Vous êtes sur la page 1sur 29

540131

research-article2014

SCUXXX10.1177/2329496514540131Social CurrentsLeahey and Moody

Sociological Innovation
through Subfield Integration

Social Currents
2014, Vol. 1(3) 228256
The Southern Sociological Society 2014
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/2329496514540131
scu.sagepub.com

Erin Leahey1 and James Moody2,3

Abstract
Is domain-spanning beneficial? Can it promote innovation? Classic research on recombinant
innovation suggests that domain-spanning fosters the accumulation of diverse information and
can thus be a springboard for fresh ideasmost of which emanate from the merger of extant
ideas from distinct realms. But domain-spanning is also challenging to produce and to evaluate.
Here, the domains of interest are subfields. We focus on subfield spanning in sociology, a
topically diverse field whose distinct subfields are still reasonably permeable. To do so, we
introduce two measures of subfield integration, one of which uniquely accounts for the novelty of
subfield combinations. We find (within the limits of observable data) the costs to be minimal but
the rewards substantial: Once published, sociology articles that integrate subfields (especially
rarely spanned subfields) garner more citations. We discuss how these results illuminate trends
in the discipline of sociology and inform theories of recombinant innovation.
Keywords
science knowledge, boundary spanning, higher education, networks, multilevel models,
innovation
How is new and useful knowledge produced?
Classic and recent literature suggests that
spanning boundaries and pooling diverse
information from distinct knowledge domains
is essential. Adam Smith ([1766] 1982:539)
argued that when the mind is employed about
a variety of objects it is somehow expanded
and enlarged. Katz and Lazarsfeld (1955:345)
noted that people who see and act on differences across groups, and bridge them, have an
advantage in detecting and developing rewarding opportunities. More recent scholarship
suggests that information pooled from disparate sources provides a (if not the) foundation
from which new combinations and ideas spring
(Abbott 2001; Fleming, Mingo, and Chen
2007; Hargadon 2002). The close link between
domain spanning and idea generation is captured in the term recombinant innovation
(Weitzman 1998).

Yet scholars are beginning to question the


benefits of domain-spanning in academe and
to document associated challenges. Jacobs and
Frickel (2009) show that the benefits of domain
spanning are largely unsubstantiated, so enthusiasm may be premature. In fact, domain-spanning ideas face numerous challenges. On the
producer side, it is difficult to search unfamiliar topics (Fleming 2001; Schilling and Green
2011), to master and adequately represent literature
from
distinct
subfields,
to
accommodate the research mores of multiple
1

University of Arizona, Tucson, USA


Duke University, Durham, NC, USA
3
King Abdulaziz University, Jeddah, Saudi Arabia
2

Corresponding Author:
Erin Leahey, Department of Sociology, University of
Arizona, P.O. Box 210027, Tucson, AZ 85721-0027,
USA.
Email: leahey@arizona.edu

Downloaded from scu.sagepub.com by guest on June 23, 2015

229

Leahey and Moody


specialty areas (Lamont, Mallard, and
Guetzkow 2006), and to produce output that is
standard in form and content. On the audience
side, reviewers drawn from the representative
domains may evaluate the merits of a work
very differently (Lamont 2009; Lamont et al.
2006; Mansilla 2006), leading to low interreviewer reliability that, in turn, increases the
likelihood of rejection.
The goal of this article is to critically apply
these ideas to academic scholarship. How do
domain-spanning academic articles fare? Do
they benefit from the diverse perspective,
tools, and theories that different domains offer,
as the recombinant innovation literature suggests? Or does their integrative character make
them challenging to produce and evaluate?
These questions are important to answer
because domain spanning is ubiquitous in academe; for example, interdisciplinary scholarship is on the rise (Brint 2005; Jacobs and
Frickel 2009), perhaps because most research
problems lie at the intersection of established
ideas (Braun and Schubert 2003). To address
these questions, we focus on the field of sociology, a diverse discipline integrating many
topics (Clemens et al. 1995) with robust yet
permeable subfields1 (Moody 2004). Our study
of inter-subdisciplinary scholarship elucidates
the process of (knowledge) product legitimation and increases our understanding of how
scientific knowledge evolves (Mulkay
1974:228). More immediately, our research
speaks to ongoing academic debates about the
state of social science disciplines, especially
concerns about increasing specialization and
fragmentation in sociology.
This article makes contributions in terms of
theory, measurement, and the substantive
topic.Theoretically, we introduce the concept
of subfield integration and demonstrate its relevance to scholarship on recombinant innovation and on the evaluation of domain-spanning
work. We then develop two measures of such
scholarly domain spanning: a binary one that
measures whether domain-crossing occurs
(which we call nominal integration) and a
more refined one that incorporates the distinctiveness of the pairing (which we call novel
integration). Use of alternate measures allows

us to assess two mechanisms that may explicate the aforementioned benefits of integration: Is it merely access to a broader audience
(i.e., members of two subfields rather than
one) that enhances value, or is the novelty of
the research implicated? While we cannot
measure innovation directly, we assess whether
novelty is one reason why subfield integration
accrues benefits. Thus, like Fleming et al.
(2007), we distinguish between novelty (a
mechanism we tap by juxtaposing two measures of subfield integration) and usefulness
(the outcome of interest). Our results show that
there are consistent, strong, and meaningful
positive returns to subfield integration.

The Concept of Integration


Integration across disciplinesinter-disciplinarityhas received much scholarly attention
(Abbott 2001:131; National Academies of
Science 2005; Rhoten and Parker 2006), but
integration within disciplineswhat we call
subfield integrationis much less studied.
While this likely reflects the topical homogeneity of many disciplines (Moody and Light
2006), it is an obvious lacuna in broad, diverse
fields like sociology where the internal driver
of new ideas might well come from integrating
subfields. This imbalance, which we begin to
rectify in this article, is surprising given that
processes that occur both between and within
disciplines are critical to adequately capturing
the the substantive heart of the academic system (Abbott 2001:148). Like fields, subfields
produce insights, provide structure to the academic labor market (where advertisements for
positions are distinguished by both field and
subfields), and prevent knowledge from
becoming too abstract and overwhelming for
scholars (Abbott 2001).
An influential working definition of interdisciplinary research is adaptable to the kind of
integrative research that interests us: that
which spans or bridges subdisciplines. The
National Academies report (National
Academies of Science 2005:188) defines interdisciplinary research as a mode of research . . .
that integrates perspectives, information, data,
techniques, tools, perspectives, concepts, and/

Downloaded from scu.sagepub.com by guest on June 23, 2015

230

Social Currents 1(3)

Figure 1. Growth in number of American Sociological Association sections.

or theories from two or more disciplines or


bodies of specialized knowledge. By analogy,
we define subfield integration as the borrowing or mixing of insights from two or more
subfields within a discipline. While perhaps
less stable than disciplines, subfields are the
building blocks of science (Small and Griffith
1974). Whereas disciplines form the teaching
domain of science, smaller intellectual units,
nestled within and between disciplines constitute the research domain of science (Chubin
1976) and thus are most relevant to a study of
scholarly innovation. Subfields are the microenvironments for research (Hagstrom
1970:93) that provide potential interconnections (Ennis 1992:260). According to our
definition, an article on student performance
might fall cleanly within the subfield sociology of education, but an article that examines
gender differences in student performance
might be drawing fromand speaking back
tothe research literature in two subfields,
sociology of education as well as gender,
and thus be considered integrative.
Internal divisions like subfields may be particularly important in low consensus fields
(Shwed and Bearman 2010) like sociology.
Subfields are largely distinguished by their
substantive focus of inquiry (more so than
methodological or theoretical approaches,
where there is overlap) and typically have their
own American Sociological Association

(ASA) section, journal(s), and conferences. As


these subfields grow in number (see Figure 1),
so do the opportunities for combination and
cross-fertilizationwhat we call subfield integration. Numerous studies in the 1990s used
data on ASA section comemberships or cocitation patterns to identify sociologys subdisciplinary structure (Cappell and Gutterbock
1992; Crane 1972; Ennis 1992). These studies
show that while subfields are clearly distinct,
some hang together more than others, forming
a loosely coupled system. This is also apparent
in our data (for the period 19852004) as
depicted in Figure 2. Proximate fields possess
cognitive overlaps that are based largely on
ideas and subject matter (Cappell and
Gutterbock 1992; Edge and Mulkay 1975). It
is near the edges of subfields, where they overlap with other subfields, that innovations are
most likely to emerge (Chubin 1976).
While most scholars simply assess whether
domain spanning has occurred (Clemens et al.
1995; Jacobs and Frickel 2009), consideration of
the novelty of the pairing is important for clarifying innovation processes. Only a few attempts
have been made to assess the uniqueness, or rarity, of combinations (Braun and Schubert 2003;
Carnabuci and Bruggeman 2009; Rosenkopf
and Almeida 2003; Schilling and Green 2011;
Uzzi et al. 2013). This is ideal because it is new
connections that characterize originality
(Guetzkow, Lamont, and Mallard 2004). The

Downloaded from scu.sagepub.com by guest on June 23, 2015

231

Leahey and Moody

Figure 2. Observed subfield integration, 19852004.

Note. Edges present if subfields share more than 1.75 (Expected value).

most fertile creative products are drawn from


domains that are far apart (Poincare [1908]
1952) and the best conceptual metaphors are
those that create ties across great distances
(Knorr Cetina 1980). Put simply, integrative
work can be more or less innovative, depending
on the relationship between the integrated entities (Carnabuci and Bruggerman 2009:608).
Thus, for our analysis of subfields, it is possible
to examine not only whether two subfields are
spanned (which we call nominal integration)
but also how rarely they are spanned, which
gives us an indication of the novelty of the pairing (which we call novel integration).2
The scholarship on domain-spanning also
tends to neglect time (see Carnabuci and

Bruggeman 2009 for an exception). When considering the effects of subfield integration, it
may be critical to assess the trajectory of each
combination. In sociology, the combination of
poverty and culture was on the wane 1985
2004, but the combination of poverty and
urban sociology was on the rise. A rare-andgetting-rarer combination may reflect something of a dead endreflecting a combination
that used to be of interest but is no longer,
while a rare-but-increasingly common combination reflects a growth area that might signal
disciplinary excitement. Bridges do decay
(Burt 2002; Ryall and Olav 2007), and the
advantages of bridging distinct domains,
which we discuss below, decline when

Downloaded from scu.sagepub.com by guest on June 23, 2015

232

Social Currents 1(3)

everyone strives for them (Buskens and van de


Rijt 2008).

The Effect of Integration


Scholarship on the effect of domain-spanning
is divided. This is likely because innovation is
a risky endeavor that is challenging but potentially beneficial. We tailor these ideas to the
context of academe (where domain-spanning
is positively valued) and the case of sociology
(where robust subfields are clearly distincta
majority of articles fall in a single subfield
but spanning is common enough to allow
exploration) (Clemens et al. 1995; Moody
2004). As elaborated below, we expect that
although challenges may be confronted in the
short term, benefits of subfield spanning likely
emerge in the long term.

