You are on page 1of 22

Evaluating Qualitative Research in Social Geography: Establishing 'Rigour' in Interview

Analysis
Author(s): Jamie Baxter and John Eyles
Source: Transactions of the Institute of British Geographers, New Series, Vol. 22, No. 4 (1997),
pp. 505-525
Published by: Blackwell Publishing on behalf of The Royal Geographical Society (with the
Institute of British Geographers)
Stable URL: http://www.jstor.org/stable/623115
Accessed: 12/01/2009 15:19

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=black.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.

Blackwell Publishing and The Royal Geographical Society (with the Institute of British Geographers) are
collaborating with JSTOR to digitize, preserve and extend access to Transactions of the Institute of British
Geographers.

http://www.jstor.org
505

Evaluating qualitative research in social


geography: establishing 'rigour' in
interview analysis
Jamie Baxter and John Eyles
A review of 31 empirical and eighteen substantive papers by qualitative social
geographers mainly using in-depth interviews reveals little explicit reference to the
principle(s) adopted to enhance 'rigour' and to ensure meaningful inference. Given
the modest explicit discussion of evaluative criteria in these papers, a scheme from
evaluation research itself is critically reviewed. A set of evaluation questions derived
from this review and their application to an empirical piece of qualitative work
frame an argument for a general set of criteria rather than rigid rules for assessing
qualitative work. Such criteria can serve as anchor points for qualitative evaluation.

key words methods qualitative evaluation social geography

Department of Geography, McMaster University, Hamilton, Ontario, Canada L8S 4K1


e-mail: baxterjw@mcmail.cis.mcmaster.ca

revised manuscript received 19 August 1996

Introduction of the researcher in shaping the interpretation of


findings have become increasingly influential as
This paper addresses the issue of evaluating the writing in social and cultural geography seeks to
designs and findings of qualitative research in re-present the subjects of inquiry (see Barnes and
social geography through establishing a set of Duncan 1992; Livingstone 1992). And yet, once
questions and criteria to be asked by and of such reflexivity is acknowledged, the important con-
work. It is not our purpose to trace the history of cerns over what and how we present has in some
the use of qualitative methods which may be seen respects overshadowed the problems connected
as part of the humanistic and cultural turns in with obtaining and interpreting interview texts.
geography (see Eyles and Smith 1988; Ley and Three starting points guide our discussions of
Samuels 1978; Smith 1984). We recognize that there evaluation. First, despite problems in the research
is an apparent tension between the creativity of the act (e.g. power relations), academic discourse is
qualitative research process - which implies con- sufficiently different from lay accounts to merit
tingent methods to capture the richness of context- attention in its own right; the analytic views of
dependent sites and situations - and evaluation - 'outsiders' are important for understanding. Sec-
which implies standardized procedures and modes ondly, we recognize that all qualitative researchers
of reporting. Despite the problems implied by this reflect actively on what they do and how they
tension, evaluation is critical if qualitative evidence relate to their subjects. The researcher is herself a
or findings are to gain acceptance outside the 'positioned subject' (Rosaldo 1989) - consciously
community of practitioners, especially given the thinking about what and where he is and what and
widespread recognition of the appropriateness of how she does things. Such reflexivity is a strength
different approaches to investigation (see, for for evaluating qualitative work, allowing a con-
example, Clark et al. 1987). Nevertheless, the scious deliberation of what we do, how we inter-
significance of authorship and the characteristics pret and how we relate to subjects. Thirdly, as
TransInst Br Geogr NS 22 505-525 1997
ISSN 0020-2754 ? Royal Geographical Society (with the Institute of British Geographers) 1997
506 JamieBaxterand JohnEyles
interpretive geographers 'concerned with the quantitative research. Yet we must not forget the
understanding and analysis of meanings in specific general context of rigour around the principles of
contexts' (Eyles 1988a, 2), we set out to learn to academic integrity (see Frost and Stablein 1992;
view the world of individuals or groups as they Knafl 1994; Teich and Frankel 1992) including
themselves see it. In Schwartz and Jacobs' (1979) responsibility and honesty: dimensions of self-
phrase, interpretative geography is in the reality reflection, essential to qualitative research. An
reconstruction business, attempting to develop important dimension of rigour concerns the extent
representations and constructions to describe to which a piece of research is believable and hence
the representations and constructions that take worthy of attention, a notion similar to Lincoln and
place within the social world. Although there are Guba's (1985, 290) trustworthiness: 'findings which
accessible literatures on the methods involved (e.g. are worth paying attention to, worth taking
Burgess 1984; Eyles and Smith 1988), little attention account of'. 'Evaluative criteria' are the basic prin-
has been paid in human and social geography to ciples used to guide the judgement of the integrity
judging what makes sense and what is plausible in or trustworthiness of a study. They are guiding
the findings and the designs on which they are principles rather than rigid standards and can be
based. This is especially true of approaches such as satisfied only in a restricted number of ways. This
in-depth interviewing and the conversational part is particularly important for assessing qualitative
of participant observation which involve talking/ research which has relatively few standardized
conversing with people on a face-to-face basis. procedures for evaluation and whose practitioners
More evaluative work has been undertaken on the are encouraged to be flexible and to utilize novel
analyses of texts (see Barnes and Duncan 1992; methodological and analytical procedures. Indeed,
Cosgrove and Daniels 1988) which may require techniques range from passive observation and
different kinds of evaluation (Eyles 1988a). These personal reflection to intervention, with a common
are not examined in this paper. theme of shared meanings and subjective under-
The purposes of this paper are three-fold: to standing (Smith 1994). Qualitative researchers are
review existing strategies for rigour and the criteria encouraged to allow the research situation to guide
of evaluation currently employed in qualitative research procedures in order that they may gain
research in social geography; to provide a set of access to human experiences. Yet for the research to
criteria and strategies which may be used for evalu- be evaluated, there must be clarity of design and
ation and a set of questions to guide implementa- transparency in the derivation of findings.
tion; and to provide a critique of an example of
qualitative research using the evaluation questions.
It is important to be clear about our terms, in Current strategies for rigour in qualitative
particular 'evaluation', 'rigour' and 'evaluative social geography
criteria'. 'Evaluation' refers to the judgement of
empirical scientific work according to what has To discover how social geographers judge their
traditionally been called 'rigour'. In general terms, qualitative findings, we carried out CD-ROM and
this is a process of critical appraisal to determine manual searches1 of the (English-language) geog-
whether or not a study is worthy of attention. raphy literature for empirical studies and discus-
Evaluation takes place in at least three ways: by sion papers published within the last twelve years
addressing the research methodology, methods which deal with qualitative research. We wanted to
and analysis (plausibility of research design); via discover what qualitative geographers, and espe-
the corroboration or refutation of research findings cially those using in-depth interviews, were doing
(plausibility of accounts); and through the fit with to make their work plausible and deserving of
an existing body of literature or theory (appeal to attention.
interpretive community). While all are essential to Table I shows that, among the 31 empirical
evaluation, it is the first two that are central to this papers identified, the most common ways to
paper, although most appeals to plausibility are ensure rigour are the provision of information on
based on the third which may be expressed cyni- the appropriateness of the methodology, the use of
cally as 'trust me, I agree with you'. 'Rigour' has multiple methods, information on respondent
come to mean the satisfaction of the conventional selection and the presentation of verbatim quota-
criteria of validity, reliability and objectivity within tions. Most of the studies include a rationale for
Table I Strategies for establishing qualitative 'rigour' in geographic work, 1984-95

gImmersionl
aRationalefor bMultiple lInterview eInterview fProceduresfor lengthy 'Ver
Source methodology methods 'Respondent quotations practices analysis fieldwork hRevisits res

Bonnett (1992) x x X
Bridge (1994) xX xX X X
Brown (1995) X X X X X
Cooper (1995) X X X X X X X
Cooper (1994) X X X xX x x
Droogleever Fortuijn and Karsten (1989) X X X X xX
Dyck (1989) X X X X xX xX
Eyles et al. (1993) x
X X x X X X
Eyles and Perri (1993) X X X x X
Eyles and Donovan (1986) xX x
X X x
Fernandez Kelly (1994) x x x X x x x
Gregson and Lowe (1995) x
Harrison and Burgess (1994) x X X
Herod (1991) xX X
Hewitt (1994) X X X
Katz (1991) X X X X X
Leckie (1993) X X X
Mackenzie (1992) xX X X X
McDowell (1994) xX X X X X
McDowell and Court (1994) xX X X X xX
Newton (1995) xX X X X
Porteous (1988) x
X X X X xX X X
Rollinson (1990) xX X X X
Rowe and Wolch (1990) xX X X xX X
Rutherford (1995) Xx X X X X
Shute and Knight (1995) x
X X X xX X
Valentine (1995) X X X
Valentine (1993) xX X X
Valentine (1989) x
X X X X
Wilson (1993) xX X X X x
Winchester and Costello (1995) xX x
Totals 23 22 26 27 13 12 7 5

