Hw.2

https://doi.org/10.1177/1049732318788379

Qualitative Health Research
2018, Vol. 28(13) 2094 –2101
© The Author(s) 2018
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/1049732318788379
journals.sagepub.com/home/qhr

Commentary

In one word, to draw the rule from experience, one must
generalize: This is a necessity that imposes itself on the most
circumspect observer.

―Henri Poincaré

Between the 1970s and the early 1990s, advancements in
computer technology facilitated the processing of large
data sets in quantitative studies (Morse, 2008; Ormston,
Spencer, Barnard, & Snape, 2014). This fact, together
with the increased interest of evidence-based decision
making in medical and health fields in the 1980s, and a
widespread diffusion of the systematic review methodol-
ogy in scientific research (Bradt, 2009), has made gener-
alizability of findings, already a valued standard for
decades, a pivotal criterion for evaluating the excellence
of quantitative research (Polit & Beck, 2010).

Since then, researchers in the qualitative tradition have
begun to pay greater attention to discussions on the feasibil-
ity of generalization in qualitative studies (Flyvbjerg, 2001).
However, this feasibility of generalizing findings could be
seen as a “phantom,” as an “illusion,” for qualitative schol-
ars (Groleau, Zelkowitz, & Cabral, 2009, p. 417). Indeed,
without due precautions and clarifications, the nature of
interpretivist paradigm, relying on a different logic to the
positivist paradigm, could make generalizability almost an
unachievable aim in qualitative research (Morse, 2008).

Trying to define the thorny and illusive concept of gen-
eralizability may represent a challenge to researchers
because of the ambiguity and polyhedral meanings associ-
ated with it (Polit & Beck, 2010). Simply put, the act of
generalizing involves forming general and broad statements
from specific cases (Schwandt, 2001), which also means
that inferences about what cannot be observed are made on
the basis of what can be observed (Polit & Beck, 2010).
Defined in this way, the notion of generalizability has
neither a quantitative nor a qualitative dimension per se, but
it has, instead, a neutral and impartial connotation.

Nevertheless, the dominant role of the positivist tradi-
tion in social sciences has led generalizability to acquire
a more quantitative nature (Kerlinger & Lee, 2000). Thus,
generalizability has become almost inherently associated
with quantitative research and its focus on finding univer-
sal laws and statistical generalizations (Delmar, 2010).
Consequently, due to this positivist echo, the application
of the same concept of generalizability to qualitative per-
spectives, grounded in an interpretivist paradigm, has

788379QHRXXX10.1177/1049732318788379Qualitative Health ResearchCarminati
article-commentary2018

1University of Surrey, Guildford, United Kingdom

Corresponding Author:
Lara Carminati, Surrey School, University of Surrey,
388 Stag Hill, Guildford GU2 7XH, UK.
Email: [email protected]

Generalizability in Qualitative
Research: A Tale of Two Traditions

Lara Carminati1

Abstract
Generalizability in qualitative research has been a controversial topic given that interpretivist scholars have resisted the
dominant role and mandate of the positivist tradition within social sciences. Aiming to find universal laws, the positivist
paradigm has made generalizability a crucial criterion for evaluating the rigor of quantitative research. This positivist
echo has led generalizability to acquire a quantitative meaning, inappropriate for describing qualitative studies. The
purpose of qualitative research has, thus, been directed toward providing in-depth explanations and meanings rather
than generalizing findings. Through a critical review of empirical and theoretical studies, this commentary seeks to
show that in qualitative domains, generalizability is possible provided that, first, generalizability is the main objective
of the study; second, due precautions concerning the philosophy and terminology selected are taken. Hence, this
commentary contributes to the literature on qualitative research by making suggestions for more consistent and
unanimous procedures to adopt in qualitative inquiries.

Keywords
qualitative and quantitative research; positivist and interpretivist tradition; probabilistic generalizability; theoretical
generalizability; transferability; critical review; the UK

http://us.sagepub.com/journals-permissions

https://journals.sagepub.com/home/qhr

mailto:[email protected]

http://crossmark.crossref.org/dialog/?doi=10.1177%2F1049732318788379&domain=pdf&date_stamp=2018-07-25

Carminati 2095

become rather controversial (Polit & Beck, 2010). As a
result, generalizability in qualitative studies has often
been misunderstood and disregarded in favor of the more
common purpose of providing in-depth understanding of
the specific topics under research (Ayres, Kavanagh, &
Knafl, 2003).