Challenges
Scholarship on academia has documented a
number of challenges associated with domainspanning research (Cummings and Kiesler
2005). On the producer side, it is difficult for
scholars to accommodate the research mores
and concepts of multiple specialty areas
(Lamont et al. 2006), to master and adequately
represent literature from distinct subfields, and
to produce output that is standard in form and
content (Bauer 1990). On the audience side,
experts in different subfields may disagree on
the merits of an article (Lamont 2009), and this
conflict may be particularly evident during the
peer-review process (Mansilla 2006).
Birnbaum (1981) found that research that does
not fit neatly within the substantive bounds of
normal science is typically received by journal editors and reviewers with irritation, confusion, and misunderstanding. This makes the
road to journal publication challenging (Ritzer
1998). Former sociology journal editors reinforced this point. Sociological Forum editor
(19931995) Stephen Cole (1993:337)
believed that if an author writes an article on a
relatively narrow subject . . . the chances of the
article being accepted are significantly greater
than if the author is more ambitious. American
Sociological Review (ASR) editor (19781980)

Rita Simon (1994:34) noted that works of


imagination, innovation, and iconoclasm may
fail to receive positive appraisals from reviewers who are good at catching errors and omissions but might miss a gem, or at least the
unusual, the provocative, the outside the mainstream submission.
These challenges are likely tempered, or
partly off-set, in academic sociology, the discipline we study here. First, academe (especially
of late) operates under the general presumption
that domain spanning is beneficial to science
and perhaps scientists as well (Jacobs and
Frickel 2009). This is evident from the recent
enthusiasm for interdisciplinarity (National
Academies of Science 2005), synthesis (Parker
and Hackett 2012), as well as cross-cutting
funding initiatives like National Science
Foundations (NSF) Creative Research Awards
for Transformative Interdisciplinary Ventures
(CREATIV).3 Second, the rise of collaboration
in academic science, including sociology
(Hunter and Leahey 2008; Moody 2004), combined with a division of labor based largely on
subfield expertise (Leahey and Reikowsky
2008), suggests that integrative sociology articles may be becoming easier to produce. Third,
unlike behavioral sciences like psychology
(which has a limited number of distinct and
highly autonomous subfields), sociology has a
plethora of subfields but comparatively weak
boundaries dividing them (Moody 2004).
To the extent that challenges remain under
these more welcoming circumstances, integrative research entails risk. Given the difficulties that integrative work experiences in
the review process, it likely gets rejected
repeatedly and ends up in a low-tier journal.
Authors, ever aware of the advantages of toptier publication for promotion and scholarly
impact, typically submit an article to the most
prestigious journal they think it can be published in and turn to less prestigious journals
if it is rejected (Calcagno et al. 2012). But
there is also a (perhaps small) chance that
integrative work will strike big and get
accepted in a top journal. In sociology, the
three top journals are generalist in nature:
They publish articles on any sociological
topic and strive to publish articles of interest

Downloaded from scu.sagepub.com by guest on June 23, 2015

233

Leahey and Moody


to many sociologists. One top journal explicitly aims to publish integrative work: ASR
asks reviewers to assess whether a submitted
manuscript speaks to two or more subfields,
and if not, how it could be revised to do so
effectively.4 The top journals also aim to publish
the very best, most innovative research. Thus,
we expect integration to have a curvilinear,
U-shaped effect on journal prestige. Integration
will generally have a negative effect on journal prestige, except for the most innovative
integrative articles that may be more likely to
appear in top journals.5
Of course, the biggest penalty an article
can incur is relegation to the notorious file
drawer: It never gets accepted for publication.
Like almost all research on science, we rely
on archival data that is subject to selection
bias (Fleming et al. 2007:464): Unpublished
articles are not observed and may be more
integrative than published articles. We thus
cannot know whether some subset of integrative articles is so difficult to frame successfully, and so challenging to review, that its
authors give up trying to publish it (or never
try at all). We alleviate this concern somewhat by our sampling technique, described
below, which captures elite as well as low-tier
journals; much research only investigates the
former (Clemens et al. 1995; Karides et al.
2001). We also perform sensitivity analyses
to mimic the effect that selection bias might
have on our results so we can gauge the magnitude and direction of bias.

Benefits
Research on creativity and innovation demonstrates that drawing on ideas from diverse
domains is advantageous. Actors and organizations that span domains are exposed to diverse,
unrelated ideas that can be recombined in new
ways (Carnabuci and Bruggerman 2009). Such
new combinations produce good ideas (Burt
2004), higher quality output (Singh 2008), and
serve as a foundation for innovation (Hargadon
2002; Schumpeter 1939; Weick 1979;
Weitzman 1998). As Uzzi and Spiro (2005:447)
summarize, We know that creativity is spurred
when diverse ideas are united or when creative

material in one domain inspires or forces fresh


thinking in another.
When applied to academic science, the
domains of interest are typically disciplines. A
recent study conducted by the National
Academies of Science (2005) argues that
breakthroughs in research will increasingly
occur at the interstices of disciplines, where
substantive areas of inquiry are blended.
According to Burt (2004), scientists are stimulated most by people outside their own discipline; the shock of the interface is what is
interesting and sparks creativity. Bringing
together two new things and seeing their correspondencewhat Knorr Cetina (1980)
refers to as making conceptual metaphors
promotes the extension of ideas and knowledge (Weisberg 2006). Fleming et al. (2007)
find that working with diverse scientists and
having diverse work experiences facilitates the
generation of new ideas.
The benefits of domain spanning should
also extend to subfields. Within academe, there
exists a tight correspondence between fields
and subfields. Processes that occur both
between and within disciplines are critical to
adequately capturing the the substantive heart
of the academic system (Abbott 2001:148),
though subfields better reflect the realm of
research (Chubin 1976). Like fields, subfields
produce the same insights (albeit through different routes), emerge via similar processes,
provide a structure to the academic labor market (where advertisements for positions are
distinguished by both field and subfields), and
help academics by preventing knowledge from
becoming too abstract and overwhelming
(Abbott 2001). Because subfields of a single
discipline are not as diverse as a set of disciplines, the benefits of subfield spanning may
not be as great as the benefits of field spanning. However, the sheer number and variety
of subfields, especially in a discipline as
diverse and loosely coupled as sociology, suggest that many opportunities for recombinant
innovation exist. By linking areas of research,
we gain tentative bridges between local
knowledges and a highly creative product
(Abbott 2001). Thus, scholarship that pools
ideas from diverse subfields should generate

Downloaded from scu.sagepub.com by guest on June 23, 2015

234

Social Currents 1(3)

new and valued ideas. Arguably, the greatest


benefit in reputational work organizations
(Whitley 2000) like academe is having ones
research recognized and highly regarded by
others. While this may be difficult for integrative, unconventional articles to achieve in the
review stage, it may be more likely once they
have been legitimated by publication in a peerreviewed journal. Thus, in the academic context we study here, we expect integrative
research to be more highly valued than non(or less novel) integrative research, at least
after publication. Even if challenges emerge in
the review stage, integrative research may
begin to accrue benefits after it is published.
One intuitive mechanism driving such benefits has been uncovered by previous research; in
this article, we propose and test another. Extant
research suggests that one route to scholarly
influence is to widen ones prospective audience by appealing to multiple communities. For
example, Lamont (1987) found that Derridas
ability to speak to several publics, to make his
scholarship adaptable, and to publish in various
outlets contributed to his influence both within
and outside of France. Scholars able to master
multiple genres are more likely to gain entry to
multiple conversations (Clemens et al. 1995);
indeed, the number of different journals a scientist publishes in is the largest predictor of subsequent H-index (Acuna, Allesina, and Kording
2012). Here, we test an additional mechanism:
Compared with non- (or less novel) integrative
research, integrative research appeals to scholars because of its novelty. We are able to access
this black box by first controlling for subfield
productivity (a proxy for audience size) and
then assessing the relative impact of our two
measures of integration: If it is not merely the
spanning of subfields (nominal integration) but
also the rarity of the combination (novel integration) that makes research articles noteworthy, then we can conclude that our proposed
mechanism is operating.
Scrutinizing the nature of the audience
helps reconcile our two hypotheses. Those
who decide whether an article gets published
in a given journal (editors and reviewers) and
those who cite published work (researchers)
are members of the same scholarly community

who fulfill multiple roles. Indeed, the manuscript reviewer in stage 1 (who may find integrative work challenging to review) may also
be the scholar in stage 2 (who finds integrative
work valuable and worth citing). How can we
argue that they may penalize integrative work
at one stage and value it at the next? The
answer relies on different role expectations.
Peer reviewers are solicited for their expertise,
so in this capacity, scholars provide a critical
evaluation of an article, emphasizing revisable
quality over inherent quality (Ellison 2002).
Outside their reviewer role, researchers choose
the scholarship they read, and they read to
engage with it and to glean its relevance to
their own research.
The overarching goals of this article are to
develop and measure the concept of subfield
integration, and to assess whether the challenges of integrative research are manifested in
the review stage (i.e., by appearing in lowprestige journals) and whether benefits accrue
after publication (i.e., by being cited heavily
by peers). If we find that all but the most novel
integrative articles tend to be published in lowtier journals, then we will feel confident that
the challenges of domain spanning extend to
academic sociology. If we find that not only
nominal integration but also novel integration
(articles that span rarely spanned, cognitively
distant subfields) is valued more highly by the
scientific community, then we will feel confident that we have empirically tapped a process
that is critical to scientific innovation.

Data and Method


Sample
To investigate the potential benefits of integrating subfields, we examine journal articles
written by a 20 percent probability sample of
tenured and tenure-track faculty members
located in sociology departments at Extensive
Research Universities6 in spring 2004. Details
are provided in Appendix A. The sampling
frame was constructed using faculty lists on
department Web sites, which were more upto-date than the ASA Guide to Graduate
Departments. We limit scholars articles to

Downloaded from scu.sagepub.com by guest on June 23, 2015

235

Leahey and Moody


those published on or after the first occurrence
of integrative research in Sociological
Abstracts (SA) in 1985 and to those published
on or before 2004, to give recent articles some
time to accumulate citations. This results in a
sample of 1,785 articles by 180 sociologists,
housed in 99 universities. Our statistical
approach capitalizes on this multistage sampling procedure, but because most scholars
produce both integrative and non- (or less
novel) integrative work,7 the research article
serves as the unit of observation.
This sampling strategy is appropriate for
the task at hand. The vast majority of peerreviewed research takes place at research universities (Levin and Stephan 1989). By
selecting only Extensive Research Universities,
we control for resource-based influences on
the outcomes of interest, such as productivity
expectations, and time and money for research
(Fox 1992). By selecting individuals and their
articles, we gain enough articles per person to
evaluate person- and department-level effects
and to account for unmeasured characteristics
at these levels.
Like many previous studies of inequality in
science (Keith et al. 2002; Long 1992; Reskin
1977), we study a single discipline. We do this
because disciplines differ in terms of their
stratification processes (Fox and Stephan
2001; Levin and Stephan 1989; Prpic 2002),
their degree of receptivity to boundary-crossing research, and the degree to which their
work is cited in other fields (Pierce 1999). We
study sociology because the social sciences
have generally been neglected by the sociology of science and knowledge (Guetzkow et
al. 2004). Moreover, sociology sits at the
crossroads of several different disciplines; it is
embedded in multiple fields, potentially
speaking to many audiences (Clemens et al.
1995; Moody and Light 2006) making an
investigation of integrative research in this discipline particularly informative. The permeability of sociologys subfields (Moody 2004)
provides a conservative test of the effects of
domain spanning: If we find effects in sociology, they are likely more pronounced in fields
like psychology and economics where subfields are more insular.

