Networks of neuroscientists: professional interactions within an interdisciplinary brain research institute.
Godley, Jenny ; Sharkey, Keith A. ; Weiss, Samuel 等
Introduction
This project aimed to assess how the formation of the Hotchkiss
Brain Institute (HBI; www. ucalgary.ca/hbi) at the University of Calgary
(UCalgary) in October 2004 affected professional interactions among
neuroscience researchers. Prior to the formation of the HBI, there was
no single administrative structure linking neuroscientists who work in
different academic departments at UCalgary. Since the formation of the
HBI, essentially all neuroscientists working at UCalgary (including new
hires) are encouraged to become members of the HBI. Thus, the
establishment of the HBI provides a case study through which to examine
whether the formation of an explicitly interdisciplinary administrative
unit affects professional interactions among scientists at one
institution. We conducted a whole network survey of the members of the
HBI in November, 2010. We asked all current HBI members (N = 95) to fill
out an online survey reporting on their professional interactions with
each of the other members since the foundation of the HBI (2005-2010).
In addition, for those members who joined the HBI in 2005, we asked
about their interactions with other members before the foundation of the
Institute. Eighty-one scientists (a response rate of 85%) filled out the
survey, indicating their working relationships with other HBI members.
We analyzed the data using social network analytic techniques,
described below, as well as descriptive statistics. We also examined
whether individual-level characteristics of the scientists such as
gender, rank, department, office location, research theme, and research
pillar affected their relationships with other scientists. Research
pillar is a term used in Canada to classify all health researchers into
one of four categories--biomedical, clinical, health services or
population health (Canadian Institutes of Health Research, 2009).
Finally, for each of the professional interaction networks, we examined
the positions of those who hold leadership roles in the HBI. We use our
results to discuss the effect that the establishment of the HBI has had
on professional relationships among neuroscientists at UCalgary. We
conclude by reflecting on the usefulness of social network analysis as
an evaluation method for interdisciplinary research institutes.
Background
The need for collaboration in science is well accepted (Adams,
Black, Clemmons, and Stephan, 2005), and many academic fields now
encourage interdisciplinary work (Hackett, 2005). Funding agencies are
increasingly mandating interdisciplinary teams on grant applications
(Bammer, 2008; Mattson, Lager, Vindefjard, and Sundberg, 2010). However,
there remains much debate in the literature over how to measure,
classify and evaluate interdisciplinary research and collaboration (see
Huutoniemi, Klein, Bruun and Hukkinen, 2010 for a recent overview).
The first large-scale interdisciplinary research centres were set
up in the United States in the 1980's, in an attempt to encourage
both cross-disciplinary and cross-sector (university and industry)
research collaboration. Most evaluations of such centres have focused on
productivity as measured by publications and patents (Geiger, 1990;
Ponomariov and Boardman, 2010). Typically, co-authorship is used as a
measure of collaboration, and journal subject categories are frequently
considered as a measure of interdisciplinarity.
Collaboration involves more than simply co-authorship, though, and
evaluations of interdisciplinary research teams need to capture the
achievements of such teams beyond publications and patents (Katerndahl,
2012; Katz and Martin, 1997). Over the past twenty years,
interdisciplinary institutes, centers and groups have been established
at many institutions, yet methods for assessing and evaluating
interdisciplinary scientific endeavors are still in their infancy
(Harris, 2010; Yang, Park & Heo, 2010; Levitt & Thelwall, 2008).
Fundamentally, interdisciplinary scientific collaboration
materializes through relationships and interactions among people from
different academic disciplines. These relationships may involve people
from different administrative units within one institution, or from many
institutions. In order to evaluate interdisciplinary collaboration
properly, we first need to understand how such relationships and
interactions emerge and change. Second, we need to assess the factors
that both encourage and hinder interactions among scientists
(Katerndahl, 2012). In this paper, we utilize the case study method to
examine how the professional interactions among scientists at one
university change with the establishment of an interdisciplinary
research institute. We draw upon the theories and techniques developed
in the field of social network analysis to guide our study.