Notes: 'Qualitative methods are argued to be the most (or only) appropriate way to address the research question; bmore than one method used for stu
of the group(s) of respondents (e.g. number and gender ratio is given); dthe words of the respondent may be read; 'details of how interviews were
schedules and autobiography are provided); fa description of how data were converted/condensed into theoretical constructs is given; git is
(self-proclaimed or stated to exceed one year) develop rapport with respondents and/or enable deep understandings of the research situation; hrevisit
(but sometimes unstated) to clarify meanings and build rapport; 'respondents were contacted to verify interpretations/meanings; jan existing theor
findings, i.e. there is more than reference to the literature; krationale for showing that there is agreement between constructs/interpretations and th
is provided.
508 JamieBaxterand JohnEyles
using a qualitative approach. For example, Brown are important for revealing how meanings are
(1995, 162) claims that ethnography expressed in the respondents' own words rather
than the words of the researcher. However, these
can reveal such geographies[of AIDS]in spite of the
distancesperpetuatedby spatialscience.The linchpin vary considerably in detail in the papers reviewed.
of my argumentis that sociospatialdistancesacrifices For those papers containing quotations, the
geographicknowledge;that ethnographyovercomes number ranges from as few as one (Brown 1995) to
distance. over 100 (Eyles and Donovan 1986); some reports
whilst Droogleever Fortuijn and Karsten (1989, provide quotations of considerable detail with less
investigator commentary3 while others provide
366) suggest shorter quotes with considerable commentary4.
It is evident thatbecauseof the selectioncriteriamen- Why? We are uncertain, although we share the
tionedabovethe researchpopulationis not representa- disquiet of Bryman (1988) and Silverman (1993) on
tive for all persons whose daily activity pattern is the anecdotal nature and assumed representative-
characterizedby combiningtasks. Insteadof making ness of such accounts. While there need not be a
generalizations,the purposeof this study is primarily model for the size and number of quotations, it
to obtainmorein-depthknowledgeaboutthe different is reasonable to expect some discussion of why
ways people combine activities and about the con-
straintsthey meet.Forthe samereasonwe have chosen particular voices are heard and others silenced
a mainlyqualitativemethodof data collection. through the selection of quotes.
There are at least seven other strategies for
But while there are numerous ways to undertake demonstrating rigour (but these are mentioned in
'ethnography' and use 'qualitative methods', they less than half of the 31 articles): details of inter-
are not self-explanatory. Elaboration is needed to view practices, discussions of the procedures for
clarify the ways in which the methodology and analysis, immersion/lengthy fieldwork, revisits to
methods are carried out to achieve things like respondents, verification by respondents, appeals
'overcoming distance' and obtaining 'in-depth to interpretive communities and the provision
knowledge'. of a rationale for verification (validity) of the
All but eight studies involve a combination of findings.
methods and many of these use in-depth inter- Research practices which may enhance rigour
views together with participant observation or include the use of standardized interview guides5
textual analysis. The use of multiple methods and attention to the power relations involved
enables triangulation (see below) but simply using in research interviews.6 The guides allow for inter-
two or three different methods does not necessarily interview comparisons of emergent phenomena,
guarantee more rigorous results. Few investigators while detailing the power relations between the
comment on the reasons for using more than interviewer and interviewee helps account for
one method. Does each method address the same the ways that interview texts are constructed.
or different research questions and what are the However, the reader is often left to interpret for
implications for evaluation? themselves statements such as
Most of the papers include a brief mention of
who was interviewed. However, only in ten We are white and middle class. One of us is thirty-
papers2 do the researchers mention how respond- something,the otherforty-something,and for the pur-
ents were recruited and several do not indicate poses of the interviewswe wore 'professionalsuits'
thatblendedin with the clothesof ourwomenrespond-
how many people were interviewed. Sample size is ents. The respondents, bar one, were white; all but six
relevant for qualitative researchers. It is the basis of were under 40. They were socially adept and articulate.
discovery and description and, while an N of 1 can (McDowell 1994, 665)
be easily justified (Dukes 1965), a rationale should
be provided. Further, a description of respondent The implications should be stated, since similarities
characteristics is critical since experiences crucial to between interviewers and interviewees may, for
the research question may be unnecessarily over- example, foster or stifle interview conversations.
looked. They offer an indication of who is allowed Similar comments may be made about pro-
to speak and, of equal importance, who is not. cedures for analysis. We argue that one of the most
All but four of the papers include (what appear important issues for analysis is the manner in
to be) verbatim respondent quotations. Quotations which interview conversations are constructed into
Evaluatingqualitativeresearchin social geography 509
theoretical concepts. Although most qualitative Such statements reveal what, for the author, are the
reports display verbatim quotations, there is rarely things about the study that make the findings
a discussion of how particular quotations are worthy of attention. Without these clues, the reader
selected for presentation from the range of avail- may judge work unfairly according to criteria
able interview texts. Since there are no standard which may not be as relevant to the research. For
procedures for analysing interview texts (but see example, the lack of interview quotations in Katz's
Willms et al. 1990), it is necessary to elaborate how study does not necessarily detract from the value
data get transformed into concepts/theory(ies) to of the findings if this represents a trade-off for
show readers whose meanings are represented helping to maintain a rapport with the children
and why. and parents with whom she spoke. Such trade-offs,
Lengthy fieldwork or immersion in the research however, should be explicit.
context of interest, evident in seven of papers, is a Appealing to an existing body of literature - the
traditional way of lending credence to the theories 'interpretative community' - our third component
that emerge from qualitative interviews. However, of evaluation, is often the overarching and implicit
while immersion may lead to more sensitivity to strategy for organizing some research projects.
the subtleties of meaning in the group(s) being In such cases, the quotes and interpretations are
studied, it can also threaten the credibility of find- usually forwarded as evidence for the adequacy
ings. For example, the researcher may 'go native', or lack thereof (e.g. Bridge 1994) of concepts and
whereby the study group, and not the community theories formulated elsewhere. For example,
of researchers, becomes the main group with Bonnet (1992), Herod (1991) and Rutherford (1995)
whom the researcher identifies. seem to use literatures on racism, deindustrializa-
One of the main threats to ensuring qualitative tion and restructuring respectively as a good pro-
validity is the misinterpretation of meanings portion of the support for the trustworthiness of
expressed through interview conversations. Re- their findings. Reference to a body of literature is
visits to respondents, undertaken by only five of a necessary but insufficient condition for plausi-
the researchers, is one way to check interpretations. bility. Verification based solely on appeals to con-
Revisits are often used to cover new ground rather ventional wisdom does not necessarily lead to
than to verify what was covered in previous inter- rigorous findings and it may be counterproductive
views and so it is important to report when they to the development of new wisdom (often a goal
are used to confirm the researchers' interpretations. of qualitative research). Valentine (1993), however,
For example, Cooper (1995, 352) points out not only incorporates feminist literature but, by
including both a commentary on her rationale for
WhenI involvedtheseindividualsin the finalselection respondent recruitment and the use of detailed
of illustrativepassages of their discourse,they were
quotations, she goes a long way towards demon-
able to identifywith the contextin which I intendedto
use the data. Similarly,I have occasionallyintroduced strating the rigour of her findings.
The discussion of the various strategies in
themes and questionsinto the interviews,which had
Table I will be revisited and reformulated in the
previouslybeenabsentfromthe adolescents'discourse.
next section of the paper which develops a check-
Throughthese interviews,where themes the adoles-
cents considered unimportantand important were list of general criteria for establishing trustworthi-
discussed, a more comprehensiveanalysis became ness. However, together the 31 papers suggest that
possible. there is insufficient mention of the practices (or
techniques) and principles for establishing rigour.
In only five cases is there explicit mention of the This is not to suggest that each rigorous study
rationale for verification of the findings. Thus Katz should have a complete row of 'Xs' in Table I since
(1991, 495-6) states reasonable argument may be made for focusing on
The populationwas selectedbased on a full enumera- only a few (e.g. Katz 1991). But, for the most part,
tion and survey of the entirevillage. This, along with such arguments are rarely evident. Readers are not
the diversity of the methodology I developed, the given much basis for judging the merits of some of
standardsI maintainedin workingwith the children, these studies since there is scant mention of the
and comparisonsof my resultswith those of the few principles of good qualitative work to which they
existing relevant studies, give me confidencein the hope to appeal. This does not necessarily mean
validityof the informationproduced. that these studies are not rigorous, merely that
510 JamieBaxterand John Eyles
researchers may not be reporting fully on what capturing the other' (ibid., 467, emphasis added).
they are in fact practising. Qualitative papers are But, if there is no 'capture', how can we re-present?
often edited substantially to satisfy journal size Also, if we do not or cannot re-present, are we
limitations and so details of the research process rendered speechless? What the voice says is as
may be forfeited in order to focus attention on the important as recognizing the existence of voice. We
particulars of the concepts/theory which emerge prefer not to 'refuse capturing the other'; rather we
from the data. Eliminating sections on research want to ensure that the other's voice is heard
practices not only leaves readers wondering about alongside that of the researcher. Deliberations on
the trustworthiness of findings but also robs us of epistemology, ontology and practice, while neces-
helpful clues to ways of doing good research. We sary for enriching the findings of qualitative
submit that there is a need - with no parallel research, are not sufficient on their own for evalu-
in quantitative research with its relatively well- ation purposes. Principles for evaluation are
established and agreed strategies7 - for such needed to bridge the gap between the philosophi-
content. cal concerns of a qualitative epistemology and
ontology on the one hand and the practice of
qualitative methods on the other. These issues are
Current evaluation criteria used in revisited in the discussions of disciplined subjec-
qualitative social geography tivity, bracketing and member checking in the next
section.
To assist in discovering more about the principles/ One criterion of evaluation - validity - has
criteria used by qualitative social geographers to received considerable attention. Eyles (1988a, 11)
guide their work, we turn to papers which focus on promotes a meaning-centred version of the 'scien-
qualitative methods and methodology rather than tific' notion of validity whereby 'principles of vali-
research findings. We found eighteen such papers.8 dation are internal to the discourse itself' and
Despite some consideration of a qualitative notion 'interpretations must be justified in terms of the
of validity (see Katz 1991), there has been relatively presented evidence, so much depends on the
little discussion of the principles which do/should coherence of argument and the reason, consistency
guide qualitative research. Calls for more explicit and honesty of the theorist'. Eyles' qualitative
principles to guide qualitative research in geogra- interpretation of the conventional (quantitative)
phy date back at least to Smith (1984) and, more face validity refers to the plausibility of connec-
implicitly, to Ley (1974) and Ley and Samuels tions between data and concepts which appeal to
(1978). In her discussion of the numerous Chicago- common sense and consensus. He distinguishes
school ethnographic studies of the city, Smith the qualitative interpretation of face validity as
(1984, 357) protests 'there are rarely any explicit involving data-to-concept links which not only
"rules of the game" lending structure to the mono- make sense to scientists but also to the lay people
graphs they produced'. She argues that the enthusi- on whose experiences interpretations are based.
asm for providing novel and rich accounts of life in This notion is advanced by engaging Schutz's
the city perhaps overshadowed any concerns about (1962, 44) postulate of adequacy which claims that
evaluation. It may even be argued that 'rules' of a construct
any sort may stifle the creativity which helps to must be understandablefor the actor himself [sic] as
make all types of qualitative work so vibrant.
well as for his fellow-men[sic] in terms of common-
Much attention has been focused recently on the sense interpretations of everyday life. Compliance with
interview process and the need for reflexive con- this postulate warrants the consistency of the con-
sideration of how knowledge is produced through structs of the social scientist with the constructs of
the social relations of the interview: a key element common-sense experience of the social reality.
in the postmodern and new cultural turns.9 For
example, Pile (1991) focuses on the power relations In this way, scientific constructs or meanings
between researcher and researched, and the inter- (second-order typifications) are validated as
subjective nature of the interpretative, interview- adequate by lay constructs (first-order typifica-
based research process. Further, he claims that 'by tions). Similarly, Rose (1982) suggests that it is
assuming that the subject can make as much sense important to distinguish participant concepts
as the researcher, we can refuse the objective of (terms used by the respondents themselves) from
Evaluatingqualitativeresearchin social geography 511
theoretical (researcher-derived) concepts. In order inferences which emphasize the relationship
to satisfy the postulate of adequacy, the former are between phenomena and constructs of these phe-
preferred to the latter where possible. nomena. Secondly, interpretations should appeal to
Jackson (1985, 170) claims that social scientists scientists, those researched and a wide array of lay
are currently wedded to a narrow (quantitative) people. The review of the empirical papers indi-
definition of validity and that 'alternative bases of cates that if there is any attention being paid to
validity are rarely even considered'. In keeping qualitative validity, it is to the first issue rather than
with Eyles' notion of face validity, Jackson (ibid., the second, since only three papers make mention
171, emphasis added) claims, of checking their findings with those researched.
While the discussion of validity will be taken up
The criterionof validity most appropriateto the case
below, it seems that the key to 'validity' is clarity -
study methodconcernsthe logicalrelationshipbetween
characteristicsratherthan their representativenessor making the implicit 'rules' explicit.
typicality.The logicalityof this connectionis in turnto Clarity about evaluation criteria is evoked by
be judged from the adequacywith which the wider Athens (1984, 260, emphasis added) who states
social contextis specified.The extent to which gener- wheneverworks in science are judged, the evaluator,
alizations may be made from case studies depends whetheror not admitted,makeshis or her evaluation
upon the adequacyof the underlyingtheory and the on the basis of some criteria.
whole corpusof relatedknowledgeof which the case
forms a part rather than on the particularinstance Although, as Athens argues, there may not be
itself. much explicit discussion of qualitative evaluative
there are implicit principles used to guide
While it is necessary to pay attention to Jackson's criteria,
and evaluate such work. However, as Merton
concern with the 'logical relationship between
(1968, 71-2) states,
characteristics', he seems to confound validity and
representativeness which, we will argue, should be the sociological analysis of qualitative data often
separate considerations.10 residesin a privateworldof penetratingbut unfathom-
McDowell (1992a) promotes a different ap- able insights and ineffableunderstandings;however,
to in defence of her claim that scienceis public,not private.
proach validity
Schoenberger (1991) (unsuccessfully) tries to apply We argue that a step towards making qualitative
the traditional principle of validity to qualitative analysis more 'public' is to be explicit about the
findings. McDowell maintains that validity is not principles for making this work rigorous. Being
so much a property of interpretations as it is the forthcoming about these criteria will better equip
collective agreement of intended audience(s) that those who do not traditionally work within the
the interpretations are convincing. This seems to qualitative paradigm to judge its approach and
align well with Eyles' promotion of Schutz's pos- findings and, perhaps more importantly, these
tulate of adequacy - that the scientific and lay criteria will be made public for constructive
communities must be convinced by these interpre- scrutiny and debate.
tations - but she perhaps understates the impor-
tance of the 'property of interpretations' and the
role of the interpretive community. We argue that Towards a set of criteria for establishing
interpretations which are rigorous, rather than qualitative rigour
merely compelling/novel, are the most 'convinc-
ing'. Smith (1984) makes similar claims in her Table II sets out Lincoln and Guba's (1985) evalu-
discussion of 'logical inference'-'the process by ation criteria. These are widely cited among quali-
which the analyst draws conclusions about the tative researchers outside geography (see Burns
essential linkage between two or more characteris- 1988; Krefting 1990; Sandelowski 1986) and thus
tics in terms of some systematic explanatory deserve to be critically reviewed here. The four
schema' (Mitchell and Draper 1982, 200). While criteria in the table extend the somewhat limited
she does not refer to this as 'validity', the connec- discussion of evaluative criteria in social geogra-
tion between data (characteristics) and theory phy which seems to focus mainly on only one:
(explanatory schema) is evident. validity/credibility. 'Strategies' for enhancing
For us, there are two common threads in these rigour are also listed against each criterion. The
discussions. First, it is important to use logical table allows the reader to connect the criteria to
512 Jamie Baxter and John Eyles
Table II Criteria for evaluating qualitative research