This study, like most studies of academic


rewards (Allison and Long 1987; Ferber and
Loeb 1973; Fox and Faver 1985; Moody 2004;
Reskin 1978; Ward and Grant 1995; Xie and
Shauman 1998), is based only on journal articles and does not include other forms of scholarly publication, most notably books. This is a
limitation imposed by our reliance on SA to
obtain keywords (discussed below), which we
use to measure integration. However, previous
studies have found that article productivity
correlates strongly with total productivity that
includes books, book reviews, and contributions to edited volumes (Clemens et al. 1995;
Leahey 2007; Reskin 1977, 1978). Articles and
books differ largely in terms of their method
and evidence, not substantive subfields
(Clemens et al. 1995), which serve as the basis
of the integration measure we use here.

Data Sources
Most of the data needed to construct our key
explanatory variable, subfield integration, and
other article-level variables (all described in
the next section) were obtained from the electronic database Sociological Abstracts (SA).
By entering sampled sociologists names, we
accessed all of their refereed journal articles.8
For each article, we collected classification
codes (keyword descriptors indicating disciplinary subfieldssee the entire list in
Appendix B), which are assigned by staff at
Cambridge Scientific Abstracts (CSAs), the
umbrella organization that manages the database.9 SA applies at least one and sometimes
two classification codes to each article. While
other aspects of articles (e.g., abstracts, text,
bibliography) also indicate its content, keywords give a good indication of each articles
substantive topic, map easily onto ASA sections that demarcate substantive areas of study,
and help structure professional identity within
the field. Because there is a fixed set of classification codes assigned by information science
experts at SA, they are easier to work with analytically than an open-ended list generated by
authors or a Web of Science algorithm (e.g.,
there is no need to construct a thesaurus for
similar terms). Classification codes have been

Downloaded from scu.sagepub.com by guest on June 23, 2015

236

Social Currents 1(3)

used extensively in previous work on sociological subfields (Leahey 2006, 2007; Moody
2004) and serve as the basis for our key
explanatory variable (subfield integration) as
well as two control variables: subfield prestige
and subfield productivity. Later, we show that
our results are robust to the use of an alternate
classification scheme.
We rely on the Thomson Reuters Web of
Science to construct two outcome variables. To
assess whether integrative research ends up in
lower-tier journals, we collected data on journal prestige from Journal Citation Reports, a
comprehensive and unique resource tool that
allows you to evaluate and compare journals
using citation data.10 To assess whether integrative research is more highly valued, we collected data on each articles citation counts as
of 2010, from the Web of Science.
Additional data about scholars, their respective departments, and their subfields were
obtained from department Web pages, professional association directories, and curriculum
vitae (CVs), which provide data comparable to
other sources of career histories (Dietz et al.
2000; Heinsler and Rosenfeld 1987). We
obtained the prestige rating of each scholars
department from the National Research
Council (NRC) (Goldberger, Maher, and
Flattau 1995).

Measures
Key explanatory variables.We use two measures of the key explanatory variable: subfield
integration. Both measures are based on extant
keywords from SA; they are not based on the
authors own classifications of their articles.
The less precise measure is binary and captures what we call nominal integration. It
indicates whether the article was assigned keywords that come from more than one keyword
family. A keyword family comprises a parent code and several child codes. In SAs classification system (Appendix B), there are 29
parent codes containing 94 child codes. For
example, the parent code complex organization contains several child codes, including
bureaucratic structure and jobs and work
organization. If an article was assigned one

code belonging to the social control family


and one code from the complex organization
family, then it would be considered integrative. If an article received two codes from the
same family, then it would not be considered
integrative. Just over one-quarter (470/1785)
of the sociologists articles were classified as
integrative according to this binary measure. A
validity check suggests that the measure we
use is quite good at distinguishing integrative
from nonintegrative articles.11
The more refined measure of integration is
continuous and captures what we call novel
integration. The measure captures not only
whether two subfields are combined but also
the rarity (and thus novelty) of the combination. Like similar measures used by others
(Schilling and Green 2011; Uzzi et al. 2013),
we control for chance occurrence based on the
size of each subfield by using the expected
number of integrative articles. For each nonzero cell, we calculate novel integration as 1
[observed/expected], and where the expected
number of cross-subfield publications is calculated in the standard manner: Eij = pi pj T,
where pi is the proportion of articles in subfield
i, and T is the total number of articles published.12 Thus, as the number of observed
cross-subfield articles approaches the number
expected by chance, the value approaches
zero, and the article is thus not very integrative. When we see more cross-subfield publications than expected by chance, the article has
a low score on the novel integration scale.
When we see fewer cross-subfield publications than expected by chance, the article has a
high score on the novel integration scale.13
Uniquely, our calculation of novel integration
is based on a time-sensitive co-classification
matrix, akin to the one presented in Appendix
Cwhich shows frequencies for all combinations during the entire time period under study
(19852004)but restricted to the five-year
window prior to each articles publication date.
Assuming that any cross-subfield publication
is more integrative than a publication that falls
only within one category, we hard-code all
single-category (i.e., nonintegrative) publications to just below the observed minimum
(7).14 Finally, for ease of interpretation, we

Downloaded from scu.sagepub.com by guest on June 23, 2015

237

Leahey and Moody


rescale the score, using percentiles, so that it
ranges from 1 to 100.
To illustrate, consider articles with two contrasting combinations, both published in 1995:
One is associated with the codes for Gender
(29) and Organizations (06), the other is associated with the codes for Political Sociology
(09) and Family (19). (Appendix C provides
the prevalence of each combination among all
integrative articles published between 1985
and 2004not just the articles in our sample.)
When we restrict it to the five years prior to
publication (19901995), we find that 267 articles are classified as both Gender and
Organizations. While these are both large subfields, with 2,119 Gender articles and 1,431
Organizations articles (out of a total of 24,521
articles), we would only expect to find 123.6
integrative articles [(2,119 1,431) / 24,521],
so the novel integration score is 1(267 /
123.6)=1.16 (84th percentile), indicating that
this is a fairly common (and not so novel) combination. On the contrary, we find 33 articles
classified as both Political Sociology and
Family (containing 1,284 and 2,200 articles,
respectively), we would expect to find 115.2
articles by chance [(1,284 2,200) / 24,521],
and so the 33 observed articles in the 1990
1995 window results in a novel integration
score of 1(33 / 115.2) = 0.71 (which would
put it in the 100th percentile).
With data covering a span of almost 20 years,
and an interest in distinguishing short- and longterm effects, it is important to account for time.
We do this in three ways. First, we construct and
use a time-sensitive co-classification matrix
(i.e., a five-year window prior to each articles
publication date) to measure novel integration,
as described above. Second, we control for year
of publication in the analyses. Third, we control
for the trend in each combinations popularity.
Two articles published in the same year could
have the same novel integration score, but their
respective combinations could be on very different trajectories. For example, the combination of poverty and culture was on the wane
19852004 (suggesting perhaps a dead end),
but the combination of poverty and urban sociology was on the rise (indicating disciplinary
excitement). To ensure that slopes are scaled

similarly across categories, we first standardize


the data to have a global mean of 0 and standard
deviation of 1, and then calculate slopes for
each category combination. We use a periodspecific slope, using two periods (before or after
1995), as a way to capture the general trend relevant to each articles publication date.15
Outcome variables.To assess potential challenges of integrative work, we examine journal prestige. This is captured by the Web of
Sciences journal impact factor (JIF), calculated as the number of citations to recent
(within the past few years) articles in the journal divided by the number of articles recently
published, and thus is the average number of
citations per article. We assign a value of
0.005 to journals that are not indexed by the
Web of Science (which indexes 1,000 social
science journals) and thus omitted from the
Journal Citation Reports.16 Impact factors are
available only from 1997 onward; because
they shift somewhat from year to year, we use
the average of the factors from 1997, 2000,
and 2003. The JIFs correlate very highly with
sociologists general perceptions, and also
with Allens (1990) assessment of journal
influence,17 which we cannot use because it is
only available for a small fraction of the journals represented in our sample.
We use a cumulative citation count to test
our second hypothesis: that integrative research
is viewed as more innovative and thus will be
more noteworthy to other scholars. Specifically,
the cumulative number of citations that each
article has received as of fall 2010 captures the
extent to which the work was useful toand
valued byother scholars.18 Although selfcitations are not eliminated from this count,
Clemens et al. (1995:455) found these to be
rare, and there are few differences in motivation for citation to self and to others (Bonzi and
Snyder 1991). Articles that have never been
cited are given a value of 0.005, and the entire
variable is then log transformed.19
Both outcome measures rely on citations,
whichdespite various criticismsare generally accepted as an indicator of an articles
impact. Certainly, citation counts may also be
reflecting the authors visibility, disciplinary

Downloaded from scu.sagepub.com by guest on June 23, 2015

238

Social Currents 1(3)

alliances, attempts to flatter potential reviewers,


or even the controversial nature of an article
(Baldi 1998; Ferber 1986; Latour 1987).
Authors may cite previous work in a casual
way, or rely on it heavily. They may think of it
highly, or dismiss it as flawed. Despite this variation in what a citation signals (van Dalen and
Henkens 2005), citation counts reflect the usefulness of an article because it contributed, in some
way, to a subsequent work. This is largely accepted
by the academic community, which continues to
rely on citation counts when making decisions
about merit raises, promotion, and tenure
(Diamond 1986; Sauer 1988; Leahey et al. 2010).
Control variables. In addition to the trend in each
combinations popularity, other characteristics
of subfields likely influence the impact of articles written in those areas. For example, even a
nonintegrative article may garner a fair number
of citations if the single subfield it addresses is
highly productive. There are also sharp differences in the frequency with which different subfields are cited (Clemens et al. 1995:472). For
these reasons, we control for the productivity
and prestige of the subfield(s) corresponding to
each article. Subfield productivity is captured by
counting the number of articles (in the population of articles appearing in SA 19852004) that
were assigned each possible keyword descriptor, which also helps alleviate concerns about
fluctuations in small-topic areas. For articles
that spanned more than one subfield (i.e., had
multiple keyword descriptors), we summed the
counts from the respective subfields. A rough
proxy for subfield prestige is the number of
times each subfield (keyword descriptor)
appeared in (or was applied to an article appearing in) American Journal of Sociology (AJS),
American Sociological Review (ASR), or Social
Forces (SF) in the year 2003.
We also control for other variables that may
confound the relationships of interest. Because
coauthored articles tend to be cited more
(Wuchty, Jones, and Uzzi 2007), we include a
control variable for sole-authored articles. We
also control for whether the article can be considered theoretical, methodological, or both
(i.e., receiving both the theory and the method
keyword). Because disciplinary lore suggests

that qualitative work might be more difficult to


publish in top-tier journals, and that quantitative work may squeeze out creativity and perhaps integrative impulses (Blalock 1991; Ritzer
1998), we control for whether the article uses
quantitative methods. Instead of relying, as
Moody (2004) did, on the number of tables presented in the article as a proxy for quantitative
methods, we use a more conservative measure,
which only classifies articles as quantitative if
they have a quantitative methods keyword
assigned to it by SA. We control for year of
publication because more recently published
articles are disadvantaged in terms of exposure
(i.e., opportunity to be cited), and also because
almost all research is cited less frequently over
time (Dogan and Pahre 1990). Last, when the
outcome of interest is an articles citation count,
we also control for JIF.
We also account for several author-level
attributes and one department-level characteristic. Because newcomers to a field are more
likely to instigate paradigm shift (Collins
1968:136) and more established professors are
most likely to seek involvement outside their
home field (Klein 1996; Pierce 1999), we control for authors professional age (years since
PhD) and its square. We also control for
authors productivity and visibility in the field,
as these factors influence access to top journals
as well as citations. An authors productivity is
captured by a cumulative, time-varying count
of published journal articles. An authors visibility is captured by a binary variable indicating
whether a scholar has published in ASR or AJS
to date. We also control for each scholars
department prestige using 1995 ratings from
the NRC; we convert the scores to deciles
(ranging from 1 to 10, where higher values
indicate greater prestige) and use regressionbased imputation for unrated departments (predictors include average journal impact, average
citations, percent of faculty publishing in top
journals, and percent of faculty publishing in
JCR-rated journals).

