See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/286317838

Inductive Top-Down Theorizing: A Source of New Theories of Organization

Article in Academy of Management Review · April 2011

DOI: 10.5465/AMR.2011.59330952

CITATIONS READS
276 1,830

2 authors:

Dean A. Shepherd Kathleen M Sutcliffe

University of Notre Dame Johns Hopkins University
504 PUBLICATIONS 53,242 CITATIONS 140 PUBLICATIONS 30,022 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Dean A. Shepherd on 16 March 2022.

The user has requested enhancement of the downloaded file.

姝 Academy of Management Review
2011, Vol. 36, No. 2, 361–380.

INDUCTIVE TOP-DOWN THEORIZING:

A SOURCE OF NEW THEORIES OF
ORGANIZATION
DEAN A. SHEPHERD
Indiana University

KATHLEEN M. SUTCLIFFE
University of Michigan

Building on coherence theory and a pragmatist tradition, we offer an inductive model

of top-down theorizing that can be a source of new theories of organization. We
explain how an initial hypothesis is refined to enhance its potential contribution
(consistent with abduction). But, unlike abduction, we explain how inquiry begins—
how research tensions are carved out of the flux of the vast literature—and how
constant comparison facilitates an abductive process.

As organizational scholars, we are often en- which promotes divergent thinking (opportunal
couraged to improve our theorizing—to develop thinking [Cameron, 1986; Rothenberg, 1976] and
more interesting and imaginative hypotheses generative potency [DiMaggio, 1995; Gergen,
(e.g., Locke, Golden-Biddle, & Feldman, 2008; 1978]) and thereby provides the potential to gen-
Weick, 1989). But these exhortations are more erate new theoretical insights in the form of
easily said than done. Some advocate that the- hypotheses. These hypotheses then can be
orists start with generalizations (imagined tested by collecting and analyzing data from the
worlds) and determine if these generalizations phenomenon of interest.
apply to specific instances, whereas others ad- An alternative approach to deductive top-
vocate that theorists start with observations of down theorizing is an inductive bottom-up ap-
specific instances (the fallen apple) and seek to proach. Bottom-up theorizing begins at the inter-
establish generalizations about the phenome- section of a theorist’s “general wonderment”
non under investigation (Hyde, 2000: 83). These and raw data. A general wonderment provides
processes of organizational inquiry are often de- the necessary motivation for theorizing but may
scribed as top-down deduction or bottom-up in- not necessarily evoke a search for compelling
duction, respectively.
questions or fancy techniques, especially if one
Deductive theorizing is perhaps the most
lets the data speak (Glaser, 1992). Indeed, theo-
prominent approach for theorizing on organiza-
rists pursuing studies using a “classic” bot-
tions. Opportunities for this top-down approach
tom-up approach often take a position of “un-
begin at the intersection of the theorist and the
knowing,” which gives openness and life to the
existing knowledge typically contained in the
concept or idea (Glaser, 2001) such that a theory
literature. The theorist discovers a problem in
the literature—tension, opposition, or contradic- can emerge from the data (Clarke, 2005). Theory
tion among divergent perspectives and expla- emerges through the process of coding data,
nations of the same phenomenon—and then classifying it into concepts, and then making the
sets out to create a solution to that problem. connections between these concepts clear (Dey,
Top-down theorizing toward a superior explana- 1993: 30). The emerging theory is then compared
tion than previously offered involves reexamin- with existing theories published in the literature
ing the taken for granted assumptions that un- to determine its contribution. This bottom-up ap-
derlie current knowledge (Krippendorff, 1984), proach can provide a credible story of organiza-
tion that is theoretically important (Dyer &
Wilkins, 1991).
We thank Karl Weick, Roy Suddaby, and two anonymous Each model of theorizing is purported to be
reviewers for their comments on earlier drafts of this paper. limited as a method for generating “new” orga-
361
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright
holder’s express written permission. Users may print, download, or email articles for individual use only.
362 Academy of Management Review April

nizational theories. For example, critics of de- with a literature review. Yet it differs from the
ductive theorizing claim that resulting theories literature review of a deductive approach, which
are sometimes sterile, debatable, and unable to relies on theorists’ knowledge and experience to
be widely tested (Weick, 1996). Critics of induc- narrow the scope of the search. Rather, induc-
tive theorizing argue that resulting theories lack tive top-down theorizing relies on the data them-
scope, abstraction (Glaser, 2001; Sandelowski, selves to speak to the theorist (through the for-
1997), and sometimes generalizability since they mation of gists) to focus attention so as to detect
are more concerned with offering a rich descrip- tensions, conflicts, or contradictions. Gists are
tion of specific cases (Eisenhardt, 1991) or a list gestalt-like, holistic representations of the data
of variables (Merton, 1967). As a result, scholars and require approaching the literature with few
have proposed that theorizing is enhanced by preconceived notions. Our model builds on co-
combining inductive with deductive ap- herence theory to integrate aspects of deductive,
proaches, or vice versa (Thompson, 1956; Weick, inductive, and abductive approaches to theoriz-
1996). Although the notion that theorizing should ing and, in doing so, makes three contributions.
involve both induction and deduction is not new First, although it has been widely acknowl-
(e.g., Thompson, 1956), progress in integrating edged that theorizing is an iterative process in-
the two has been slow. Organization scholars volving both deduction and induction (Fine,
continue to call for theorists to “drop their heavy 2004; Gunter, 2005; Thompson, 1956; Weick, 1996),
tools” and use both induction and deduction in substantially more progress has been made in
their theorizing (Weick, 1996: 310). describing this “combination” in bottom-up the-
Some progress, in particular, has been made orizing using data from the environment (Eisen-
in combining deductive approaches with bot- hardt, 1989; Eisenhardt & Graebner, 2007; Yin,
tom-up theorizing. For example, Eisenhardt 2003, 2004) than in top-down theorizing using
(1989; see also Eisenhardt & Graebner, 2007), Yin existing studies representative of current knowl-
(2003, 2004), and others (e.g., Siggelkow, 2001, edge. Consistent with Weick, who argues that
2002) have proposed a bottom-up approach that “we cannot improve the theorizing process until
begins with a definition of the research ques- we describe it more explicitly, operate it more
tion, the type of organization and data that will self-consciously, and decouple it from valida-
form the case(s), and perhaps an a priori speci- tion more deliberately” (1989: 516), we take a
fication of constructs before approaching the step toward explicitly describing an inductive
data. These “preconceived” notions aid theoriz- top-down model of theorizing so that it can be
ing by reducing the likelihood that one will be used more self-consciously. We hope that by
overwhelmed by the volume of the data. They offering a deeper explanation of an underex-
also help to provide the most appropriate cases plored form of theorizing, we can collectively
for building the theory depending on the re- develop and improve this process to generate
search question (Eisenhardt, 1989). Although new theories of organization.
progress has been made in introducing deduc- Second, we extend abductive theorizing. Ab-
tion to bottom-up theorizing on organization, duction refers to a process of thought and in-
less progress has been made in introducing in- quiry that proceeds from “an anomaly to the
duction to top-down theorizing. delineation of a kind of explanatory hypothesis
Our goal then is to offer an inductive top- which fits into an organized pattern of concepts”
down model of theorizing as a source of new (Paavola, 2004: 279). It is triggered by doubt
theories of organization. We ground our theoriz- (Anderson, 2005; Hildebrand, 1996; Paavola,
ing in a coherence framework and a pragma- 2004). Although recent work on abduction high-
tism perspective. Our model is top down in that lights how scholars can engage doubt (Gross,
it is informed by the literature, but it is inductive 2009; Locke et al., 2008), left unanswered is how
in that it begins with the data (where, in this theorists generate the doubt necessary for trig-
case, the data are the flux of papers, books, gering abduction. Dewey (1989/1938) pointed out
presentations, working papers, and so on repre- that there is an indeterminate situation that
senting the collective knowledge of the intellec- gives rise to the focusing of attention on and/or
tual community) from which a theory is built. On formulation of a problem, which becomes the
the surface it may seem similar to typical de- impetus for abductive inquiry. But what is this
ductive approaches in which a theorist starts antecedent condition for abductive inquiry? To
2011 Shepherd and Sutcliffe 363

explain how theorists generate doubts from Weick, 1996). For example, when theorists focus
which they theorize, in the first stage of our exclusively at the level of variables rather than
inductive top-down model we use coherence the level of constructs, the outcome is not suffi-
theory as a way to focus on attention allocation ciently connected to our current understanding
processes. (or lack thereof) of the topic. Consequently, such
Finally, we contribute to the understanding of theories offer an insufficient “explanation of
theorizing by offering a different perspective why the theory or approach leads to a new or
both on the role of knowledge and experience in unanswered theoretical question” (Sutton &
the process and on what constitutes the data of Staw, 1995: 373). Examples of this form of incom-
interest. Deduction and abduction emphasize plete theorizing include relying on references as
the role of knowledge and experience in detect- theory and/or naming prevailing theories (Sut-
ing an enigma from which to begin theorizing. ton & Staw, 1995), but these are not theories
As we propose in our model, it is the absence of because they remain unconnected to current
(strong) preconceptions that allows the data to knowledge (of the community of scholars) and
speak to form a gist. This gist enables the theo- they do not provide an explanation of the causal
rist to focus attention on salient aspects of the reasoning of the proposed relationships. If a the-
literature that signify a research opportunity ory is not abstract enough to connect to the com-
and triggers subsequent stages of theorizing. munity of scholars’ broader understanding of
Further, inductive and abductive reasoning fo- the topic, that theory’s potential contribution is
cus on data about a phenomenon. We focus on restricted.
data as represented by the Literature— collec- A top-down approach assumes an under-
tive knowledge of the whole intellectual com- standing of the literature sufficient to discover
munity across time to this point, contained in, tensions from which to theorize. However, the
for example, all books, research articles, chap- relevant literature is vast, highly dynamic, and
ters, conference presentations, and so on. This scattered (Thompson, 1956), and theorists are of-
definition of Literature is consistent with the ten drawn to overworked localities in organiza-
notion of “the truth as is known today,” where tions, to ready-made problems, to fashionable
truth is defined as “the ideal limit of indefinitely styles of thinking (Weick, 1980), and to taking
continued inquiry” (Dewey, following Peirce, positions that are hardened over time (such as
quoted in Campbell, 1995: 15) and is consistent macro and micro [Weick, 1996]). As a result, re-
with the “intellectual property” notion of prior search on organizations has been described as
art—“the entire body of knowledge from the be- relatively free of “theoretical tension” (Weick,
ginning of time to the present” (Blenko, 1991: 45). 1980: 179), as idle speculation (Van Maanen, So-
The Literature represents an amorphous stream renson, & Mitchell, 2007), and as sterile (Weick,
from which a theorist must carve out paradigms, 1996). We acknowledge that these criticisms
research questions, and explanations. can be directed at the theorist and not the ap-
proach; however, the approach is silent on the
discovery (or creation) of a tension from which to
LIMITATIONS OF DEDUCTIVE AND
theorize—silent on how a theorist discovers or
INDUCTIVE THEORIZING
creates tensions, oppositions, and/or contradic-
Deduction is a method that involves “going tions among discrepant perspectives and expla-
from generals to particulars; deriving conclu- nations of the same phenomena.
sions based on premises through the use of a Induction is a method that involves “going
system of logic” (Samuels, 2000: 214) and, in gen- from particulars to generals; deriving knowl-
erating inferences, is typically associated with a edge from empirical experience based upon a
top-down approach to theory building that be- system of handling sense data” (Samuels, 2000:
gins with the existing knowledge. One of the 214) and is typically associated with a bot-
most common criticisms of deductive top-down tom-up approach to theory building that starts
theorizing is that theorists often forgo “abstrac- with data. A common criticism of inductive bot-
tion in favor of the immediately applicable and tom-up theorizing is that the theorizing has not
settle for ‘common sense’ hypotheses framed at started sufficiently close to the phenomenon.
low levels of abstraction, without regard for When theorists do not start with data close to
general theory” (Thompson, 1956: 110; see also the phenomenon, their preconceived notions
364 Academy of Management Review April