Criteria Definition Assumptions Strategies/practicesto satisfy criteria

Credibility Authentic representations of Multiple realities Purposeful sampling


experience Causes not distinguishable from Disciplined subjectivity/bracketing
effects Prolonged engagement
Empathetic researcher Persistent observation
Researcher as instrument Triangulation
Emphasis of the research endeavour Peer debriefing
Negative case analysis
Referential adequacy
Member checking

Transferability Fit within contexts outside Time and context-bound experiences Purposeful sampling
the study situation Not responsibility of 'sending' Thick description
researcher
Provision of information for
'receiving' researcher
Dependability Minimization of Researcher as instrument Low-inference descriptors,
idiosyncrasies in Consistency in interpretation (same mechanically recorded data
interpretation phenomena always matched with the Multiple researchers
Variability tracked to same constructs) Participant researchers
identifiable sources Multiple realities Peer examination
Idiosyncrasy of behaviour and Triangulation, inquiry audit
context

Confirmability Extent to which biases, Biases, motivations, interests or Audit trail products
motivations, interests or perspectives of the inquirer can Thick description of the audit process
perspectives of the inquirer influence interpretation Autobiography
influence interpretations Focus on investigator and Journal/notebook
interpretations

Source:Lincoln and Guba (1985)

philosophical concerns about epistemology, ontol- the concepts which the social scientist uses to
ogy and methodology as well as to the research recreate and simplify them through interpretation.
process itself. Note that plausibility of design and The parallels with some of the recent discussions
accounts are often dealt with simultaneously since surrounding 'validity' within geography are evi-
many strategies employ overlapping criteria (e.g. dent. In particular, Eyles' (1988a) and McDowell's
triangulation). This is similar to quantitative (1992a, 1992b) concern that interpretations be cred-
research in that a validity issue is also one of ible both to those within the research endeavour
reliability and vice versa. Yet in qualitative research (including investigators and respondents) and to
there are fewer conventional procedures and more the wider scientific and lay communities is com-
resourceful ones. In fact, qualitative researchers mensurate with the notion that 'those having the
tend to applaud new and innovative ways for experience would recognize it immediately'. It is
making work rigorous. apparent that research practices should involve
strategies for returning interpretations to respond-
Credibility ents for commentary (and perhaps revision).
The most important principle for guiding qualita- Credibility is based on the assumption that there
tive studies is the notion of credibility. This may be is no single reality but rather multiple realities,
defined as the degree to which a description of mentally constructed by ourselves. It is not confir-
human experience is such that those having the mation that is required from respondents as much
experience would recognize it immediately and as a commentary from them on the plausibility of
those outside the experience can understand it the interpretations offered. There may be disagree-
(Lincoln and Guba 1985). Credibility refers to the ments between respondents and scientists (Borland
connection between the experiences of groups and 1991) or the former may deny, repress, falsify or
Evaluatingqualitativeresearchin social geography 513
otherwise give only partial accounts of their reported (see King et al. 1994). Detachment is often
experiences (Miles and Crush 1993). Since outsider seen as being more problematic than observer bias
accounts may well differ from insider accounts, and may itself lead to indifferent responses
Eyles and Perri (1993) report unease with some of (LeCompte and Goetz 1982).
their comments on the Italian-Canadian family Patton (1990) identifies sixteen purposeful sam-
studied by them, although the 'insiders' were pling strategies. Not all are equally useful. For
asked to confirm that (academic) accounts made example, 'stratified purposeful sampling' is consid-
sense. Hence, the goal of the researcher is to ered useful in ensuring that all sub-groups within a
represent adequately the realities of groups in such research setting are given a voice so that compari-
a way that not only does the scientific community sons can be used to construct commonalities and
but also the people who constructed the reality in differences in interpretations across groups. On the
the first place understand the (re)construction of other hand, 'convenience sampling' - interviewing
that reality. The problem of participants providing only those who are easy to access (e.g. all people in
only partial accounts of their experiences and one amiable community group) - is problematic
meanings is more problematic as it is often difficult since easy contacts are not necessarily the most
for the researcher to detect gaps between what informative contacts and such people may have
is reported and what 'actually occurred'. Since only limited capacity to comment on issues rel-
realities are assumed to be multiple and flexible, it evant to the research question. Further,Wax (1971)
is also assumed that there is no way to distinguish urges against choosing participants merely on the
between things like 'causes', 'effects' and 'truth' basis of personal preference and to select respond-
within the social world. This serves to distin- ents as broadly as possible. In this sense, the
guish credibility from its quantitative counterpart: qualitative researcher will often try to establish the
internal validity. range of possible respondents (an ongoing process)
In order to enhance credibility, researchers focus by selecting proportionally from all groups/types.
on respondent selection procedures, interview Thus sampling is important. At the very least,
practices and strategies for analysis. Patton (1990) the essential characteristics of qualitative sampling
identifies two major strategies for recruiting strategies should be considered, even if the sample
respondents: random sampling and purposeful design - although preconceived (at least enough to
sampling. Random sampling, based on statistical answer the question 'where and with whom do I
representativeness, is used only rarely in qualita- start?')- is flexible and evolves as the study
tive research (e.g. Eyles and Donovan 1990) but progresses; who and what comes next depends on
may be employed if there is no conceptual reason who and what came before. The sample is selected
for directing attention to particular informants at serially or adjusted continuously or 'focused' by
the outset of the research process. Purposeful sam- the concurrent development of theory. Selection
pling - the strategy used most often by qualitative continues to a point of thematic saturation and
researchers - stresses the search for 'information- sampling includes a search for negative cases in
rich cases'. Such respondents are at ease and talk order to give developing theory greater breadth
freely with the researcher such that a great deal can and strength (Kuzel 1992; Lincoln and Guba 1985).
be learned about the research question. Sample size As indicated in the previous section, the prac-
is determined largely by the need to involve as tices of interviewing have probably received the
many experiences as possible in the development most attention in recent papers on qualitative
of a conceptual framework/theory. Recruitment methods in social geography. As the researcher is
then often occurs until 'redundancy' or 'satura- the active instrument in qualitative research, inter-
tion'; that is, until no new themes or constructs views raise questions about the role of the 'author'
emerge. Thus credibility need not be threatened and how her/his characteristics become a forma-
by low sample sizes. Yet there is still a requirement tive influence upon them. More generally, it is the
for qualitative researchers to be mindful of self- interviewer's skill at developing a rapport with
selection 'biases' which may come from certain respondents and his/her ability to use this to
strategies like snowball-sampling. While bias may develop information-rich conversations which
be used to advantage, the skewing of sample shape the data gathered. But, beyond this, power
characteristics and types of questions answered relations and the presentation of self in the
and not answered needs to be recognized and interview are crucial determinants. Age, gender,
514 JamieBaxterand JohnEyles
ethnicity and other outward appearances can theories. Source triangulation, the most common of
potentially affect how respondents react in the the four, refers to the use of more than one report
interview (Pile 1991). Vigilance over these issues from a data set to corroborate a construct. One of
has long been advocated by survey researchers the most frequently used (often implicit) forms of
(Moser and Kalton 1981) and field workers source triangulation is the use of quotations from
(Burgess 1984). The practice of being mindful of several different respondents (e.g. Eyles and
one's own ethnocentricity and biases has been Donovan 1986). Method triangulation involves cor-
called 'disciplined subjectivity' (Erickson 1973) and roboration of constructs based on information
'bracketing' (Lincoln and Guba 1985). Further po- derived from at least two different methods. Vari-
tential interviewer effects arise from the researcher's ous combinations of qualitative methods are often
socio-demographic characteristics in relation to used and, increasingly, qualitative and quantitative
those of the respondents (a relationship usually methods are combined in the same study (e.g.
represented through a biographical sketch)." Harrison and Burgess 1994; Leckie 1993). To
Many of the strategies for conducting credible counter methodological eclecticism, Fielding and
analyses are closely wedded to interview practices. Fielding (1986) suggest that the methods chosen
In particular, both share the problem of dealing should make sense within one theoretical perspec-
with the implications for credibility of analyses of tive. Investigator triangulation involves multiple
social relations between the researcher and the investigators investigating the same phenomenon
researched. Lincoln and Guba (ibid.) propose three and comparing results. This may work well if
'preventative' techniques for dealing with this: investigators are part of the 'same team', such that
prolonged engagement, persistent observation and they are, in effect, looking at the same phenomena
triangulation. Prolonged engagement involves with similar perspectives, e.g. graduate students
spending sufficient time in the field to build trust and their supervisor(s). Problems may arise, how-
and rapport with the respondents, to learn the ever, when the subtle nuances of the interview (e.g.
'culture' of the relevant group(s) and to investigate body gestures) are known only to the researcher
for possible misinformation/distortions intro- who conducted the interview and helped construct
duced by self or respondents. It may result, how- the interview text. Variations in interpretation
ever, in the problem of going native discussed resulting from differential access to what happened
earlier (see also Vidich et al. 1964), although the in the interview can, however, be resolved by
so-called 'contamination' of going native can pro- negotiation between researchers. Theoretical trian-
vide the basis of vibrant accounts (see Plummer gulation is usually reported as epistemologically
1983). unsound and counter to appeals to the interpretive
Persistent observation is complementary to pro- community.
longed engagement in the sense that the latter There are also analytical techniques which may
provides scope while the former provides depth. be used after data have been collected. We high-
While prolonged engagement involves being light four: peer debriefing, negative case analysis,
aware of the 'multiple influences and mutual referential adequacy and member checking. Peer
shapers and contextual factors' (Lincoln and Guba debriefing involves exposing data and interpreta-
1985, 304), persistent observation involves focusing tions to a respected colleague in order to point up
on the 'things that count' in terms of the research possible sources of misinterpretation and the 'sup-
questions being asked. It is closely linked to pur- pression' of themes or voices that do not 'fit' the
posive sampling since both may involve seeking 'storyline'. There are dangers, however, that 'un-
out a diversity of (informative) respondents to resolvable' sources of disagreement may arise or
ensure that relevant experiences are not omitted. that one person may defer to the other on the basis
Triangulation is one of the most powerful tech- of unequal power/authority relations (Risteen
niques for strengthening credibility. It is based Hasselkus 1991). Negative case analysis involves a
on convergence: when multiple sources provide largely inductive process of constantly revising an
similar findings their credibility is considerably hypothesis by comparing it with all interview texts
strengthened (Knafl and Breitmayer 1989; Krefting until it accounts for all known cases. This process
1990). Denzin (1978) suggests that there are four serves to explore numerous dimensions of a theme
major types of triangulation involving the use in order to make it robust and is particularly
of multiple sources, methods, investigators and recommended for the development of constructs
Evaluatingqualitativeresearchin social geography 515
intended for central theoretical status (Kidder (Baxter et al. 1992). These 'contracts' are now often
1981). Lincoln and Guba (1985) assert that account- demanded by ethics committees that sanction
ing for all negative cases is 'too rigid' a criterion to research on human subjects.
satisfy, since some cases may be so obscure that Ethics aside, a key substantive question remains.
they are of little conceptual/theoretical conse- How much analytical refinement is appropriate for
quence. Referential adequacy is the practice of the information which is to be submitted for