Statistical Approach
Given the structure of our data, we use multilevel modeling to derive estimates. The data

Downloaded from scu.sagepub.com by guest on June 23, 2015

239

Leahey and Moody


are inherently clusteredarticles within scholars, and scholars within departmentsand
thus suitable for multilevel (i.e., hierarchical)
modeling (Raudenbush and Bryk 2002).
Specifically, the level 1 unit is the article, the
level 2 unit is the scholar, and the level 3 unit
is the department. This modeling strategy
accounts for clustering by incorporating both
fixed and random effects.20 Our hypotheses
concern main effects, and we explore two
same-level interactions, but we do not specify
cross-level interactions. The models we estimate (using PROC MIXED in SAS 9.3) can be
expressed as a combination of submodels, one
representing each level:
Level 1 ( articles ) : ln (Yijk ) = 0 jk + jk ARTIC + eijk

Level 2 ( authors ) : 0 jk = 0 k + k AUTH + rjk


Level 3 ( departments ) : 0 k = 0 + DEPT

where there are i articles belonging to j


authors in k departments. The intercept from
the article-level equation (0jk) is not specified
as a fixed effect. Rather, it is allowed to vary
across authors (a random effect), and that variation is then explained using author-level characteristics, such as gender or productivity, in
the author-level (level 2) equation. Similarly,
the intercept from the author-level equation
(0k) is allowed to vary across departments (a
random effect), and that variation is then
explained using department-level characteristics, such as prestige. To estimate the model,
we combine these three equations and rearrange terms to distinguish the fixed and random components:
ln (Yijk ) = 0

+ jk ARTIC + k AUTH + DEPT

= intercept +

all fixed effects

+ eijk + rjk
+ random effects.

The combined equation looks very similar


to single-level regression model, with the
exception of the last term, which is a random
effect associated with authors. This can be
thought of as an error term, or unexplained
variance, associated with the author level. A
random effect at the department level was not
specified.

Results
Among the 1,785 articles of interest, there is
wide variation in both outcomes of interest:
JIF and the number of cumulative citations as
of 2010 (see Table 1). The JIF ranges from 0
(for articles not indexed in the Web of Science)
to 23.87 (the highest impact journal is
Science), and the citation count ranges from 0
to 340. Three-quarters of the articles appear in
a journal that is rated by the Web of Sciences
JCR, and just over a quarter are integrative
according to our binary measure. Authors of
these articles are predominantly men (36 percent are women), in early- to mid-career stage,
with nine published articles; over one-third
have experience publishing in the disciplines
top journals (ASR and AJS). Authors are disproportionately employed at prestigious universities. Within this elite sample, however,
lies some significant variation between integrative and nonintegrative articles (results not
shown). First, integrative pieces are written by
authors who are professionally younger than
authors of nonintegrative work. Second,
integrative articles are more likely to be published in large and prestigious subfields.
Importantly, this may buffer them from the
penalties typically associated with domainspanning work.
Contrary to expectations, subfield integration does not have a curvilinear U-shaped
effect on journal prestige. In fact, novel integration has no significant effect on JIF (see
Table 2), indicating that integrative research is
no less likely to appear in prestigious journals
than nonintegrative research. This suggests
that, despite theoretical reasons to expect difficulty publishing integrative work, there
appears to be no publication-prestige penalty.21 The trend in the popularity of the underlying category or combination, however,
negatively affects impact factor, suggesting
that as topics become more popular, articles
on such topics appear less frequently in top
journals. We also find that sole-authored articles and theory articles tend to be published in
lower-tier journals. Few other variables are
significantly associated with JIF. Exceptions

Downloaded from scu.sagepub.com by guest on June 23, 2015

240

Social Currents 1(3)

Table 1. Descriptive Statistics of 1,785 Articles Published by 180 Sociologists in 99 Departments.


Mean
Outcomes variables
Journal impact factor (mean of 1997,
0.86
2000, and 2003 scores)
Cumulative citation count as of 2010
19.34
Article attributes
Novel integration score (percentiles)
24.69
Nominal integration score (binary
0.26
measure)
Year of publication
1995
Trend
0.006
Sole authored
0.23
Theory
0.07
Method
0.06
Theory and method
0.002
Quantitative methods topic
0.03
Subfield(s) productivity
13,055
Subfield(s) prestige
0.68
Author attributesfor last year in data-file
Professional age (2004-PhD year)
8.91
Productivity (number of articles
8.79
published)
Ever published in ASR or AJS
0.37
(yes = 1, no = 0)
Gender (female = 1, male = 0)
0.36
Department attributes
NRC rating of department prestige
5.85

SD

Percent 0s

Minimum

Maximum

1.02

23.8

23.87

30.68

18.3

340

39.7

73.7a
73.7

1
0

100
1

5.36
0.09
0.42
0.26
0.25
0.05

5,399

77.1
92
93
99.7
97.5

31.7

1985
0.15
0
0
0
0
0
1,149
0

9.65
9.75

9
1

39
69

63

64

2.66

12c

10

2004
1.26
1
1
1
1
1
21,293
1

This is the percent of observations that are nonintegrative, so have a score of 1.


This is the percent of observations with the imputed score of 7, equivalent to percentile score of 1.
This is the percent of observations whose department prestige score was predicted based on their average journal
impact, citations, percent of faculty publishing in top journals and in JCR-rated journals. This predicted value was
imputed.
ASR = American Sociological Review; AJS = American Journal of Sociology; NRC = National Research Council
b
c

include department prestige and author experience publishing in top journals like ASR and
AJS, demonstrating some effective path
dependency. These results are robust to the
substitution of the binary measure of nominal
integration: The sign and significance of the
coefficient remain the same, and model fit is
comparable.
Consistent with expectations, we find that
integrative research garners a greater number
of citations than non- (and less novel) integrative work. The models presented in Table 3
suggest that even after controlling for

characteristics of authors, departments, and


subfields, articles that are more integrative
receive a greater number of citations. In the
model with all controls (model 3c), the coefficient for novel integration is positive and
significant (+0.005**) and, considering the
scale of the novel integration measure (1 to
100), represents a reasonably sized effect. For
a 10 percent increase in novel integration, an
articles citations increase by 5 percent, and
for a 50 percent increase in novel integration,
an articles citations increase by 25 percent.
Controlling for year of publication and the

Downloaded from scu.sagepub.com by guest on June 23, 2015

241

Leahey and Moody


Table 2. Assessing Risks: Effects on Journal Impact Factor (Multilevel Model Estimates).
Model 2a

Fixed effects
Article attributes
Novel integration scorea,b a,b
Year of publication
Trend
Sole authored
Theory
Method
Theory and method
Quantitative methods topic
Subfield(s) productivity
Subfield(s) prestige
Author attributes
Professional age (years since PhD)b
Professional age2
Productivity (number of articles)b
Ever published in ASR or AJSb
Gender (female = 1, male = 0)
Department attributes
Department prestige (NRC rating)
Intercept
Random effects
Variance of level 1 random effect
Variance of level 2 random effect
Sample size
2 Res log pseudo-likelihood
BIC

Model 2b

Coefficient

SE

0.0008
0.009
0.56*
0.18**
0.21*
0.08
0.02
0.12
0.008****
0.07

0.0006
0.0004
0.26
0.06
0.09
0.12
0.45
0.19
0.005
0.05

19.0*

9.17

0.83***
0.02
0.27***
0.04
1,785
5,019.4
5,029.7

Coefficient

Model 2c
SE

Coefficient

0.0008
0.007
0.47****
0.17**
0.26**
0.06
0.03
0.13
0.0009*
0.06

0.0006
0.005
0.26
0.06
0.09
0.13
0.45
0.18
0.0004
0.05

0.02*
0.0005*
0.009*
0.72***
0.03

0.008 0.02*
0.0002 0.0004****
0.004 0.009*
0.08
0.62***
0.09
0.05

14.36

11.22

0.82***
0.03
0.18***
0.04
1,785
4,977.7
4,988.1

0.0009
0.006
0.47****
0.21***
0.28**
0.05
0.001
0.13
0.0009*
0.06

0.07***
13.09

SE

0.0006
0.005
0.26
0.06
0.09
0.12
0.44
0.18
0.0005
0.05
0.008
0.0002
0.004
0.08
0.09
0.01
10.77

0.82***
0.03
0.14***
0.03
1,785
4,965.8
4,976.2

Note. ASR = American Sociological Review; AJS = American Journal of Sociology; BIC = Bayesian Information Criterion ;
NRC = National Research Council.
a
One-tailed tests (two-tailed otherwise).
b
Time-varying variable.
*p < 0.05. **p < 0.01. ***p < 0.001. ****p 0.10.

prestige and productivity of subfields, we find


that articles on topics that are increasing in
popularity actually receive fewer citations
(the coefficient for the trend variable is 2.22
and statistically significant). These effects
hold even when we control for journal prestige, arguably the largest determinant of an
articles citation count. To provide a sense of
the size of the effects in model 3c, we provide
predicted value plots in Figure 3 that stratify

integration (x axis) by selected other variables


(different lines spanning data range). For ease
of comparison, we have equated the y axis
range across all panels.
The size of integrations effect is even
more apparent when we substitute in the
binary measure of nominal integration. In
model 3d, the coefficient for nominal integration (+0.41**) demonstrates that integrative
articles receive almost 41 percent more

Downloaded from scu.sagepub.com by guest on June 23, 2015

242

Downloaded from scu.sagepub.com by guest on June 23, 2015

Note. BIC = Bayesian Information Criterion


One-tailed tests (two-tailed otherwise).
b
Time-varying variable.
c
Not significant, removed.
*p < 0.05. **p < 0.01. ***p < 0.001. ****p 0.10.

Fixed effects
Article attributes
Novel integration score (percentiles)a,b
Nominal integration scorea
Year of publication
Trend
Sole authored
Theory
Method
Theory and method
Quantitative methods topic
Subfield(s) productivity
Subfield(s) prestige
Journal prestige (impact factor)
Author attributes
Professional age (years since PhD)b
Professional age2 (years since PhD)b,c
Productivity (# articles published)b
Ever published in ASR or AJS (yes = 1, no = 0)b
Gender (female = 1, male = 0)
Department Attributes
Department prestige (NRC rating)
Interaction terms
Novel integration score Journal prestige
Novel integration score Trend
Intercept
Random effects
Variance of level 1 random effect (residual)
Variance of level 2 random effect
Sample size
2 Res log pseudo-likelihood
BIC

25.4

133.8***
6.601***
0.23
1.3***
0.24
1,785
8,640.4
8,650.7

0.002

0.01
0.74
0.18
0.27
0.35
1.27
0.53
0.00001
0.16
0.06

SE

0.005**

0.06***
2.29**
0.18
0.96
0.16
2.18****
0.04
0.00004**
0.12
1.28***

Coefficient

Model 3a

1,785
8,644.5
8,654.9

6.62***
1.17***

109.1***

0.04**

0.01
0.49*
0.09

0.005**

0.05***
2.21**
0.15
0.99***
0.12
2.09****
0.12
0.0003**
0.11
1.25***

Coefficient

Model 3b

0.23
0.23

30.4

0.01

0.01
0.22
0.26

0.002

0.02
0.74
0.17
0.27
0.35
0.27
0.53
0.00001
0.16
0.07

SE

0.03

0.01

0.01
0.22
0.26

0.002

0.02
0.74
0.18
0.27
0.35
1.27
0.43
0.00001
0.16
0.07

6.601***
0.23
1.3***
0.24
1,785
8,647.3
8,657.7

108.6***

0.02

0.04**

0.02
0.47*
0.19

0.005**

0.05***
2.22**
0.12
1.01***
0.12
2.11****
0.17
0.00004*
.11
1.24***

SE

30.37

Model 3c
Coefficient

Table 3. Assessing Benefits: Effects on Logged Cumulative Citations (Multilevel Model Estimates).

