mentoring process and evaluate it.
so this is followed by a summary of thinking about the credibility of qualitative
research. The last section of this chapter concentrates upon the way in which
findingscan be generalized from surveys, action research and case studiesto more
general contexts.
Validity
We take validity to raise the question of whether you are actually investigating
what you claim to be investigating. Campbell and Stanley (1963) give the classic
account of threats to validity. Many of them we just have to live with, on the
principle that flawed information is better than none, and there is sometimes little
that can be done about some threats. So, while survey research aims to be reliable
and valid, because the survey format itself constrainsrespondents,it compromises
validity. It is a matter for judgement whether the cost to validity is large (and it

52 Interviewingforsocialscientists
will be least where the survey instrument has grown out of piloting that has
includedqualitative interviews) and acceptable.
In both quantitative and qualitative approaches, threats to validity are legion,
although some qualitative researchers would use alternative terms to 'validity'.
For example, your study may claim to look at the attitudes of young people
towards the police. Research based on interviews with 15-year-olds at school
might have validity problems because those playing truant, whose attitudes
towards the police might be quite distinctive, would be missed. Another threat to
validityis when we fail to create a situation in which the respondent feels easy and
able to talk. One report tells of a researcher who talked to business executives on
trainjourneys:
What struck [the researcher] was the extent to which the views and opinions of the
managers,off-guard and to aperson they were unlikely to meet again, contradictedthe
'reality' contained in much contemporary management literature. Had the interview
taken place in the manager's office, the results might well have been quite different.
(Easterby-Smithet al., 1991:78)
Validity is enhancedby:
Interviewing techniques that build rapport, trust and openness and which give
informants scope to express the way they see things.
Schedulesthat contain questions drawn from the literature and frompilot work
with r&.pondents.
A set of questions that fully covers the issues raised by the research question-
key aspects are not ignored.
Not asking questionsthat are irrelevantto the research topic -a waste of scarce
inteikieuctime.
Promptsthat encourage informantsto illustrate, expand and clarify their initial
responses, talking in detail and about specifics.
A sample that is fit for the purpose of the research. If the work is preliminary,
opportunity samples and snowballing (see below) are acceptable. If the aim
is to make claims about a group, or to give a rounded account of an event,
sampling needs to ensure that all points of view are appreciated. Big samples,
preferably selected at random, are needed if you want to claim that your
findings are very likely to hold good for the population.
Thinking about the possible effects of interview times and settings. Ideally,
respondents should be interviewed more than once, with the setting changing.
Interviews that are long enough. Inte~iewersoften find themselves pressed for
time (seeBox 3.4).However,no matterhow longaninterview,wehave foundthat
'one of the most basic rules of interviewingis that the most interestingmaterial
emergeswhen the recorder is switched off (Powney and Watts, 1987: 139).
Reliability
If it is assumed that there is a stable reality 'out there', which is to be precisely
measured and described, then it is important that the findings are not corruptedby
Designing an interview-based study 53
the research process: the design and tools need to be reliable. Some qualitative
researchers would want to talk in terms of the consistency, trustworthiness and
authenticity of the research, although they would still find themselvesaddressing
many of the issues covered here. In survey research, reliability is mainly about
trying to reduce interviewer bias so that we can trust that the findingsare neither
the product of the research instiuments, nor of the interviewer's quirks and
improvisations. It is likely to be important to make sure that all informants
are asked exactly the same questions and given similar sorts of clarification.
The findings would be unreliable if it turned out that some questions were
explained to some respondents who were puzzled by them but not -?other
puzzled respondents. This is especially important where several interviewers
might be involved in a large-scale survey.
Complete reliability is not attainable, although the guidelines for survey
researchers (Box 4.5) show how it may be maximized by limiting variations in
interviewingpractice (seealso Brenneret al., 1985).Thereis also furtherdiscussion
of reliability in the context of telephone interviewing, which is described in
Chapter 6. Reliable data analysis is covered in Chapter 11.
Box 4.5 Typical guidelines for survey interviewers
Look interested.
Stick to the schedule: read the questions exactly asprinted, in that order;
read all of them.
If there arepictures, cue cards or other apparatus,use them as scheduled.
Try not to signal approval or disapproval of any answer.
Do repeat the question if asked.
* If the respondent refuses to answer a question, that has to be accepted,
and without any signal of irritation.
Make sure that you understand a response.
If not, probe in a non-directive manner ('Would you tell me more about
that, please?'); if need be repeat (but do not alter) the question.
If an answer is not adequate (it is an answer to a different question),
thank the respondent for the answer and repeat the original question
('Thank you. And could you tell me ... ').
Do not answer for the respondent, explain the question, or give any
other new information, unless the schedule allows it.