and naive theories focus their attention on spe- Indeed, there exists a debate in the literature
cific aspects of the data such that they constrain on classical versus hybrid versions of bottom-up
the theorist’s vision (Poole & Van de Ven, 1989). theorizing, and most of the debate is over using
Consequently, the data have less of an opportu- deduction to facilitate an inductive process. For
nity to speak to the researcher (Glaser, 1992), example, Fine concluded that
reducing the openness and life of a concept or
idea (Glaser, 2001). If theorists using a bot- it is not possible to separate deduction and in-
tom-up approach have not started close to the duction in the way that has been suggested, par-
phenomenon, they can miss constructs and re- ticularly as regard to field research. . . . the induc-
tive and deductive models of research can never
lationships important in describing and ex- be disentangled. . . . Theoretical analysis is not
plaining the phenomenon, and they risk being something that occurs only before entering the
accused of imposing their own agenda or “pet field, but is a continuing and recursive process.
theory” on the data (Wilson & Hutchinson, 1996). . . . Researchers should always be engaged in
Another common criticism of inductive bot- theory building— before, during and after the
gathering of ethnographic data (2004: 5, 11; cited
tom-up theorizing is that theorists often stop
in Gunter, 2005: 672).
with a description of the phenomenon, simply
listing variables or constructs. Glaser (1992), in
Similarly, Eisenhardt (1989) acknowledged that
fact, noted that theorists using a grounded the-
the bottom-up approach that requires starting
ory method can fall into the trap of finishing
with the data without any consideration of a
their theorizing after developing a full concep-
theory under construction is laudable but im-
tual description. Indeed, numerous scholars
possible to achieve in its purest form. Rather, as
have warned about the dangers of terminating
we noted earlier, she suggests that theorists us-
theorizing based on an outcome of only descrip-
ing case studies for the purpose of theory build-
tion (Glaser, 2001; Morse, 1994; Sandelowski,
ing start by defining their research question as
1997) or lists of variables (Merton, 1967; Sutton &
Staw, 1995). Strategy researchers, such as Eisen- well as the type of organization and data that
hardt, for example, have claimed that “some will form the case(s), and that they even specify
grounded theorists are more concerned with de- relevant constructs a priori. These preconcep-
veloping a rich description of the specific cases tions can help a theorist in analyzing data in
under study than developing generalizable the- part because they decrease the possibility that
ory” (Eisenhardt, 1989: 546). Similarly, Black, he or she will be overwhelmed by the volume of
Carlile, and Repenning have expressed a con- the data and because they also provide the most
cern with such theorizing on technology and ac- appropriate cases for building the theory (de-
tivity: pending on the research question, to replicate,
extend emergent theory, fill thematical catego-
Although, as Glaser and Strauss argued, the pro- ries, and/or provide polar opposites).
gressive consolidation of substantive analyses
into more formalized, general categories is a key
step in developing theory that spans multiple
inquiries, few scholars of technology implemen- INDUCTIVE TOP-DOWN THEORIZING
tation have made such an attempt to provide a
more general account of the influences between Given the assorted limitations of traditional
new technology and social action. Until such the- deductive and inductive approaches noted
ories are produced, it will be difficult to test em- above, it is not surprising to find repeated calls
pirically our understanding of the impact of new
technology to improve theory or provide useful for improvements to theorizing. Our aim in this
advice to practitioners (2004: 574). section is to show how bringing inductive meth-
ods to bear on a top-down approach to theoriz-
A good theory both describes and explains ing can circumvent several of these limitations.
(Whetten, 1989). Studies that only describe fail to We start by providing a brief introduction to
offer a generalizable theory because they lack coherence theory, which we use as an underly-
the necessary scope and abstraction (Glaser, ing framework, and we follow this by articulat-
2001) to explain why persistent findings have ing some assumptions and boundary conditions
been observed (Glaser & Strauss, 1967; Sutton & of our theorizing. We then dig more deeply into
Staw, 1995). our model.
2011 Shepherd and Sutcliffe 365

Introduction of Coherence Theory As an herd, McMullen, & Jennings, 2007; Simon, Snow,
Underlying Framework & Read, 2004; Skyrms, 1992).
Coherence theory relies on a connectionist
We use coherence theory as a framework in
network with a parallel constraint satisfaction
developing our model and briefly introduce it
process: explanations are represented as nodes
here. Coherence theory has two main branches:
in a network and are interconnected via links to
perception and explanation.
the decision alternatives and to other explana-
Coherence theory of perception explains how
tions (Simon, Pham, Le, & Holyoak, 2001). The
data as quanta are turned into a representation
greater the activation of an inference’s network,
that is used to “see” objects or relationships
the more likely it is to be used to interpret sub-
between objects (Rensink, 2002). The nature of sequent information (Read, 1987: 293). Impor-
this sensory representation depends on where tantly, the movement toward coherence is bidi-
people focus their attention. The focus of atten- rectional—the sensory image of the data
tion is guided by processes such as knowledge informs the conceptual representation, and the
based on past experience (e.g., preconceptions) conceptual representation informs the formation
or the formation of a gist. A gist is a gestalt-like, of the sensory representation; both the sensory
holistic representation of the environment that representation (different data attended) and the
does not require attention to stabilize a subset of conceptual representation can change to close
the environmental stimuli (Oliva & Torralba, the gap between the two (Holyoak & Simon, 1999;
2001; Rensink, 2002). Once attention is focused Simon & Holyoak, 2002; Simon, Krawczyk, &
(either by knowledge or a gist) on specific as- Holyoak, 2004; Simon, Snow, & Read, 2004). For
pects of the environment, these sources are example, in an experiment where participants
“held” in a “coherence field,” which provides a were asked to reach a “verdict in a complex
sensory representation of the environment. Be- legal case involving multiple conflicting argu-
cause attention is limited (Kahneman, Treis- ments,” the individual arguments informed an
man, & Gibbs, 1992; Simon, 1947), only specific emerging verdict, and the emerging verdict led
data can be attended to at one time. to changes in the assessment of the individual
These coherence structures (and their corre- arguments in a way that increased coherence
sponding sensory representation) can only be between the verdict and the arguments (Holyoak
maintained as long as attention remains fo- & Simon, 1999: 3).
cused (Wolfe, 1999). Therefore, attention con- Evidence of this constraint satisfaction pro-
structs a limited and temporary coherence field cessing has been found in inference-based judg-
to capture the nature of the phenomenon. Coher- ments (Holyoak & Simon, 1999; Read & Marcus-
ence theory of perception typically concludes Newhall, 1993), preference-based choice tasks
with a sensory representation of the environ- (Simon, Krawczyk, & Holyoak, 2004), probabilis-
ment to explain detection, such as visual or au- tic judgments (Simon, Snow, & Read, 2004), and
dio changes (e.g., Horváth, Czigler, Sussman, & analogical reasoning (Spellman & Holyoak,
Winkler, 2001; Rensink, 2000). However, the 1992). The end result of this process is a coherent
maintenance of sensory representations over an story. The more coherent a story, the more it is
extended period requires that individuals rely accepted as a plausible explanation of the phe-
on a system other than attention—a system nomenon (Pennington & Hastie, 1992; Read, 1987;
based on beliefs. Read & Marcus-Newhall, 1993).
Coherence theory of explanation explains From the coherence theory of perception we
how perception becomes represented in (and is build on two notions: (1) the nature of sensory
influenced by) beliefs (Lehrer, 1992; Merricks, representations of the environment depends on
1995; Walker, 1989). People attend to data and where attention is focused, and (2) attention is a
then make inferences in order to tie the data into limited resource that can be focused by the en-
a coherent mental representation that they can vironment itself or by prior knowledge and con-
use to explain the phenomenon (Read, 1987). textual influences. From the coherence theory of
This involves an iterative process of the cogni- explanation we build on two additional ideas:
tive system, where incoherent representations (1) a conceptual representation is created and
are replaced by increasingly more coherent rep- modified to offer a coherent, lasting explanation
resentations (Millgram & Thagard, 1996; Shep- of sensory perceptions (which can also be mod-
366 Academy of Management Review April

ified), and (2) a coherent explanation can be (tensions, conflicts, and/or contradictions) and
replaced by a more coherent explanation. To potential solutions (literature, theories, con-
this coherence framework we introduce the Lit- structs, relationships) emerge to offer a descrip-
erature, induction, and abduction to develop a tion and then a coherent resolution of a research
model of inductive top-down theorizing. problem, ultimately constituting a new theory of
organization.
Building on the central role of attention in
A Coherence Explanation of Inductive
the coherence theory of perception (Rensink,
Top-Down Theorizing
2000, 2002), we explore how the Literature itself
Inductive approaches to theorizing typically (as an undifferentiated flux of data) can focus
begin with data about the organizational phe- the theorist’s attention on specific works to
nomenon from which concepts and relation- form a sensory representation. However, ad-
ships emerge to offer a description and then an mittedly, the role of the Literature in focusing
explanation of the phenomenon, ultimately con- attention to form a sensory representation is
stituting a theory of organizing. The theorist in- limited when the theorist’s attention is also
fers relationships from the data. The process of focused by his or her knowledge and organi-
induction is described as bottom-up theorizing zational context. A theorist’s sensory represen-
because it generally starts with a case that is tation is a perception of tensions, oppositions,
“close” to the raw data of a phenomenon (e.g., and/or contradictions among divergent per-
grounded theory and case studies; Dey, 1993; spectives and explanations of the same phe-
Dyer & Wilkins, 1991; Glaser, 2001). Yet we can nomena in the Literature. Consequent to the
also apply the notion of induction to top-down development of a sensory representation, the
theorizing. In our model the data are not the theorist develops a conceptual representation
rapidly generated volatile structures that con- (representing a tentative solution to the prob-
tain information about phenomena (Daft & lem—a solution that includes concepts not
Weick, 1984); rather, the data are the array of given in mere data [Peirce, as described in
rapidly generated volatile structures that con- Dyer, 1986]). This conceptual representation
tain information that exists in the Literature. refers to general abstract statements of rela-
As shown in Figure 1, we propose that induc- tionships between constructs—incorporating
tive top-down theorizing begins with data con- explanations of “how” and “why,” boundary
tained in the Literature from which problems conditions of values, and assumptions of time

