Answers to frequently asked questions
about thematic analysis
Reflexive TA | the basics
»What’s the difference between a code and a theme?
»What’s the difference between a subtheme and a theme?
»What’s the difference between a domain summary and a theme?
»What is a central organising concept and why is it important in thematic
analysis?
Reflexive TA in context: contrasts with other types of
thematic analysis
»What’s the difference between reflexive thematic analysis (e.g., ‘Braun &
Clarke’) and other approaches?
»What’s the difference between thematic coding and TA?
Reflexive TA in context: contrasts with other analytic
approaches
»What’s the difference thematic analysis and IPA?
»What’s the difference between thematic analysis and grounded theory?
»What’s the difference between constructionist thematic analysis and discourse
analysis?
»What’s the difference between thematic analysis and (qualitative) content
analysis?
Doing Reflexive TA
»I’ve collected five interviews – is that enough for a TA?
»Why are we critical of the notion that ‘themes emerged’ from data?
»Why don’t we advocate multiple-coders and inter-rater reliability for reflexive
TA?
»How many themes should I have?
»Are latent codes/themes better than semantic ones?
Reflexive TA: What is it good for?
»What types of data is reflexive thematic analysis suitable for?
»What types of research questions is reflexive thematic analysis suitable for?
Quality in reflexive TA research
»What are some of the common problems I should avoid when using reflexive
TA?
»How do I know if I have done a good reflexive thematic analysis?
»Should someone else check my coding?
»Should I use numbers when reporting themes?
Teaching and supervision
»Should I teach my students thematic analysis?
»Is thematic analysis sophisticated enough for a doctoral project?
Reflexive TA| the basics
»What’s the difference between a code and a theme?
»What’s the difference between a subtheme and a theme?
»What’s the difference between a domain summary and a theme?
»What is a central organising concept and why is it important in thematic
analysis?
What’s the difference between a code and a theme?
A theme captures a common, recurring pattern across a dataset, clustered
around a central organising concept. A theme tends to describe the different
facets of that singular idea, demonstrating the theme’s patterning in the dataset.
Codes tend to be more specific than themes. They capture a single idea
associated with a segment of data, and consist of pithy labels identifying what is
of interest in the data (in relation to the research question). Codes can be
conceptualised as the building-blocks that combine to create themes – so
multiple codes typically are combined to create themes, during the process of
TA.
Top
What’s the difference between a subtheme and a theme?
A theme captures a common, recurring pattern across a dataset, organised
around a central organising concept. A theme tends to describe the different
facets of a pattern across the dataset. A subtheme exists ‘underneath’ the
umbrella of a theme. It shares the same central organising concept as the
theme, but focuses on one notable specific element. Subthemes generally should
be used sparingly, only when there is one particular element of a theme that has
a particular focus, is notable, and/or is particularly important for the research
question. Through naming and analysing a specific subtheme, that aspect of the
theme becomes particularly salient.
Top
What’s the difference between a domain summary and a theme?
The difference between a theme and a domain summary is a source of frequent
confusion in much published TA research. A domain summary is a summary of
an area (domain) of the data; for example, a summary of everything the
participants said in relation to a particular topic or interview question. Unlike
themes, there isn’t anything that unifies the description of what participants said
about this topic – there is no underlying concept that ties everything together
and organises the analytic observations. In our approach to TA, themes are
conceptualised as patterns in the data underpinned by a central concept that
organises the analytic observations; this is rather different from a domain
summary, and the two should ideally not be confused when using our approach.
More simply put, a theme identifies an area of the data and tells the reader
something about the shared meaning in it, whereas a domain summary simply
summarises participant’s responses relating to a particular topic (so shared topic
but not shared meaning). To make things complicated, some approaches to TA
do conceptualise themes as domain summaries – this conceptualisation of
themes is evident in both coding reliability approaches (see Boyatzis, 1998;
Guest et al., 2012) and codebook approaches, such as template, framework and
matrix analysis. Sometimes the confusion between domain summaries and
themes is simply an issue of poorly named themes – the theme itself is a
conceptually founded pattern, but the theme name does not reflect this.
‘Experiences of Y’ or ‘Benefits of X’ are classic examples of domain-summary
type theme names. These theme names identify that, for example, ‘benefits of
X’ was an important area of the data in relation to the research question(s), but
they don’t communicate the essence of this theme; they don’t tell the reader
something specific about these benefits and what underlying concept
underpinned what the participants had to say about the benefits of X.
Literature
To understand more about the differences between domain summaries and fully
realised themes, we recommend the following three papers:
Connelly, L. M. & Peltzer, J. N. (2016). Underdeveloped themes in qualitative
research: Relationships with interviews and analysis. Clinical Nurse Specialist,
January/February, 51-57.
DeSantis, L. & Ugarriza, D. N. (2000). The concept of theme as used in
qualitative nursing research. Western Journal of Nursing Research, 22(3), 351-
372.
Sandleowski, M. & Leeman, J. (2012). Writing usable qualitative health research
findings. Qualitative Health Research, 22(10), 1404-1413.
We have written about this in this chapter:
Clarke, V., Braun, V., Terry, G & Hayfield N. (2019). Thematic analysis. In
Liamputtong, P. (Ed.), Handbook of research methods in health and social
sciences (pp. 843-860). Springer.
We discuss it in this lecture.
Top
What is a central organising concept and why is it important in
thematic analysis?
A central organising concept captures the essence of a theme. It is an idea or
concept that captures and summarises the core point of a coherent and
meaningful pattern in the data. If you can identify the central organising concept
of a theme, you can capture the core of what your theme is about. If you cannot
do this, your theme may lack coherence.
Top
Reflexive TA in context: contrasts with other
analytic approaches
»What’s the difference between reflexive thematic analysis (e.g., ‘Braun &
Clarke’) and other approaches?
»What’s the difference between thematic coding and TA?
What’s the difference between reflexive thematic analysis (e.g.,
‘Braun & Clarke’) and other approaches?
We distinguish between three main types of TA – our reflexive approach, coding
reliability TA and codebook approaches, which include methods like template
analysis and framework analysis.