However, dismissing a priori the ability to generalize
from qualitative studies on the bases of biased premises
due to “quantitative contamination” can significantly
limit the strength and beauty of qualitative research, espe-
cially in the eyes of early-career researchers. A clear con-
ceptualization of the meanings associated with the
concept of generalizability in qualitative research is
needed, taking into account both methodological and
philosophical boundaries.

Therefore, this commentary aims to show that even in
qualitative domains, generalizability is possible provided
two paramount conditions exist. First, generalizability is the
main objective of the study. Second, due precautions related
to the terminology selected are taken into consideration.
Concerning these due precautions, on one hand, if research-
ers want to preserve the term generalizability in qualitative
enquiries, the philosophical traditions underpinning the
research have to be clearly specified. In so doing, the kind
of generalizability that can be attained, namely, analytical/
theoretical, and the type of knowledge that can be pursued
in qualitative research would be straightforward and
unequivocal. On the other hand, if new terms were to be
introduced, a suitable vocabulary should be unanimously
shared by all qualitative scholars to unmistakably describe
quality in qualitative enquiries. Hence, this commentary
contributes to the literature on qualitative research
approaches by making suggestions on those paramount
conditions that may lead to more consistent and unanimous
procedures to describe and elaborate qualitative research.

The commentary unfolds in four main parts. In the first
section, “Generalizability in the Positivist, Quantitative
Tradition,” after providing a brief definition of the term,
the value of generalizability within the positivist tradition
is analyzed. The second section “Generalizability in the
Interpretivist, Qualitative Tradition” investigates the nature
of generalizability in qualitative approaches in light of
interpretivist perspectives. In the third section, “Grounded
Theory, a Theory-Generating Research Method,” an exam-
ple of a theory-generating research method, that is,
grounded theory, within the interpretivist tradition is out-
lined. The fourth section “Due Precautions in Qualitative
Investigations” tackles two main precautions researchers
should take when conducting qualitative investigations.
This part specifically proposes the adoption of the notion
of contextual knowledge as phronesis and the diffusion of
a terminology for qualitative research universally approved
by the scientific community. The section “Implications and
Conclusion” closes the commentary.

Generalizability in the Positivist,
Quantitative Tradition

The positivist tradition in social sciences has strongly
polarized the notion of generalizability and influenced
the terminology in scientific research (Flyvbjerg, 2001;
Ormston et al., 2014). Since the Enlightenment, natural
sciences, which are nomothetic by essence (Thomae,
1999), have indisputably dominated Western culture,
drawing attention to instrumental rationality and the
focus on generalizable and context-independent theories
(Falk, Rocha, & Warnick, 2009). In an effort to emulate
natural science (Flyvbjerg, 2001), the positivist tradition
in social sciences, which traces back to philosophers such
as Comte, Hume, and Bacon, has argued and acknowl-
edged the possibility of reaching one, monolithic truth
(Lee & Baskerville, 2003). Positivism assumes that “sci-
entific knowledge is the paradigm of valid knowledge”
(Larrain, 1979, p. 197), and, as such, authentic knowl-
edge can be verified through empirical evidence, which is
in turn interpreted through reason and logic (Macionis &
Gerber, 2010).

Consequently, the ultimate aim of quantitative research
defined by the positivist tradition is to produce laws able
to explain and govern every observed phenomenon and to
determine a universal knowledge (Delmar, 2010) that
holds true, and is invariable in, all places and at all times
(Falk et al., 2009). In this sense, as Morse (2008) has
underlined, generalizability is at the heart of usefulness.
Hence, by assuming an impartial and objective research-
er’s involvement (Winter, 2000) to limit any possible
forms of influence in the research process (Davies &
Dodd, 2002), quantitative experimental methods have
been employed to test hypotheses and infer generalizable
conclusions (Dingwall, Murphy, Watson, Greatbatch, &
Parker, 1998; Hoepfl, 1997).