FIGURE 1
Model of Inductive Top-Down Theorizing

Theorist’s The
prior Literature
knowledge

Limits Focusing attention

Scholarly
context

Sensory representation
Constant
comparison

Conceptual representation
Potential
contribution

A new theory of
organization
2011 Shepherd and Sutcliffe 367

and space—that allow for a more coherent res- 2007)—for example, the Literature is assumed
olution of the theorist’s sensory representation. to exist independent of any one individual’s in-
Building on the notions of bidirectional up- quiry.
dating of representations (Holyoak & Simon,
1999, Simon, Snow, & Read, 2004) and the prin-
Focusing Attention and Inductive Top-Down
ciple of induction (Glaser, 1978, 1992, 2001; Gla-
Theorizing
ser & Strauss, 1967), our model explains how a
theory emerges through the constant compar- Attention and a theorist’s sensory representa-
ison of conceptual and sensory representa- tion. Inquiry is preceded by the perception of a
tions and is assessed in terms of its potential problem or a specific doubt. Dewey referred to
contribution. Finally, a theorizing outcome this stage as an indeterminate situation and
represents a new addition to the literature and characterized it as “disturbed, troubled, ambig-
a new stimulus for subsequent theorizing. Be- uous, confused, full of conflicting tendencies,
fore describing the model in detail, we estab- obscure, etc” (1989/1938: 105). The theorist fo-
lish some key assumptions and boundary con- cuses his or her attention on some aspect of this
ditions. indeterminate situation and environment that
gives rise to a problem/doubt. Dewey (1896) em-
phasized that “data” are never “given” but,
Assumptions and Boundary Conditions
rather, are “taken” by the researcher, and what
Building on the pragmatic tradition underly- is taken to form an observation depends on lim-
ing abduction (e.g., Peirce, Dewey), we make a ited sense organs, habits, perspectives, and pur-
number of assumptions. First, theorists are pri- poses of observation (Webb, 2007: 1067). In fact,
marily motivated to solve problems through a understanding where attention is focused is
process of inquiry—they are driven to find solu- central to theories of perception. From a coher-
tions to problems where problems represent ence theory perspective (e.g., Rensink, 2000,
doubt over the applicability of existing habits 2002; Thagard, 1989), the environment is an un-
(such as mental models for guiding thought and differentiated flux of data from which constructs
action; Locke et al., 2008; Peirce, Hartshorne, and relationships between constructs have to be
Weiss, & Burks, 1958). Second, antecedent to in- “forcibly carved out” (Chia, 2000: 517; see also
quiry is an indeterminate situation where atten- Weick, Sutcliffe, & Obstfeld, 2005).
tion comes to be sufficiently focused on a prob- To the top-down theorist, the Literature is a
lematic situation to ask a question or specify a vast and continually changing undifferentiated
doubt (Dewey, 1989/1938: 118). Third, the outcome flux of papers, books, presentations, working pa-
of theorizing is fallible (acknowledged to be pos- pers, and so forth from which he or she must
sibly false) and can be revised (by the focal “carve out” paradigms, tensions, oppositions,
theorist or others) as new information is re- and contradictions. Carving out these founda-
vealed (consistent with Peirce, James, and tional building blocks for theorizing requires at-
Dewey [Almeder, 2007]). Fourth, theorists’ recog- tention. Given that theorists’ attention is limited,
nition, perception, and interpretation of prob- as with all people (Kahneman et al., 1992;
lems and solutions can be influenced by their Rensink, 2000), they can attend only to specific
mental models and the cultural environments in aspects of the Literature at one time. Therefore,
which they are embedded (Alexander, 1988). the theorist’s sensory representation of the Lit-
Fifth, theorizing involves alternating between erature is based on a subset of data—a subset of
habits (which can be unconscious, such as au- the papers, books, presentations, working pa-
tomatic activation of a schema) and (more con- pers, and so forth, which depends on where the
scious) creativity (Gross, 2009; see also Dewey theorist’s attention is focused. Attention can be
and Peirce, as described in Webb, 2007: 1070). focused by the Literature data itself (inductive)
Sixth, theorizing outcomes become meaningful or it can be focused by the theorist’s knowledge
in larger cultural systems and structures (Gross, and scholarly context. The more attention is fo-
2009: 370). Finally, consistent with Peirce and cused by the theorist’s knowledge and/or schol-
Dewey, we assume that knowledge is the (falli- arly context, the less inductive the top-down the-
ble) understanding of reality where reality ex- orizing and the more traditionally deductive the
ists independent of human inquiry (Almeder, process is. We now investigate each.
368 Academy of Management Review April

The Literature and a theorist’s focused atten- Clarke, 2005; Glaser, 2001) or minimal theory
tion. The Literature itself can focus a theorist’s (hybrid; Eisenhardt, 1989; Suddaby, 2006). In an
attention by “speaking” to the researcher and inductive top-down approach, where the data
directing his or her attention to salient papers, are from the Literature (including theories),
books, presentations, working papers, and so openness requires that the theorist refrain from
on. In the extreme form of inductive top-down attending too closely to specific literature, theo-
theorizing, the theorist begins close to the ideal: ries, constructs, methods, and so on, and also
no theory under construction and no hypotheses remain open to alternative routes of interpreta-
to allow the data of the phenomena to speak to tion and analysis (Alvesson & Karreman, 2007;
the theorist. In less extreme forms the theorist Rorty, 1989). He or she does this by becoming
has a tentative research question and perhaps familiar with a variety of literature to broaden
tentative constructs but is open to shifts in re- his or her repertoire of issues, theories, and vo-
search question and to exploring difficult is- cabularies (Alvesson & Karreman, 2007; Rorty,
sues—and doing so while avoiding thinking 1989) and by holding off on categorizing these
about specific theories. data (focusing on specific aspects of the litera-
Similar to bottom-up theorists who are asked ture) to provide an opportunity for gists to de-
to approach the data with an open mind so that velop, which then can focus attention on specific
the data can speak to them (Glaser, 1992), induc- aspects of the data. This openness necessary for
tive top-down theorists have the opportunity to the inductive aspect of inductive top-down the-
approach their data (Literature) with an open orizing is consistent with Dewey’s characteriza-
mind and to allow this “undifferentiated flux of tion of the “unreflective situation,” where the
information” to generate a gist (a holistic repre- individual has a “vague awareness of the total-
sentation of the Literature) that subsequently ity of surroundings—rather than differentiated
directs attention toward specific literature, the- features” (quoted in Webb, 2007: 1070). Gists
ories, and constructs that cohere. These aspects work in the cognitive context of being embedded
of the Literature cohere to the extent that they in the data of the Literature (by becoming famil-
offer explanations of the same phenomena but iar with theories, constructs, vocabularies) with-
represent a research opportunity because the out preconceived notions of which of those bod-
theorist perceives a tension, opposition, and/or ies of literature, theories, and/or methods will
contradiction among these divergent perspec- have precedence. This provides the attentive
tives. A gist of the Literature directs the theo- conditions for the theorist to discover (or create)
rist’s attention to salient bodies of work and a tension, opposition, and/or contradiction to
certain sources within those bodies. From a co- trigger the inquiry process.
herence theory perspective, gists are formed in To document an ongoing stream of conscious-
the absence of attention (but not absent wonder- ness about what is being read, a theorist can
ment or commitment) and, once formed, guide use literature notes. Building on Eisenhardt’s
focused attention (Oliva & Schyns, 2000; Oliva, (1989) ideas regarding field notes, literature
Torralba, Castelhano, & Henderson, 2003; notes refer to documentation of whatever im-
Rensink, 2002; Torralba & Oliva, 2003). They offer pressions arise to the researcher while reading
sources of spontaneous variation that can en- aspects of the literature. The goal is for the the-
hance the novelty of a resulting theory consis- orist to allow these impressions to flow freely
tent with the analogy of a kaleidoscope. As Nord from his or her mind without simultaneously
and Connell remark, “Turning a kaleidoscope filtering them for relevance. Relevance is deter-
can: (1) dislodge old patterns, (2) generate new mined retrospectively and is difficult to do a
patterns, and (3) foster awareness that numer- priori. In constructing literature notes, theorists
ous configurations are possible” (1993: 117). use questions such as “What am I learning?”
Approaching data with openness to allow for and “How does this article, theme, literature, or
the generation of a gist in inductive top-down paradigm differ from the last?” This intermin-
theorizing differs from that required in bot- gling of content with impressions (a potential
tom-up approaches. In classic or hybrid bot- source of departure for theory development) pro-
tom-up approaches, where the data are of the vides embeddedness in the data without pre-
phenomena, approaching data with openness conceived notions of preeminence, which en-
means that the theorist uses no theory (classic; ables the formation of gists that then focus
2011 Shepherd and Sutcliffe 369