Let’s consider coding reliability approaches first, which are exemplified by the
work of US researchers such as Boyatzis (1998) and Guest et al. (2012) and
Joffe (2011) in the UK. Unsurprisingly, given the name ‘coding reliability’
approaches, this type of TA emphasises the measurement of the accuracy or
reliability of coding through the use of a structured codebook and multiple
independent coders, and the measurement of coding reliability (the level of
‘agreement’ between coders, usually calculated with Cohen’s Kappa – a Kappa of
>.80 indicates a very good level of coding agreement and supposedly reliable
coding). While we would agree that it can be helpful to code data with another
researcher (to bounce around ideas, and to reflect on your assumptions and
what you might have overlooked in the data), this doesn’t necessarily result in
‘better’ coding (particularly in the sense of ‘more accurate’ or ‘more reliable’
coding), just different coding. This is why we don’t advocate the use of a coding
frame, or the calculation of inter-rater reliability scores associated with coding
reliability approaches. Coding reliability approaches to TA recommend the use of
a coding frame precisely because this enables the calculation of inter-reliability
scores. The use of inter-rater reliability scores is underpinned by the
(realist/positivist) assumption that there is a reality in the data that can be
accurately captured through coding, if appropriate techniques are used.
Our approach to coding is flexible and organic, and coding should evolve
throughout the coding process – to some extent our approach to coding is
similar to initial coding in grounded theory (Charmaz, 2006). We understand
coding as an active and reflexive process that inevitably and inescapably bears
the mark of the researcher(s). With no one ‘accurate’ way to code data, the logic
behind inter-rater reliability (and multi-independent coders) disappears. We
argue that inter-rater reliability scores can be understood as showing that two
researchers have been trained to code data in the same way, rather than that
their coding is ‘accurate’. Furthermore, the use of a codebook often results in
relatively superficial codes (precisely because of the need to facilitate multiple
coders applying codes in the same way). The positivist assumptions
underpinning coding reliability approaches often limit the claimed theoretical
flexibility of these types of TA.
Another important difference between coding reliability approaches and reflexive
TA relates to the conceptualisation of themes as analytic inputs in coding
reliability TA rather than analytic outputs (as in reflexive TA). In coding reliability
TA, themes are developed early in the analytic process, either before the
analytic process (and often based on data collection questions) or after data
familiarisation. Coding is conceptualised as a process of searching for evidence
for themes (rather than codes being the building blocks for themes as in
reflexive TA). Furthermore, themes are often understood as domain summaries
(as summaries of what participants said in relation to a particular topic or data
collection question), rather than patterns of shared meaning underpinned by a
central organising concept. We’ve also noticed that coding reliability proponents
tend to describe themes as if they exist within the data prior to analysis and the
researcher’s role is to find the themes within the data. This is very different from
how themes are conceptualised in reflexive TA – as analytic outputs, created
from codes and through the researcher’s active engagement with their data.
Coding reliability proponents often describe TA as an approach that bridges the
positivist/quantitative and interpretative/qualitative divide. Whether or not this
‘bridging’ is viewed as possible, seems to hinge on how qualitative research is
defined – as techniques and tools for collecting (and analysing) qualitative data
(often underpinned by a positivist philosophy or paradigm), or as both
techniques and an underlying qualitative philosophy. For coding reliability
‘bridgers’, qualitative research is implicitly defined as (merely) techniques and
tools, and TA bridges the divide not by bringing two (arguably incommensurate)
paradigms together (positivism and a qualitative paradigm), but by bringing
together a positivist paradigm and qualitative techniques.
We turning now to codebook approaches – exemplified by approaches such as
template analysis (e.g. King & Brooks, 2017), framework analysis (e.g. Ritchie &
Spencer, 1994) and matrix analysis (Miles & Huberman, 1994). If we imagine TA
on continuum with positivist, structured coding reliability approaches at one end,
and reflexive TA at the other, these approaches sit somewhere in the middle.
They share with coding reliability approaches the use of a structured codebook
and the conceptualisation of themes as analytic inputs and domain summaries
but not the positivist-inflected concern for measuring coding reliability. Instead
codebook approaches share with reflexive TA a broadly qualitative philosophy or
paradigm.
Literature
Boyatzis, R. E. (1998). Transforming qualitative information: Thematic analysis
and code development. Thousand Oaks, CA: Sage.
Charmaz, K. (2006). Constructing grounded theory: A practice guide through
qualitative analysis. Thousand Oaks, CA: Sage.
Guest, G., MacQueen, K. M. & Namey, E. E. (2012). Applied thematic analysis.
Thousand Oaks, CA: Sage.
Joffe, H. (2011). Thematic analysis. In D. Harper & A. R. Thompson (Eds.),
Qualitative methods in mental health and psychotherapy: A guide for students
and practitioners (pp. 209-223). Chichester: Wiley.
King, N. & Brooks, J. M. (2017). Template analysis for business and
management students. London: Sage.
Miles, M. B. & Huberman, A. M. (1994). Qualitative data analysis: An expanded
sourcebook, 2
nd
ed. Thousand Oaks, CA: Sage.
Ritchie, J. & Spencer, L. (1994). Qualitative data analysis for applied policy
research. In A. Bryman & R. G. Burgess (Ed.), Analysing qualitative data (pp.
173-194). London: Taylor & Francis.
Yardley, L. (2008). Demonstrating validity in qualitative psychology. In J. A.
Smith (Ed.), Qualitative psychology: A practical guide to research methods (pp.
235-251). London: Sage.
Top
What’s the difference between thematic coding and TA?
There is an approach to TA known as ‘thematic coding’ (e.g. Gibbs, 2007; Flick,
2014; Rivas, 2018), which most often involves the use of ground theory
techniques to develop themes from data (rather than the categories and
concepts associated with a grounded theory analysis). This approach to TA was
particularly common before there were widely cited procedures for TA, but some
qualitative researchers still advocate the use of thematic coding (e.g. Flick,
2014; Rivas, 2018).