These generalizable conclusions stem from a research
process based on methodological rigor. Rigor in the posi-
tivist tradition refers to the soundness, exactitude, and
accuracy of a study with regard to its planning, data col-
lection, analysis, and reporting (Marquart, 2017). The
assessment of objective scores, such as reliable and valid
measurements, are indicators of such exactitude, so that
results from one study can be replicated in other studies
and then generalized (Claydon, 2015).

Results in quantitative research can be generalized
through probabilistic generalization (Polit & Beck, 2010),
which is based on randomly selected samples representa-
tive of the population (Lee & Baskerville, 2003). Although
some authors have claimed that convenience sampling bias
could hamper the validity of findings in many studies
(Gheondea-Eladi, 2014), statistical generalizability remains
a crucial tenet and a prerogative in quantitative approaches
(Winter, 2000). This probabilistic generalization is also

2096 Qualitative Health Research 28(13)

termed external validity (Gheondea-Eladi, 2014). External
validity, together with internal validity, namely, the degree
to which it is possible to infer causality between two vari-
ables (Bryman, Becker, & Sempik, 2008; Joppe, 2000), and
reliability, which describes the extent to which results are
consistent over time and accurately represent the total pop-
ulation (Bryman et al., 2008; Joppe, 2000), is a paramount
criterion for evaluating quality in quantitative studies
(Gheondea-Eladi, 2014; Polit & Beck, 2010).

As a result, generalizability has become associated
with the notion of external validity, which represents only
a specific aspect of the original term, namely, probabilistic
generalizability (Flyvbjerg, 2001; Lee & Baskerville,
2003). Hence, generalizability has been pulled toward the
positivist approaches, and their quantitative influence has
been instilled. It is no surprise that when the term general-
izability is mentioned, statistical generalization is the first
notion to come into one’s mind, as if synonymous with it.

Generalizability in the Interpretivist,
Qualitative Tradition

The highlighted importance of the word generalizability
in quantitative studies has led many qualitative scholars
to question the possibility of generalizing results from
qualitative research (Davies & Dodd, 2002) and, conse-
quently, ponder on how to judge its quality (Hallberg,
2013). As Ormston et al. (2014) have underlined, qualita-
tive approaches in social sciences have started to emerge
as recent phenomena alongside the widespread diffusion
of sophisticated statistical methods, which were framed
within positivist principles to imitate natural sciences
methodology.

Nonetheless, the philosophical tradition behind qualita-
tive research, drawing from authors such as Weber, Kant,
and Dilthey is the interpretivist paradigm (Bryman et al.,
2008; Morse, Barrett, Mayan, Olson, & Spiers, 2002) and,
in an antipodal way to the positivist perspective, is idio-
graphic in nature (Thomae, 1999). Interpretivism priori-
tizes the understanding of human behaviour over the
prediction and generalization of causes and effects
(Macionis & Gerber, 2010).

As such, the strength of qualitative inquiries defined
by the interpretivist tradition is the understanding of how
individuals, through their narratives, perceive and experi-
ence their lives, constructing meanings within their social
and cultural contexts (Groleau et al., 2009; Mishler,
2000). In this sense, interpretivist research emphasizes
the hermeneutics and perception of the social world, and
the interactions between individuals and the surrounding
context (Ormston et al., 2014). It follows that findings are
intrinsically linked to the research context and that the
interfacial distance between the researcher and the
researched is minimized (Davies & Dodd, 2002).

Consequently, the researcher is an agentic instrument
for collecting and analyzing data and an essential tool for
actively constructing concepts (Hallberg, 2013). Thus, by
using either theoretical or nonprobabilistic samples
(Gheondea-Eladi, 2014), qualitative studies aim to
explore meanings and processes of people’s everyday
lives to gain an in-depth understanding of situations and
actions (Groleau et al., 2009; Hallberg, 2013). Because of
this stress on values and subjective meanings, qualitative
research has often been criticized as being relativistic,
soft, or unscientific compared with quantitative enquiries
(Ormston et al., 2014).