attention on specific aspects of the Literature. rist’s knowledge before approaching the data, the
Thus, a feature of inductive top-down theorizing more this “research situation” focuses attention on
is its dynamism and the opportunity to make aspects of the Literature believed to be salient
adjustments during problem (contradiction, con- (preconceptions) for detecting research problems
flict, or tension) detection and theory formation. (tensions, conflicts, and/or contradictions) and po-
This dynamism is constrained when attention is tential solutions (literature, theories, constructs,
focused more by the theorist’s prior knowledge relationships). Consequently, there is less of an
and/or scholarly context. opportunity for the research problems and solu-
Prior knowledge and a theorist’s focused at- tions to be generated from the Literature itself.
tention. Preconceived notions about what is im- Organizational/institutional contexts and a
portant direct and focus attention (Rensink, 2000, theorist’s focused attention. The theorist’s atten-
2002). Some scholars (e.g., Eisenhardt, 1989; tion is influenced by the contexts (e.g., organi-
Suddaby, 2006) argue that theorists who wish to zational and institutional) in which he or she is
build theory from case studies will be more ef- embedded. Simon (1947) argued that organiza-
fective if they begin with a preconceived ques- tions influence individuals’ decision processes
tion, sampling strategy, and even ideas of main by channeling their attention to specific aspects
constructs. For example, Ferlie, Fitzgerald, of a situation and specific situations, and orga-
Wood, and Hawkins (2005) approached their bot- nizations also influence the aspects of situa-
tom-up study of the nonspread of innovations in tions that individuals ignore (filter out). Simi-
health care in the United Kingdom with a pre- larly, in his attention-based view of the firm,
conceived research question and corresponding Ocasio (1997) argued that a decision maker is
sample: situated in the procedural and communication
channels of the organization, and these gener-
We developed the theme as defined by the NHS
commissioners into the following lead research ate a set of values that makes some issues and
questions as we went into the field: (1) Are inno- answers more salient than others.
vation pathways in health care linear or messy? Most theorists are members of an organization
(2) Is robust scientific evidence sufficient to lead (e.g., university or research institution), as well
to successful diffusion? (3) What impact does as an organizational subunit (e.g., school, de-
greater innovation complexity have? . . . [we did
this] to avoid “drowning in the data” (2005: 119). partment). They also are members of profes-
sional communities of practice. Although these
To select the sample, the researchers “con- organizations on some dimensions are similar
structed a two-by-two cell design to explore ef- (teaching and research functions), they vary on
fects of stronger/weaker scientific evidence and other dimensions. Universities and their sub-
the degree of innovation complexity on speed units have organizational norms and proce-
pathways” (2005: 119). Such preconceptions aid dures that focus their members’ attention on
theorizing by reducing the aspects of the envi- specific aspects of research output, including
ronment to be processed to a more cognitively procedures and practices that assign research-
manageable level and by providing the most ers to a department within a school, procedures
appropriate cases for building the theory (Eisen- for promotion and tenure, and procedures for
hardt, 1989). enhancing the research ranking of the school as
The theorist’s idiosyncratic knowledge of the represented by national lists of excellence
literature can also direct attention. This idiosyn- (Gomez-Mejia & Balkin, 1992; Jarley, Chandler, &
cratic knowledge is developed as a conse- Faulk, 1998). These may focus the theorist’s at-
quence of the theorist’s own interests, competen- tention on specific topics (e.g., management re-
cies and expertise, body of work (e.g., search when in a management department), on
publications), and experiences (e.g., conference specific journals characterized as “A” and “A
presentations or conferences attended, articles minus” (publication in which can lead to promo-
he or she has reviewed for journals, doctoral tion, tenure, pay raises, recognition, and so on;
courses taken and taught). This idiosyncratic Johnson & Podsakoff, 1994), and on more or less
knowledge shapes the theorist’s research strat- consensus-challenging research (McMullen &
egy, such as the research question, literature of Shepherd, 2006). The journals themselves have
interest, and types of journals sampled. The more procedures that may influence the theorist’s at-
the research strategy is “informed” by the theo- tention by rewarding the theorist with publica-
370 Academy of Management Review April

tion of certain research questions and programs that are already buried in the identification of the
of study (Pfeffer, 1993). problem (Weick, 1980: 181).
Similarly, these organizations also differ in That is, blind spots can lead to confirmation
their communication channels. Some universi- bias, strategic myopia, and other forms of
ties may provide electronic access to a vast ar- change blindness. The more research strategy is
ray of journal articles, working papers, and directed by the theorist’s knowledge and/or
books, whereas others have a more narrow scholarly context, the less opportunity there is
range. Some universities encourage greater for inductive top-down theorizing—that is, top-
communication (formal and informal) across de- down theorizing becomes less inductive and
partments within a school and across schools more like traditional deductive theorizing.
within the university. These differences can in-
fluence the focus of a theorist’s attention. For
Constant Comparison and Inductive Top-Down
example, the theorist may focus more heavily on
Theorizing
those scholarly works accessible via the Inter-
net from his or her office or may have his or her A recognized problem and abduction. As we
“eyes opened” to a stream of literature high- proposed above, attention allocation processes
lighted by a sociology professor at a university- focus theorists’ attention on aspects of the Liter-
wide faculty picnic. Organization procedures ature that represent tensions, oppositions,
and communication channels can focus a theo- and/or contradictions. It is these anomalies that
rist’s attention and thereby influence the forma- provoke theorists to form new ideas (Czar-
tion of the sensory representation central to the- niawska, 1999; Hansen, 2008; Locke et al., 2008;
orizing. Paavola, 2004). Anomalies can trigger abduc-
The use of preconceived notions (whether gen- tion—that is, they can trigger abductive inquiry
erated consciously or unconsciously, or inter- because they represent doubt, experienced as
nally by idiosyncratic knowledge or externally not knowing, and motivate the allocation of at-
by the organizational and institutional contexts) tention to search for its resolution (belief; Locke
to focus attention is a double-edged sword. On et al., 2008; Peirce et al., 1958). Doubt represents
the one hand, knowledge and/or contextual and an irritation that excites thought (Peirce et al.,
institutional norms and practices and communi- 1958) to “generate inventive solutions, new
cation channels focus attention on specific data ideas, exploratory propositions, and theoretical
that are presumed to be most relevant to the elements” (Locke et al., 2008: 908-909).
theory-building process; on the other hand, fo- Abduction begins with and is sustained by
cusing attention on some data means that other doubt (Anderson, 2005; Hildebrand, 1996; Locke
data are ignored. This limits theorizing because et al., 2008; Peirce et al., 1958). It offers a guess at
data that are ignored may be as or more impor- an explanatory hypothesis (hunch), which ini-
tant for understanding the phenomenon. Eisen- tially appears to fit into an organized pattern of
hardt, for example, suggests that theorists concepts (Paavola, 2004). As conceived by Peirce
should “formulate a research question and pos- et al. (1958), abduction merely suggests the pos-
sibly specify some potentially important vari- sibility of something. It represents an “imagina-
ables based on reference to the literature but tive effort of understanding beginning with an
should avoid thinking about specific relation- aesthetic-hypothetic response (Alexander, 1990:
ships as much as possible” (1989: 536). In this 329) to the world” (Locke et al., 2008: 913). After
case it is more difficult to theorize about rela- detecting a problem, this abductive approach
tionships involving variables not initially at- represents a next step in the theorizing process;
tended to. Indeed, preconceived notions (theo- the theorist allocates attention to resolve doubt
rists’ prior knowledge and organization and to form a stable belief (Burks, 1946). Indeed, an
institutional contextual influences) can lead to exploratory explanation (hypothesis) is put to
blind spots, which the test by forcing it to make verifiable predic-
tions (Peirce et al., 1958) and then using that
divert attention from the dynamics of problem information to modify the explanation (Alexan-
finding; they carry along all of the assumptions,
surplus meanings, and arbitrary boundaries in der, 1990).
the algebra problem; and they bias theoretical From a coherence theory perspective, the res-
explanations toward commonplace variables olution of doubt to form a stable belief occurs
2011 Shepherd and Sutcliffe 371

with a plausible explanation of the phenome- fers to the process by which the theorist “checks”
non—that is, a coherent story (Pennington & and “tests” the emerging conceptualization with
Hastie, 1992; Read, 1987). A coherent story sensory representations to determine the extent
emerges by putting an initial explanation to the to which the conceptual representation offers a
test. By being put to the test, the conceptual coherent explanation of the data. Consistent
representation can be changed to better account with the notion of bidirectional processing from
for the sensory representation, and the emerg- coherence theory (Holyoak & Simon, 1999, Simon
ing conceptual representation can then inform & Holyoak, 2002; Simon, Krawczyk, & Holyoak,
the allocation of attention to specific data to 2004; Simon et al., 2001; Simon, Snow, & Read,
provide other “testing” sensory representations, 2004), both the conceptual and sensory represen-
which can lead to an emerging explanation of tations can change in this iterative process. For
greater coherence (Holyoak & Simon, 1999; example, a gap between the sensory represen-
Simon & Holyoak, 2002; Simon et al., 2001). tation and the initial conceptual representation
Abductive theorizing in practice often occurs informs adjustments to the conceptual represen-
through the constant comparative method (Cof- tation. The newly adjusted conceptual represen-
fey & Atkinson, 1996; Richardson & Kramer, 2006; tation directs attention to specific data (con-
Suddaby, 2006). We now turn to constant com- structs and relationships in the Literature) that
parison for its application to the constant flux of generate a new sensory representation. This test
data that is the Literature. reveals the need to make further adjustments,
Constant comparison of sensory and concep- which leads to the focusing of attention on dif-
tual representations. Whether shaped to a ferent data (and the generation of a different
greater or lesser extent by gists (depending on sensory representation) as a new test to inform
the focusing of attention by prior knowledge and further adjustments to the emerging theory. It is
context), coherence theory proposes that individ- through this process of changes to the concep-
uals develop a sensory representation of as- tual and sensory representations that the
sorted data (Rensink, 2000, 2002). A conceptual emerging explanation becomes more coherent
representation is then applied to this sensory (Pennington & Hastie, 1992; Read, 1987; Thagard,
representation of the data to “test” the extent to 1989).
which it rectifies the “current state of knowl- Whetten (2002) offers an example of such a
edge” in a way that allows for a more coherent dynamic comparison process. He suggests that
explanation of what has been sensed (Dietrich & when reviewing the Literature the researcher
Moretti, 2005; Lehrer, 1999). This process of com- write the names of the constructs that catch his
parison and adjustment is consistent with an or her attention on “sticky notes.” This allows
important threshold of grounded theory. Glaser the theorist to configure and reconfigure the
(1978) proposed that open coding is complete sticky notes as the conceptual representation
when the theorist begins to see the possibility emerges so as to accommodate new constructs
of a theory—when variables have begun to or new interpretations of the literature. There
emerge. It is once this sensory representation is are numerous other tools and techniques that
formed that the theorist begins “selectively” can be used to facilitate comparison; we note
coding around emergent variables to “conceptu- (but do not fully explain) two such examples:
alize how the substantive codes may relate to thought experiments and metaphorical reason-
each other as hypotheses to be integrated into a ing.
theory” (Glaser, 1978: 72). The story being woven Thought experiments to compare representa-
together (at its various stages) is a conceptual tions. Thought experiments can be used to focus
representation of the phenomenon, which Gla- theorists’ attention on comparisons between
ser (2001) referred to as concepts that are ab- sensory and conceptual representations, lead-
stract in terms of time, place, and people in ing to adjustments in order to achieve a greater
psychosocial patterns. coherence between the two. For example, Weick
The theorist continues to compare representa- (1989) proposed the notion of theorizing as “dis-
tions to refine this initial conceptualization—to ciplined imagination,” “where the ‘discipline’ in
refine definitions, determine boundary condi- theorizing comes from consistent application of
tions, understand why relationships likely exist, selection criteria to trial-and-error thinking and
and so on. Specifically, constant comparison re- the ‘imagination’ in theorizing comes from de-
372 Academy of Management Review April