verifying the constructs developed through an respondent review? For example, the information
interpretation of the bulk of the data by the subse- may be interpretations of single interviews (low
quent analysis of a selection of data which has been level of refinement) or interpretations of multiple
archived (i.e. not part of the original analysis). interviews (high level of refinement). While
Member checking is arguably one of the most respondents may feel that they are 'qualified' to
important strategies for enhancing credibility since comment only on their own transcripts, the higher-
it involves checking the adequacy of analytic level interpretations are often more meaningful for
categories/constructs/hypotheses with members theory development. Porteous (1988) used member
of the group(s) from which the data were obtained. checking by giving each respondent a copy of
But Hammersley (1992, 65) is cautious about this chapters with their own quotes highlighted. In this
strategy: way, not only does the author verify the highest
level of interpretation, he shows each respondent
Toassumethatrespondentscan validateor even falsify how their comments fit into the analysis.
accountsin some definitiveway is to forge the social
characterof the relationshipbetween researcherand
participantsand to assume that they have privileged Transferability
access to the truth. Neither of these assumptionsis Transferability refers to the degree to which find-
sustainable. ings fit within contexts outside the study. Elements
of research produced in one context may be trans-
Whilst respondents do not have privileged access ferred to others. However, such a strategy is clearly
to the truth, they do have privileged access to their dependent on the correspondence between 'send-
own opinions and meanings. It is the adequate ing' and 'receiving'12 contexts. It is analogous - in
representation of these that should be the goal for principle at least - to the more familiar notion of
member checking. The appeal of this strategy is generalizability or external validity. The qualitative
that it is implicit in the qualitative notions of researcher is rarely as concerned about transfer-
validity (in particular, Schutz's (1962) postulate of ability as she is about the credibility of the findings.
adequacy), whereby interpretations are more cred- Within the qualitative paradigm, experiences and
ible if they are meaningful for both academia and meanings are assumed to be largely bound to the
the group studied. There is no definitive procedure time, people and setting of the particular study.
for returning information to the respondents for Most qualitative researchers focus on one context
such a credibility check and there is also the risk of in order to discover, describe, hypothesize and
what Borland (1991) calls 'interpretive [sic] con- otherwise reconstruct the things that are meaning-
flict', whereby the participants will largely disagree ful to the people within it. It is assumed, therefore,
with the researcher's interpretation of the inter- that statements will be idiographic rather than
view text. We agree with Borland who suggests nomothetic and it is perhaps not surprising that
that checking should be done in the spirit of an none of the empirical papers reviewed here makes
'exchange of ideas' and, if we do not check our any claims about the transferability (or generaliz-
interpretations with participants, we are in danger ability) of findings.13
of merely fitting data into the preconceived This does not mean, however, that qualitative
theories/frameworks with which we are comfort- research need pertain only to the specific cases
able. There is also an ethical imperative to let under investigation. Meanings are often shared by
participants know how their interviews are being many individuals. While a qualitative study may
used. It is increasingly the case that 'contracts' reconstruct meanings as they apply to the experi-
with groups and communities that agree to be ences of only a very small sub-group, it is possible
researched include feedback of some sort (if that these experiences may be common to a larger
wanted). Thus reporting back occurred in research group. However, there is nothing built into most
on the effects of the Hagersville (Ontario) tyre fire qualitative research designs to assess the degree to
516 JamieBaxterand JohnEyles
which findings are transferable beyond the single researched. Since qualitative researchers accept the
case. Lieberson (1992) points to the problems of, inevitability of changes in 'reality', they tend to
and relatively rare circumstances in which, 'gener- focus rather on design/researcher-induced changes.
alized' conclusions may be drawn from case Yet, while quantitative researchers focus on the
studies. Since most qualitative researchers adhere survey items themselves (e.g. wordings) and the
to the notion of idiographic knowledge, claims manner in which they are administered, qualitative
about transferability are rarely made (at least researchers focus on the interpretations and their
convincingly so) by the 'sending' researcher, who consistency from one interview transcript to an-
may enhance transferability by careful ascription other. Thus dependability refers to the plausibility
of findings to specific sub-groups in the study. of accounts; reliability to the plausibility of design.
LeCompte and Goetz (1982) also argue that the So, although both types of research are concerned
multi-site study is one strategy for increasing the with plausibility, it is seen and treated differently.
likelihood that findings will transfer. There are at least two phenomena which may
It is, however, important to recognize the quali- threaten the dependability of interpretations in
tative researcher's responsibility with respect to qualitative research: poorly defined analytical con-
transferability. There is an onus on qualitative structs and premises (LeCompte and Goetz 1982),
researchers to provide data which allow for trans- and premature closure (Lincoln and Guba 1985).
ferability, rather than necessarily demonstrating its When analytical constructs are poorly delineated,
existence, by providing the database on which such they may be subject to variable interpretation by
judgements may be made by others. The original both researchers and those being researched. For
researcher must describe the study context as com- example, the concepts of 'class' or 'culture' are
pletely as possible because, at root, transferability known to have numerous definitions, all with their
involves the degree to which constructs are mean- own attendant assumptions and implications. Pre-
ingful to other groups (as yet unstudied or not yet mature closure occurs when the researcher final-
compared with the original group). Detailed, thick izes analytical constructs sooner than the available
description (Geertz 1973) - as a methodological as data warrant.
well as interpretative strategy - of how constructs/ LeCompte and Goetz (1982) suggest five strate-
hypotheses are developed and what they mean, gies for guarding against threats to dependability:
will be of use to the researcher or layperson who low-inference descriptors, mechanically recorded
wishes to determine the degree to which they may data, multiple researchers, participant researchers
be transferred to other contexts. and peer examination. Analogous to some of these
are triangulation and the inquiry audit (Lincoln
Dependability and Guba 1985). Low-inference descriptors and
Dependability is the degree to which it is possible mechanically recorded data are two interrelated
to deal with instability/idiosyncrasy and design- methods, whereby the degree to which the inter-
induced change. Kirk and Miller (1986) see this pretations 'agree' with the data are authenticated
criterion as being as important as credibility if by others (Pelto and Pelto 1978; Schatzman and
qualitative research findings are to be taken seri- Strauss 1973). The most common types of low-
ously. We assert that dependability includes the inference descriptors are fieldnotes and audio
consistency with which the same constructs may recordings which include verbatim accounts and
be matched with the same phenomena over space narratives of behaviours, activities and events.
and time (see LeCompte and Goetz 1982) but is These may be used by other researchers to compare
largely concerned with documenting the research with the interpretations of the original researcher.
context. In this way, there are similarities with Multiple researchers (investigator triangulation),
reliability, although the latter involves standards of participant researchers (a form of member check-
stability, consistency and predictability, whereby ing) and peer examination (peer debriefing) are
multiple applications of the same research instru- all methods, already discussed, of introducing
ment are expected to yield similar findings alternative perspectives in data analysis prior to
(Streiner and Norman 1989). Few qualitative re- finalizing the set of theoretical constructs. None of
searchers are willing to concede the 'unreliability' the empirical papers gives any indication of the use
of a study based on changes that have occurred of such strategies to check for the dependability of
'naturally' within the study group(s) being construct-to-data matching.
Evaluatingqualitativeresearchin social geography 517
Some may argue that the similarity in strategies investigator and the interpretations. Conventional
for improving dependability and credibility objectivity is usually associated with the following
implies that they should be collapsed into one assumptions: there is a single, largely unchanging
criterion. While the strategies to satisfy both are reality; good data may reflect only that reality;
similar, the criteria themselves are quite different. and, when the researcher disturbs the data or the
Credibility refers to the accurate representation of reality, objectivity is compromised. By contrast,
experiences while dependability focuses attention confirmability may be defined as
on the researcher-as-instrument and the degree to
which interpretation is made in a consistent man- the degree to which findings are determinedby the
ner. In support of this argument against merging respondentsand conditionsof the inquiryand not by
the biases,motivations,interestsor perspectivesof the
dependability and credibility, Lincoln and Guba inquirer.(Lincolnand Guba1985,290)
(1985, 317) claim that
Scriven (1971, 95-6) adds that, to be objective, data
Since there can be no validity without reliability(and
must be 'reliable, factual, confirmable or con-
thus no credibilitywithout dependability),a demon-
strationof the formeris sufficientto establishthe latter. firmed and so forth'. By incorporating concerns
If it is possibleusing the techniquesoutlinedin relation about the character of the data, confirmability is
to credibilityto show that a study has that quality,it more broadly based than the principle(s) of objec-
oughtnot to be necessaryto demonstratedependability tivity which focus solely on the accountability of
separately.But,while this argumenthas merit,it is also the inquirer. Thus qualitative researchers are
very weak. It may serve to establishdependabilityin expected to account for their interests and
practice,but does not deal with it in principle.A strong motivations by showing how they have affected
solutionmust deal with dependabilitydirectly.
interpretations.
The inquiry audit combines elements of thick There are similarities in the techniques used for
description with those of peer examination and is enhancing rigour between confirmability and cred-
analogous to a fiscal audit (Halpern 1983). The ibility, transferability and dependability. Confirm-
auditor maintains checks on the status of the ability highlights the research audit which includes
research to ensure that appropriate decisions are audit trail products such as raw data, data reduc-
made along the way. It is intended to produce a tion and analysis products, data reconstruction and
detailed account of how the research was done synthesis products, process notes, materials relat-
and, like the peer examiner, the auditor should ing to intentions and dispositions, and instrument
be someone intimately familiar with qualitative development information. Confirmability is also
research and/or the topic area. The inquiry audit is an account of the audit process, including how
used to ensure that the appropriate decisions are decisions were made regarding the determination
made a priori. Unlike the peer examiner who of credibility, transferability and dependability (see
enters near the end of the inquiry, the auditor is Halpern 1983). It is questionable how often this
expected to be involved at the outset to look for does or can occur in practice due to the rather
and evaluate an 'audit trail' of how and why elaborate and apparently strict audit criteria
various decisions are made regarding such things suggested by Halpern (ibid.). At the very least,
as respondent selection, methods used and data qualitative researchers need to ask some basic
interpretation techniques. Since the auditor is in a questions of all their work so as to assist in evalu-
position to advise on these decisions, he can help to ating design and findings. Rose (1982), for
keep idiosyncrasy in the design and interpretation example, suggests eight (see Table III). A detailed
to a minimum. To a certain extent, the graduate journal with notes on findings, problems and
student-professor supervisory relationship func- interpretations (see Spradley 1979) is one way of
tions as a convenient, often implicit form of providing material for an audit. The geography
auditee-auditor research relationship, albeit much literature reviewed in this paper gives an indica-
less formally than Halpern suggests. tion of how infrequently audits, however scant in
detail, are included in published reports. Rarely is
there mention of the biases, motivations and inter-
Confirmability ests of the researcher in relation to the questions
Confirmability, similar to the conventional notion asked and the decisions made throughout the
of objectivity, focuses attention on both the research process. One exception is Cornwell (1988),
518 Jamie Baxter and John Eyles
Table III Checklist for evaluating qualitative interview research

