Concept Creation, Coherence, and Cohesion
University of Chicago Booth School of Business
5807 South Woodlawn Ave.
Chicago, IL 60637
MIT Sloan School of Management
50 Memorial Drive, E52
Cambridge, MA 02142-1347 USA
Please do not distribute without permission of the authors
Concept Creation, Coherence, and Cohesion
This paper explores why actors may develop different interpretations of a new concept. It
leverages work from cognitive science to define this interpretive process in terms of creating
coherence: fitting the new concept into the existing conceptual schema. We argue that variation
in conceptualization results in part from cognitively cohering the concept in different ways. We
appeal to the social structure of the group to explain this variance. Cohesive groups favor
establishing similarities between the new concept and something familiar in the existing schema.
Diffuse groups lack a clear, agreed-upon schema to hook a new concept into, and therefore favor
making sense of the new concept without directly integrating it into the schema. We illustrate
the relationship between social cohesion and cognitive practices by analyzing the different
conceptualizations of the computer during its introduction in the insurance industry from 1940-
Contemporary sociology has become increasingly interested in how new concepts get
formed and developed. As actors encounter new things in the world, they form conceptual
representations of them which, in turn, influence future interactions. Often actors develop
different conceptualizations of the same new object. In the sociology of technology, for
instance, Pinch and Bijker (1987) show that some groups interpreted the new bicycle in terms of
safety; whereas, others focused on speed. Given these differences in interpretation, sociologists
have generally tried to explain the mechanisms that select a particular interpretation—for
example, why the safety bicycle emerged as the dominant conceptualization. Or, absent a
selection process, what enables translation between alternative interpretations, for example
boundary objects (Bowker and Star 1999). Sociological research has paid less attention to how
these different interpretations form in the first place. Yet, to fully comprehend the subsequent
selection and translation processes requires an understanding of the socio-cognitive processes
that develop the initial interpretations.
How different social groups interpret a new concept is partly a function of the cognitive
processes that the groups use to construct these interpretations. Durkheim explains that these
cognitive processes entail fitting the new concept with existing beliefs and conceptions. He
asserts, “If they [new concepts] are not in harmony with other beliefs, other opinions, in a word
with the whole gamut of collective representations, they will be denied; minds will be closed to
them; they will be as if they never were” (Durkheim 1912 , pp.333-4). DiMaggio (1997)
provides a more modern interpretation of Durkheim’s assertion, describing “the whole gamut of
collective representations” as a schema into which a new concept must fit. Following this
schema image, it is quite conceivable that actors vary in how they fit the same new concept
within their conceptual schema. For instance, the concept of safety bicycle fits the bike into the
conceptual schema by focusing on the “safety” relation, but the speed bicycle focuses on the
“speed” relation. This possibility of different ways to fit in the new concept lies at the heart of
Swidler’s (1986) notion of culture as a “tool kit” – actors may differ in how they apply their
In this article, we begin to develop an explanation of why actors may vary in how they
fit a new concept within a conceptual schema. We call Durkheim’s “fitting in” process creating
coherence. Recently, cultural sociologists have been building closer ties with cognitive science
(Cerulo 2002; DiMaggio 1997) and we believe research into the various cognitive devices actors
use to make sense of concepts provides a helpful framework to specify the coherence process. In
particular, cognitive science has developed individual theories about how cognitive processes
like analogies or conceptual combinations work such that, taken collectively, they show the
various ways actors can fit a new concept within an existing schema. We use this work to
further define coherence along two dimensions: the aspect of the schema to which the new
concept is linked, and whether that link emphasizes similarities or differences. These
dimensions, in turn, express the different ways a new concept can cohere. Analogies, for
instance, map relational structures between schemas and highlight similarities; whereas,
classification works at the attribute level and focuses on differences. Therefore, using an
analogy leads to a different kind of coherence than using classification.
While cognitive scientists help explain what cognitive coherence is, they do not pay as
much attention to why actors may vary in how they cohere. To address this important concern,
we leverage the long standing work that relates social with conceptual structure to argue that the
social structure of the group doing the interpretation places different constraints on the cognitive
processes of coherence. In particular, we focus on the cohesion of the social group, invoking
Martin’s (2002) concepts of tightness and consensus to measure the level of cohesion. We
propose that cohesive groups require something familiar within the existing schema to pull the
new concept in, consequently favoring coherence processes that emphasize similarities over
differences. In contrast, diffuse groups generate a different kind of constraint. Because there is
no clear schema to hook a new concept into, these conditions favor making sense of the new
concept without integrating it directly into the schema.