liberate diversity introduced into the problem emerging from the sensory representation. Mis-
statements, thought trials and selection criteria matches signal lines for extended theorizing
that comprise the thinking” (Weick, 1989: 516). (Bacharach, 1989).
Consistent with the principles of abduction, Second, metaphors inherently provide an
thought experiments represent creative probes “openness of meaning” (Cornelissen, 2005: 753)
into the unknown, recognizing that these probes that encourages new and multiple ways of con-
are highly fallible yet inform further theorizing ceptualizing the problem (incompleteness,
and recognizing that the emerging theory in- inadequacy, or incommensurability; Locke &
forms which thought experiments will be most Golden-Biddle, 1997) and its resolution. It en-
probative. courages the theorist to take a different perspec-
In their simplest form thought experiments tive, which directs attention to different aspects
(also referred to as thought trials) are “abstract of the literature so as to recognize new concep-
hypothetical scenarios” (Harggqvist, 1996: 9) that tual distinctions and bridges (cf. Morgan, 1983).
act like imaginary experiments (Weick, 1989). For example, the metaphor of the organizational
For example, Weick described his theorizing mind (Weick & Roberts, 1993) has led to in-
about the Mann Gulch disaster as a process creases in our understanding of learning within
through which the organization (e.g., the learning organization;
Senge, 1990) and managing different aspects of
the argument developed partially by taking the organizational culture (e.g., managing multiple
Mann Gulch data as symptoms and, through a
series of thought trials corresponding to treat- organizational identities [e.g., Pratt & Foreman,
ments, seeing which made a difference in those 2000]).
symptoms. This exercise in disciplined imagi- Finally, metaphors provide sources of analo-
nation resulted eventually in the theory that gous information (Fauconnier & Turner, 1998;
sensemaking collapses when role structures Tourangeau & Sternberg, 1982). Cornelissen ex-
collapse. . . . The differential “responsiveness” of
data to changes in treatment is frequently an emplifies these ideas in using a metaphor to
informative precursor to theorizing (1995: 388). explain the role of metaphor in theorizing:

The data Weick (1993) used as symptoms for his One way to interpret this is in terms of a skeleton
analogy, whereby each domain has a skele-
thought experiments came from Norman Mac- ton—a framework—that is directly represented:
lean’s 1992 account of the fire at Mann Gulch, in the inherent structure of the domain. This frame-
a book entitled Young Men and Fire. Brown work identifies the general structure of the con-
(1986) also noted the importance of thought trials joined concepts (although, in some cases, this is
in the creation of quantum mechanics and rela- quite fragmentary), but because they are difficult
to directly conceptualize, it leaves many of the
tivity (see also mental simulations [Klein, 1998] details blank. The framework, then, is the direct
and dramatic rehearsal [Dewey, 1933]). There- representation, and the metaphor adds the
fore, a thought trial can be considered a hypo- “flesh” to this by filling in information that is
thetical (and possibly imagined) scenario that initially seen as dissimilar and not directly rep-
provides a sensory representation that can be resented—in many cases, the bulk of the content
of the concepts. In this way, the framework does
compared to the more abstract and general rep- not explicitly assert a likeness but again pro-
resentation to refine and improve the conceptu- vides us with a template for constructing or
alization. “blending” correspondences (Fauconnier &
Metaphorical reasoning to compare represen- Turner, 1998), invites us to make comparisons for
tations. Metaphorical reasoning as a theorizing a further “filling in” or specification (Scheffler,
1979: 129), and directs our attention to “unex-
tool engages comparisons to build an abstract pected or subtle parallels and analogies” (David-
generalizable representation (conceptual repre- son, 1984: 256) (2005: 756 –757).
sentation) from a sensory one. First, the use of
metaphors provides the theorist the vocabulary Therefore, constant comparison (through
to express and map a conceptual representation thought experiments, metaphorical reasoning,
of the research opportunity (Cornelissen, 2005; or other tools) “tests” the “fit” of the emerging
Tsoukas, 1991). By comparing the sources and conceptualization with the theorist’s sensory
the target of the metaphor, theorizing is en- representations to determine the extent of the
hanced because it offers a provisional way of gap between the two and provides the basis for
organizing the constructs and relationships bringing the conceptual and sensory represen-
2011 Shepherd and Sutcliffe 373

tations “into line.” The greater the mismatch the factors can be deleted because they add little
more likely it will trigger further theorizing (see additional value to our understanding (Dubin,
coherence theory [Simons & Levin, 1998; Simons 1978; Whetten, 1989). A good theory finds a bal-
& Wang, 1998] and theorizing [Folger & Turillo, ance between being overly exhaustive and
1999; Weick, 1995]). overly exclusive (Feldman, 2004: 566).
Third, a theory that is explicit about intercon-
nections is more likely to make a contribution.
A Theory and Further Inductive Top-Down
An explanatory hypothesis for a set of facts will
Theorizing
be assessed as better when those facts “are part
Theorizing continues until the theorist be- of a network of interconnected explanations
lieves that the sensory and conceptual represen- rather than when they are in isolation” (Read &
tations cohere. Coherence has been operation- Marcus-Newhall, 1993: 438). In describing a good
alized in terms of the extent to which theory, Whetten suggested the theory must offer
explanations account for more evidence, are an explanation of “why”—“What are the under-
simpler, are themselves explained, and are su- lying psychological, economic, and social dy-
perior to alternative explanations (Read & Mar- namics that justify the selection of factors and
cus-Newhall, 1993). These attributes of a coher- the proposed causal relationships? This ratio-
ent explanation correspond to attributes of a nale constitutes the theory’s assumptions—the
strong theory from a philosophy of science per- theoretical glue that welds the model together”
spective (Dubin, 1978; Whetten, 1989) and a so- (1989: 491).
cially constructed perspective of contribution Finally, a theory with fewer alternative expla-
(Locke & Golden-Biddle, 1997). We offer the fol- nations is more likely to make a contribution.
lowing as a point of convergence on assessing The evaluation of a theory is partly comparative
the strength of a theorizing outcome and ac- in that a judge is partially influenced by the
knowledge that these dimensions are relatively availability of alternate explanations and how
independent—that is, an outcome may be stron- good they are (Read & Marcus-Newhall, 1993;
ger on some dimensions and weaker on others. Thagard, 1989). Indeed, Read and Marcus-
First, a theory that is broader is more likely to Newhall found that “the perceived goodness of a
make a contribution. The breadth of a theory set of explanations decreased when a strong
refers to the range of phenomena encompassed alternative explanation was available to sub-
by it (Bacharach, 1989: 509). A broader theory jects” (1993: 443). A better theory is one that is
explains more facts (Peirce et al., 1958) and, in more interesting (Davis, 1971)—a theory that
doing so, provides a more coherent explanation loosens “the normal science straightjacket”
than one that explains fewer facts. For example, (Daft & Lewin, 1990) to offer something new
Read and Marcus-Newhall (1993) found that a (Feldman, 2004; Mone & McKinley, 1993) that
broader explanation was considered “better” challenges and extends existing knowledge
than a single, narrow explanation. This is evi- (Whetten, 1989). To the extent that a theory is
dent in the judge’s comments regarding Ferraro, provocative, Sutton and Staw proposed that “the
Pfeffer, and Sutton’s (2005) Best Article Award for standards used to evaluate how well it is tested
the Academy of Management Review. As the or grounded need to be relaxed, not strength-
judges claimed, “The insights are fundamental, ened” (1995: 382). Moreover, novel theories are
surprising, and well argued. . . . Its potential more likely to be published (Beyer, Chanove, &
goes far beyond any one literature area” Fox, 1995; Zuckerman, 1987). In commenting on
(Kilduff, 2006: 792). This quote highlights the im- Moore, Tetlock, Tanlu, and Bazerman’s paper
portance of a theory’s breadth. (2006) as a Best Article finalist for the Academy
Second, a theory that is simpler is more likely of Management Review, the committee noted,
to make a contribution. A simpler theory is one “This model has profound practical implica-
that requires fewer assumptions. For example, tions, clearly suggesting that self-regulation, or
the more propositions required to explain a phe- self-policing, is unlikely to resolve the potential
nomenon, the less coherent the “story” is (Read for conflicts of interest to arise” (Kilduff, 2007:
& Marcus-Newhall, 1993). This principle is con- 1012). This highlights the attractiveness of the
sistent with the well-known notion of parsi- proposed explanation over self-regulation or
mony. A less parsimonious theory is one where self-policing.
374 Academy of Management Review April

An outcome of theorizing—for example, a jour- namic bodies of literature, in which the “state-
nal article, book, or conference proceeding— can of-the-art” knowledge is rapidly shifting, pose
form part of the Literature and can represent a difficulties for theorists to track (and compre-
stimulus for new theorizing. That is, the theory hend) all changes and remain “current.” Com-
represents a new data point in the flux of papers plex bodies of literature, in which constructs,
making up the Literature. Its visibility and legit- relationships between constructs, the nature of
imacy make it more likely that this theory will the paradox, and potential resolutions of the
“catch” theorists’ attention, stimulating another paradox are not simple, require considerable
round of inductive top-down theorizing (or de- cognitive effort to understand. A body of re-
ductive theorizing). Indeed, producing the theory search is disparate when the sources of infor-
outcome (paper) likely stimulates the original mation differ greatly such that its various liter-
researcher to further theorize on the topic. ature is at different levels of analysis, comes
A single theoretical paper is often not the final from different disciplines, has different para-
outcome of theorizing but one step in an ongo- digms, and has unique information contained in
ing, dynamic process. As Weick (1995: 385) different forms, such as double-blind peer-
pointed out, the products of the theorizing pro- reviewed articles, books, presentations, and
cess (e.g., published journal articles) seldom conference proceedings. When the body of re-
emerge as full-blown theories and mostly repre- search work is disparate, the superficial simi-
sent approximations, but these approximations larities across its bodies of literature are low,
represent “interim struggles” that may help the- requiring theorizing to develop/find structural
relationships.
orists inch toward stronger theories. This is con-
The reverse also holds. Inductive top-down
sistent with the pragmatic tradition (Peirce and
theorizing may be less appropriate when bodies
Dewey) that inquiry is an open-ended process
of literature are narrow, stable, simple, and well
with endless looping, self-questioning, and re-
integrated. A more appropriate approach is a
capitulation (Van Orden, 2008; Webb, 2007). The
typical top-down deduction to review the rele-
products of theorizing can be considered on a
vant aspects of the Literature as part of the the-
continuum (Weick, 1995), and subsequent theo-
orizing process. Alternatively, if the focal body
rizing (and its outcomes) can progress along that
of work is too narrow (or nonexistent), then a
continuum by producing outcomes that repre-
typical inductive bottom-up approach (or a de-
sent increasingly coherent (stronger) theories—
ductive bottom-up approach [e.g., Eisenhardt,
theories that are broader, simpler, more explicit
1989]) can be used to build theory from data
about interconnections, and have fewer alter-
about the phenomenon.
nate explanations.