Literature Review
Social network analysis, which prioritizes the structure of social
relationships over the attributes of individuals, would suggest that,
fundamentally, scientific output is the product of social relations
(Emirbayer, 1997). While this claim may be too broad, as it neglects
additional factors that influence scientific productivity (such as
discipline and funding) (Quinlan, Kane, and Trochim, 2008), social
network analysis does provide data collection methods and data analytic
techniques that have been developed specifically for relational data
(Scott, 2000; Wasserman and Faust, 1994). Thus, we employ social network
analytic techniques to examine and assess the professional interactions
among the members of the HBI.
Social network analysis has been used in a variety of ways to
examine collaboration in academic research. Most commonly, scholars have
applied whole network analytic techniques to bibliographic databases to
investigate how networks of co-authorship and citation change over time
(Barabasi, Jeong, Neda, Ravasz, Schubert, and Vicsek Borgatti, 2002).
Findings from these co-authorship studies highlight several factors
which facilitate collaboration among scientists. Mattson et al. (2010),
examining European research collaboration networks in the life sciences,
found that while co-authorship is affected by geographical proximity and
language, funding mechanisms also have a large impact on scientific
collaboration. Wagner and Leydesdorff (2005), on the other hand,
concluded that the recent growth in international research collaboration
in six scientific fields was not driven by funding, but rather by the
scientific interest of researchers.
Johnson, Christian, Brunt, Hickman, and Waide (2010) examined the
US Long Term Ecological Research (LTER) Program using intersite
publications as a measure of collaboration. They found the most
important predictors of collaboration were common research theme and
communication at meetings and conferences. Sun and Manson (2011)
examined the growth of co-authorship in geographic information science
from 1992 to 2007. They found language and country of practice to be
important in predicting co-authorship. Braun, Glanzel, and Schubert.
(2001) examined co-authorship in the field of neurosciences. They found
that scientists in the middle of their career are most likely to work
collaboratively than both newcomers and senior authors. Thus, findings
from co-authorship studies suggest that while geography, language and
funding all affect collaboration, shared research interests,
communication channels, and career stage are also important.
Other researchers have used network survey data to highlight
barriers and facilitators to interdisciplinary research both within and
across universities (Aboelela, Merrill, Carley and Larson, 2007; Haines
et al., 2010; Katerndahl, 2012; Godley, Barron & Sharma, 2011).
These studies demonstrate that although researchers are still most
likely to collaborate with those in their own discipline, established
collaborations of researchers organized into institutes or research
groups can increase interdisciplinary work. Thus, the results from both
network analyses of co-authorship and network surveys of researchers
attempting to engage in interdisciplinary research would seem to suggest
that scientific collaboration should be enhanced by the establishment of
an interdisciplinary research centre.
Most previous evaluations of interdisciplinary research centres
have not used social network analytic techniques, but rather have relied
on survey data and measures of productivity such as funding received and
peer-reviewed publications to assess the success of the centres (Bozeman
et al., 2010; Ponomaniov and Boardman, 2010; Wagner, Roessner, Bobb,
Klein, Boyack, et al., 2011). Discussing methodological issues that
arise in studying large research initiatives, Quinlan et al. argue that
ideally evaluators should utilize multiple methods (including
bibliometric analysis, surveys, and interviews) to gage the success of
interdisciplinary research centers (Quinlan et al., 2008). Findings from
these studies highlight several additional factors which appear to be
important for the success of an interdisciplinary research centre,
including infrastructure, shared vision, communication, and leadership.
Meyer, Fabor, and Hesselbrock (1988), in a study of an
interdisciplinary research centre for alcohol addiction, found that
success was determined by four factors: the strength of the
infrastructure (including resources, facilities and personnel); the
articulation of a shared vision; efficient management; and clear
communication networks. Hagen et al. (2011) studied a multicenter
clinical research network and also argue that the following are key to
success: shared vision; governance; infrastructural support; and
communication. Studying several university research centers in science
and engineering, Boardman and Corley (2008) argue that while
multidisciplinarity within a centre is necessary for research
collaboration, funding had the biggest effect on the amount of time
scientists allocate to collaborative work.