Finally, we illustrate this argument by analyzing the discursive strategies among different
groups within the insurance industry use to interpret the new concept of the computer when it
was introduced in the 1940s – 1960. The coherence processes varied from using an analogy with
human thinking to using the conceptual combination “information processing machine.” We
explain these differences in creating coherence in terms of the level of cohesion of the social
groups to which these interpretations were targeted. We believe this work extends the renewed
interest in the structural characteristics of culture and knowledge by exploring how social and
conceptual structures of existing schemas influence the production of new concepts and
knowledge. We also believe that this analysis contributes to the growing sociological literatures
that invoke social cognition, such as organization theory and sociological studies of technology,
by showing the need for greater appreciation of the variation in the cognitive processes actors
use to make interpretations.
We begin by applying insights from cognitive science to define coherence.
Different ways to cohere – Insights from Cognitive Science
The term “coherence” is vague. Philosophers describe coherence as a characteristic of
belief systems and focus on its role in the justification of beliefs (Thagard 2000). However,
while conceptual schemas vary in their level of internal coherence, each has the potential to form
a new concept. Thus, in this paper, we are not interested in coherence as a criterion for
justification, but in coherence as the process for fitting in a new concept. While cognitive
scientists do not typically use the word “coherence,” they more generally address how actors
leverage different aspects of the cognitive repertoire to form new concepts. These explanations,
in turn, form the basis for a more formal definition of conceptual coherence.
Cognitive scientists define conceptual schemas in terms of the concepts, the attributes
that define them, and the relations between the concepts (Gentner, Holyoak and Kokinov 2001).
In general, the coherence of a new concept involves retrieving one of these aspects with which to
fit the new concept. Specifically, a new concept can be fit in at the conceptual, attribute, or
relational level. In addition, cognitive scientists specify the nature of fitting in as showing how
the concept is either similar to or different from the retrieved aspect of the conceptual schema
(Gentner, Holyoak and Kokinov 2001). For example, fitting the new concept in at the attribute
level could involve showing how it shares that attribute or how it differs from that attribute.
These processes characterize two important dimensions of coherence – what in the conceptual
schema the concept is fit into and whether the fitting in involves establishing similarities or
differences. Cognitive scientists also explain the kinds of cognitive devices actors use to form
new concepts, such as analogies, metaphors, or classification, in terms of these two dimensions
(Gentner, Holyoak and Kokinov 2001). These cognitive devices, in turn, represent different
ways actors can cohere a new concept into their conceptual schema. Consequently, an actor who
uses an analogy understands the same concept differently than an actor who uses classification.
Cognitive scientists tend to consider these devices piecemeal to develop theories about
how they function; however, for our purposes we consider them collectively to show differences
in how they cohere. To make this discussion more concrete, we use a thought experiment to
explain the different ways to cohere a new vegetable X into an existing vegetable schema. We
have purposefully chosen to use the example of vegetables because it is significantly different
than our empirical example of computers within the insurance industry. We focus on three
common cognitive devices discussed in the cognitive science literature – analogy, classification,
and conceptual combination – to show how they cohere the new vegetable X. While this list
certainly is not exhaustive, it illustrates the substantial differences in how a concept can cohere.
Since coherence involves relating the new concept to an element of the conceptual
schema, it is first necessary to further define its elements. We begin with concepts. Concepts
denote the different categories that identify similar instances and differentiate them from other
concepts (Zerubavel 1996). Examples of concepts for the vegetable schema may include pepper,
romaine lettuce, and root vegetable. Eleanor Rosch and Mervis (1975) recognized that concepts
occur at different levels of aggregation. Some concepts, like romaine lettuce, are quite specific,
whereas others, such as pepper, are more general, including among its members other concepts
like bell peppers and jalapeno peppers.
For each concept, a set of attributes characterizes its members (instances of that concept).
For example, among the many attributes that root vegetables have, some include growing in the
ground, having tough outer skin, and being harvested mostly in the fall. Sociologists typically
follow Rosch and Wittgenstein in the view that there is no one set of attributes that all members
have; rather, members may share different clusters of attributes (Hannan, Polos and Carroll
2007; Rosch and Mervis 1975; Wittgenstein 1958). For instance, jalapeno and habanero peppers
both are hot, but bell and apple peppers are sweet; nevertheless, they all are thought of as
peppers. This implies an additional process that identifies which of the shared attributes defines
the concepts. Sociologists call this process the institutionalization of the concept and have
emphasized the social processes that determine the salient attributes (Berger and Luckmann
1966; Hannan, Polos and Carroll 2007). This view also implies that there is an inherent structure
to each concept, where some instances are more prototypical than others (Rosch and Mervis
1975). 1 Some types of peppers may be seen as more core than others. Given that the salient
attributes are socially defined, groups may differ in their definition of prototypical members. In
Italy, the sweet bell peppers may be seen as the prototypical pepper, but in Mexico, the hot
jalapeno may be the prototypical pepper.