Limitations/Challenges of Inductive
DISCUSSION
Top-Down Theorizing
When to Use Inductive Top-Down Theorizing
As with other types of theorizing, inductive
Inductive top-down theorizing is a mode of top-down theorizing has some limitations and
theorizing that may enhance the discovery or challenges. First, scholars are trained in certain
creation of a paradox (within or across para- disciplines, fields, and methods, and it is diffi-
digms) and is especially appropriate when the cult to put aside preconceived notions of theory
body of previous research is vast, dynamic, com- and method in order to approach the Literature
plex, and/or from disparate sources. Inductive with an openness that allows for the generation
top-down theorizing is most appropriate for un- of gists that direct attention to salient aspects
derstanding and resolving paradoxes that exist within it. Of course, this challenge is similar to
in bodies of literature that are extensive and the one faced by scholars who use an inductive
difficult for any one researcher to “know” at any approach to understanding a phenomenon. Per-
one point in time. Vast bodies of literature, con- haps those well versed in methods of inductive
taining large volumes of journal articles, books, theory building may be better equipped to deal
presentations, and working papers, represent a with these challenges. However, such a recog-
substantial amount of information for any single nition also implies that this is a skill that can be
researcher to be aware of and comprehend. Dy- learned. Theorists can learn to engage gists to
2011 Shepherd and Sutcliffe 375

direct attention to the discovery (or creation) of bottom-up approaches use deduction to provide
tensions, oppositions, contradictions, and expla- some initial direction for the bottom-up pro-
nations. cesses that still allow the theory to emerge from
Second, scholars who face the same contex- the data of the phenomena. Less progress has
tual environments—access to parts of the Liter- been made for top-down theorizing. Indeed, un-
ature, university policies for promotion and ten- like for scholars using bottom-up theorizing,
ure, desired journals for output, accepted scholars using a top-down approach appear to
practices by reviewers—are likely to focus their have fewer resources that describe the tools of
attention on similar aspects of those environ- the craft. In this article we offer a top-down
ments. This limits sources for new theories. Yet hybrid that focuses on induction to direct the
prescriptions to overcome the challenges of top-down process of theorizing. There is consid-
context are more difficult since theorists often erable literature describing what represents a
cannot change the context in which they are research opportunity—for example, the nature
embedded. Still, one can try to become less “vul- of a paradox and the means for its resolution
nerable” to contextual influences—for example, (e.g., Poole & Van de Ven, 1989). Building on
by obtaining tenure, using sabbaticals to work coherence theory’s notion of gists, our inductive
in different contexts, and/or imaginatively em- top-down model helps to answer the question of
bedding oneself in a different context for a par- how one finds a paradox or recognizes a re-
ticular line of theorizing. These strategies may search opportunity. That is, we explain how
help overcome the limitations of inductive top- gists are generated from the Literature itself and
down theorizing directing the theorist down op- focus attention. We also explain how their gen-
portunity paths common to other theorists. eration is obstructed by the strong use of the
Finally, inductive top-down theorizing is a dy- theorist’s knowledge and/or the strong influence
namic process, with an underlying tenet that of scholarly context.
recognizes that all theories are fallible, yet Specifically, we build on coherence theory to
knowledge is socially constructed. It is a chal- examine three attention allocation processes.
lenge to know when to stop theorizing and offer Attention allocation processes explain how a
an outcome to others (e.g., submit a paper to a theorist’s attention is focused on certain sources
journal for publication). Perhaps this limitation of information within the flux of data contained
is overcome by the notion of “saturation” in papers, books, working papers, and presenta-
(Strauss & Corbin, 1990: 212) or the comfort of tions that make up the Literature. A gist gener-
knowing that all theories are fallible and that ated from the Literature itself, the theorist’s
offering an outcome is simply one step in the knowledge, and the organizational and institu-
ongoing path of theorizing. Inductive top-down tional contexts (comprising norms, practices,
theorizing has its limitations and challenges. procedures, and communication channels) can
Some of the solutions may come from adapting all focus a theorist’s attention on specific bodies
existing deductive or inductive techniques, of work to form a sensory representation of the
some may require new solutions, and others research opportunity. We focused on gists for
may not be resolvable, but we believe that this inductive top-down theorizing and detailed how
approach offers a new source of theories of or- the other attention allocation processes obstruct
ganization. this form of theorizing. That is, the more the
theorist’s attention or scholarly context influ-
ences the allocation of attention, the less it re-
Implications for Scholars Theorizing
sembles inductive top-down theorizing and the
on Organizations
more it resembles traditional deductive top-
It is by gaining a deeper understanding of down theorizing. Future research can explore
theorizing (explanation) that we develop and the fine-grained nature of and interrelationship
improve it (perhaps generate prescriptions). Al- between these attention allocators. For exam-
though there have been numerous calls for the- ple, how can the theorist’s search heuristics
orizing to involve both induction and deduction (based on training and knowledge) be put on
and to use abduction, most of the progress to- hold to allow for the generation of a gist? Can
ward meeting these calls has been made by organizational or institutional norms, proce-
scholars using a bottom-up approach. These dures, and practices facilitate a theorist’s gen-
376 Academy of Management Review April

eration and use of gists in focusing attention? cles, books, presentations, and so on represents
There is ample opportunity for future research an amorphous stream of data from which a the-
on the attention allocation phase of inductive orist must carve out paradigms, research ques-
top-down theorizing. This is critical for a deeper tions, and explanations.
understanding of the recognition of research op- Second, abductive reasoning follows the the-
portunities. orist’s observation (perception/experience) to
Constant comparisons help explain how a build hypotheses about the experiential phe-
conceptual representation develops from bidi- nomenon (Fann, 1970; Hansen, 2008), which as-
rectional interaction with sensory representa- sumes knowledge of an anomaly (a doubt) that
tions of the Literature. From these bidirectional, requires a solution (to form a belief). Building on
iterative adjustments to the theorist’s conceptu- coherence theory, our model of inductive top-
alization, a theory emerges. We highlighted two down theorizing offers an antecedent stage in
“tools” for facilitating this comparison process: that we offer an explanation for how (and where)
thought experiments and metaphorical reason- the theorist’s initial attention is focused and,
ing. No doubt there are many others. One impor- thus, on what he or she is theorizing. Specifi-
tant avenue for future research on methods is to cally, we detail how a gist generated from the
adapt existing techniques from, say, grounded Literature itself can focus his or her attention on
theory for coding data of the phenomena to cod- specific aspects of the Literature to recognize a
ing data of the Literature (papers, books, presen- tension, opposition, and/or contradiction. This
tations, working papers, and so on). Such work tension, opposition, and/or contradiction is the
that develops “open” and “closed” coding tech- starting point for the theorizing path. We offer
niques for vast, dynamic, complex, and/or dis-
an explanation for why theorizing starts where
parate bodies of literature will make an impor-
it does and different means by which scholars’
tant contribution to theorizing.
theorizing is stimulated.
Given the fallibility of any theory, theorists
Future research can further investigate the
could continue to theorize indefinitely without
relationship between the formation and use of
sharing the outcomes of the process with others.
gists and how an “abductive suggestion comes
The decision to share likely depends on the as-
to us like a flash” (Peirce, 1955: 304). Such re-
sessment that the theory will make a contribu-
search might build on the literature on problem-
tion to the literature. A theory is more likely to
solving incubation or the literature on percep-
make a contribution when it is broader, simpler,
explicit about how and why, and has fewer ac- tual and/or conceptual fluency. For example, a
ceptable alternative explanations. There has flash of insight toward the solution of an unre-
been considerable discussion about the “qual- solved problem can come without conscious
ity” of the outcomes of theorizing (Kilduff, 2006; cognitive effort after a temporary shift away
Powell, 2001; Sutton & Staw, 1995; Whetten, 1989). from that problem (Knoblich, Ohlsson, Haider, &
Our purpose is not to offer a different set of Rhenius, 1999; Metcalfe & Weibe, 1987). This in-
criteria—indeed, ours are consistent with previ- cubation effect appears stronger for divergent
ous research. Rather, we want to highlight the thinking tasks and longer preparation periods
conditions under which theorizing is stopped (Sio & Ormerod, 2009), such as those often re-
(temporarily) to offer a theory—that is, when the quired for theorizing. Similarly, the ease with
outcome of theorizing is more likely to make a which stimuli (e.g., terms, constructs, relation-
contribution to knowledge. ships, etc., of the literature) come to the theorist’s
We recognize that some of our ideas are con- mind and influence the formulation of a solu-
sistent with abductive reasoning, particularly tion—perceptual and conceptual fluency (Ha-
around the notion of establishing a guess and mann, 1990; Jacoby & Dallas, 1981)—is influ-
then starting a journey of refinement, yet we enced by prior exposure (Cabeza & Ohta, 1993).
offer two major points of distinction. First, ab- Indeed, aspects of inductive top-down theorizing
ductive (and inductive) reasoning focuses on are likely to be conscious, such as the coding
data about a phenomenon. We focus on data as of data and constant comparison, and other as-
represented by the Literature. That is, we as- pects involve unconscious information process-
sume that collective knowledge of the intellec- ing, such as “in a flash of insight” that may
tual community as represented in journal arti- occur after a period of incubation.
2011 Shepherd and Sutcliffe 377

Although it is beyond the scope of the current Burks, A. 1946. Peirce’s theory of abduction. Philosophy of
article, future research can make an important Science, 13: 301–306.
contribution by investigating conscious and un- Cabeza, R., & Ohta, N. 1993. Dissociating conceptual prim-
conscious information processing and how the ing, perceptual priming, and explicit memory. European
Journal of Cognitive Psychology, 5: 35–53.
two are linked in inductive top-down theorizing.
It could be that unconscious processing is nec- Cameron, K. S. 1986. Effectiveness as paradox: Consensus
essary to achieve insight (Knoblich et al., 1999) and conflict in conceptions of organizational effective-
ness. Management Science, 32: 539 –553.
and that this is facilitated by conscious informa-
tion processing in the preceding stage of the Campbell, J. 1995. Understanding John Dewey. Chicago:
Open Court Press.
inductive top-down theorizing process. For ex-
ample, perhaps the conscious information pro- Chia, R. 2000. Discourse analysis as organizational analysis.
cessing involved in constant comparison pro- Organization, 7: 513–518.
vides the foundation on which unconscious Clarke, A. E. 2005. Situational analysis: Grounded theory
processing can build an insight. Furthermore, after the postmodern turn. Thousand Oaks, CA: Sage.
inductive top-down theorizing may rely on con- Coffey, A., & Atkinson, P. 1996. Making sense of qualitative
scious information processing to transform the data: Complementary research strategies. London: Sage.
insight into an articulated contribution (Locke & Cornelissen, J. 2005. Beyond compare: Metaphor in organiza-
Golden-Biddle, 1997). There are numerous re- tion theory. Academy of Management Review, 30: 751–
search opportunities to explore. 764.
In this article we offered an inductive top- Czarniawska, B. 1999. Writing management: Organization
down model of theorizing as a new explanation theory as a literary genre. New York: Oxford University
of theorizing about organizations and hopefully Press.
a useful tool for those looking to develop new Daft, R., & Lewin, A. 1990. Can organization studies begin to
theories of organization. break out of the normal science straitjacket? Organiza-
tion Science, 1: 1–9.