Gibbs’ (2007) account of thematic coding presents it as a generic technique for
qualitative analysis, that informs many different analytic approaches, centred on
the development of “a framework of thematic ideas” (p. 38) about data. As such
there are many similarities between this approach and TA. Gibbs distinguishes
between data-driven (descriptive) codes (close to the respondent’s terms) and
concept-driven (analytic and theoretical) codes, a distinction that maps onto the
semantic/latent coding distinction in TA (he also mentions ‘categories’ as a ‘mid-
point’ between descriptive and analytic codes). Similarly, his distinction between
data-driven and concept-driven coding maps on to inductive and deductive
approaches to TA. He argues that these approaches to coding are not exclusive
and most researchers move backwards and forwards between these different
approaches. Like coding reliability and codebook forms of TA, he advocates for
the use of coding memos – a staple of grounded theory – that record the label or
name of the code, who coded it (if working in a team), the definition of the code
(to ensure coding is systematic and consistent) and other notes of thinking
about the code (including a record of any changes). But he also emphasises the
importance of a flexible approach and warns researchers “not to become too tied
to the initial codes you construct.” (p. 46) He discusses grounded theory as one
of the most commonly used approaches to thematic coding and identifies line-
by-line coding and constant comparison as particularly useful techniques.
Flick (2014) argued that his development of the thematic coding method in the
1990s pre-dates the development of TA, but it seems that both TA and thematic
coding have been in use since at least the 1980s. Flick (2014) described his
method for thematic coding as combining ideas from TA and grounded theory
(specifically Strauss, 1987) to develop an approach that has both inductive and
a-priori features and involves processes of simultaneous data collection and
analysis and the analysis of each case (e.g. interview) in turn before developing
an overall theory or thematic structure. The steps involve first producing a short
description of each case, which is continuously checked and modified throughout
the analysis, and then developing a thematic structure of “thematic domains and
categories” (p. 425) for each case using grounded theory processes of open and
selective coding. This process is repeated for each of the initial cases and then
the researcher checks and compares the thematic structure developed for each
case to produce an overall structure for the first few cases, which is then
assessed for each further case and modified if necessary. The approach Flick
describes is very similar to codebook types of TA – in that there is some
delimiting of focus before the analysis, and the thematic structure, like a
framework or matrix, allows for an emphasis on both cases and themes (and a
comparison of cases and groups of participants).
Rivas (2018) describes thematic coding as similar to the early stages of a
grounded theory, but with a focus on participants’ lived experience in and of
itself rather than as a precursor to theory development. Like grounded theory,
Rivas’ version of thematic coding involves memo writing, constant comparative
analysis, and it can also involve some concurrent data collection and analysis, it
begins with data familiarisation before moving to open coding and then the more
interpretative stages of category formation (through grouping open codes and
defining the categories) and finally the most interpretative stage of theme
development (which can involve the use of concept or thematic mapping to
review the themes). Like TA proper, this version of thematic coding can involve
either or both inductive and deductive coding.
Does it matter that researchers claim to use (some) grounded theory
procedures, but actually produce something more akin to a TA, than a grounded
theory? Particularly now that TA is a well-developed method? Is insisting on the
‘proper’ use of methods (doing what you claim to do) simply a preoccupation
with the purity of method – or proceduralism (King & Brooks, 2017: 231) –
“where analysts focus on meticulously following set procedures rather than
responding creatively, imaginatively (and systematically) to data”? We certainly
agree with critics of proceduralism that rigid rules for good practice, insisting on
one true way of applying a method, and avoiding interpretation, theory and
taking reflexivity seriously, often leads to qualitative research that is rather
limited – merely descriptive/summative, with little or no interpretation of the
meaning of the data, implicitly peppered with positivism, and lacking depth of
engagement, thoughtfulness and creativity. However, muddling grounded theory
and TA like this doesn’t seem to result from thoughtfulness and creativity, or
indeed theoretical knowingness and reflexivity, rather it seems too often stem
from the researchers’ lack of knowledge, and narrow methodological range.
Methodological practice is contested and debated, but not necessarily in a way
that is helpful for those learning about qualitative research or to develop
understanding of particular techniques or philosophical concerns; productive
debate requires that researchers use techniques knowingly and are able to
clearly articulate the assumptions underpinning these. On balance, we think this
type of muddling only contributes to further confusion about qualitative
research.
Literature
Flick, U. (2014)
Gibbs, G. R. (2007). Analysing qualitative data. London: Sage.
King, N. & Brooks, J. (2018). Thematic analysis in organisational research. In C.
Cassell, A. L. Cunliffe & G. Grandy (Eds.), The Sage handbook of qualitative
business and management research methods (pp. 219-236). London: Sage.
Rivas, C. (2018). Finding themes in qualitative data. In C. Seale (Ed.),
Researching society and culture, 4
th
ed. (pp. 429-453). London: Sage.
Strauss, A. L. (1987). Qualitative analysis for social scientists. Cambridge:
Cambridge University Press.
Top
Reflexive TA in context: contrasts with other
types of thematic analysis
»What’s the difference thematic analysis and IPA?
»What’s the difference between thematic analysis and grounded theory?
»What’s the difference between constructionist thematic analysis and
discourse analysis?
»What’s the difference between thematic analysis and (qualitative)
content analysis?
What’s the difference between thematic analysis and IPA?
There are some differences between interpretative phenomenological analysis
(IPA) and TA; however, the end result of an IPA and a TA analysis can often be
very similar. Often the end result of an IPA doesn’t seem to be any different
from a TA on a small sample.
Let’s consider the divergences between IPA and TA first, which result mainly
from:
1. The fact that IPA is better thought of as a methodology (a theoretically
informed framework for how you do research) rather than a method (a
technique for collecting/analysing data), whereas TA is just a method.
2. Differences of procedure between IPA and TA.
We can think of IPA as like a piece of ready-made furniture – all the design
choices (colour-scheme, dimensions, materials etc.) have been made for you. As
well as outlining a range of analytic procedures, IPA specifies:
• what the ontological and epistemological underpinnings of your research are
(critical realism and contextualism; see Larkin, Watts & Clifton, 2006),
• what theoretical framework should inform your research (phenomenology),
• what types of research questions you can ask (about people’s experiences
and perspectives),
• what sampling strategy you should use (homogenous, small N), and
• how (ideally) you should collect data (qualitative interviews) (see Smith,
Flowers & Larkin, 2009).
Obviously, we are simplifying the characteristics of IPA a bit, but the main point
to take away is that IPA provides an entire framework for conducting research.
By contrast, TA is like the piece of furniture you build yourself, where you
choose your colour scheme, the dimensions of the piece, the materials, etc.