To avoid this subordinated position, the same stan-
dards used to judge quality in quantitative research have
been transposed and employed to evaluate quality in
qualitative studies (Delmar, 2010), so that generalizabil-
ity has become a rather pivotal matter even in an area
where it is beset by indistinctness (Delmar, 2010;
Golafshani, 2003). In fact, this osmotic transfer has raised
numerous issues related to its fundamental inadequacy in
depicting the true nature of qualitative research processes
and aims (Kitto, Chesters, & Grbich, 2008). Winter
(2000) has stated that “qualitative research sets itself up
for failure when it attempts to follow established proce-
dures of quantitative research” (pp. 11). Indeed, by doing
so, interpretivist social sciences have accepted terms that
are by nature self-defeating (Flyvbjerg, 2001).

As a result, the importance of generalizability of find-
ings in the qualitative field has been frequently dismissed,
considered unattainable or irrelevant (Kitto et al., 2008)
and not the purpose of quality enquiries (Morse, 1999).
Researchers have, thus, preferred to invert their focus to
the other fundamental aim of qualitative studies, which is
to engage in research able to produce thorough descrip-
tions as well as a deep, rich, and contextualized under-
standing of human experience (Polit & Beck, 2010).

However, (a) if the purpose of a study and its research
questions aim to build a new theory to bridge a gap in the
literature (Gheondea-Eladi, 2014) or (b) if due precau-
tions of the conceptualization of generalizability are
respected (Hallberg, 2013), then discussing generaliz-
ability of findings even in qualitative research makes
sense (Delmar, 2010; Krefting, 1991; Morse, 1999; Polit
& Beck, 2010; Winter, 2000). The following two sections
expand on these two paramount conditions to clarify gen-
eralization in the interpretivist paradigm.

Grounded Theory, a Theory-
Generating Research Method

With regard to the first point of building a new theory,
within the many different qualitative methods (Krefting,
1991), theory-generating procedures seek to generalize
findings (Eisenhardt, 1989; Glaser & Strauss, 1967;

Carminati 2097

Hallberg, 2013; Ormston et al., 2014). An example of
these theory-generating procedures is grounded theory
(Corbin & Strauss, 2008; Eisenhardt, 1989; Glaser &
Strauss, 1967). Although the aim of this commentary is
far beyond providing an exhaustive description of
grounded theory, some key features of this theory-gener-
ating method are now discussed to provide evidence and
strength to the argument that generalizability is achiev-
able also in qualitative research. This is feasible since the
purpose of grounded theory is to inductively construct a
general theory that is able to answer the research question
in the absence or incompleteness of alternative existing
frameworks (Gheondea-Eladi, 2014).

Starting from thick descriptions and then going
beyond through to high levels of abstraction (Goulding,
2002), grounded theory develops an iterative process
based on four main steps, namely, coding, conceptualiza-
tion, classification, and categorization, to build a sub-
stantiated theory (Gheondea-Eladi, 2014). After this first
theoretical result, the research is replicated in other set-
tings so that several other substantiated theories can be
yielded (Goulding, 2002). Indeed, as Firestone (1993)
has noted, replication in different conditions reinforces
generalizability in qualitative investigations, since con-
sistency of results when conditions vary indicates that
findings are robust. This also explains the reason why
extreme cases, those that due to their extreme character-
istics accentuate the dynamics being investigated facili-
tating the emergence of the aspects of interest (Eisenhardt,
1989), are so critical for enhancing qualitative research
(Polit &Beck, 2010).

When every observation has been incorporated and no
unexplained variance has been left out, then saturation is
reached (Dingwall et al., 1998). Saturation implies that
sufficient and redundant information for all aspects of the
phenomenon under research has been gathered (Morse
et al., 2002), and thus, no further themes or concepts can
emerge (Glaser & Strauss, 1967). At this point, a formal,
final theory can be elaborated (Corbin & Strauss, 2008).

The formulation of a new theory is possible due to the
action of abstracting the “general” from the “particular”
in terms of similarities and differences in individuals and
circumstances (Delmar, 2010). From this perspective, the
power of generalization of the final theory not only
depends on the researcher’s thorough immersion and
engagement with the data (Polit & Beck, 2010) but also
on the intensity of within-case and across-case analyses
(Ayres et al., 2003). As Morse (1999) has argued, “if
qualitative research was considered not generalizable,
then it would be of little use, insignificant and hardly
worth doing” (p. 5). Hence, there would be a high risk of
falling into an endemic relativism (Davies & Dodd,
2002). Nevertheless, as grounded theory has shown, gen-
eralizability of findings in qualitative studies is possible:

It is simply a different kind of generalizability compared
with that of quantitative research (Polit & Beck, 2010). It
is an analytic, or theoretical, form of generalization.