REFERENCES Daft, R., & Weick, K. 1984. Toward a model of organizations

and interpretation systems. Academy of Management
Alexander, J. C. 1988. Action and its environments: Toward a Review, 9: 284 –296.
new synthesis. New York: Columbia University Press.
Davidson, D. 1984. Inquiries into truth and interpretation.
Alexander, T. M. 1990. The pragmatic imagination. Transac- Oxford: Clarendon Press.
tions of the Charles S. Peirce Society, 26: 325–348.
Davis, M. W. 1971. That’s interesting! Philosophy of the Social
Almeder, R. 2007. Pragmatism and philosophy of science: A Sciences, 1: 309 –344.
critical survey. International Studies in the Philosophy
of Science, 21: 171–195. Dewey, J. 1896. The reflex arc concept in psychology. Psycho-
logical Review, 3: 357–370.
Alvesson, M., & Karreman, D. 2007. Constructing mystery:
Empirical matters in theory development. Academy Dewey, J. 1933. How we think. New York: Heath.
Management Review, 32: 1265–1281. Dewey, J. 1989. (First published in 1938.) Experience, knowl-
Anderson, D. 2005. Peirce and the art of reasoning. Studies in edge and value: A rejoinder. In J. A. Schilpp & L. E. Hahn
Philosophy and Education, 24: 277–289. (Eds.), Philosophy of John Dewey: 517– 608. La Salle, IL:
Bacharach, S. B. 1989. Organizational theories: Some criteria Open Court.
for evaluation. Academy of Management Review, 14: Dey, I. 1993. Qualitative data analysis. London: Routledge.
496 –515.
Dietrich, F., & Moretti, L. 2005. On coherent sets and the
Beyer, J. M., Chanove, R. G., & Fox, W. B. 1995. The review transmission of confirmation. Philosophy of Science, 72:
process and the fates of manuscripts submitted to AMJ. 403– 424.
Academy of Management Journal, 38: 1219 –1260.
DiMaggio, P. J. 1995. Comments on “What theory is not.”
Black, L. J., Carlile, P. R., & Repenning, N. P. 2004. A dynamic Administrative Science Quarterly, 40: 391–397.
theory of expertise and occupational boundaries in new
technology implementation: Building on Barley’s study Dubin, R. 1978. Theory development. New York: Free Press.
of CT scanning. Administrative Science Quarterly, 49: Dyer, A. A. 1986. Veblen on scientific creativity: The influence
572– 607. of Charles S. Peirce. Journal of Economic Issues, XX:
Blenko, W. J. 1991. Considering what constitutes prior art in 21–37.
the United States. JOM, 43: 45. Dyer, W. G., & Wilkins, A. L. 1991. Better stories, not better
Brown, J. R. 1986. Thought experiments since the scientific constructs, to generate better theory: A rejoinder to
revolution. International Studies in the Philosophy of Eisenhardt. Academy of Management Review, 16: 613–
Science, 1: 1–15. 619.
378 Academy of Management Review April

Eisenhardt, K. M. 1989. Building theories from case study Stockholm studies in philosophy. Stockholm: Almqvist &
research. Academy of Management Review, 14: 532–550. Wiksel International.
Eisenhardt, K. M. 1991. Better stories and better constructs: Hildebrand, D. L. 1996. Genuine doubt and the community in
The case for rigor and comparative logic. Academy of Peirce’s theory of inquiry. Southwest Philosophy Review,
Management Review, 16: 620 – 627. 12(1): 33– 43.
Eisenhardt, K. M., & Graebner, M. E. 2007. Theory building Holyoak, K. J., & Simon, D. 1999. Bidirectional reasoning in
from cases: Opportunities and challenges. Academy of decision making by constraint satisfaction. Journal of
Management Journal, 50: 25–32. Experimental Psychology: General, 128: 3–31.

Fann, K. T. 1970. Peirce’s theory of abduction. The Hague: Horváth, J., Czigler, I., Sussman, E., & Winkler, I. 2001. Simul-
Martinus Nijhoff. taneously active pre-attentive representations of local
and global rules for sound sequences in the human
Fauconnier, G., & Turner, M. 1998. Conceptual integration brain. Cognitive Brain Research, 12: 131–144.
networks. Cognitive Science, 22: 133–187.
Hyde, K. F. 2000. Recognising deductive processes in quali-
Feldman, D. 2004. What are we talking about when we talk tative research. Qualitative Market Research: An Inter-
about theory? Journal of Management, 30: 565–567. national Journal, 3: 82–90.
Ferlie, E., Fitzgerald, L., Wood, M., & Hawkins, C. 2005. The Jacoby, L. L., & Dallas, M. 1981. On the relationship between
nonspread of innovations: The mediating role of profes- auto-biographical memory and perceptual learning.
sionals. Academy of Management Journal, 48: 117–134. Journal of Experimental Psychology: General, 110: 306 –
Ferraro, F., Pfeffer, J., & Sutton, R. I. 2005. Economics lan- 340.
guage and assumptions: How theories can become self- Jarley, P., Chandler, T. D., & Faulk, L. H. 1998. Are we playing
fulfilling. Academy of Management Review, 30: 8 –24. the same game? Publishing task environments and re-
search productivity among management specialists.
Fine, G. A. 2004. The when of ethnographic theory. Perspec-
Human Relations, 51: 799 – 824.
tives (Newsletter of the ASA Theory Section), 27: 4, 5, 11.
Johnson, J. L., & Podsakoff, P. M. 1994. Journal influence in the
Folger, R., & Turillo, C. J. 1999. Theorizing as the thickness of
field of management: An analysis using Salancik’s in-
thin abstraction. Academy of Management Review, 24:
dex in a dependency network. Academy of Management
742–758.
Journal, 37: 1392–1407.
Gergen, K. J. 1978. Toward generative theory. Journal of Per-
Kahneman, D., Treisman, A., & Gibbs, B. 1992. The reviewing
sonality and Social Psychology, 36: 1344 –1360.
of object files: Object-specific integration of information.
Glaser, B. G. 1978. Theoretical sensitivity: Advances in the Cognitive Psychology, 24: 175–219.
methodology of grounded theory. Mill Valley, CA: Soci- Kilduff, M. 2006. Editor’s comments: Publishing theory. Acad-
ology Press. emy of Management Review, 31: 252–255.
Glaser, B. G. 1992. Basics of grounded theory analysis: Kilduff, M. 2007. Editor’s comments: Award winners for 2006 –
Emerging versus forcing. Mill Valley, CA: Sociology 2007. Academy of Management Review, 32: 1012–1013.
Press.
Klein, G. 1998. Sources of power: How people make decisions.
Glaser, B. G. 2001. The grounded theory perspective: Concep- Cambridge, MA: MIT Press.
tualization contrasted with description. Mill Valley, CA:
Knoblich, G., Ohlsson, S., Haider, H., & Rhenius, D. 1999.
Sociology Press.
Constraint relaxation and chunk decomposition in in-
Glaser, B. G., & Strauss, A. 1967. The discovery of grounded sight problem solving. Journal of Experimental Psychol-
theory: Strategies for qualitative research. London: Wie- ogy: Learning, Memory, and Cognition, 25: 1534 –1555.
denfeld and Nicholson.
Krippendorff, K. 1984. An epistemological foundation for
Gomez-Mejia, L. R., & Balkin, D. B. 1992. Determinants of communication. Journal of Communication, 34: 21–36.
faculty pay: An agency theory perspective. Academy of Lehrer, K. 1992. Coherence and skepticism. Philosophical
Management Journal, 35: 921–955. Issues, 2: 171–192.
Gross, N. 2009. A pragmatist theory of social mechanisms. Lehrer, K. 1999. Knowledge, skepticism and coherence. Phil-
American Sociological Review, 74: 358 –379. osophical Perspectives, 13: 131–139.
Gunter, V. 2005. News media and technological risks: The Locke, K., & Golden-Biddle, K. 1997. Constructing opportuni-
case of pesticides after Silent Spring. Sociological Quar- ties for contribution: Structuring intertextual coherence
terly, 46: 671– 698. and “problematizing” in organization studies. Academy
Hamann, S. B. 1990. Level-of-processing effects in conceptu- of Management Journal, 40: 1023–1062.
ally driven implicit tasks. Journal of Experimental Psy- Locke, K., Golden-Biddle, K., & Feldman, M. S. 2008. Making
chology: Learning, Memory, and Cognition, 16: 970 –977. doubt generative: Rethinking the role of doubt in the
Hansen, H. 2008. Abduction. In D. Barry & H. Hansen (Eds.), research process. Organization Science, 19: 907–918.
New approaches in management and organization: 454 – Maclean, N. 1992. Young men and fire. Chicago: University of
463. London: Sage. Chicago Press.
Harggqvist, S. 1996. Thought experiments in philosophy. McMullen, J. S., & Shepherd, D. A. 2006. Encouraging consen-
2011 Shepherd and Sutcliffe 379

sus-challenging research in universities. Journal of management and organization theories. Academy of