Because TA is just a method and the hallmark of TA is its flexibility:
• It can be used widely across the epistemological and ontological spectrum
(TA can be [critical] realist or constructionist),
• It can be underpinned by phenomenology, as well as by any number of
other theories,
• It can be used to address a wide range of research questions (including
questions about people’s experiences and perspectives)
• There are no specific requirements for sampling in TA (we would generally
recommend a larger N than for IPA studies because TA does not share IPA’s
ideographic focus – more on this below – but some degree of homogeneity
in sampling is helpful in smaller studies), and
• It can be used to analyse most types of qualitative data (interviews, focus
groups, diaries, qualitative surveys, secondary sources, vignettes, story
completion tasks etc.)
IPA has a dual focus on the unique characteristics of individual participants (the
idiographic focus mentioned above) and on patterning of meaning across
participants. In contrast, TA focuses mainly on patterning of meaning across
participants (this is not to say it can’t capture difference and divergence in data).
In terms of analytic procedures, both IPA and TA involve coding and theme
development, but these processes are somewhat different for each method.
Coding in TA begins after a process of data familiarisation, in which the
researcher notes any initial analytic observations about each data item and the
entire data-set. The researcher then codes across all of the data items. The
researcher either collates the data relevant to each code as they code, or they
collate all the relevant data at the end of the coding process. By contrast, coding
in IPA consists of a process of ‘initial commenting’ or ‘initial noting,’ in which the
researcher writes their initial analytic observations about the data on the data
item (if working with interview transcripts, initial notes are usually recorded in a
wide margin on the right hand side of the transcript). These initial notes are brief
commentaries on the data (rather than succinct codes). This means initial noting
in IPA lies somewhere between data familiarisation and coding in TA.
Another difference is that in IPA, the researcher codes their first data item then
progresses to developing themes for that data item, rather than coding across
the entire dataset, and then progressing to theme development. So IPA focuses
on developing each stage of the analysis for each data item, before moving to
the next; whereas TA involves developing each stage of analysis across the
whole dataset.
With regard to types of code, IPA refers to both ‘descriptive’ and ‘conceptual’
comments and these are very similar to ‘semantic’ and ‘latent’ codes in TA.
In terms of procedures for theme development, there are two levels of theme
development in IPA and one level in TA. In IPA, these are referred to as
‘emergent’ and ‘superordinate’ themes. Emergent themes are noted on the data
item (if working with interview transcripts, emergent themes are usually
recorded in a wide margin on the left hand side of the transcript). Superordinate
themes are developed from emergent themes. Once coding and theme
development are complete for each data item, the researcher develops
superordinate themes across the dataset. In TA, themes are developed from the
codes (and collated data), across all data items.
As a general rule, there will be a lot more emergent themes generated from an
IPA compared to the number of themes generated from a TA; this means that
emergent themes in IPA lie somewhere between codes and themes in a TA
(somewhat akin to subthemes in a TA). An IPA will generate roughly the same
number of superordinate themes as the number of themes generated from a TA.
However, another difference is that superordinate themes in an IPA simply
provide an organising framework for the analysis, it is the emergent themes that
are discussed in detail in the write-up.
Overall, IPA procedures help the researcher to stay close to the data, because
you develop codes and themes on the actual data item, and to focus on the
unique characteristics of each individual participant, because you code and
develop themes for each data item in turn. By contrast, the procedures of TA
help the researcher to identify patterns across the entire data-set.
Despite these procedural differences, the end result of an IPA and a
phenomenologically-informed TA of interview data would potentially be quite
similar – analytic procedures play some role in shaping analysis, but, in general,
our conceptual lenses play a much greater role.
So what does this all mean for choosing whether to use IPA or TA? We’d
recommend using TA when you want to address research questions that are not
about people’s experiences or perspectives, when you want to use data that do
not capture first-person accounts of personal experiences (such as focus groups
or story completion tasks), and when working with larger samples. If you are
asking an experiential-type question within a phenomenological framework
based on the analysis of interview data – how do you choose between IPA and
TA? There’s not a lot in it, but we’d recommend using IPA if you are working
with a smaller sample and want to maintain a more idiographic focus, and TA if
you are working with a larger sample and want to focus more on patterned
meaning across the data-set. We sometimes encounter the assumption that if
you’re doing phenomenological research, you MUST use IPA. Therefore, it’s
important to note that there is a long and diverse history of phenomenological
research in the social and health sciences (see Langdridge, 2007 for an
introduction to phenomenological research in psychology). Indeed, TA has a long
history as a phenomenological method that predates the development of IPA.
Literature
Langdridge, D. (2007). Phenomenological psychology: Theory, research and
method. Harlow, UK: Pearson Education.
Larkin, M., Watts, S. & Clifton, E. (2006) Giving voice and making sense in
interpretative phenomenological analysis. Qualitative Research in Psychology,
3, 102-120.
Smith, J. A., Flowers, P. & Larkin, M. (2009) Interpretative phenomenological
analysis: Theory, method and research. London: Sage.
Top
What’s the difference between thematic analysis and grounded
theory?
For a start, that depends on what is meant by grounded theory! There are
seemingly endless varieties and versions of grounded theory, with different
theoretical underpinnings and different procedures (for a useful overview, see
Birks & Mills, 2011).
One reason for the existence of many difference approaches to grounded theory
is that it represents one of the first attempts to develop a systematic method for
analysing qualitative data. Grounded theory was developed in the 1960s (Glaser
& Strauss, 1965, 1967) in the very early days of the (re)emergence of
qualitative methods in the social sciences, and a lot has changed since then! For
example, one famous recommended practice in grounded theory is not engaging
with the relevant literature prior to beginning data analysis, to avoid the analysis
being shaped by preconceptions from existing research, rather than being truly
grounded in the data (this practice has rather assumed the status of a
commandment ‘Thou shalt not...’).
More than four decades later, there’s a lot of grounded theory out there! This
means that at the very least a researcher should determine what grounded
theories exist relevant to their topic. Furthermore, in the 1960s, when this
recommendation was developed, researchers did not routinely submit ethics
applications prior to beginning their research. Today, ethics applications (and
often other forms of research proposals) are mandatory, and therefore some
engagement with the literature is required to produce such applications.
Unsurprisingly, there is now a lot of discussion and debate about this and other
elements of grounded theory procedure, with lots of different (competing)
recommendations for how to perform a grounded theory.
We find that it’s useful to make a distinction between a full grounded theory and
what we call grounded theory-lite. A full grounded theory requires the
implementation of the full range of grounded theory procedures, including
theoretical sampling, with the aim of producing a theory grounded in data.