Due Precautions in Qualitative
Investigations

Given the fact that research undertaken in the interpretiv-
ist tradition can aim for theoretical generalizability, how
are researchers able to avoid confusion when using such
an ambiguous and misused term? What can be done to
properly describe qualitative studies, releasing them from
the pervasive influence of positivist perspectives?

First Precaution

The first due precaution is related to the fact that if
researchers want to preserve the term generalizability
within the qualitative enquiries, then the interpretivist
paradigm should be clearly specified at the beginning of
a study (Hallberg, 2013). Stating the philosophical tradi-
tion underpinning a research determines the kind of gen-
eralizability, that is, analytical/theoretical, that can be
obtained and pursued and prevents misunderstandings
(Hallberg, 2013). In this sense, by recognizing and
acknowledging the existence of this other specific form
of generalization, the quantitative meaning can be bal-
anced out and the originally neutral connotation of the
broad term restored. Restoring the neutral connotation of
the term generalizability emphasizes the view that gener-
alizability in social sciences can never be certain (Polit
& Beck, 2010).

Because of the focus on and importance of the context,
interactions, and hermeneutics, the aim of the interpretiv-
ist tradition is to predict a theoretical understanding of the
topic under examination (Yin, 1994). As such, results are
not reached through statistical procedures or other means
of quantifications (Corbin & Strauss, 2008), and general-
ization is interpreted as generalization toward a theory
rather than toward the population (Polit & Beck, 2010).

This is why the notion of sample size in relation to
generalizability is rather controversial in qualitative
investigations (Malterud, Siersma, & Guassora, 2016).
On one hand, sample sizes may be too large to permit a
deep and detailed analysis; on the other hand, they may
be too small to support claims of having achieved either
theoretical saturation or information redundancy (Boddy,
2016). In addition, data saturation and information redun-
dancy are elusive concepts, and no clear explanations of
how they should be understood or implemented to justify
the number of participants exist (Morse, 1995). Although,
recently, some guidelines have been proposed to deter-
mine “information power” (Malterud et al., 2016), defin-
ing the adequate sample size in qualitative research is

2098 Qualitative Health Research 28(13)

ultimately a matter of the researcher’s experience and
judgment of the quality of the data collected and the
research method employed (Boddy, 2016).

Thus, generalizability in qualitative studies is focused
on the researcher’s analysis and understanding of circum-
stances rather than on the collection of representative data
(Delmar, 2010; Morse, 1999). In this sense, generalization
entails inferring the potential extrapolations, or transfer-
ability, of those results on the basis of both a theoretical
analysis of the aspects generating the outcomes and the
effects of the context (Eisenhardt, 1989; Yin, 1994).

Such contextual effects are crucial in qualitative
enquiries. Since they seek to understand phenomena
unfolding naturally in specific environments (Patton,
2002), declaring the philosophical tradition underpin-
ning a study also has an impact on the type of knowledge
that can be acquired. Within the philosophical tradition
of interpretivism, a theory that pertains only to the set-
ting in which it was developed would not detract from its
scientific status nor would it prohibit researchers from
extending that theory to additional settings (Lee &
Baskerville, 2003). On the contrary, what may work for
people in a certain circumstance is likely to work for
other people placed in a similar situation (Hallberg,
2013; Hyde, 2000).

Therefore, since qualitative understanding relies on a
comprehensive knowledge of the case and its nuances in
each specific context (Hyde, 2000), talking about “situ-
ated” knowledge rather than epistemological knowledge
would then be more appropriate (Davies & Dodd, 2002;
Haraway, 1988). As such, as Flyvbjerg (2001) has sug-
gested, it would be interesting to introduce the Aristotelian
concept of phronesis, value-based, context-dependent
knowledge. Words such as episteme (pure knowledge)
and techne (applied knowledge) have informed the cur-
rent language (i.e., epistemology, technic, technology).
While episteme indicates generalizable-across-settings
knowledge, usually called science, techne refers to practi-
cal, specific knowledge. However, occurrences of the
word phronesis, which sits between episteme and techne,
and encompasses both scientific and technical knowledge
by adding values, experiences, context, and reflexivity,
do not exist in the present language (Flyvbjerg, 2001).