Question Elaboration/examples Evaluationof Eyles et al. (1993)

1 What was the natural Originalpurpose(s)of the research Stated- researchresponsesto a tyre fire
history of the research? Rationalefor methodology Stated- to study residentresponsesin context

How researchdeveloped over time Little discussion


Fieldworkrelations Little discussion
2 Whatdata were collected Methodof note-keeping; None apparent
and by what methods? Methodof tape-recording Five interviewers, over one month, three months
after event (fire)
Debriefing and training of interviewers
Verbatim transcription
3 How was the sampling Explicit delineation of sample frame 1300-1400 sample frame, 173 who talked with
done? (working universe) community outreach programme, 66 who agreed
to follow-up=43 respondents
Random or purposeful? Purposeful - opportunistic
Rationale for type of sampling used Stated - mutual trust
4 How was the data analysis Procedures for summarizing and presenting Not apparent (implicit - inter-interview themes)
done? data
How data were selected for presentation Not apparent
5 Whatresultsare presented? Description of researcher's objective for Descriptions of how the findings fit in with
results presentation (e.g. theory-building or existing theory and outline policy implications
description)
Differentiation of data-derived as opposed to Relate findings to literatures on environmental
pre-existing constructs stress and risk perception
Differentiation of participant concepts as Participant (quotations), data derived (researcher
opposed to theoretical (researcher-derived) commentary), theoretical (Figure 1 in the paper)
constructs
6 How credible and Details of the relationship(s) between the Analytical validation through investigator
dependable are the data and constructs/concepts derived from triangulation
data-construct links? data (e.g. member checking)
7 How credible is the Specification of the relationship between Yes, see note 5
theory / hypothesis? constructs/concepts and theory/hypotheses Relate literature to this case study

8 How transferable are the Recognition of the limits imposed by the Acknowledged case study
findings? sampling strategy Depends largely on credibility of research
constructs