Here’s one notable difference from TA; although TA can produce conceptually-
informed interpretations of data, it does not attempt to develop a theory. A full
grounded theory then is something that is only achievable in a large research
project (Pidgeon & Henwood, 1997), and is rarely used, even when a grounded
theory method is claimed. A grounded theory-lite involves using the techniques
of grounded theory for the development of categories (and concepts), and an
understanding of the relationship between the various categories (and concepts)
(Pidgeon & Henwood, 1997). This is probably the most frequently used form of
grounded theory. The practice of TA is similar to grounded theory-lite. Both
involve coding and the generation (and interpretation) of broader patterns in
data. Differences in terminology – for example, categories versus themes – can
mask the underlying similarities.
Grounded theorists are also rather keen on the development of graphic or
textual models (to represent their analytic findings), and that is not something
we particularly advocate for TA (which is not to say it can’t be done – this paper
provides a nice example of a tentative model developed from a TA: Anderson &
Clarke, 2017). (We do recommend the use of thematic maps in the theme
development process, and these can be included in your appendices, if writing a
dissertation/thesis.)
A key difference between grounded theory (full or lite) and TA is that like most
other analytic approaches grounded theory is a methodology, not just a method
(which TA is). All versions of grounded theory have an inbuilt theoretical
framework (and ontological/epistemological assumptions), advocate the use of
particular types of research question (a focus on social processes or the factors
that influence particular phenomenon is common in grounded theory), and the
use of particular methods of data collection (qualitative interviews are a key
method, but there are other possibilities, see Charmaz, 2006), as well as the use
of a particular set of analytic procedures (coding, memoing, etc.). However, the
results of a contextualist grounded theory-lite and a contextualist TA could be
rather similar.
So, what does all this mean for choosing grounded theory(-lite) or TA? We’d
recommend using TA:
• when you want to address research questions that are not focused on social
processes or influencing factors,
• when using methods other than interviewing, and/or
• when you are new to qualitative research (we think TA is more accessible
and easier to implement than grounded theory).
Literature
Anderson, S. & Clarke, V. (2017). Disgust, shame and the psychosocial impact
of skin picking: Evidence from an online support forum. Journal of Health
Psychology.
Birks, M., & Mills, J. (2011). Grounded theory: A practical guide. London: Sage.
Charmaz, K. (2006). Constructing grounded theory: A practice guide through
qualitative analysis. Thousand Oaks, CA: Sage.
Glaser, B. G., & Strauss, A. L. (1965). Awareness of dying. Chicago: Aldine.
Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded theory:
Strategies for qualitative research. Chicago: Aldine.
Pidgeon, N., & Henwood, K. (1997a). Using grounded theory in psychological
research. In N. Hayes (Ed.), Doing qualitative analysis in psychology (pp. 245-
273). Hove, UK: Psychology Press.
Top
What’s the difference between constructionist thematic analysis and
discourse analysis?
There is potentially little difference between a constructionist – or ‘critical’ – TA
and a thematic discourse analysis (DA) – such labels could be used to describe
very similar (even identical) analytic approaches.
In general, however, we view a constructionist TA as the application of our
recommended analytic procedures (including the identification of codes and
themes) for TA within a constructionist theoretical framework, whereas in
thematic DA, codes, themes and discourses - “underlying systems of meaning”-
are identified (Taylor & Ussher, 2001: 297), and the analytic process might be
more fluid and organic. Furthermore, as with most DA approaches, thematic DA
involves attention to the constructive role of language, and with multiple and
shifting meanings around social objects (Taylor & Ussher, 2001). But it retains a
specific interest in patterned meaning (discourses) within the dataset.
For the novice ‘critical’ researcher, TA is a far more accessible method, as it has
a clearly defined set of procedures, rather than relying on less clearly defined
analytic practices such as ‘craft skills’ (Potter & Wetherell, 1987). Furthermore,
although constructionist or critical TA recognises the constitutive nature of
language and discourse, it does not generally involve a micro analysis of
language use. Therefore, it does not require a technical knowledge of language
practice which is a feature of some forms of DA, and which are useful for
developing the nuances of a wide range of discursive approaches including
thematic discourse analysis.
Critical TA can be applied flexibly within a wide range of critical frameworks,
from symbolic interactionism and psychodynamism (e.g., Nicolson & Burr, 2003)
to feminist poststructuralism (e.g., Farvid & Braun, 2006). It can be used to
analyse almost all forms of qualitative data (and there is no ideal form of data,
unlike in some versions of DA), both small and large data-sets, and to address
most of the different types of research question posed by constructionist and
critical researchers. This means that critical TA potentially has a wide range of
uses than thematic DA.
Literature
Farvid, P. & Braun, V. (2006). ‘Most of us guys are raring to go anytime,
anyplace, anywhere’: Male and female sexuality in Cleo and Cosmo. Sex Roles,
55, 295-310.
Nicolson, P., & Burr, J. (2003). What is 'normal' about women's (hetero)sexual
desire and orgasm?: a report of an in-depth interview study. Social Science &
Medicine, 57, 1735-1745.
Potter, J., & Wetherell, M. (1987). Discourse and social psychology: Beyond
attitudes and behaviour. London: Sage.
Taylor, G. W. & Ussher, J. M. (2001). Making sense of S&M: A discourse
analytic account. Sexualities, 4(3), 293-314.
We have also written about the differences between thematic discourse analysis
and constructionist TA here:
Clarke, V. & Braun, V. (2014) Thematic analysis. In T. Teo (Ed.), Encyclopedia of
Critical Psychology (pp. 1947-1952). New York: Springer.
Top
What’s the difference between thematic analysis and (qualitative)
content analysis?
As a starting caveat, it’s important to recognise that TA is not one approach –
there are three different broad approaches (which we call ‘coding reliability’,
‘codebook’ and ‘reflexive’); Our approach is a form of reflexive TA. So how you
make sense of the differences between content analysis and TA, in part depends
on how you define TA (and content analysis).