It follows that the idea according to which true knowl-
edge is only that which can be replicated in different
places and times has to be rejected. Instead, context-
dependent knowledge, with its different modes of expres-
sion, can offer a likewise true understanding, but in
another form in terms of its replicability and applicability
(Delmar, 2010). Thus, in the interpretivist tradition of
social science, for which the constant interaction between
human agents and the surrounding context is essential, it
would be more suitable to talk about phronetic knowl-
edge (Flyvbjerg, 2001). Because of the link with the

setting, a context-dependent knowledge would do justice
to the essence of qualitative enquiries and their conceptu-
alization of generalizability.

Second Precaution

The second precaution is related to the introduction of
new terms. It refers to some qualitative scholars’ plea to
select a novel and unanimously accepted vocabulary to
better indicate the principles underpinning qualitative
approaches and evaluate their quality (Agar, 1986;
Golafshani, 2003; Seale, 1999). Quality is in fact a hall-
mark of every piece of research (Hallberg, 2013), and
“each paradigm should be judged by its own paradigm’s
terms” (Golafshani, 2003, p. 601). This calls into ques-
tion the use of the term generalizability itself.

Some researchers, echoing Shakespeare’s “What’s in a
name?”1 (Romeo & Juliet, Act II, Scene II, v. 47), have
argued that the positivist terminology, such as the head-
ings reliability, validity, and rigor, could be still used in
interpretivist qualitative research and just redefined in
terms of meaning (Guba & Lincoln, 1981; Morse et al.,
2002). For instance, while reliability could indicate the
“negotiation of truths through a series of subjective
accounts,” validity could reside “with the representation
of the actors, the purposes of the research and appropriate-
ness of the processes involved” (Winter, 2000, pp. 9–10).

Nonetheless, some new labels have also been pro-
posed to cut ties with the positivist tradition and under-
line the independent identity of the interpretivist
perspective (Agar, 1986). Indeed, qualitative research
often breaks with and challenges many of the tenets of
so-called good quantitative research, as it is the case
with the notion of reliability and validity (Davies &
Dodd, 2002; Golafshani, 2003). Hence, a plethora of
alternative concepts have arisen to mark the initially
moderate and then radical paradigm shift (Seale, 1999),
and an alternative terminology, able to express qualita-
tive connotations, has been supported. For example, the
word trustworthiness could be considered as an alterna-
tive to the qualitative term for rigor (Davies & Dodd,
2002; Morse et al., 2002). Trustworthiness has been pro-
posed and initially defined by Guba and Lincoln (1981)
through four criteria: truth value (corresponding to
quantitative internal validity), applicability (quantita-
tive external validity or generalizability), consistency
(quantitative reliability), and neutrality (quantitative
researcher’s objectivity and distance).

To relate this term even more closely to the interpretiv-
ist tradition, the same authors have further modified the
four-point list of criteria (Lincoln & Guba, 1985). This
modified and well-established model comprises concepts
such as credibility (prior truth value—quantitative internal
validity), transferability (prior applicability—quantitative

Carminati 2099

external validity or generalizability), dependability (prior
consistency—quantitative reliability), and confirmability
(prior neutrality—quantitative researcher’s objectivity
and distance; Lincoln & Guba, 1985). In subsequent semi-
nal works, Guba and Lincoln (1989, 1994) also introduced
the notion of authenticity, to represent a range of multiple
realities and truths in line with the interpretivist perspec-
tive. In line with this alternative terminology, transferabil-
ity should be used and preferred in qualitative research
instead of theoretical generalizability.

However, even if the original aim of the changes in the
terminology was to bring clarity and a universally
acknowledged terminology in the field, these guidelines
still lack shared agreement between academics (Morse
et al., 2002). A bewildering proliferation of new terms has
hindered the selection of a vocabulary unanimously
accepted by scholars (Seale, 1999). The absence of a rec-
ognized canon, through which to appropriately evaluate
the quality and trustworthiness of the study may be one of
the reasons for the skepticism …