Source:Rose (1982);and Lincolnand Guba (1985)

who argues that there are many false trails in of our own papers (Eyles et al. 1993) concerning
exploratory qualitative research, whilst the 'warts- the social construction of risk in the community
and-all' of research are often told in autobiogra- surrounding the Hagersville tyre fire in February
phies of the research process.14 1990. We do not choose one of our own works in
order to paint it in a favourable light (or, con-
versely, as an example of what not to do) but
A case study of rigour and trustworthiness because we are in the unique position of being able
to comment on the editorial exchanges between the
In our review of qualitative research in social journal editor and the authors which influenced
geography, we indicate that there is inadequate what was included in the published paper.
mention of the practices and criteria for producing While the purpose of the research is stated as
trustworthy results. But can the criteria in Table III 'identifying the effects of the fire, particularly
be used reasonably and fairly to evaluate such through the ways in which people responded to
work? We will address this issue by reviewing one and coped with the event' (ibid., 283), and the
Evaluatingqualitativeresearchin social geography 519
rationale for the methodology as 'a qualitative traumatizing circumstances, and the researchers.
research design using depth interviews to allow us For this reason, a purposeful/opportunistic sam-
to examine the types and forms of residents' pling strategy was used which took advantage of
responses to the tire fire in their societal context' existing community remediation efforts, instead
(ibid., 282), there is little else mentioned of the of a more 'rigorous' sampling strategy. The inter-
natural history of the research (see Table III). There views were conducted as a follow-up to visits by
is a discussion of how the study fits in with the public health nurses. The following excerpt details
history of community remediation efforts in that how the sample was achieved:
the sample of 43 respondents was drawn from
those who were visited by public health nurses As a result, our sample was doubly selected, being
from the Community Outreach Programme (COP). drawn from those families who talked to the COP
(N=173)and agreedto a follow-up(N=66).Wetriedto
However, there is not much discussion of what contactall 66. Some had moved and some were away.
happened after that in terms of the fieldwork We talked to all those we could contact (N=43) and
relations between the researchers/interviewers their stories form the basis of our account of the
and respondents. There is passing mention of inter- impacts of the fire.
viewer debriefing sessions intended for 'identify-
ing issues arising from the interviews which are There is, clearly, the potential to miss the accounts
not self evident from the transcripts' (ibid., 283) of certain groups. This is critical for a study claim-
but there are no examples of such issues, which ing to provide stories which represent most groups
may well have affected the manner in which the within the community. Do those who did not have
transcripts were constructed. time for a follow-up comprise a group which has a
There is substantial information about the data different understanding of the event from those
and methods of collection. The interviews were who did? The authors claim that the double sam-
conducted over a four-week period, three months pling is not a threat to providing representative
after the fire, by five interviewers who taped each stories (ibid.):
conversation. The interviews were transcribed for
These 43 provide a cross-section of people within the
thematic analysis. However, the implications of
evacuation zone, some being upwind of the fire, some
using five interviewers (each with her/his own in the red zone (closest to the site), some on the
interviewing 'style' which influences what and neighbouring concession lines, and some downwind
how things get said) are not discussed in much but further away.
detail other than the mention of the existence of
interviewer training and debriefing sessions. These Yet there is no evidence that the people who were
sessions seem to have been put in place to ensure not interviewed have stories of the fire which are
consistency in the way residents were interviewed similar to those of the respondents. Those who did
but it is not clear how this was (or was not) not agree to a COP visit or were not visited at all by
accomplished. Consistency is also addressed the COP (N=1200), did not agree to a follow-up
through the use of an interview checklist: (N=107) interview or could not be followed-up by
one of the five research assistants (N=23) may
Depth interviewswere guided by a checklistof topics
to be coveredwith all respondentswhile allowingthem speak differently of the fire. It is important to
considerablefreedomto describetheirexperienceand recognize that, while this study provides consider-
storiesin theirown terms.(ibid.) able detail about sampling when compared with
many of the 30 other studies, more information
Although it is not stated explicitly, covering the about who was and who was not interviewed
same (minimum) of topics facilitates analyses that would be useful for revealing where gaps may
may compare themes across interview texts. While exist. It is impractical to expect that all stories can
the checklist itself is not provided - an editorial be told in any given research situation but readers
decision - it is apparently brief enough to allow should know from whom they are and are not
respondents the power to direct the conversation. hearing.
The paper also includes considerable detail of As is the case with many qualitative papers in
sampling. One of the main concerns in conducting social geography, there is scant detail about how
the study was the establishment of 'mutual trust' the data are transformed, in this case from 43
(ibid.)between the residents, who faced potentially interviews to a framework for understanding
520 JamieBaxterand JohnEyles
anxiety, uncertainty and risk in the Hagersville the process of triangulation/interpretation. There
community (ibid., Figure 1). The data are con- is also an almost complete lack of reference to the
densed to eleven quotations from only eight of the relationships between the interviewers and inter-
original respondents, with no indication of how viewees which, as argued above, can have pro-
this reduction occurred. What is offered, however, found implications for what becomes the 'data' in
is the following statement regarding the validity of the first place.
the constructs in the 'framework': The issue of transferability in the paper is
The analyses and interpretationsof the stories have shaped by the fact that it was a case study of the
risk, anxiety, uncertainty and coping experiences of
been undertakenand discussedby severalmembersof
the community surrounding the Hagersville tyre
the researchteam so that the end-productis a set of
themesvalidatedby multipleanalysts.(ibid.) fire. The authors admit that the results are context-
specific. In fact, that is the point of their endeavour.
Although such triangulation by multiple analysts The limits to transferability are also apparent in
may lead to credible interpretations, there is no the explicit statements regarding sampling. The
mention of how these analysts managed the data. discussion of the policy implications, however,
To return to an issue raised earlier, the editorial/ suggests that the authors consider that some issues
peer review process resulted in the elimination of a should be considered transferable to other events
section of the original paper which specified the which require similar emergency responses:
strategies for analysing/condensing the transcript TheHagersvillecasestudypointsto a need for authori-
data. As is often the case, the authors compromised
to recognizeand caterto
ties, in difficultcircumstances,
by directing the reader to a companion methodo- the ways in which lay publicsact, think,and talk.The
logical piece in another journal which explains the use of local officialswith such knowledgemay be an
data management and analysis in considerable answer. (Eyles et al. 1993, 288-9)
detail (see also Porteous 1988).
The results appear in the forms of quotations, There is no audit trail of the research process as
author commentary concerning the quotations and suggested by Lincoln and Guba (1981), or excerpts
the 'framework' outlined in Figure 1 of the paper from field diaries, for ascertaining how research
which shows how theoretical concepts are linked. decisions were made. There is, however, consider-
The authors do differentiate lay concepts (the quo- able detail in the stories and descriptions them-
tations), data-derived constructs (the commentary) selves to determine the similarity between the
and those constructs which already exist in the Hagersville 'event' and other environmental
literature on environmental stress and risk percep- events, so revealing how credibility and transfer-
tion (various parts of the framework). Further ability may become intertwined. Yet all that is
distinctions are made through a review of the available to determine the transferability of the
literature at the beginning of the paper and a findings is the credibility of the constructs/theories
subsequent revisit to this literature at the end of presented.
the paper insofar as it relates to the constructs for This review of Eyles et al. (1993) indicates that
the Hagersville case. These reviews reveal how the the eight questions in Table III are general enough
researchers relate constructs developed from the to provide a reasonable assessment of qualitative
stories in the study to a broader literature and how work which uses interviews as the main method of
this needed to be adapted in order to understand data collection. It serves not only as a guide for
the Hagersville context. For example, the issues what to look for but reveals where gaps exist in
related to evacuation, latency, coping strategies reporting information necessary for ascertaining
and financial resources 'loomed large' in the after- rigour. The table is not intended to provide specific
math of the fire but are not as prominent in the standardized rules for what should be done to
literature which were reviewed. produce trustworthy findings. It indicates the basic
The credibility of both data-to-construct and information requirements for appraisal. The four
construct-to-theory links is implicitly ensured by criteria and the strategies outlined in Table II
the investigator triangulation mentioned above provide more detailed ways of assessing rigour
and appeals to the literature of environmental risk and we suggest that they should be incorporated
and coping. The paper suffers from a lack of into the research process as a basis for answering
convincing elaboration/discussion, particularly of the questions posed in Table III.
Evaluatingqualitativeresearchin social geography 521