The schedule should give guidance on the clarification you can give
if asked, and that should be based on the experience of piloting the
schedule. Practice varies from being allowed to repeat questions but not
to clarify, to being able to explain the meaning of words or phrases, to,
sometimes, being allowed to re-phrase the question (which compromises
reliability).

54 Interviewing for social scientists
The assumptions that underpin qualitative research mean that classic concepts
of reliability do not sit very well with this approach:
Qualitative research is not looking for principles that are true all the time and in all
conditions, like laws of physics; rather the goal is understanding of specific circum-
stances, how and why things actually happen in a complex world. Knowledge in
qualitativeinterviewingis situational and conditional.(Rubin and Rubin, 1995:38)
,
The quest for situational and conditional understanding is quite different from
the quest of positivist social science.Consequently, the concepts of reliability and
validity cannot be imported from positivist approaches to qualitative ones.
'Reliability', based on assumptions that phenomena are regular and unchanging,
is particularly inappropriate in epistemologies that see situated cognition,
complepity and change as pervasive and normal features. Therefore, it is import-
ant to recogbize the futility of imagining that 'if you could strip the interview of
all of these [biasing] factors, the "real" or "true" or "unbiased" response would
emerge' (Briggs, 1986: 21). Yet, it can be a useful exercise for the qualitative
researcher to ask traditional reliability and validity questions as a way of getting
a freshperspe&ve on the research design.
Thisviewof the inappropriatenessof traditionalnotions of reliabilityandvalidity
inmuch qualitative research does not implythat qualitativeresearchersare castinto
the pit of relativism:
The ethnographeris not committedto 'any old story', but wants to provide an account
that communicateswith the reader the truth about the setting and the situation,as the
ethnographerhas come to understand it.(Altheide and Johnson, 1994:496)
By and large, qualitative researchers have responded to the problem of demon-
strating the credibility or trustworthiness of non-positivist research in two ways.
First, by recognizing the value of continuing to ask, but not to be bound by, the
questionsraised by classic accounts of validity; and secondly, by developingtheir
own criteria (Lincoln and Guba, 1985),which have a lot of common ground with
the rules of historical enquiry. The three most relevant concepts here are:
Consistency: this is akin to 'reliability'. The requirement is that the researcher
shows how the research has been done and decisions have been made, so that
the reader could conduct an 'audit trail', examining the good sense and plaus-
ibility of the researcher's thought and actions. There should be evidence that
the inevitable inconsistencies in the data had been considered and there should
be some account of how they have been handled. This largely embraceswhat
Rubin and Rubin (1995) call transparency and consistency. It implies that
researchers will describe their findings and their analysis, and their own work
in detail - 'thick description' as it is sometimes called.
* Truth value: this involves providing evidence that what the researcher has
captured is recognizable as a fair representation of things as informants see
them. It involves triangulation of methods and triangulation by examining
Designing an interview-based study 55
different perspectives on the research topic (see Chapter 2). It involves
checking with interviewees that what they said is what they meant, and that
interpretations make sense to them. It should come from immersion in the
research and, to some, it implies participant research. Inevitably, differences
and inconsistencies will be highlighted, and, following the principle of consis-
tency, the researcher should show how they have been dealt with.
Neutrality: this is a requirement that the researcher considers their own role in
the research. Patton (1990) has remarked that in qualitative research, the
research is the instrument. Clearly,the researcher cannot be the objectiveagent
depicted in some positivist views of scientific enquiry. The researcher has an
influence, and a differentresearcherwould have a somewhat different influence.
The aim is not to try to standardize researchers, but to have them reflect on
the ways in which their background (class, gender, race, special concerns),
personality (which is critical to achieving rapport and trust), mind set (assum-
ptions and preconceptions), and actions have contributedto their account.The
aim is to check the 'rampant subjectivity inherentin more phenomenologically
based paradigms that will prove to be the nemesis of new paradigm research'
(Lather, 1986: 68): the aim is to check, not to obliterate.
Reliability and validity in qualitative research
In a nutshell, the qualitative response to the issue of reliability and validity is to
require researchers to demonstrate that what they do is fit for their research
purpose. Invariably, this shades into questions about the characteristics of good
research. It is not possible fully to define these in general terms, for good research
consists of adequate response, in the research setting, to the questions raised in
the checklist at the back of this book (pages 185-191). Nevertheless, a number of
features have been identified and should be borne in mind, including:
The value of the outcomes for action (in action research), for programme
development(in evaluationresearch),forpolicy and for the research community
in general. This implies evidence of a clear answer to the 'so what?' question.
Consideration of issues of power and of structural inequalities in the research
setting(especiallyin feministresearch [Blackmore, 19961and studies influenced
by critical theory [Robinson, 19961).