30.36

0.03

0.01

0.01
0.23
0.26

0.16
0.02
0.75
0.17
0.27
0.35
1.26
0.53
0.00001
0.16
0.07

SE

6.62***
0.23
1.15***
0.23
1,785
8,638.5
8,648.9

108.37***

0.02

0.04**

0.02
0.47*
0.19

0.41**
0.05***
2.27**
0.12
1.00***
0.12
2.08****
0.13
0.00004*
0.11
1.24***

Coefficient

Model 3d

0.04

0.01

0.010
0.22
0.25

0.002

0.02
1.38
0.18
0.27
0.35
1.26
0.53
0.00001
0.16
0.07

SE

0.23
0.22

0.001
0.02
30.0

1,785
8,644.4
8,650.7

6.56***
1.01***

0.005**
0.08***
114.56***

0.06

0.04**

0.012
0.36****
0.04

0.0005

0.06***
6.91***
0.12
1.08***
0.14
1.94
0.07
0.00004*
0.19
1.14***

Coefficient

Model 3e

243

Leahey and Moody

Figure 3. Relative effect of novel integration (x axis) on log citations, compared across selected
variables.
Note. Predicted values from model 3c, Table 3; all models estimated holding other variable at sample mean.

citations than nonintegrative articles.


Comparing models 3c and 3d, we note that
both measures of integration (nominal in
model 3d, and novel in model 3c) reach the
same level of statistical significance, and both
models fit the data well: The Bayesian
Information Criterion (BIC) and log likelihood values are almost indistinguishable (a

0.009 percent difference). Not only domainspanning itself but also the distance between
domains is beneficial.22
The main effect of novel integration varies
by two other variables, which we add in model
3e.23 First, there is a positive and significant
interaction between novel integration and our
trend variable (+0.08*). Because the main

Downloaded from scu.sagepub.com by guest on June 23, 2015

244

Social Currents 1(3)

Figure 4. Effect of novel integration on citations, interacted with subfield popularity trend.

Note. Predicted values from model 3e, plus interaction with trend; all other variables held at sample mean.

effect of trend is negative, this suggests that


integration is more beneficial when the combination is on the rise but is less beneficial when
the combinations popularity is declining.
Figure 4 demonstrates this strong interaction
effect across the range of trend variable values
in our data (0.16 to +1.26). This finding
reflects a general fact about innovation: It is
valuable to be an early mover in novel areas,
but the discipline quickly ignores integrative
work that spans areas that are falling out of
fashion.
Second, we find that the interaction of novel
integration and journal prestige is statistically
significant and positive (+0.005**), suggesting that integrations positive effect on citations is amplified when the article is published
in a prestigious journal (and, intuitively, integrations effect is effectively null when the
article is published in a journal with impact
factor of zero [i.e., in a journal not indexed in
the Web of Science]). Figure 5 provides the
model predicted effects (and 95% CLMs24) for
integration stratified across our observed range
of JIF. Because the intercept effect of JIF is so
large (justifiably so because the impact factor

is calculated from citations), the inset to Figure


5 provides the differences in slope when the
intercepts are equated, to make the comparison
easier to see. The generally positive returns to
integration are even stronger if one is able to
get integrative work published in prestigious
outlets.
Importantly, these findings appear to be
robust to the influence of selection effects.
First, a comparison of our sample with the
population of articles published 19852004
and indexed in SA suggests that a more inclusive sample of articles would be slightly less
integrative: with 22 percent of the population
integrative, compared with 25 percent of our
sample. Including a broader array of scholars
outside elite research universities, who publish in lower-tier journals and garner fewer
citations, would support the effects of integration we document. Second, once we control
for article and subfield characteristics, women
are no longer disadvantaged with regard to
cumulative citations. If anything, we expect
the relationships between gender and these
article and subfield characteristics to be stronger within the more general population of

Downloaded from scu.sagepub.com by guest on June 23, 2015

245

Leahey and Moody

Figure 5. Effect of novel integration on citations, interacted with journal prestige.

Note. Predicted values from model 3e, plus interaction with Journal impact factor; all other variables held at sample
mean.

female sociologists, and thus genders lack of


significance to hold beyond our elite sample.
Third, we crudely mimic the process of selection via publication: The distinction between
unpublished articles (for which we have no
data) and published articles may be akin to the
distinction between low-tier articles and hightier articles. Thus, we respecified our models,
limiting the sample to articles in journals with
a minimum JIF (we tried the median and the
25th percentile); results are robust to this sample adjustment.
These results are also independent of the
classification system that serves as the basis
for our integration measures. Our results rely on
SA classification codes (Appendix B). When
we match these codes with those provided by an
ASA task force on sociological specialty areas
(http://www2.asanet.org/footnotes/septoct05/
fn7.html) and recalculate the nominal and
novel integration scores, there is no change in
the direction or significance of the hypothesized effects (results available upon request).
Data from other sources also support these
findings. First, a list of the most highly cited
ASR articles (Jacobs 2005) reveals that since
1986, 43 percent (3/7) were integrative. This is

high, especially considering that only 26 percent of the ASR articles in our sample are integrative. Second, a number of authors of
sociology Citation Classics commented,
years later, on the integrative nature of their
articles and how it contributed to its visibility
and impact (Leahey & Cain 2013).25 To provide just one example, Wilson (Sociobiology
1975) tied together a great deal of disparate
information acquired by many specialists in
different topics and provided the first synthesis of its kind that cast a lot of already
familiar material in a new, scientifically better
form while suggesting a way to bridge biology
and the social sciences.
Results are also robust to model respecification. We find that integrations effect is not
curvilinear: When we square the novel integration score and add it to the models, it does
not reach statistical significance. This difference between our results and Uzzi and colleagues (2013) may be partly attributable to a
difference in domains (we study subfields,
they study fields) and the sheer number of
domains (we work with 29 sociological subfields, they work with 240+ fields). Once we
control for other influences, we find that

Downloaded from scu.sagepub.com by guest on June 23, 2015

246

Social Currents 1(3)

integrations effect is the same for scholars of


all ages (i.e., the interaction of novel integration with professional age is not statistically
significant). When we add year-squared to the
models, we find that it is statistically significant and negative, suggesting that times positive effect on both citations and impact factor
is not sustainedbut this effect is very small.
Results also obtain for the subsample of soleauthored articles.

Discussion
In this article, we contributed methodologically and theoretically to a sociological understanding of the process of scientific innovation.
We began by introducing a new concept, subfield integration, and arguing for its relevance
to scientific innovation, thereby expanding
previous research on disciplinary integration
(Clemens et al. 1995) and innovation in academic work (Guetzkow et al. 2004; Simonton
1991). We then developed two empirical measures of subfield integration, one of which
improves upon extant measures by incorporating the novelty of the pairing. We drew upon
two bodies of scholarship: one that documents
challenges associated with domain-spanning
and the other that theorizes benefits. We theorized a reconciliation: Challenges may be
apparent in the short term (review stage), and
benefits in the long term (postpublication
stage). Tests of these hypotheses, using data
from a probability sample of sociologists and
multilevel modeling, reveal negligible risks
(integrative articles are no less likely to be
published in low-tier journals) and tangible
rewards (higher citation rates) associated with
subfield integration.
Our lack of support for challenges is likely
attributable to multiple factors. Based on our
sensitivity analyses, we do not think it is attributable to selection. But other explanations may
obtain. One explanation is that we only studied
one possible penalty (being published in a
low-tier journal); perhaps challenges of integration manifest themselves in another way
(e.g., the work takes longer to produce, or

[unobservable in these data], the article is


reframed to fit a single subdisciplinary niche).
A second explanation rests on our incorporation of time: We expected penalties to be
incurred early on (in the review process), but
perhaps they manifest later, even after publication. For example, an integrative article may
be the prototypical flash in the pan that
receives a lot of attention initially but fails to
accrue a steady stream of followers. Extensions
of this work with time-sensitive citation data
could assess this possibility. Perhaps the third
explanation, which rests on the uniqueness of
sociology, is most likely. Scholars in a field as
diverse as sociology may be particularly
skilled at targeting their work for the intended
audience26, and editors and reviewers in this
field may even expect such integration, given
the permeable boundaries between sociological subfields that scholars commonly bridge
(Moody 2004). Recent editors of the flagship
sociology journal (ASR) have indicated that
they value subfield integration. A former editor argued that one of the principal functions
of a general sociology journal is to enhance
the cross-fertilization of otherwise disparate
specialties . . . [and to] . . . serve as a counterweight against undue fragmentation and segmentation in the discipline (Jacobs 2004),
and more recent editors ask reviewers to consider the number of specialty areas an article
contributes to when submitting their recommendation (Roscigno and Hodson 2007).
Once published, however, integrative
research garners a greater number of citations
than non- (or less novel) integrative research.
Importantly, this is not solely a function of its
broad appeal; that is, integrative research does
not receive more citations simply because it
spans productive subfields and thus attracts a
larger audience, as Lamont (1987) and
Clemens et al. (1995) found. Because we control for the level and growth of subfield productivity, our results suggest that integrative
research garners more attention because it is
viewed as more innovative than research that
fits neatly within a single specialty area. We are
confident of this interpretation for two reasons.