However, Bammer (2008) argues that the key to research
collaboration is effectively harnessing differences through strong
leadership. Using the Human Genome Project as a case study, he
demonstrates that strong leadership is needed to encourage productive
differences and discourage unproductive differences in interdisciplinary
research collaborations. Chompalov, Genuth, and Shrum (2002) further
argue that the structure of the interdisciplinary research organization,
and the type of leadership, are dependent on the disciplines involved.
Others suggest that the success of an interdisciplinary research
centre is largely dependent on the individuals involved, and the fields
they represent. Rijnsoever and Hessels (2011) found that scientists in
Basic fields are more likely to collaborate within their own
disciplines, while scientists in Applied fields are more likely to work
with those from other disciplines. Neuroscience involves both Basic and
Applied researchers. Neuroscience not only involved both Basic and
Applied researchers, it is also inherently interdisciplinary, as it
emerged in the 1970's as an amalgamation of brain research in many
fields, including anatomy, physiology, pharmacology, neurology,
psychology and psychiatry (Doty, 1987). Birnholtz (2007) conducted a
survey to measure scientists' "propensities to
collaborate." He found that neuroscientists have a
'medium' propensity to collaborate, lying somewhere between
physicists (high propensity to collaborate) and engineers (low
propensity to collaborate).
To summarize, while results from both network analyses of
co-authorship and network surveys of interdisciplinary researchers
suggest that scientific collaboration may be enhanced by the
establishment of an interdisciplinary research centre, previous
evaluations suggest that a strong infrastructure (including funding), a
shared vision, a clear means of communication, and solid leadership will
be necessary for a centre's success. Additionally, previous
research suggests that the success of any interdisciplinary endeavor is
largely dependent on the individuals involved, and the disciplines they
represent.
Setting--the Hotchkiss Brain Institute
This paper takes a case study approach, examining the effect of the
establishment of an interdisciplinary research centre on the
collaborative activities of a group of neuroscientists at one Canadian
University. We do not attempt to generalize from this case study to any
other institutions or research centres, but rather present it as an
example of the effect of administrative endorsement of interdisciplinary
work on the professional interactions of researchers at one institution.
The Hotchkiss Brain Institute (HBI) was established in 2004 at
UCalgary with a generous philanthropic donation from a local family. All
of the faculty members who were working in neuroscience at the UCalgary
in 2004 were encouraged to become HBI members when the Institute was
established. The original membership consisted of 64 faculty members,
representing ten different disciplines. All new Faculty members who were
hired in neuroscience after 2004 became members of the HBI when they
joined UCalgary. By 2010 the number of members had risen to 95. The HBI
has operated under the leadership of the same Director since its
establishment, and has several sub-committees and programs led by other
members.
Initially, researchers in the HBI were organized by department
(discipline) and research pillar, and some of the researchers were in
translational research programs. In 2010, following an extensive
external review and evaluation, the leadership of the HBI formalized
three research themes within the Institute: Axon Biology and
Regeneration; Cerebral Circulation; and Neural Systems and Behaviour.
These research areas cross disciplinary boundaries, as stated on the HBI
website (www.ucalgary.ca/hbi).
"The mission of the Hotchkiss Brain Institute (HBI) is to be a
centre of excellence in neurological and mental health research,
translating discoveries into innovative health care solutions. This
mission will aim to support and conduct research on the healthy and
diseased brain, spinal cord and peripheral nerves to assess, understand
and disseminate knowledge about the diseases affecting the nervous
system...It is the aim of the HBI to have a collective expertise in the
field of neurosciences, increased collaboration and elevated synergism
in our research efforts."
From the outset, the HBI appeared to incorporate many of the
elements which comprise a successful interdisciplinary research centre.