The incorporation of ethical considerations into the work (is it possible for
unfair work to be good work?).
The achievement of a situational equality between interviewer and informant
(Benney and Hughes, 1956) or of closeness between interviewer and inter-
viewee (especially in much feminist research - but compare this with the
distance that the historian Lummis [I9871 advocates). This is important if the
researcher is to hear 'private talk', which discloses meanings that are absent
from 'public talk', to hear an authenticvoice that might tell of things subversive
or discreditable,rather than to hear the socially sanitized accountsof the public
voice (Cornwell, 1984).
Sampling adequacy.

56 interviewing for social scientists
The appropriateness and power of the interview questions.
Design flexibility: this means that the researcher was responsive to unanti-
cipated meanings and opportunities.
Transcript quality (see Chapter 10).
A po&nodern problem
Questions of the Eredibilityof the research become problematic once it is assumed
that informants' accounts are shifting and shaped by the time and circumstances
of the interview, and if it is accepted that the interviewer will construct the
meani&s'of%e interviews. It is difficult to give advice onjudging the credibility
of such research. In the literature, the emphasis is on the researcher providing
plenty of data on how the research was done so that readers can judge whether
this appears to be anhonest attempt to explore it systematically. There are no rules
for judging, only the requirement that the researcher make the research process
transparent, being explicit about what was done and why:
While the transferof findingsfrom one case study to anotheris done by the reader,the
researcher has an obligationto provide rich, detailed, thick description about the case.
(Firestone, 1993: 18)
When such research is presented in non-conventional forms (Chapter 12),
problems multiply, since there is no consensus onjudging these research 'reports',
although many academics suspect that they - and postmodem research stances-
can be licences for the slack exercise of subjectivity and relativism. In such cases
it is prudent to establish, in advance, what your sponsor or supervisor would
regard as evidence that the research is credible.
Sampling and generaliz~ngfrom the study
The question of the ways in which generalizations can be made from interview
studies has been touched upon in Chapter 1. Table 4.2 extends that discussion by
showing how different interviewing purposes are associated with different
sampling strategies and thence with the confidence with which the researcher
might make claims about the generalizability of the findings from the sample
to the population. (A population is the group from which the sample is drawn:
on the basis of samples of trainee midwives, football supporters, or retired men,
researchers might make claims about the populations of trainee midwives, foot-
ball supporters, and retired men.)
Sampling always needs to be done thoughtfully, since the sample of respon-
dents or informants affects the information that will be collected and determines
the sort of claims that can be made about the meaning of that information. For
example,if the aim is to generalize from a sampleto a population, then the sample
must be representative of the population and large enough for generalizations
to be made with confidence. There is a considerable technical literature on the
confidencewith which generalizations can be made from a sample to a population
Designing an interview-based study 57
TABLE 4.2 Interviews, sampling and generalizingfrom the sample to apopulation
Interview purpose Sampling strategy and generalization
makes claims
about
Survey
Survey
Survey plus qualitative
questions
Qualitative -
researching an event or
time
Qualitative -cultural
interviews (finding out
about beliefs,
understandings and
feelings)
Exploratory -getting
into a field
Random sample. Size as large as possible. See
Oppenheim (1992) for example, for advice on
sample size. Researcher makes strong claims
about the generalizability of the findings
Reader makes
inferences about
generalizability
Structured sample. Used where there is a
danger that random sampling might lead to key
groups being unrepresented. The larger the
sample, the greater the confidence when
generalizing from it, although some consider
that structured sampling is not as powerful a
basis for generalization as is random sampling
Exceptions to the rule
If the survey is the most important element,
follow rules above. If the exploration is the
most important, see below. Take advice on
generalizability from appropriate section of
this table
For example, intention is to survey and
generalize but not possible to control sampling
-use opportunity sample. Researchers are on
shaky ground if they try to claim the findings
are generalizable
Assuming no great concern to generalize, use
opportunity sampling, look for good
informants, increase sample size by
snowballing. Keep adding to sample until you
are hearing nothing new. But, also take care to
hear the story from different perspectives -
seek out people who may have a different slant
on what happened
As above, taking care not to concentrate on
high-status informants and those who readily
come forward to be interviewed - danger of
not hearing the private or silent voices
Converse with anyone who might be able to
help you get oriented. At best, the researcher
can suggest that readersmight consider
implications of the findings for a population
A sample of one (for example, Boyle and
Woods, 1996) is enough to show that some
research generalizations can be too sweeping

58 Interviewing for socia/ scientists
(for example,Oppenheim, 1992).Research undertaken in the 1970s looked at the
characteristicsofpeople volunteering forresearch (quoted in Holden et al., 1993).