Downloaded from scu.sagepub.com by guest on June 23, 2015

247

Leahey and Moody


First, we account for such prospective audiences by including controls for subfield productivity and prestige in the statistical models.
Second, we use two measures of subfield integrationa binary one that captures nominal
integration (simply the spanning of any two
subfields) and a continuous one that captures
novel integration (which considers rare pairings to be more novel)and both are statistically significant and positively related to
citation counts. In fact, the model including
novel integration score fits the data just as well
as the model including nominal integration
score. Because it is not merely the mixing of
areas but also the uniqueness of that mixing
that makes an article more noteworthy to other
researchers, we are confident that it is the novelty of subfield integration that enhances its
usefulness, not merely audience size. A recent
article in Science (Uzzi et al. 2013) corroborates this finding and also suggests such novelty, when combined with convention, is most
predictive of hit article status (having citations in the top 5 percent of the entire
distribution).
Our findings support and extend recombinant innovation theory. As the theory suggests,
published articles that span subfields are more
valued by the scholarly community. But we
also extend this theory in two ways. First, we
make a methodological improvement by constructing a measure integration that considers
the novelty of the pairing, to better tap into
theoretical ideas about the importance of the
relatedness, or cognitive distance, between
subfields (Hargadon 2002; Knorr Cetina 1980;
Poincare [1908] 1952; Weitzman 1998).
Second, we find that not only the rarity of the
combination (at a given point in time) but also
its trajectory matters. Integration matters most
when the combination is increasing in popularity. Authors who aim for noteworthy, highly
cited articles should seek not only to combine
rarely combined subfields but also to choose

combinations that are just beginning to be


explored by others.27
The results of this article speak to broader
trends of specialization in science, particularly
concerns about fragmentation within sociology
(Collins 1994; Davis 1994; Gouldner 1970;
Steinmetz and Chae 2002). Just as Calhoun
(1992) sees focusing on sociology to the exclusion of other social science disciplines as
impoverishing, so might focusing on a single
subfield be detrimental in the long run.
Although research that falls within a single
subfield is no more likely than integrative
research to appear in prestigious journals, it
may in the end garner less attention from the
scholarly community. It is integrative work that
spans subfields, especially rarely spanned subfields, which is truly appreciated by subsequent
scholars. At this point, integrative research is
still rare, but perhaps recognizing its long-term
fruitfulness is a first step to help overcome the
ethnocentric insularity of mainstream sociology (Calhoun 1992).
Although we investigated sociologyan
inherently interstitial discipline that has been
neglected in the sociology of science and
innovation studies (Guetzkow et al. 2004)
extensions to other disciplines will be fruitful.
The penalties and benefits of integrative
research may vary across fields, which differ
in terms of their stratification processes (Fox
and Stephan 2001; Prpic 2002), their interdisciplinary reach (Pierce 1999), and the permeability of their subfields. Natural science
fields like physics have sharply divided subfields, such as theoretical and experimental
physics (Traweek 1992). Even other social
science fields like economics and psychology
have largely autonomous subfields that are
less open to integration. While our study cannot speak to fields that lack sociologys breadth
and tendency toward cross-fertilization, they
will hopefully serve as a springboard for such
investigations.

Downloaded from scu.sagepub.com by guest on June 23, 2015

248

Social Currents 1(3)

Appendix A
Details of Sampling Procedures
We began with the population of (n = 151)
Extensive universities, which award 50 or
more doctoral degrees per year in at least 15
disciplines. The Carnegie Foundation has classified such institutions as Research I in the
past and RU/VH: Research Universities (very
high research activity) since our data were
collected. We constructed a complete list, as of
2004, of tenured and tenure-track faculty
members housed in sociology departments at
these universities, excluding affiliated and
emeritus faculty. From this list, we selected a

20 percent systematic sample (n = 181) of faculty members. We then obtained these faculty
members publication histories (and keywords
associated with each publication) from SAs.
As noted in the text, we only include articles
published between 1985 (when the first integrative article appeared in SAs) and 2004 (to
allow even more recent articles time to accrue
citations). This results in a sample of 1,785
articles by 180 sociologists, housed in 99
universities.

Appendix B
Classification Codes Assigned by Sociological Abstracts to Journal Articles
0100

0200

0300

0400

0500

0600

methodology and research technology


0103
methodology (conceptual and
epistemological)
0104
research methods/tools
0105
statistical methods
0161
models: mathematical and other
0188
computer methods, media, and
applications
sociology: history and theory
0202
of professional interest (teaching
sociology)
0206
history and present state of sociology
0207
theories, ideas, and systems
0267
macrosociology: analysis of whole
societies
0285
comparative and historical sociology
social psychology
0309
interaction within (small) groups
0312
personality and social roles
0373
cognitive/interpretive sociologies
0394
life cycle and biography
group interactions
0410
social group identity, intergroup
relations (race, age, and sexuality)
0491
refugees
culture and social structure
0513
culture
0514
social anthropology
complex organization

1500

sociology of religion
1535
sociology of religion

1600

social control
1636
sociology of law
1653
police, penology, and correctional
problems
sociology of science
1734
sociology of science
1772
sociology of technology

1700

1800

demography and human biology


1837
demography (population studies)
1844
human biology/sociobiology
1864
genetic engineering/reproductive
biotechnology

1900

the family and socialization


1938
sociology of the child
1939
adolescence and youth
1940
sociology of sexual behavior
1941
sociology of the family, marriage, and
divorce
1976
socialization
1977
birth control (abortion,
contraception, fertility)
1978
sociology of death and dying
sociology of health and medicine
2046
social psychiatry (mental health)
2045
sociology of medicine and health care
2079
substance use/abuse and compulsive
behaviors

2000

(continued)

Downloaded from scu.sagepub.com by guest on June 23, 2015

249

Leahey and Moody

Appendix B (continued)

0700

0800

0900

1000

1100

1200

1300

1400

0621

jobs, work organization, workplaces,


and unions
0623
military sociology
0624
bureaucratic structure/organizational
sociology
social network analysis
0665
sociology of business and
0671
entrepreneurism
voluntary associations/philanthropy
0674
social change and economic development
0715
social change and economic
development
market structures and consumer
0749
behavior
capitalism/socialism - world systems
0770
mass phenomena
0826
social movements
0827
public opinion
0828
0829
0842
0850
0868
0869

2100

social problems and social welfare


2143
social gerontology
2147
sociology of crime
2148
social work and welfare services
2151
juvenile delinquency
2187
social service programs/delivery
systems
2190
victimology (rape, family violence,
and child abuse)
sociological practice (clinical and
2192
applied)

2200

sociology of knowledge
2233
sociology of knowledge

2300

communication
2400
collective behavior
sociology of leisure/tourism
popular culture
transportation systems and behaviors
sociology of sports

political sociology/interactions
0911
interactions between societies,
nations, and states
0925
sociology of political systems, politics,
and power
0989
welfare state
0995
nationalism
social differentiation
1019
social stratification/mobility
1020
sociology of occupations and
professions
1022
generations/intergenerational
relations
rural sociology and agriculture
1116
rural sociology (village, agriculture)
urban sociology
1218
urban sociology
sociology of language and the arts
1330
sociology of language/sociolinguistics
1331
sociology of art (creative and
performing)
1375
sociology of literature
sociology of education
1432
sociology of education

2500

2252
history of ideas
community/regional development
2317
sociology of communities and
regions
policy, planning, forecasting
2454
planning and forecasting
2460
social indicators
2462
policy sciences
2496
negotiation, dispute settlements
2499
sociology of ethics and ethical
decision making
radical sociology
2555
Marxist and radical sociologies
2580
critical sociology

2600

environmental interactions
2656
environmental interactions
2681
disaster studies
2682
social geography
2697
famine, hunger, and malnutrition

2700

studies in poverty
2757
2793

2800

2900

studies in poverty
homelessness
studies in violence
2858
studies in violence
2884
terrorism
2898
genocide
feminist/gender studies
2959
2983

Downloaded from scu.sagepub.com by guest on June 23, 2015

feminist studies
sociology of gender and gender relations

250

Downloaded from scu.sagepub.com by guest on June 23, 2015

0100
0200
0300
0400
0500
0600
0700
0800
0900
1000
1100
1200
1300
1400
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
2500
2600
2700
2800
2900

methods
theory
social psych
groups
culture
orgs
social change
mass society
political
stratification
rural
urban
language
education
religion
social control
science
demography
family
health/medicine
social problems
knowledge
community
policy
radical
environment
poverty
violence
gender

Methods

Theory

234
454
270
323
303
284
650
139
83
137
86
155
233
199
149
180
353
230
199
189
15
94
186
93
26
69
599

Groups

Social psych

309
125 210
227 721
31 124
242 859
65 1087
78 322
9
37
15 213
244 413
90 757
86 389
52 388
61
47
33 262
422 755
265 562
132 326
14
28
2
58
27 151
4
21
32
96
10
52
39 106
405 400

Culture

55
179
164
208
45
24
39
176
51
113
37
42
81
239
300
48
37
9
24
9
60
12
41
222

Orgs
394
308
346
277
75
41
93
235
59
130
231
126
732
514
290
16
32
186
39
76
45
25
1055

Mass society

Social change
306
497 1093
124
83
275
83
131
90
32 156
82 118
80 187
50 154
116 222
115
48
106 464
124 358
67 254
29
23
32
32
107
81
89
45
193 341
77
43
15 135
265 900

Political
120
59
125
247
111
618
237
109
64
167
79
106
38
34
236
62
158
43
121
578

Rural

Stratification
38
43
8
53
4
357
17
61
11
135
7
53
28
74
52
188
67
406
59
86
47
17
3
9
44
29
40
23
8
23
82
43
52
10
4
470 165

Urban
19
8
15
9
25
31
34
41
59
2
52
85
7
97
121
14
31

Language
154
30
66
34
22
106
88
41
27
2
21
23
11
2
7
269

Education
97
29
99
27
380
76
97
31
16
69
36
7
37
35
475

Religion
63
39
31
296
155
103
29
24
17
23
43
6
32
200

Social control
42
28
367
333
494
17
7
70
20
23
19
51
292

Social problems

Health/medicine

Family

Demography

Science
44
41 409
80 182 1468
34 122 619 792
157
18
17
20
3
12
6
24
17 30
94 151 234 412 131
15
5
10
9
9
46
71
54 111 57
1
21 164 127 127
1
18 205
94 120
210 164 888 486 361

Knowledge
0
8
67
7
0
3
80

Policy

Community
17
3
5
31 127
4 47
9
8
8 95

Radical
68
2 21
4
9
57 167

Environment

14
38 108

0100 0200 0300 0400 0500 0600 0700 0800 0900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 2900

435
197
121
171
249
63
181
101
75
25
19
51
79
67
46
80
96
210
268
240
41
11
59
24
53
40
20
209

Poverty

0100 0200 0300 0400 0500 0600 0700 0800 0900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 2900

Co-Classification of Specialty Areas (n = 57,442 Integrative Articles Indexed in Sociological Abstracts 19852004)

Violence

Appendix C

Gender

251

Leahey and Moody


Acknowledgments
The authors thank sociology audiences at Cornell,
University of Washington, Duke, and North
Carolina State University and organizational behavior groups at MIT, Stanford, and Erasmus University
for their valuable feedback. Ronald Breiger, Omar
Lizardo, Michele Lamont, Linda Molm, Mark
Pachuki, Kyle Siler, and Ezra Zuckerman kindly
provided detailed comments on earlier drafts of this
article. Ludo Waltman and Nees Van Eck at Leiden
University provided citation count data.

Declaration of Conflicting Interests


The author(s) declared no potential conflicts of
interest with respect to the research, authorship,
and/or publication of this article.

Funding
The author(s) received no financial support for the
research, authorship, and/or publication of this
article.