Different concepts may overlap with each other to the extent that they share attributes.
For example, the concepts of romaine lettuce and peppers share the attributes of green and
crunchy. As a result, within a schema at a given point in time, concepts vary in how precisely
they are defined and how they are demarcated from other concepts at the attribute level.
Population ecologists are beginning to develop explanations of how the internal structure and
overlap between categories at the attribute level influence market behavior (Hannan, Polos and
1 Cognitive scientists propose several different ways to determine the core attributes of a concept beyond typicality
(see Cerulo 2002) for a review). For instance, there is the exemplar and theoretical view. While these approaches
differ significantly in explaining the identification of the salient attributes, they share the same fundamental belief
that only certain attributes define a concept. We are agnostic to this debate and use the prototypical approach simply
as an example.
Finally, concepts are connected through a series of relations. Like attributes, there are
many different kinds of relations: causation, functional, comparison, and opposition to name a
few. For instance, root vegetables has the relation of “is denser than” with peppers, whereas
jalapeno has the relation of “is member of” with peppers, rain has the functional relationship of
“provides nutrients to” with each of the mentioned vegetable concepts. There is a complex
relational structure that relates different concepts together independent of the attributes that
define each concept. And, just as certain attributes may be favored over others, so too may
certain relations be favored.
With the elements of a conceptual schema better specified, now consider the emergence
of a new vegetable X that requires actors to cohere it with this existing schema. For the sake of
illustration, imagine that this new vegetable has the following attributes and relations: mealy
texture, sweet taste, yellow and red in color, it depletes minerals in the ground preventing other
plants from growing and thrives in desert conditions. We can now consider some of the different
ways in which vegetable X can be fit into the schema.
Actors could cohere vegetable X through an analogy, for example with a lion. The
structure-mapping theory of analogies asserts that analogical thinking involves the relational
mapping between the base and the target (Gentner 1983). In this case, the analogy involves
mapping the relations the lion has with other concepts to vegetable X and its relations with other
concepts. For example, the lion eats it prey just as the vegetable depletes the minerals from the
soil or the lion thrives within the Sahara just as the vegetable thrives within its desert climate.
Attributes of the concepts play a lesser role; in fact, vegetable X looks nothing like a lion. The
analogy of vegetable X with the lion tells us how the vegetable relates to its environment and
sources of nourishment as opposed to defining the salient attributes that define and differentiate
the vegetable. Therefore, from a coherence perspective, analogical thinking focuses on
similarities in relational structure. However, in a pure sense, analogies do not fit the new
concept into the base concept’s conceptual schema. Vegetable X is not assimilated within the
lion’s schema, but shown to have a similar relational structure. In this respect, DiMaggio (1997)
postulates that analogies are useful tools to link different schemas together.
In another form of coherence, actors could classify vegetable X within the existing
conceptual schema of vegetables. Zerubavel (1996) describes this process as lumping and
splitting: a process of identifying similar attributes among the different instances of vegetable X
and then differentiating these attributes from other concepts. For instance, a set of actors,
perhaps plant experts, could determine that of vegetable X’s many attributes, mealy texture, is
the salient characteristic which defines it. Beyond identifying similar attributes within vegetable
X, these actors may also compare it to other familiar vegetables. In this process, they recognize
that other vegetables share these attributes such that vegetable X overlaps with them, and must
determine whether it should be assimilated within an existing category or exist independently.
While there may be other specific processes that do not involve experts, classification in general
focuses on differentiation at the attribute level. Therefore, where analogies work mostly at the
relational level and focus on similarities, classification works mostly at the attribute level and
focuses on differentiation.
Finally, conceptual combination is a third kind of coherence which involves combining
aspects of two familiar concepts to define the new concept. In one form, this process combines
an attribute from one concept to another. For example, we could describe vegetable X as a
“peach pepper” which maps the properties of mealy, sweet, and yellowish red from the
modifying concept, “peach”, to the head concept, “pepper.” According to Wisniewski (1996),