Historically the terms 'content analysis', 'qualitative content analysis' and
'thematic analysis' have been used interchangeably to refer to often rather
similar approaches to qualitative data analysis. The terms content
analysis/qualitative content analysis are less popular among qualitative social
researchers, particularly in countries where qualitative approaches have
flourished in the last few decades, and where 'branded' approaches to qualitative
analysis (e.g., grounded theory; interpretative phenomenological analysis) have
developed and provided researchers with systematic procedures for analysing
qualitative data. The term 'thematic analysis' is now associated with a distinctive
set of procedures - the Braun and Clarke approach is the most widely used, but
the coding reliability approach advocated by US researchers such as Boyatzis
(1998) and Guest et al. (2012) (and Joffe in the UK) has also grown in
popularity over the last decade or so. Content analysis is less clearly branded,
with lots of different versions and varieties (none of which are particularly widely
used, within the qualitative social/health research fields). CA and TA can be very
similar/identical, or they can be very different: it depends on how people make
sense of, and use, both of these methods. In practice, CA seems to have most in
common with more structured coding reliability and codebook approaches to TA.
This means that by and large reflexive TA offers researchers a distinctive
method. Like TA, CA is seen as a method rather than a methodology; however,
the theoretical underpinnings of (particular enactments of) CA are rarely
discussed. Thus, CA is often understood as an atheoretical method (at the same
time as positivist assumptions are often imported into the method through the
use of coding reliability measures). Our view is that qualitative analysis can
never be atheoretical, we always make theoretical assumptions whether
acknowledged or not. For this reason we advocate for the use of TA over CA as it
is understood as a theoretically flexible method not an atheoretical one.
Top
Doing Reflexive TA
»I’ve collected five interviews – is that enough for a TA?
»Why are we critical of the notion that ‘themes emerged’ from data?
»Why don’t we advocate multiple-coders and inter-rater reliability for
reflexive TA?
»How many themes should I have?
»Are latent codes/themes better than semantic ones?
I’ve collected five interviews – is that enough for a TA?
The size of your sample does depend on the size and scope of your project, but
generally 5 interviews (however long and detailed) is too small a sample size for
a TA. This is because TA focuses on the identification of patterns across data. We
provide detailed information about sample size in our book Successful qualitative
research – but in brief, we recommend a sample size of:
• 6-10 interviews for a small TA project (e.g., UK undergraduate; NZ Honours)
• 10-20 for a medium TA project (e.g., UK or NZ Masters; UK Professional
Doctorate)
• 30+ for a large TA project (e.g., NZ or UK PhD; NZ Professional Doctorate).
You also need to keep in mind that if you are interested in publishing your
research, a smaller sample size may limit your choice of publication outlets to
less prestigious journals, and journals that specialise in publishing qualitative
research. More prestigious journals, North American journals and journals that
don’t specialise in publishing qualitative research are likely to expect larger
samples. We’ve had experiences where TA sample sizes of around 20
participants have been rejected outright from journals as being ‘too small a
sample’.
Top
Why are we critical of the notion that ‘themes emerged’ from data?
Within our reflexive TA approach, we are very clear that we do not conceptualise
themes as ‘emerging’ from data and the idea that they do is problematic (e.g.
Braun & Clarke, 2006, 2013). This language suggests that meaning is self-
evident and somehow ‘within’ the data waiting to be revealed, and that the
researcher is a neutral conduit for the revelation of said meaning. In contrast,
we conceptualise analysis as a situated and interactive process, reflecting both
the data, the positionality of the researcher, and the context of the research
itself. Our position around this is not unique or particularly radical. Researchers
within a qualitative paradigm tend to treat research as a subjective process.
Given that understanding, it is disingenuous to evoke a process whereby themes
simply emerge, instead of being active co-productions on the part of the
researcher, the data/participants and context.
Our objection applies specifically to people who claim that ‘themes emerged’
when using our reflexive TA approach. Although this position reflects our
conceptualisation around qualitative research process, it’s not universal; other
qualitative analytic approaches do use the language of emergence (e.g. IPA) in a
specific way, often to express the idea that themes were generated inductively
from data.
Top
Why don’t we advocate multiple-coders and inter-rater reliability for
reflexive TA?
The use of inter-rater reliability is underpinned by the (realist/positivist)
assumption that there is an accurate reality in the data that can be captured
through coding. Our approach to TA sees coding as flexible and organic, and
coding should evolve throughout the coding process. We understand coding as
an active and reflexive process that inevitably and inescapably bears the mark of
the researcher(s). With no one ‘accurate’ way to code data, the logic behind
inter-rater reliability (and multi-independent coders) disappears. We argue that
inter-rater reliability scores can be understood as showing that two researchers
have been trained to code data in the same way, rather than that their coding is
‘accurate’.
Top
How many themes should I have?
Unfortunately, there is no precise formula for determining the number of themes
in a TA, as it depends on the data, the research question, the length of the
analytic report, and the focus of the analysis…
A TA can range from a detailed reporting of one theme (usually with subthemes)
to a comprehensive overview of dataset, in relation to your research question. In
the former, you will go into more depth; in the latter, probably less.
But regardless, you don’t want your analysis to be thin, and fall victim to
common errors in TA, such as paraphrasing the data, and not actually providing
an interpretative analysis. If you have a report of about 10,000 words, an
overview is unlikely to be able to sufficiently cover more than six themes in any
depth. In general, 2-6 themes (and subthemes) is about right for a single
journal article, an undergraduate project, an Honours or Masters dissertation,
and a single analytic chapter in a doctoral thesis.
Top
Are latent codes/themes better than semantic ones?
TA comes in lots of different shapes and forms. None are inherently better than
any other, but they serve different purposes. If, in your analysis, you want to
present a more realist and descriptive account of participants’ experiences of
eating, for example, then semantic codes are likely better than latent ones.
If you want to present a more constructionist account of the assumptions
underpinning media reports of women and weight gain, then latent codes are
likely better. Latent codes allow researchers to move away from the explicit and
obvious content of the data. Regardless of approach however, your analysis
needs to interpret the data and make sense of them for the reader.
Top
Reflexive TA: What is it good for?
»What types of data is reflexive thematic analysis suitable for?
»What types of research questions is reflexive thematic analysis suitable
for?
What types of data is reflexive thematic analysis suitable for?
There is no ideal type of data for a TA analysis. TA can be used to analyse most
types of qualitative data including interviews, focus groups, qualitative surveys,
story completion tasks, diaries, vignettes, and a wide range of secondary
sources including printed materials, online and electronic materials, and
broadcast media and film.