Notes
1. In this article, we use the terms specialty area,
subfield, and subdiscipline inter-changeably.
2. While our concept of subfield integration
advances the literature on disciplines and their
subfields and moves beyond simple dichotomies, it is also subject to limitations. First, our
concept of integration captures information
transfer across subfields but cannot distinguish among possible sources, which include
borrowing, collaboration, and publishing in
other subfields journals (Pierce 1999) as all
of these sources are consistent with our measure of integration, which relies on keyword
descriptors. Second, we cannot distinguish
between integration that is instrumental (where
the goal is to enhance the home subfield) and
that which is conceptual (where the goal is to
be critical of the home subfields assumptions
and omissions) as Salter and Hearn (1996) do.
Third, we focus on the integration of substantive, not methodological or theoretical, subfields. This aligns well with the definition of
interdisciplinarity upon which we draw and
our data source (which cannot discern whether
an article uses multiple methods or theories).
Cappell and Gutterbock (1992:271) found
that neither the Theoretical Sociology nor
the Methodological Sociology section was
central to the structure of the discipline, and

in our analyses, the inclusion of a binary variable indicating whether an article spanned the
theory or methods subfield did not reach
statistical significance or alter any results.
Substantive subfields are the most prominent
boundaries within sociology (http://www2.
asanet.org/footnotes/septoct05/fn7.html).
3. CREATIV is the Creative Research Awards
for Transformative Interdisciplinary Ventures.
4. Despite this pressure, our data do not suggest
that articles in American Sociological Review
(ASR), American Journal of Sociology (AJS),
and Social Forces (SF) are more integrative
than other articles: Their respective percentages
are 22 percent, 22 percent, and 25 percentall
very close to the overall average of 26 percent.
In fact, other specialty journals, like Gender
& Society and Work & Occupations have, in
certain years, a higher proportion of integrative
articles than these three top journals.
5. Another hypothesis that could be justified
from this discussion is that integrative articles
have a higher variance in terms of the prestige
of their outlets, compared with non- (or less
novel) integrative articles. When we aggregate
data up to the person level and test whether a
persons average integration score impacts the
variance of their journal impact factors, the
results corroborate what we find for the curvilinear relationship.
6. This Carnegie classification denotes universities that awarded 50 or more doctorates in
15 or more fields and is roughly equivalent to
Research I universities.
7. Seventy-eight percent of the sociologists in our
sample (140/180) produced both integrative
and nonintegrative articles between 1985 and
2004.
8. Sociological Abstracts (SAs) covers over
5,000 social science journals (compared with
only 1,000 in the Web of Science); the list of
serials is available at http://search.proquest.
com/socabs. Previous research documents a
strong overlap between secondary sources like
SA and scholars CVs (Heinsler and Rosenfeld
1987), and with SAs retrospective additions
and increased coverage in recent years, the
overlap is likely greater.
9. Cambridge Scientific Abstracts (CSA) staff
members are highly trained in keyword application, and the current list of keywords is upto-date and consistent across the time period we
study. SA is more comprehensive in its application of keyword descriptors than the Web of

Downloaded from scu.sagepub.com by guest on June 23, 2015

252

Social Currents 1(3)

Science, which determines keywords for only a


subset of articles and has no extant set of keywordsrelying instead on a formal algorithm
that searches titles, abstracts, and references for
key themes and on author-supplied keywords
that favor validity over reliability.
10. See http://isi17.isiknowledge.com/portal.cgi?
DestApp=JCR&Func=Frame
11. Among articles sharing one keyword, we randomly selected integrative and non-integrative
articles to compare. For example, a nonintegrative article assigned only to education was
about the teacherstudent relationship among
at-risk students, whereas integrative articles on
the topic of education were joined with methods when the topic was the adequacy of educational data resources and joined with gender
when the topic was gender differences in
parental involvement in education. Similarly,
an article on religious involvement received
only the religion code, whereas an article on
the Promise Keepers also received the social
movements code.
12. Of course, there are other ways one could capture subfield integration but many of these are
difficult in our modeling context because they
are either more subjective (requiring author
self-report) or more labor-intensive (requiring
not only data collection but also hand coding
by the researcher) than the approach used here,
or they are confounded with an important control variable (e.g., author productivity).
13. One potential trade-off from expected valuebased metrics is that small subfields may be
sensitive to small-n fluctuations. To help alleviate concerns here, we control for the size of
subfields in all models.
14. The results are robust to other values of the
hard-coded minimum, even those that are
slightly above the off-diagonal values (e.g.,
the 10th percentile and the 1st quartile of the
distribution). The raw scoring produces a small
number of large negative values; our concern
over these outliers is alleviated by rescaling
using percentile values.
15. For nonintegrative articles, we regressed the
single subfields prevalence on time. Slopes
for within-subfield change are relatively flat.
Using the slope gives us both the sign (positive
or negative) and speed of change (near or far
from zero), which may capture some element
of overall excitement or boredom with a combination. Alternative specifications that are not
period specific show comparable results.

16. Results are robust to other imputed values,


including the lowest observed value (0.017).
17. See http://www.asanet.org/footnotes/dec03/
fn11.html. In the subsample of journals represented by our sample for which both impact
factors and influence scores are available, the
correlation is 0.70.
18. We do not have access to year-specific citation
totals; future work would do well to examine
how subfield spanning affects citation trends.
19. Using a zero value for uncited articles, we also
estimated zero-inflated Poisson models; results
are robust. Results are also robust to limiting
the analysis to the subset of ever-cited articles.
20. A scholars department (i.e., institutional affiliation) was captured at the time of sampling,
in 2004. We chose to recognize the multilevel
structure of the data, transform the outcomes
variables, and run hierarchical linear models. However, results are robust to alternative
estimation procedures, such as zero-inflated
Poisson, which disregard the hierarchical
nature of the data and focus on the distribution of the outcomes (e.g., the plethora of zero
values for both impact factors and citations
counts) as well as a fixed effects model that
better controls for author quality/style.
21. Although this finding counters the general
expectation that integrative work (including
interdisciplinary work) suffers in the review
process, empirically judging a nonsignificant
result is difficult. As a reviewer notes, alternative mechanisms exist (e.g., authors may carefully reframe rejected articles to better target
single subfields) but cannot be assessed with
our data.
22. Because only 25 percent of the articles in our
sample were integrative according to the binary
measure, we do not have the same amount of
statistical power to identify novel integrations
effect among just the subset of (nominally)
integrative articles. Despite this low power, we
find a positive effect of novel integration that
is significant at the 0.10 level.
23. Note that in this model, integrations main
effect is not significant. That is, when impact
factor (among other variables) is zero, the
effect of integration is flat but increases
steadily with journal impact factor.
24. Because low-impact journals outnumber highimpact journals, the confidence limits are
wider as journal impact factor increases.
25. These are available at http://www.garfield.
library.upenn.edu/classics.html.

Downloaded from scu.sagepub.com by guest on June 23, 2015

253

Leahey and Moody


26. While this may be especially true for collaborative teams that include specialists from the
represented subfields, we found that our results
held for the subset of (n = 564) sole-authored
articles as well.
27. Because we do not have time-sensitive citation data, we cannot push this analysis further
and assess the degree to which citation success from novel integrations led to a steeper or
accelerated trend.

References
Abbott, Andrew. 2001. Chaos of Disciplines.
Chicago, IL: University of Chicago Press.
Acuna, Daniel E., Stefano Allesina, and Kongrad P.
Kording. 2012. Predicting Scientific Success.
Science 489:20102.
Allen, M. P. (1990). The Centrality of Journals
in Sociology. Foonotes November.
Allison, Paul D. and J. Scott Long. 1987.
Interuniversity Mobility of Academic
Scientists. American Sociological Review
52:64352.
Baldi, Stephane. 1998. Normative versus Social
Constructivist Processes in the Allocation
of Citations: A Network-analytic Model.
American Sociological Review 63:82946.
Bauer, Henry H. 1990. Barriers against
Interdisciplinarity: Implications for Studies
of Science, Technology, and Society (STS).
Science, Technology, & Human Values
15:10519.
Birnbaum, Philip H. 1981. Integration and
Specialization in Academic Research. The
Academy of Management Journal 24:487503.
Blalock, Hubert M. 1991. Providing Opportunities
for Disciplined Creativity. Teaching Sociology
19:40307.
Bonzi, Susan and H. W. Snyder. 1991. Motivations
for Citation: A Comparison of Self Citation and
Citation to Others. Scientometrics 21:24554.
Braun, Tibor and Andreas Schubert. 2003. A
Quantitative View on the Coming of Age of
Interdisciplinarity in the Sciences: 1980-1999.
Scientometrics 58:18389.
Brint, Steven. 2005. Creating the Future:
New Directions in American Research
Universities. Minerva 43:2350.
Burt, Ronald S. 2002. Bridge Decay. Social
Networks 24:33363.
Burt, Ronald S. 2004. Structural Holes and
Good Ideas. American Journal of Sociology
110:34999.

Buskens, Vincent and Arnout van de Rijt. 2008.


Dynamics of Networks if Everyone Strives
for Structural Holes. American Journal of
Sociology 114:371407.
Calcagno, V., E. Demoinet, K. Gollner, L. Guidi,
D. Ruths, and C. de Mazancourt. 2012. Flows
of Research Manuscripts among Scientific
Journals Reveal Hidden Submission Patterns.
Science 338:106569.
Calhoun, Craig. 1992. Sociology, Other
Disciplines, and the Project of a General
Understanding of Social Life. Pp. 13796
in Sociology and Its Publics, edited by T. C.
Halliday and M. Janowitz. Chicago, IL: The
University of Chicago Press.
Cappell, Charles and Thomas M. Gutterbock. 1992.
Visible Colleges: The Social and Conceptual
Structure of Scientific Specialties. American
Sociological Review 57:26673.
Carnabuci, Gianluca and Jeroen Bruggeman.
2009. Knowledge Specialization, Knowledge
Brokerage and the Uneven Growth of Technology
Domains. Social Forces 88:60741.
Chubin, Daryl. 1976. The Conceptualization
of Scientific Specialties. The Sociological
Quarterly 17:44876.
Clemens, Elisabeth S., Walter W. Powell, Kris
McIlwaine, and Dina Okamoto. 1995. Careers
in Print: Books, Journals, and Scholarly
Reputation. American Journal of Sociology
101:43394.
Cole, Stephen. 1993. The Editors View: Why
Most of the Articles in Sociology Journals
are Boring and What Can Be Done About It.
Sociological Forum 3: 337-39.
Collins, Randall. 1968. Competition and Social
Control in Science: An Essay in Theory
Construction. Sociology of Education 41:
12340.
Collins, Randall. 1994. Why the Social Sciences
Wont Become High-consensus, Rapid-discovery
Science. Sociological Forum 9:15577.
Crane, Diana. 1972. Invisible Colleges: Diffusion of
Knowledge in Scientific Communities. Chicago,
IL: University of Chicago Press.
Cummings, J. N. and S. Kiesler (2005).
Collaborative Research across Disciplinary
and Organizational Boundaries. Social Studies
of Science 35(5): 703-722.
Davis, James A. 1994. Whats Wrong with
Sociology? Sociological Forum 9:17997.
Diamond, Arthur M. 1986. What is a Citation
Worth? Journal of Human Resources 21:
20015.

Downloaded from scu.sagepub.com by guest on June 23, 2015

254

Social Currents 1(3)

Dietz, James S., Ivan Chompalov, Barry Bozeman,


Eliesh ONeil Lane, and Jongwon Park.
2000. Using the Curriculum Vita to Study
the Career Paths of Scientists and Engineers:
An Exploratory Assessment. Scientometrics
49:41942.
Dogan, Mattei and Robert Pahre. 1990. Scholarly
Reputation and Obsolescence in the Social
Sciences. International Social Science Journal
42:41727.
Edge, David O. and Michael J. Mulkay. 1975.
Case Studies of Scientific Specialties. Kolner
Zeitschrift fur Soziologie und Sozialpsychologie
18(Suppl.):197229.
Ellison, Glenn. 2002. Evolving Standards for
Academic Publishing: A q-r Theory. Journal
of Political Economy 110:9941034.
Ennis, James G. 1992. The Social Organization
of Sociological Knowledge: Modeling the
Intersection of Specialties. American Sociological Review 57:25965.
Ferber, Marianne. 1986. Citations: Are They an
Objective Measure of Scholarly Merit? Signs
11:38189.
Ferber, Marianne A. and Jane W. Loeb. 1973.
Performance, Rewards and Perceptions of Sex
Discrimination among Male and Female Faculty.
American Journal of Sociology 78:9951002.
Fleming, Lee. 2001. Recombinant Uncertainty in
Technological Search. Management Science
47:11732.
Fleming, Lee, Santiago Mingo, and David Chen.
2007. Collaborative Brokerage, Generative
Creativity, and Creative Success. Administrative Science Quarterly 52:44375.
Fox, Mary Frank. 1992. Research, Teaching,
and Publication Productivity: Mutuality versus Competition in Academia. Sociology of
Education 65:293305.
Fox, Mary Frank and Catherine A. Faver. 1985.
Men, Women, and Publication Productivity:
Patterns among Social Work Academics. The
Sociological Quarterly 26:53749.
Fox, Mary Frank and Paula E. Stephan. 2001.
Careers of Young Scientists: Preferences,
Prospects, and Realities by Gender and Field.
Social Studies of Science 31:10922.
Goldberger, Marvin L, Brendan A. Maher, and
Pamela Ebert Flattau. 1995. Research-doctorate
Programs in the United States. Washington,
DC: National Academy Press.
Gouldner, Alvin Ward. 1970. The Coming Crisis
of Western Sociology. New York: Basic
Books.