Management Studies, 43: 1643–1669. Management Review, 14: 562–578.
Merricks, T. 1995. On behalf of the coherentist. Analysis, 55: Powell, T. C. 2001. Competitive advantage: Logical and phil-
306 –309. osophical considerations. Strategic Management Jour-
nal, 22: 875– 888.
Merton, R. K. 1967. On theoretical sociology. New York: Free
Press. Pratt, M. G., & Foreman, P. O. 2000. Classifying managerial
responses to multiple organizational identities. Acad-
Metcalfe, J., & Weibe, D. 1987. Intuition in insight and non-
emy of Management Review, 25: 18 – 42.
insight problem-solving. Memory & Cognition, 15: 238 –
246. Read, S. J. 1987. Constructing causal scenarios: A knowledge
structure approach to causal reasoning. Journal of Per-
Millgram, E., & Thagard, P. 1996. Deliberative coherence.
sonality and Social Psychology, 52: 288 –302.
Synthese, 108: 63– 88.
Read, S. J., & Marcus-Newhall, A. 1993. Explanatory coher-
Mone, M. A., & McKinley, W. 1993. The uniqueness value and
ence in social explanations: A parallel distributed pro-
its consequences for organization studies. Journal of
cessing account. Journal of Personality and Social Psy-
Management Inquiry, 2: 284 –296.
chology, 65: 429 – 447.
Moore, D. A., Tetlock, P. E., Tanlu, L., & Bazerman, M. H. 2006.
Rensink, R. A. 2000. Visual search for change: A probe into
Conflicts of interest and the case of auditor indepen-
the nature of attentional processing. Visual Cognition, 7:
dence: Moral seduction and strategic issue cycling.
345–376.
Academy of Management Review, 31: 10 –29.
Rensink, R. A. 2002. Change detection. Annual Review of
Morgan, G. 1983. More on metaphor: Why we cannot control
Psychology, 53: 245–277.
tropes in administrative science. Administrative Sci-
ence Quarterly, 28: 601– 607. Richardson, R., & Kramer, E. H. 2006. Abduction as the type of
inference that characterizes the development of a
Morse, J. M. 1994. Emerging from the data: The cognitive
grounded theory. Qualitative Research, 6: 497–513.
processes of analysis in qualitative inquiry. In J. M.
Morse (Ed.), Critical issues in qualitative research meth- Rorty, R. 1989. Contingency, irony and solidarity. Cambridge:
ods: 23– 43. London: Sage. Cambridge University Press.
Nord, W. R., & Connell, A. F. 1993. From quicksand to cross- Rothenberg, A. 1976. Janusian thinking and creativity. Psy-
roads: An agnostic perspective on conversation. Orga- choanalytic Study of Society, 7: 1–30.
nization Science, 4: 108 –120. Samuels, W. J. 2000. Signs, pragmatism, and abduction: The
Ocasio, W. 1997. Towards an attention-based view of the tragedy, irony, and promise of Charles Sanders Peirce.
firm. Strategic Management Journal, 18: 187–206. Journal of Economic Issues, 34: 207–217.
Oliva, A., & Schyns, P. G. 2000. Diagnostic colors mediate Sandelowski, M. 1997. “To be of use”: Enhancing the utility of
scene recognition. Cognitive Psychology, 41: 176 –210. qualitative research. Nursing Outlook, 45: 125–132.
Oliva, A., & Torralba, A. 2001. Modeling the shape of the Scheffler, I. 1979. Beyond the letter: A philosophical inquiry
scene: A holistic representation of the spatial envelope. into ambiguity, vagueness, and metaphor in language.
International Journal in Computer Vision, 42: 145–175. London: Routledge and Kegan Paul.
Oliva, A., Torralba, A., Castelhano, M. S., & Henderson, J. M. Senge, P. 1990. The fifth discipline: The art and practice of the
2003. Top-down control of visual attention in object de- learning organization. London: Century Business.
tection. Proceedings of the IEEE International Confer- Shepherd, D. A., McMullen, J. S., & Jennings, P. D. 2007. The
ence on Image Processing, 1: 253–256. formation of opportunity beliefs: Overcoming ignorance
Paavola, S. 2004. Abduction as a logic and methodology of and reducing doubt. Strategic Entrepreneurship Journal,
discovery: The importance of strategies. Foundations of 1: 75–95.
Science, 9: 267–283. Siggelkow, N. 2001. Change in the presence of fit: The rise,
Peirce, C. S. 1955. Logic as semiotic: The theory of signs. In J. the fall, and the renaissance of Liz Claiborne. Academy
Buchler (Ed.), The philosophical writings of Peirce: 98 – of Management Journal, 44: 838 – 857.
119. New York: Dover. Siggelkow, N. 2002. Evolution toward fit. Administrative Sci-
Peirce, C. S., Hartshorne, C., Weiss, P., & Burks, A. 1958. ence Quarterly, 47: 125–159.
Collected papers of Charles Sanders Peirce, vols. 1– 8. Simon, D., & Holyoak, K. J. 2002. Structural dynamics of cog-
Harvard, MA: Cambridge University Press. nition: From consistency theories to constraint satisfac-
Pennington, N., & Hastie, R. 1992. Explaining the evidence: tion. Personality and Social Psychology Review, 6: 283–
Tests of the story model for juror decision making. Jour- 294.
nal of Personality and Social Psychology, 62: 189 –206. Simon, D., Krawczyk, D. C., & Holyoak, K. J. 2004. Construction
Pfeffer, J. 1993. Barriers to the advance of organizational of preferences by constraint satisfaction. Psychological
science: Paradigm development as a dependable vari- Science, 15: 331–336.
able. Academy of Management Review, 18: 599 – 620. Simon, D., Pham, L. B., Le, Q. A., & Holyoak, K. J. 2001. The
Poole, M. S., & Van de Ven, A. H. 1989. Using paradox to build emergence of coherence over the course of decision
380 Academy of Management Review April

making. Journal of Experimental Psychology: Learning, Van Orden, G. C. 2008. Peirce’s abduction and cognition as
Memory, and Cognition, 27: 1250 –1260. we know it. Journal of Experimental & Theoretical Arti-
ficial Intelligence, 20: 219 –229.
Simon, D., Snow, C. J., & Read, S. J. 2004. The redux of cog-
nitive consistency theories: Evidence judgments by con- Walker, R. C. 1989. The coherence theory of truth: Realism,
straint satisfaction. Journal of Personality and Social anti-realism, idealism. London: Routledge.
Psychology, 86: 814 – 837. Webb, J. L. 2007. Pramatisms (plural) Part 1: Classical prag-
Simon, H. A. 1947. Administrative behavior: A study of deci- matism and some implications for empirical inquiry.
sion making processes in administrative organizations. Journal of Economic Issues, XLI: 1063–1086.
Chicago: Macmillan. Weick. K. E. 1980. Blind spots in organizational theorizing.
Simons, D. J., & Levin, D. T. 1998. Failure to detect changes to Group & Organization Studies, 5: 178 –188.
people in a real-world interaction. Psychonomic Bulletin Weick, K. E. 1989. Theory construction as disciplined imagi-
and Review, 5: 644 – 649. nation. Academy of Management Review, 14: 516 –531.
Simons, D. J., & Wang, R. F. 1998. Perceiving real-world view- Weick, K. E. 1993. The collapse of sensemaking in organiza-
point changes. Psychological Science, 9: 315–320. tions: The Mann Gulch disaster. Administrative Science
Sio, U., & Ormerod, T. C. 2009. Does incubation enhance Quarterly, 38: 628 – 652.
problem solving? A meta-analytic review. Psychological Weick, K. 1995. What theory is not, theorizing is. Administra-
Bulletin, 135: 94 –120. tive Science Quarterly, 40: 385–390.
Skyrms, B. 1992. Coherence, probability and induction. Phil- Weick, K. E. 1996. Drop your tools: An allegory for organiza-
osophical Issues, 2: 215–226. tional studies. Administrative Science Quarterly, 41:
301–313.
Spellman, B. A., & Holyoak, K. J. 1992. If Saddam is Hitler then
who is George Bush? Analogical mapping between sys- Weick, K. E., & Roberts, K. H. 1993. Collective mind in orga-
tems of social roles. Journal of Personality and Social nizations: Heedful interrelating on flight decks. Admin-
Psychology, 62: 913–933. istrative Science Quarterly, 38: 357–382.
Strauss, A. L., & Corbin, J. 1990. Basics of qualitative research: Weick, K. E., Sutcliffe, K. M., & Obstfeld, D. 2005 Organizing
Techniques and procedures for developing grounded and the process of sensemaking. Organization Science,
theory. Newbury Park, CA: Sage. 16: 409 – 421.
Suddaby, R. 2006. From the editors: What grounded theory is Whetten, D. A. 1989. What constitutes a theoretical contribu-
not. Academy of Management Journal, 49: 633– 642. tion? Academy of Management Review, 14: 490 – 495.

Sutton, R. I., & Staw, B. M. 1995. What theory is not. Admin- Whetten, D. A. 2002. Modeling-as-theorizing: A methodology
istrative Science Quarterly, 40: 371–384. for theory development. In D. Partington (Ed.), Essential
skills for management research: 45–71. Thousand Oaks,
Thagard, P. 1989. Explanatory coherence. Behavioral and CA: Sage.
Brain Sciences, 12: 435–502.
Wilson, H. S., & Hutchinson, S. A. 1996. Methodologic mis-
Thompson, J. D. 1956. On building an administrative science. takes in grounded theory. Nursing Research, 45: 122–124.
Administrative Science Quarterly, 1: 102–111.
Wolfe, J. 1999. Inattentional amnesia. In V. Coltheart (Ed.),
Torralba, A., & Oliva, A. 2003. Statistics of natural images Fleeting memories: 71–94. Cambridge, MA: MIT Press.
categories. Network: Computation in Neural Systems, 14:
Yin, R. 2003. Case study research: Design and methods. Thou-
391– 412.
sand Oaks, CA: Sage.
Tourangeau, R., & Sternberg, R. 1982. Understanding and
Yin, R. 2004. Applications of case study research. Thousand
appreciating metaphors. Cognition, 11: 203–244.
Oaks, CA: Sage.
Tsoukas, H. 1991. The missing link: A transformational view
Zuckerman, H. 1987. Persistence and change in the careers of
of metaphors in organizational science. Academy of
American men and women scientists and engineers: A
Management Review, 16: 566 –585.
review of current research. In L. S. Dix (Ed.), Women:
Van Maanen, J., Sorensen, J. B., & Mitchell, T. R. 2007. The Their underrepresentation and career differentials in
interplay between theory and method. Academy of Man- science and engineering: 127–162. Washington, DC: Na-
agement Review, 32: 1145–1154. tional Academy Press.

Dean A. Shepherd (shepherd@indiana.edu) is the Randall L. Tobias Chair in Entre-

preneurial Leadership and professor of entrepreneurship at the Kelley School of
Business, Indiana University. He earned his Ph.D. at Bond University (Australia). His
research focuses on entrepreneurial decision making, emotions, and failure.

Kathleen M. Sutcliffe (ksutclif@umich.edu) is the Gilbert and Ruth Whitaker Professor

of Business Administration and professor of management and organizations at the
Ross School of Business, University of Michigan. She earned her Ph.D. at University of
Texas at Austin. Her research focuses on processes of adaptability, resilience, and
managing unexpected events.
Copyright of Academy of Management Review is the property of Academy of Management and its content may
not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.

View publication stats