The authors came to the conclusion that in comparison with non-volunteers,
volunteers tended to be: better educated; of higher social class status; more
intelligent;in greaterneed for social approval;more sociable. Although nothingis
said relating specifically to volunteers for interview studies, it is important to
realize that the act of volunteering to be interviewed is, in itself, a sign that the
person may be in a minority in the group in which you are interested.
For qualitative researchers in particular, sampling is an exercise of judgement
which balances practical concerns (time, money, access), with the research foci,
and with the degree to which the researcher wants to generalize from the data.
Two importantprinciples to use in making those judgements are:
Try to get a sample that allows you to see things from all relevant perspectives
(what is relevant will be closely related to your research foci).
Keep trying to increase the sample size, or the size of sub-samples that
represent different perspectives, until you are not hearing any new points. For
intensive interviews designed to explore a topic, a sample of eight is often
sufficient, according to McCracken (1988), although survey methods should
then be used to check out the findings.
Although most discussions of generalization are about the conditions under
which researchers can claim that findings from a sample are likely to hold good
for a population, readers also make inferences about the significanceof a research
report for situations and groups that are of interest to them. It is not possible to
prevent readers from generalizing, since that sort of thinking is embedded in the
act of reading itself. It is, though, desirable to draw their attention to the general-
izations that you regard as secure and - or - significant. In addition, Firestone
(1993)argues that it is possible to generalize to theories, as well as to populations;
that is to say that your research may give insights into the strength of theoretical
positions. For example, the study in Box 4.4 provided data that permitted a
re-appraisal of the theory that socialization is a one-way process in which the
social environment shapes the new group member.
Generalizing from cases In qualitative research, especially if it involves one case
study, or is an action research project, the researcher may be reluctant to suggest
that it is wise to generalize to a population. However, that does not mean that no
generalization is possible, since the general is always present in the particular.
So, a study of a single high school (for example Ball, 1981) can show processes
at work that may or may not be present in all high schools but which are likely
to be ;resent in many if not all of them. Readers could be prompted to consider
whether they recognize these processes as ones that are at work in schools they
know, although it is a further research task to explore the extent to which those
processes operate in high schools in general. The considerate researcher identifies
findingqh~teemto have implications for other settings, and, as often happens
with case studies, suggests that it is for the reader to generalize on the basis of
Designing an interview-based study 59
correspondence between the research and the reader's understandings and expe-
riences (Stenhouse, 1980; Walker, 1983).
Similarly, even a single case study can call into question the assumptions of
a theory, as Boyle and Woods' study of one primary headteacher shows (1996).
Some researchers had suggested that increasing and increasingly complex work-
loads, the growth of managerialism and an emphasis on bureaucractictasks had
transformed primary headteachers' roles at the expense of the caring and educa-
tional priorities that used to characterizethejob. The study of a singleheadteacher
showed that these generalizations were too sweeping and was the basis for a
critique of theories that say that changes to social structures determine people's
behaviour. Boyle and Woods invoked social interactionist theory to argue that
people have choices in the ways in which they respond to structural changes.All
of this is based on the study of one headteacher.
Furthermore, if it ispossible to do several case studies on a topic,then a process
of analytical generalization (Firestone, 1993) can be used to test theories more
strenuously. For example, the study of educational markets in three areas, which
centred on case studies of 12 schools, provided a very powerful test of existing
theories about the effects of what the authors called 'public-markets' on
educational practices and has generated new theories (Woods et al., 1998).
All research, then, has some generalizability. Problems come when researchers
try to make generalizations that go beyond what the research design can support.
They also arise when, through failing to answer the 'so what?' question, the
researcher gives readers little help in seeing how they, not the researcher, might
generalize from the findings.
Conclusion
Early thinking about the design of the research governs the sorts of claims that
can be made on the basis of the evidence collected through interviewing. It is
one thing to speculate on the claims that might have been made had the reliability
and validity (or their equivalents) been differently framed, or if another sampling
strategy had been employed, or if the focus were somewhat altered. That is
legitimate and desirable. It is another thing entirely to make claims based on a
research design that is not fit for the purpose.
However, research design, especially in qualitative research, is not just an
armchair, pre-fieldwork matter. It has to be fitted to the research setting, which
frequently curbs the best intentions (as when it is impossible to get a samplewith
the right characteristics), and sometimes offers fresh opportunities (as when an
informant offers access to hitherto unknown sources).A research design that is fit
for the purpose will be one that has emerged as the best response to practical
considerations. Practicalities whittle away the best research intentions. That is
not your failure. It is a fact of research life, as the next chapter argues. Good
researchers explain clearly how their project is worthwhile and representsthe best
response to a problem in particular circumstances which are never, ever, ideal.