Top
What types of research questions is reflexive thematic analysis
suitable for?
We can understand qualitative research as addressing different types of research
questions, and TA can be used to address most of these, including questions
about:
• Individual experiences (e.g. what is it like to experience depression? How do
people make sense of being fat?)
• People’s views and opinions (e.g. what are people’s views on living near
wind farms? How do people make sense of transracial adoption?)
• People’s behaviours or practices – the things they do in the world, or their
accounts or perceptions of their practices (e.g., How do married
heterosexual couples manage their household finances? What role does
clothing play in the development of a gay or lesbian identity?)
• The reasons why people think or feel or do particular things and the factors
or processes that underpin and shape particular experiences or decisions
(e.g., what factors influence the decision to become a vegetarian? How do
social norms shape the experience of chronic pain?)
• Identifying and exploring the rules and norms that govern particular social
practices (e.g., what are the expectations and conventions governing the
behaviour of new fathers? What norms and expectations do people navigate
in using online ‘dating’ sites?)
• How particular social objects are represented in particular contexts (NB a
social object can be both something concrete and something abstract) (e.g.,
How is ‘female sexuality’ represented in men’s magazines? How is ‘the
family’ represented in Christmas advertising?)
• How social objects are constructed/the discourses surrounding a particular
social object (How is the notion of a ‘healthy body weight’ constructed in
focus groups with young women? How do young men construct
‘masculinity’?).
• What all these questions have in common is a focus on what people say, on
the content of the language used in the data. The only type of research
question that does not suit a TA is one focused on language practice – how
people say things. Language practice questions are typically associated with
discursive psychological and conversation analytic approaches (and some
forms of narrative analysis).
Top
Quality in reflexive TA research
»What are some of the common problems I should avoid when using
reflexive TA?
»How do I know if I have done a good reflexive thematic analysis?
»Should someone else check my coding?
»Should I use numbers when reporting themes?
What are some of the common problems I should avoid when using
reflexive TA?
There is now a lot of published TA. In looking at your own, and evaluating the TA
of others, the following questions will help in identifying whether common
problems in TA have been avoided:
• Have the data actually been analysed, or have they just been reported and
paraphrased? Does the analysis go beyond the content of the data to tell the
reader something about what the data might mean, and the implications of
the patterns identified? Data need to actually be analysed.
• Does the analysis go beyond data collection questions? Does the analysis
report patterns that don’t simply translate from the questions participants
were asked? In general, patterns should not map directly from data
collection questions.
• Does the researcher claim the themes ‘emerged’? This is a key problem, as
it suggests themes ‘exist’ in the data waiting to be discovered, when
analysis is really a more active process of developing themes through our
interaction with the data.
• Are there a lot of data extracts and very little analytic narrative? Or are
analytic claims made without supporting evidence from the data? Aim for a
good balance of data and analytic narrative.
• Are themes weak and poorly realised? Are they thin and sketchy or overly
complex and unwieldy? Do they lack coherence? Aim for themes that are
coherent and have a clear and identifiable central organising concept.
• Is there lots of overlap between the themes? Is the relationship between the
themes unclear? Do the themes fail to work together to tell a story about
the data? Aim for a set of themes in which each theme distinctive from the
other themes, but there is also a clear relationship between the different
themes.
• Overall, do the themes provide a rich and coherent analysis that answers
the research questions? The TA needs to provide a compelling and plausible
answer to the research question.
• Is the analysis plausible? Will the reader be able to see what you claim to
see in the data? Your analytic claims need to be able to be ‘seen’ in and
evidenced by the data.
• Are the assumptions around TA – and the form of TA used – clearly spelled
out? The approach you use needs to be explained. Keep in mind, there is
more than one type of TA; avoid ‘mashing-up’ approaches that have
different philosophies and procedures.
• Is the analysis theoretically coherent? Do the claims made about the
approach to TA used (inductive vs. deductive; latent vs. semantic;
essentialist vs. constructionist) and how the data are actually analysed, fit?
Your use of TA needs to be theoretically coherent and consistent (e.g., if you
say your analysis is deductive and constructionist, it needs to be that).
Simply put, you need to do what you say you do!
We have written a checklist for editors and reviewers, but also for researchers,
to help people evaluate TA papers submitted for publication and to ensure high
standards in published (reflexive) TA. This checklist will also be useful for
anyone critically evaluating a TA report.
Top
How do I know if I have done a good reflexive thematic analysis?
If you follow the guidance we have provided for doing reflexive TA (e.g., Braun &
Clarke 2006, 2012, 2013) and use the checklist below to avoid ‘common
problems’, your will have followed a robust process, and you are well on your
way to doing a good TA. However, there are no guarantees. Qualitative analysis
of all types relies on interpretative critical analytic skills of the researcher.
Therefore, the quality of a TA also depends on your analytic insights… So doing a
good TA is a combination of following a robust process, applying an analytic eye
to the data, and interpreting it light of what we already know about the issue(s)
being explored.
A 15-point checklist of criteria for good thematic analysis (reproduced from
Braun & Clarke, 2006)
1. Transcription: The data have been transcribed to an appropriate level of
detail, and the transcripts have been checked against the tapes for
‘accuracy’
2. Coding: Each data item has been given equal attention in the coding process
3. Themes have not been generated from a few vivid examples (an anecdotal
approach), but instead the coding process has been thorough, inclusive and
comprehensive
4. All relevant extracts for all each theme have been collated
5. Themes have been checked against each other and back to the original data
set
6. Themes are internally coherent, consistent, and distinctive
7. Analysis: Data have been analysed – interpreted, made sense of - rather
than just paraphrased or described
8. Analysis and data match each other – the extracts illustrate the analytic
claims
9. Analysis tells a convincing and well-organised story about the data and topic
10. A good balance between analytic narrative and illustrative extracts is
provided
11. Overall: Enough time has been allocated to complete all phases of the
analysis adequately, without rushing a phase or giving it a once-over-lightly
12. Written report: The assumptions about, and specific approach to, thematic
analysis are clearly explicated
13. There is a good fit between what you claim you do, and what you show you
have done – i.e., described method and reported analysis are consistent
14. The language and concepts used in the report are consistent with the
epistemological position of the analysis
15. The researcher is positioned as active in the research process; themes do
not just ‘emerge’.
In addition, our more recently developed guidelines for editors and reviewers to
assess submitted manuscripts provides another guideline towards quality
reflexive TA.