Summary and conclusions considerable differences between the criteria them-


selves, the strategies used to strengthen qualitative
Until recently, qualitative researchers have tended rigour are quite different from their quantitative
to focus more on what criteria should not be used counterparts. It is important to distinguish be-
to evaluate their work - the standards used to tween principles which may be similar and prac-
judge positivistic-quantitative work - and less on tices (see Table II) which are quite different when
what they should be looking for to determine the evaluating rigour for qualitative as opposed to
rigour of qualitative research. In social geography, quantitative work.
this is reflected in both empirical and substantive There remain, however, objections to establish-
work. The sample of qualitative geographic ing qualitative principles for evaluating rigour.
research reviewed here indicates scarce explicit Thus Risteen Hasselkus' (1991, 3-4) claims that
mention of the principle(s) which have guided its
We seem to be circumscribing qualitative research with
concern for rigour and ensure meaningful infer- an orthodoxy of rules to which it must conform. In my
ence. But its strengths should also be recognized in view this bend toward a dogma of qualitative inquiry
the use of multiple methods, numerous detailed can potentially smother the creative elegance of such
quotations, discussions of validity and appeals research. Further, I do not believe that the writers who
to recognized bodies of literature. Nonetheless, have brought forward these guidelines and criteria for
researchers need to be more explicit about the judging qualitative research ever intended for those
research process including the rationale(s) for, guidelines to be incorporated as mandatory rules and
among other things, respondent selection, key regulations in the qualitative research process.
changes in research direction and analytical pro- We agree that there should be no mandatory rules
cedures. This may prove problematic if journal but there should be criteria that enable a judge-
editors continue to emphasize results at the ment to be made concerning honesty, integrity and
expense of equally important accounts of strategies plausibility of design and accounts. These criteria/
for maintaining rigour. principles are intentionally general and, conse-
While it may be expected that the criteria which quently, may be satisfied in numerous ways.
these strategies are intended to address may be Indeed, the list of strategies for strengthening them
found in the substantive literature on qualitative is quite extensive but should by no means be
methods, this tends to focus on distinguishing the considered all-encompassing. It is when limits are
epistemological and ontological foundations of placed on the types of strategies that may be used
qualitative from quantitative research as well as to achieve rigour that there is a danger of a 'dogma
being reflexive about qualitative interview prac- of qualitative inquiry'. But questioning how things
tices. There is also appeal, often uncritical, to inter- are done - an essential component of self-reflection
pretive communities. Less attention has been paid - allows qualitative research to demonstrate the
to why some theories seem to be more appealing relevance of the single case (credibility) and to
than others. Why is their time right? For example, move beyond it (transferability) with a degree of
Lamont (1984) gives a fascinating assessment of the certainty (dependability and confirmability). Con-
plausibility of Derrida and his interpretations and text, contingency and the specific positioning of
accounts while Fish (1979) writes of the authority subjects (including researcher-as-instrument) are
of interpretative communities in texts. central to qualitative inquiry and are not threat-
The criteria of credibility, transferability, ened by the application of a general set of criteria
dependability and confirmability for establishing for evaluating rigour. These criteria provide
rigour are useful general principles for guiding reasonable anchor points for a paradigm which
qualitative evaluation. These criteria are analogous is often inappropriately accused of engaging in
to the traditional quantitative standards of validity, 'anything goes' science:
generalizability, reliability and objectivity, and As long as we strive to base our claims and interpreta-
much of the debate is framed in this way (e.g.
tions of social life on data of any kind, we must have a
Bryman 1988; Silverman 1993). Yet similarities
logic for assessing and communicating the interactive
regarding the principles to which both sets of process through which the investigator acquired the
evaluative criteria appeal should not be interpreted research experience and information. If we are to
as a licence to use quantitative criteria to evalu- understand the detailed means through which human
ate qualitative work (or vice versa). Beyond the beings engage in meaningful action and create a world
522 Jamie Baxter and John Eyles
of their own or one that is shared with others. (Morgan 3 For example, Cooper (1994); Dyck (1989); Leckie
1983, 397) (1993); Rowe and Wolch (1990); Valentine (1993);
and Wilson (1993).
We must acknowledge that
4 For example, Bonnett (1992); Bridge (1994); Brown
insufficient attention has as yet been devoted to evolv- (1995); McDowell and Court (1994); and Rollinson
ing criteria for assessing the general quality and rigour (1990).
of interpretive research. (ibid., 399) 5 For example, Dyck (1989); Eyles et al. (1993);
Fernandez Kelly (1994); Hewitt (1994); and
This task still largely awaits qualitative research in
McDowell (1994).
social geography. Criteria (establishing ways of
6 Cooper (1995) and McDowell (1994) provide some
thinking) and detailed questioning will help us detail on how their role in the research interviews
accomplish this task. may have affected responses.
7 While there is lively methodological debate among
quantitative social scientists, there is consensus
Acknowledgements about many basic strategies for establishing
Both authors acknowledge the advice given at rigour, particularly those involving basic statistical
presentations of this manuscript in draft at meet- procedures.
ings of the Medical Geography Specialty Group
8 Eyles (1988b); Eyles and Smith (1988); Jackson
of the Association of American Geographers, the (1985); Keith (1992); Ley and Samuels (1978);
Sixth International Symposium in Medical Geogra- Livingstone (1992); Lowe and Short (1990);
McDowell (1992a, 1992b); Miles and Crush (1993);
phy, and by the three anonymous referees. Jamie Mitchell and Draper (1981); Moss (1993); Nast
Baxter also wishes to acknowledge the support of (1994); Pickles (1988); Pile (1991); Rose (1993);
an Ontario Graduate Scholarship and a Social Schoenberger (1991); and Smith (1984).
Science and Humanities Research Council 9 For example, Keith (1992); Miles and Crush (1993);
(Canada) Research Fellowship. John Eyles is sup- Moss (1993); Nast (1994); Rose (1993); and Smith
ported by the Ecoresearch Program in Environ- (1988).
mental Health with contributions from the Social 10 We concur with Lincoln and Guba that validity
Science, and Medical Research (credibility) is quite different from representative-
Engineering ness (transferability) in the sense that the former is
Councils, Environment Canada and the Regional about representations of experiences while the latter
Municipality of Hamilton-Wentworth. concerns the applicability of the findings to other
contexts.
11 For example, Buttimer (1974); Donovan (1986);
Notes
Eyles (1985); and McDowell (1994).
12 The 'sending context' is one in which an original
1 Searching the Wilson social sciences, Wilson humani-
ties and Sociofileindexes as well as twelve geogra- study has been undertaken while the 'receiving
context' is one to which the sending context
phy journals manually, we found 49 papers, 31
empirical and eighteen commentary. While there findings may apply. Such an application must be
are undoubtedly books which include studies determined empirically.
involving qualitative interview methods, these
13 Droogleever Fortuijn and Karsten do quite the
could not be located in the computer and manual opposite. See their quotation earlier in this paper.
searches of the McMaster libraries. We acknowledge 14 See Bell and Newby (1977); Bell and Roberts (1984);
that the longer formats of books/monographs/ Eyles (1988b); and Roberts (1984).
chapters may allow writers more opportunity to
elaborate methodological and analytical procedures References
and that they may be less subject to the criticisms in
this paper but they are seldom subject to as critical Athens L 1984 Scientific criteria for evaluating qualitative
peer review as journal articles. Peer-reviewed jour- studies in Denzin N ed. Studies in symbolicinteraction
nal articles thus represent the pinnacle of successful volume5 Jai Press, London
incorporation of qualitative approaches in social Barnes T and Duncan J 1992 Writing worlds Routledge,
geography. London
2 The following articles detail how people were Baxter J Eyles J and Willms D 1992 The Hagersville tire
selected: Bridge (1994); Cooper (1994); Droogleever fire: interpreting risk through a qualitative research
Fortuijn and Karsten (1989); Dyck (1989); Eyles et al. design QualitativeHealth Research2 2 208-37
(1993); Leckie (1993); McDowell (1994); Rollinson Bell C and Newby H 1977 Doing sociologicalresearchAllen
(1990); Rutherford (1995); and Valentine (1995). Unwin, London
Evaluating qualitative research in social geography 523
Bell C and Roberts N 1984 SocialresearchingRKP,London Eyles J and Donovan J 1986 Making sense of sickness and
Bonnett A 1992 Anti-racism in 'white' areas: the example care: an ethnography of health in a West Midlands
of Tyneside Antipode24 1 1-15 town Transactionsof the Instituteof BritishGeographers11
Borland K 1991 That's not what I said: interpretive 4 15-27
conflict in oral narrative research in Gluck S B and Eyles J and Donovan J 1990 Thesocialeffectsof healthpolicy
Patai D eds Women'swords:the feminist practiceof oral Avebury Press, Aldershot
history Routledge, New York Eyles J and Perri E 1993 Life history as method: an
Bridge G 1994 Gentrification, class and residence: a re- Italian-Canadian family in an industrial city The
appraisal Environmentand Planning D: Societyand Space CanadianGeographer37 2 104-19
12 31-51 Eyles J and Smith D 1988 Qualitative methodsin human
Brown M 1995 Ironies of distance: an ongoing critique of geographyPolity Press, Cambridge
the geographies of AIDS Environmentand Planning D: Eyles J Taylor S M Baxter J Sider D Willms D 1993 The
Societyand Space 13 159-83 social construction of risk in a rural community:
Bryman A 1988 Quantity and quality in social research responses of local residents to the 1990 Hagersville
Unwin-Hyman, London (Ontario) tire fire Risk Analysis 13 3 281-90
Burgess R 1984 Field researchAllen Unwin, London Fernandez Kelly M P 1994 Towanda's triumph: social
Burns N 1988 Standards for qualitative research Nursing and cultural capital in the transition to adulthood in
ScienceQuarterly2 1 44-52 the urban ghetto International Journal of Urban and