Guetzkow, Joshua, Michele Lamont, and Gregoire


Mallard. 2004. What Is Originality in the
Humanities and the Social Sciences? American
Sociological Review 69:190212.
Hagstrom, Warren O. 1970. Factors Related to the
Use of Different Modes of Publishing Research
in Four Scientific Fields. in Communication
among Scientists and Engineers, edited by C.
E. Nelson and D. K. Pollack. Lexington, MA:
Lexington Books.
Hargadon, Andrew B. 2002. Brokering Knowledge:
Linking Learning and Innovation. Research in
Organizational Behavior 24:4185.
Heinsler, Janet M. and Rachel A. Rosenfeld. 1987.
Charting Academic Careers: Does Data
Source Make a Difference? Social Studies of
Science 17:13544.
Hunter, Laura A. and Erin Leahey. 2008.
Collaborative Research in Sociology: Trends
and Contributing Factors. The American
Sociologist 39:290306.
Jacobs, Jerry A. 2004. One Hundred ASR Papers
and Counting. Online Research Note,
American Sociological Review, 2004, Vol. 70
(February:13).
Jacobs, Jerry A. 2005. ASRs Greatest Hits.
American Sociological Review 70:13.
Jacobs, Jerry A. and Scott Frickel. 2009.
Interdisciplinarity: A Critical Assessment.
Annual Review of Sociology 35:4365.
Karides, Marina, Joya Misra, Ivy Kennelly, and
Stephanie Moller. 2001. Representing the
Discipline: Social Problems Compared to ASR
and AJS. Social Problems 48:11128.
Katz, Elihi and Paul F. Lazarsfeld. 1955. Personal
Influence. New York: Free Press.
Keith, Bruce, Jenny Sundra Layne, Nicholas
Babchuk, and Kurt Johnson. 2002. The
Context of Scientific Achievement: Sex Status,
Organizational Environments, and the Timing
of Publication on Scholarship Outcomes.
Social Forces 80:125382.
Klein, Julie Thompson. 1996. Crossing
Boundaries: Knowledge, Disciplinarities,
and Interdisciplinarities. Charlottesville, VA:
University Press of Virginia.
Knorr Cetina, Karin. 1980. The Scientist as an
Analogical Reasoner. in The Social Process
of Scientific Investigation, edited by K. Knorr
Cetina, R. G. Krohn, and R. Whitley (pp. 2552) . Boston, MA: Kluwer.
Leahey, Erin (2006). Gender Differences in
Productivity: Research Specialization as a
Missing Link. Gender & Society 20(6): 754-780.

Downloaded from scu.sagepub.com by guest on June 23, 2015

255

Leahey and Moody


Leahey, Erin, Bruce Keith, and Jason L. Crockett
(2010). Specialization and Promotion in an
Academic Discipline. Research in Social
Stratification and Mobility 28(2):135155.
Leahey, Erin, and Cindy L. Cain (2013). Straight
from the Source: Accounting for Scientific
Success. Social Studies of Science 43(6):
927-951.
Lamont, Michele. 1987. How to Become a
Dominant French Philosopher: The Case
of Jacques Derrida. American Journal of
Sociology 93:584622.
Lamont, Michele. 2009. How Professors Think:
Inside the Curious World of Academic
Judgment. Cambridge, MA: Harvard University
Press.
Lamont, Michele, Gregoire Mallard, and Joshua
Guetzkow. 2006. Beyond Blind Faith:
Overcoming Obstacles to Interdisciplinary
Evaluation. Research Evaluation 15:4355.
Latour, Bruno. 1987. Science in Action: How to
Follow Scientists and Engineers through
Society. Cambridge, MA: Harvard University
Press.
Leahey, Erin. 2007. Not by Productivity Alone:
How Visibility and Specialization Contribute
to Academic Earnings. American Sociological
Review 72:53361.
Leahey, Erin and Ryan Reikowsky. 2008. Research
Specialization and Collaboration Patterns in
Sociology. Social Studies of Science 38:425
40.
Levin, Sharon G. and Paula E. Stephan. 1989.
Age and Research Productivity of Academic
Scientists. Research in Higher Education
30:53149.
Long, J. Scott. 1992. Measures of Sex Differences
in Scientific Productivity. Social Forces
71:15978.
Mansilla, Veronica Boix. 2006. Assessing Expert
Interdisciplinary Work at the Frontier: An
Empirical Exploration. Research Evaluation
15:1729.
Moody, James. 2004. The Structure of a Social
Science Collaboration Network: Disciplinary
Cohesion from 1963 to 1999. American
Sociological Review 69:21338.
Moody, James and Ryan Light. 2006. A View From
Above: The Evolving Sociological Landscape.
The American Sociologist 38:6786.
Mulkay, Michael. 1974. Conceptual Displacement
and Migration in Science: A Prefatory Paper.
Science Studies 4:20534.

National Academies of Science. 2005. Facilitating


Interdisciplinary Research. Washington, DC:
The National Academies Press.
Parker, John N. and Edward J. Hackett. 2012.
Hot Spots and Hot Moments in Scientific
Collaborations and Social Movements.
American Sociological Review 77:2144.
Pierce, Sydney J. 1999. Boundary Crossing
in Research Literatures as a Means of
Interdisciplinary Information Transfer. Journal
of the American Society for Information Science
50:27179.
Poincare, Henri. [1908] 1952. Mathematical
Creation. in The Creative Process, edited
by B. Ghiselin (pp. 22-31). Berkeley, CA:
University of California Press.
Prpic, Katarina. 2002. Gender and Productivity
Differentials in Science. Scientometrics
55:2758.
Raudenbush, Stephen W. and Anthony S. Bryk.
2002. Hierarchical Linear Models: Applications and Data Analysis Methods. Newbury
Park, CA: Sage Publications.
Reskin, Barbara. 1977. Scientific Productivity and
the Reward Structure of Science. American
Sociological Review 42:491504.
Reskin, Barbara. 1978. Scientific Productivity,
Sex, and Location in the Institution of Science.
American Journal of Sociology 83:123543.
Rhoten, Diana and Andrew Parker. 2006. Risks
and Rewards of an Interdisciplinary Path.
Science 306:2046.
Ritzer, George. 1998. The McDonaldization of
American Sociology: A Metasociology Analysis.
Pp. 3551 in The McDonaldization Thesis:
Explorations and Extensions, edited by George
Ritzer. London, England: Sage Publications.
Roscigno, Vincent J. and Randy Hodson. 2007.
Engaging Sociological Audiences: Editors
Comment. American Sociological Review
72:iiiv.
Rosenkopf, Lori and Paul Almeida. 2003.
Overcoming Local Search through Alliances
and Mobility. Management Science 49:75166.
Ryall, Michael D. and Sorenson Olav. 2007. Brokers
and Competitive Advantage. Management
Science 53:56683.
Salter, Liora and Alison Hearn. 1996. Outside the
Lines: Issues in Interdisciplinary Research.
Montreal, Quebec, Canada: McGill-Queens
University Press.
Sauer, Raymond D. 1988. Estimates of the Returns
to Quality and Coauthorship in Economic

Downloaded from scu.sagepub.com by guest on June 23, 2015

256

Social Currents 1(3)

Academia. Journal of Political Economy


96:85566.
Schilling, Melissa A. and Elad Green. 2011.
Recombinant Search and Breakthrough Idea
Generation: An Analysis of High Impact
Papers in the Social Sciences. Research Policy
40:132131.
Schumpeter, Joseph. 1939. Business Cycles. New
York: McGraw-Hill.
Shwed, Uri and Peter S. Bearman. 2010. The
Temporal Structure of Scientific Consensus
Formation. American Sociological Review
75:81740.
Simon, Rita J. and James J. Fyfe. 1994. Editors
as Gatekeepers: Getting Published in the
Social Sciences. Lanham, MD: Rowman &
Littlefield.
Simonton, Dean Keith. 1991. Career Landmarks
in Science. Developmental Psychology 27:
11930.
Singh, Jasjit. 2008. Distributed R&D, Crossregional Knowledge Integration and Quality
of Innovative Output. Research Policy
37:7796.
Small, H. G. and B. C. Griffith (1974). The
Structure of Scientific Literatures: Identifying
and Graphing Specialties. Science Studies 1:
445-461.
Smith, Adam. [1766] 1982. Lectures on
Jurisprudence. Indianapolis, IN: Liberty Fund.
Steinmetz, George and Ou-Byung Chae. 2002.
Sociology in an Era of Fragmentation: From
the Sociology of Knowledge to the Philosophy
of Science, and Back Again. The Sociological
Quarterly 43:11137.

Traweek, Sharon. 1992. Beamtimes and Lifetimes:


The World of High Energy Physicists.
Cambridge, MA: Harvard University Press.
Uzzi, Brian, Satyam Muhkerjee, Michael Stringer,
and Ben Jones. 2013. Atypical Combinations
and Scientific Impact. Science 342:46872.
Uzzi, Brian and Jarrett Spiro. 2005. Collaboration
and Creativity: The Small World Problem.
American Journal of Sociology 111:447504.
van Dalen, Hendrik and Kene Henkens. 2005.
Signals in ScienceOn the Importance of
Signaling in Gaining Attention in Science.
Scientometrics 64:20933.
Ward, Kathryn B. and Linda Grant. 1995. Gender
and Academic Publishing. Higher Education:
Handbook of Theory and Research 11:172212.
Weick, Karl E. 1979. The Social Psychology of
Organizing. Reading, MA: Addison-Wesley.
Weisberg, Robert W. 2006. Creativity: Understanding Innovation in Problem Solving,
Science, Invention, and the Arts. Hoboken, NJ:
John Wiley.
Weitzman, Martin L. 1998. Recombinant Growth.
The Quarterly Journal of Economics 113:
33160.
Whitley, R. 2000. The Intellectual and Social
Organization of the Sciences. Oxford, England:
Oxford University Press.
Wuchty, Stefan, Benjamin F. Jones, and Brian Uzzi.
2007. The Increasing Dominance of Teams in
Production of Knowledge. Science 316:103639.
Xie, Yu and Kimberlee A. Shauman. 1998. Sex
Difference in Research Productivity: New
Evidence about an Old Puzzle. American
Sociological Review 63:84770.

Downloaded from scu.sagepub.com by guest on June 23, 2015

Vous aimerez peut-être aussi