Literature
Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology.
Qualitative Research in Psychology, 3(2), 77-101.
Braun, V., & Clarke, V. (2012). Thematic analysis. In H. Cooper (Ed.),
Handbook of Research Methods in Psychology (Vol 2: Research Designs, pp. 57-
71). Washington, DC: APA books.
Braun, V. & Clarke, V. (2013). Successful qualitative research: A practical guide
for beginners. London: Sage.
Top
Should someone else check my coding?
It depends on what you mean by ‘check’! If you mean check to see if it is
accurate or reliable, or that the person checking ‘agrees’ with how you have
coding your data, then we would say no. The assumptions underpinning that
type of ‘check’ are not consistent with our reflexive TA approach, which
foregrounds researcher subjectivity in the coding and theme development
processes. From this perspective, coding can’t be right or wrong, but it can be
weaker (superficial) or stronger (complex, nuanced, insightful). Coding is always
reflective of the subjectivity of the researcher, and one’s person’s subjectivity
can’t be correct and another’s incorrect.
However, if you conceptualise ‘check’ as sharing your coding with an
experienced supervisor (say) and using this as an opportunity to reflect on how
you have coded the data, the assumptions you’re making in coding the data, and
things you might have overlooked, then this is a good idea. Working with a more
experienced researcher can help us to develop our skills in coding and theme
development.
Top
Should I use numbers when reporting themes?
Qualitative researchers often use expressions like ‘a common theme…’, ‘in a
majority of the texts analysed…’, or ‘many participants commented that…’ when
reporting TA. But wouldn’t it be better to use expressions like ‘25/34
participants…’ or ‘45% of the texts…’? Although occasionally reporting
percentages or frequencies is useful, in general we argue no, it not better to
report percentages or frequencies. As the reason for this stance isn’t necessarily
obvious, and can be a big issue, especially as those outside a qualitative
paradigm often think that indicating the actual number of participants/data
items or proportion of the dataset reporting a theme provides more robust
evidence and validity for a qualitative analysis, we will explain our position.
There are lots of reasons:
• It reflects an anxiety about the validity of qualitative research practice, to
some extent suggesting that somehow our analysis might not be real (i.e., it
might be ‘made up’), or our themes might be ‘anecdotal’ rather than
patterned. But all research (qualitative or quantitative) relies on trust,
honesty, and good research practices. Reporting actual numbers does not
circumvent this issue.
• We agree with Australian health researcher Prisicilla Pyett (2003) who
argued that “counting responses misses the point of qualitative research” (p.
1174), as frequency does not determine value.
• Moreover, whether something is insightful or important for answering our
research questions is not necessarily determined by whether large numbers
of people said it.
• Finally, because of the nature of qualitative data, we cannot assume what
the absence of a certain meaning or theme in the data actually means.
Consider the difference between a quantitative survey, and qualitative data
collected in interview or focus group. In a quantitative survey, you ask
people to select from a number of options. Reporting, and comparing, the
proportions who select each of a series of response options is meaningful,
because they have all been asked the same thing, and given the same
response options. With an interview or focus group, the data generated from
each participant can be quite different. Because, for example interviews are
fluid, flexible, and interactive data collection tools, it’s not the case that
every participant in an interview study discusses exactly the same issues. So
if someone doing an interview study with 20 men reported that ‘12 of the
men thought …’, we can’t assume that the remaining eight men didn’t think
this, or thought the opposite – they may have just not discussed it. So we
have no way of interpreting what is not reported in qualitative data, and this
makes reporting numerical proportions somewhat deceptive and
disingenuous.
Literature
Pyett, P. M. (2003). Validation of Qualitative Research in the “Real World”.
Qualitative Health Research, 13(8), 1170-1179.
Top
Teaching and supervision
»Should I teach my students thematic analysis?
»Is thematic analysis sophisticated enough for a doctoral project?
Should I teach my students thematic analysis?
Yes! We teach TA to both our undergraduate and postgraduate students, and our
interest in TA as a method developed from our experiences of teaching TA to our
students.
In many ways, TA is an ideal ‘starter’ analytic method for those new to
qualitative research because it is accessible, flexible, and involves analytic
processes common to most forms of qualitative research. The theoretical
flexibility of TA means it can be learned without some of the potentially
bewildering (for new students) theoretical knowledge essential to many other
qualitative analytic approaches. In our experience, understanding the practice of
qualitative (thematic) analysis first seems to allow space for the theory of
qualitative research to make sense for students new to qualitative methods.
Students can progress from learning TA to learning other analytic approaches
such as grounded theory, IPA and discourse analysis, or progress from
producing largely descriptive TA to producing rich and complex, conceptually
informed TA.
We provide some guidance on teaching TA in the following paper: Clarke, V. &
Braun, V. (2013) Teaching thematic analysis: Overcoming challenges and
developing strategies for effective learning. The Psychologist, 26(2), 120-123.
Top
Is thematic analysis sophisticated enough for a doctoral project?
We sometimes hear from students and supervisors that TA is not perceived as
sophisticated enough for doctoral level research, which is why we have included
a FAQ on this issue. TA isn’t inherently sophisticated or unsophisticated. Like
with all qualitative analytic approaches, what really matters, is how you
implement it. We emphasise the word ‘you’ because the active role of the
researcher is key to the successful implementation of any qualitative analytic
approach.
We have puzzled over why people perceive TA as unsophisticated and we think
this view is based on a fundamental misunderstanding of TA. TA is often
assumed to be a realist/essentialist method or simply an atheoretical method for
identifying patterns in data. But this is not the case! TA is theoretically flexible,
which means it can be used across a wide range of theoretical approaches. What
is the key here is that the researcher chooses which theoretical framework will
inform their use of TA. TA can be used to produce relatively straightforward
descriptive overviews of the key features of the semantic content of data (within
a critical realist/essentialist framework) or complex and sophisticated conceptual
interrogations of the underlying meaning in data (within say a constructionist
framework).
As this indicates, TA is not one thing – it can be used to produce a relatively
straightforward (what some might call ‘unsophisticated’) analysis or one that is
as sophisticated as an IPA, grounded theory or discourse analysis. So yes, TA
definitely provides the tools to do analysis sophisticated enough for a doctoral
project (assuming it’s used well!).
Top