Buttimer A 1974 Values in geography Association of
RegionalResearch18 1 88-111
American Geographers, Washington
Fielding N and Fielding L 1986 Linking data Sage,
Clark M Gregory K J and Gurnell A M 1987 Horizons in
Newbury Park
physicalgeographyBarnes and Noble, Totowa, NJ Fish S 1979 Is therea text in this class? Harvard University
Cooper A 1994 Negotiating dilemmas of landscape, place Press, Cambridge
and Christian commitment in a Suffolk parish Transac- Frost P and Stablein R 1992 Doing exemplary research
tions of the Institute of British Geographers19 202-12
Sage, Newbury Park
Cooper A 1995 Adolescent dilemmas of landscape, place Geertz C 1973 The interpretationof culture:selectedessays
and religious experience in a Suffolk parish Environ-
Basic Books, New York
ment and Planning D: Societyand Space 13 349-63
Cornwell J 1988 A case-study approach to lay health Gregson N and Lowe M 1995 'Home'-making: on the
beliefs: reconsidering the research process in Eyles J spatiality of daily social reproduction in contemporary
middle-class Britain Transactionsof the Instituteof British
and Smith D eds Qualitativemethodsin human geogra-
Geographers20 224-35
phy Basil Blackwell, Oxford 219-32
Halpern E 1983 Auditing naturalistic inquiries: the devel-
Cosgrove D and Daniels S 1988 The iconographyof land-
opment and application of a model Unpubl. PhD
scape Cambridge University Press, Cambridge
Denzin N 1978 The researchact McGraw-Hill, New York dissertation, Indiana University, Bloomington, IN
Donovan J 1986 Youdon't buy sickness,it just comesGower, Hammersley M 1992 What's wrong with ethnography
Aldershot Routledge, London
Donovan J 1988 'When you're ill you've gotta carry it': Harrison C and Burgess J 1994 Social constructions of
health and illness in the lives of black people in Eyles nature: a case study of conflicts over the development
of Rainham Marshes Transactionsof the Institute of
J and Smith D eds Qualitativemethodsin human geog-
British Geographers19 291-310
raphyPolity Press, Cambridge
Herod A 1991 Local political practice in response to a
Droogleever Fortuijn J and Karsten L 1989 Daily activity
patterns of working parents in the Netherlands Area21 manufacturing plant closure: how geography compli-
4 265-76 cates class analysis Antipode23 4 385-402
Dukes W 1965 N=1 PsychologicalBulletin 64 74-9 Hewitt K 1994 'When the great planes came and made
ashes of our city': towards an oral geography of the
Dyck I 1989 Integrating home and wage workplace:
women's daily lives in a Canadian suburb TheCanadian disasters of war Antipode26 1 1-34
Geographer33 4 329-41 Jackson P 1985 Urban ethnography Progress in Human
Erickson F 1973 What makes school ethnography 'ethno- Geography9 2 159-76
graphic'? Anthropology and Education Quarterly 4 2 Katz C 1991 Sow what you know: the struggle for social
10-99 reproduction in rural Sudan Annals of the Associationof
Eyles J 1985 Senses of place Silverbrook Press, Warrington AmericanGeographers81 3 488-514
Eyles J 1988a Interpreting the geographical world in Keith M 1992 Angry writing: (re)presenting the unethical
Eyles J and Smith D eds Qualitativemethodsin human world of the ethnographer Environmentand Planning D:
geographyPolity Press, Cambridge Society and Space 10 551-68
Eyles J 1988b Research in human geography Blackwell, Kidder L 1981 Qualitative research and quasi-
Oxford experimental frameworks in Brewer B and Collins B
524 Jamie Baxter and John Eyles
eds Scientificinquiry and the social sciences Jossey-Bass, McDowell L and Court G 1994 Performing work: bodily
San Francisco representations in merchant banks Environment and
King G Keohane R and Verba S 1994 Designing social Planning D: Societyand Space 12 727-50
inquiry: scientific inference in qualitative research Merton R 1968 Social theoryand social structureFree Press,
Princeton University Press, Princeton New York
Kirk J and Miller M 1986 Reliability and validity in Miles M and Crush J 1993 Personal narratives as inter-
qualitativeresearchSage, Newbury Park, CA active texts: collecting and interpreting life histories
Knafl K 1994 Promoting academic integrity in qualitative ProfessionalGeographer45 1 84-94
research in Morse J ed. Critical issues in qualitative Mitchell B and Draper D 1982 Relevance of ethics in
researchmethodsSage, Thousand Oaks, CA geographyLongman, London
Knafl K and Breitmayer B 1989 Triangulation in qualita- Morgan G 1983 BeyondmethodSage, Newbury Park, CA
tive research: issues of conceptual clarity and purpose Moser C and Kalton G 1981 Survey methods in social
in Morse J ed. Qualitative nursing research:a contem-
investigationHeineman, London
porarydialogueAspen, Rockville, MD 193-203 Moss P 1993 Focus: feminism as method The Canadian
Krefting L 1990 Rigor in qualitative research: the assess- Geographer37 1 48-61
ment of trustworthiness The AmericanJournalof Occu- Nast H 1994 Women in the field: critical feminist meth-
pational Therapy45 3 214-22 odologies and theoretical perspectives Professional
Kuzel A 1992 Sampling in qualitative inquiry in Crabtree
Geographer46 1 54-66
B and Miller W eds Doing qualitative researchSage, Newton J 1995 An assessment of coping with environ-
Newbury Park mental hazards in northern aboriginal communities
Lamont M 1984 How to become a dominant French The CanadianGeographer39 2 112-20
philosopher AmericanJournalof Sociology93 584-622 Patton M 1990 Qualitativeevaluationand researchmethods
Leckie G 1993 Female farmers in Canada and gender
relations of a restructuring agricultural system The Sage, Newbury Park, CA
Pelto P and Pelto G 1978 Anthropologicalresearch:the
CanadianGeographer37 3 212-30
structure of inquiry Cambridge University Press,
LeCompte M and Goetz J 1982 Problems of reliability
and validity in ethnographic research Review of Educa- Cambridge
Pickles J 1988 From fact-world to life-world in Eyles J
tional Research52 1 31-60
and Smith D eds Qualitativemethodsin humangeogra-
Ley D 1974 The blackinner city as frontieroutpost Associ-
ation of American Geographers, Washington phy Polity Press, Cambridge
Pile S 1991 Practising interpretative geography Transac-
Ley D and Samuels M eds 1978 Humanistic geography tions of the Institute of British Geographers16 458-69
Maaroufa Press, Chicago
Lieberson S 1992 Small N's and big conclusions in Plummer K 1983 Documentsof life Allen Unwin, London
Porteous J 1988 Topocide: the annihilation of place in
Ragin C and Becker H eds Whatis a case? Cambridge
University Press, New York Eyles J and Smith D eds Qualitativemethodsin human
Lincoln Y and Guba E 1981 Strategies for insuring the geographyBasil Blackwell, Oxford 75-93
Risteen Hasselkus B 1991 Qualitative research: not
dependability (reliability) of naturalistic studies Paper
another orthodoxy The OccupationalTherapyJournalof
presented at the joint annual meeting of the Evaluation
Network and the Evaluation ResearchSociety,Austin, TX Research11 1 3-7
Lincoln Y and Guba E 1985 Naturalistic inquiry Sage, Roberts H 1984 Doing feminist researchRKP,London
Beverly Hills, CA Rollinson P 1990 The everyday geography of poor
elderly hotel tenants in Chicago Geografiska Annaler72B
Livingstone D 1992 The geographical tradition Oxford
2-3 47-57
University Press, Cambridge
Lowe M and Short J 1990 Progressive human geography Rosaldo R 1989 Cultureand truth Beacon Press, Boston
Progressin Human Geography14 1 1-11 Rose D 1993 Feminism, method and methods in human
Mackenzie F 1992 'The worse it got, the more we geography: an idiosyncratic overview CanadianGeogra-
laughed': a discourse of resistance among farmers of pher 37 57-61
eastern Ontario Environmentand Planning D: Societyand Rose G 1982 DecipheringsociologicalresearchMacMillan,
Space 10 691-713 London
McDowell L 1992a Valid games? A response to Erica Rowe S and Wolch J 1990 Social networks in time and
Schoenberger ProfessionalGeographer44 2 212-15 space: homeless women in skid row Los Angeles
McDowell L 1992b Doing gender: feminism, feminists Annals of the Association of American Geographers80 2
and research methods in human geography Transac- 184-204
tions of the Institute of British Geographers17 399-416 Rutherford T 1995 'Control the ones you can': production
McDowell L 1994 Social justice, organizational culture restructuring selection and training in Kitchener
and workplace democracy: cultural imperialism in the Region Manufacturing 1987-1992 The CanadianGeogra-
city of London Urban Geography15 7 661-80 pher 39 1 30-45
Evaluating qualitative research in social geography 525
Sandelowski M 1986 The problem of rigor in qualitative Streiner D and Norman G 1989 Healthmeasurementscales:
research Advancesin Nursing Science8 3 27-37 a practical guide to their developmentand use Oxford
Schatzman L and Strauss A 1973 Problemsin participant University Press, New York
observationPrentice-Hall, Englewood Cliffs, NJ Teich A and Frankel M 1992 Good science and respon-
Schoenberger E 1991 The corporate interview as a sible scientists Association of American Geographers,
research method in economic geography Professional Washington
Geographer43 2 180-9 Valentine G 1989 The geography of women's fear Area21
Schutz A 1962 Commonsense and scientific interpretation 4 385-90
of human action in Natson M ed. Collected papers Valentine G 1993 (Hetero)sexing space: lesbian percep-
volume1 Nijhoff, The Hague tions and experiences of everyday spaces Environment
Schwartz H and Jacobs J 1979 Qualitative methodsFree and Planning D: Societyand Space 11 395-413
Press, New York Valentine G 1995 Out and about: geographies of lesbian
Scriven M 1971 Objectivity and subjectivity in educa-
landscapes InternationalJournalof Urban and Regional
tional research in Thomas L ed. Philosophicalredirection Research19 1 96-111
of educational research.The seventy-first yearbookof the Vidich P Bensman J and Stein M 1964 Reflectionson
National Societyfor the Study of EducationUniversity of
communitystudies Free Press, New York
Chicago Press, Chicago Wax R 1971 Doing fieldwork:warnings and advice Univer-
Shute J and Knight D 1995 Obtaining an understanding
of environmental knowledge: Wendaban Stewardship sity of Chicago Press, Chicago
Willms D Best J Taylor D Gilbert J Wilson D Lindsay E
Authority The CanadianGeographer39 2 101-11
Silverman D 1993 Interpreting qualitative data Sage, and Singer J 1990 A systematic approach for using
London qualitative methods in primary prevention research
Smith S 1984 Practising humanistic geography Annals of MedicalAnthropologyQuarterly4391-409
the Associationof AmericanGeographers74 3 353-74 Wilson D 1993 Everyday life, spatiality and inner city
Smith S 1988 Constructing local knowledge: the analysis disinvestment in a US city InternationalJournalof Urban
of self in everyday life in Eyles J and Smith D eds and RegionalResearch17 4 576-94
Qualitativemethodsin humangeographyBasil Blackwell, Winchester H and Costello L 1995 Living on the street:
Oxford 17-38 social organization and gender relations of Australian
Smith S 1994 Qualitative methods in Johnston R ed. The street kids Environment and Planning D: Society and
dictionaryof humangeographyBlackwell, Oxford Space 13 329-48
Spradley J 1979 The ethnographicinterview Holt Reinhart
and Winston, New York