The HBI has strong funding support, maintains a strong collective
vision, and benefits from strong, consistent leadership. Communication
is encouraged among members through weekly seminars, mentoring programs,
and internal peer review panels.
Quantitative measures such as publication output and research
funding indicate that the HBI has had considerable success since 2004.
The HBI is now recognized internationally as a top research Institute in
neuroscience. It is one of the 100 most active neuroscience research
organizations in the world, as measured by publication output (Haustein,
Cote, and Beaudet, 2013). The success rate of HBI scientists in terms of
peer reviewed grant support over the last few years has been about twice
that of the national average. Grant support has increased 120% from
2004-2010, with only a 50% increase in faculty members over that time
period. Additionally, the HBI has successfully raised money from
private, community, and governmental sources for ongoing research and
educational activities.
In 2010, as part of an extensive evaluation process, and in an
effort to supplement the traditional measures of success such as
publication output and research funding, HBI leaders set out to assess
how the formation of the HBI had affected working relationships among
neuroscience researchers at UCalgary. One of the aims of the HBI was to
build on existing collegial interactions and enable natural linkages to
form amongst scientists. In order to evaluate this process, HBI leaders
turned to social network analysis.
Network analysts typically conduct one of two types of relational
analyses: whole network studies, where the boundaries of the population
are known and information is gathered from the whole population; or
ego-centred network studies, where individuals are asked to report on
others with whom they have certain relationships (Wasserman & Faust,
1994). For the current study, an online whole network survey was
designed and administered to all members of the HBI.
In the fall of 2010, HBI members were asked about their
professional relationships with one another. For the original members of
the institute, data was collected on their relationships both before and
since the founding of the HBI, as well as on individual-level attributes
of the HBI members, including gender, home department, and academic
rank. Using social network analytic techniques, this data was assessed
to determine the extent of professional interactions, and the predictors
of (and barriers to) professional interactions within the HBI.
Methods
Study Description
A complete list of all the current members of the HBI was compiled
in the fall 2010. Ninety- five members were identified. An email message
was sent to each of these researchers to ask them to complete the online
network survey. Each researcher was given a list of the other 94 HBI
members, and asked to indicate if they had any of the following
relationships with the other researchers over the past five years (from
January 2005 to the present): "went to for advice or
mentorship"; "organized a conference with"; "held a
grant with"; "co-authored papers with";
"co-supervised students with"; "co-taught a course
with." The sixty-four original members, who had ,joined the HBI
when it was founded (in October 2004), were also asked to report on
relationships with other original members prior to January 2005.
Descriptive information collected on the respondents included: gender,
department, professorial rank, CIHR Pillar membership, HBI Theme
membership (three thematic areas, as described above), and office
location. Data collection continued for two months. Eighty-one
questionnaires were returned, fifty-two of which were fiom original HBI
members (response rates of 85% for the current members, and 81% for the
original members). A copy of the survey is provided in the Appendix.
"The study design was approved by the Research Ethics Board at
UCalgary. In order to comply with recommendations from the Ethics Board,
all individuals who did not respond to the survey were completely
removed from the data. Thus, the complete networks consist only of the
81 individuals who returned their surveys for the second time period and
the 52 individuals who returned their surveys for the first time period.
Research Questions
This paper is organized around the following four central research
questions:
Research Question 1: How many and what type of professional
interactions existed among HBI members before the formation of the HBI,
and how many have occurred since the formation of the HBI? Has the
extent of involvement in professional interactions among the HBI members
changed over time?
Research Question 2: What is the composition of the professional
interaction networks within the HBI, with regards to individual-level
attributes such as gender, department, research theme and research
pillar?
Research Question 3: Which individual-level characteristics of
respondents predict professional interactions between researchers? Have
these predictors changed over time?
Research Question 4: Are the currently appointed HBI leaders
occupying leadership positions in the professional interaction networks?
Exploring these research questions will enable an assessment of the
success of the HBI that goes beyond traditional measures of success such
as publication output and research funding. This research will evaluate
the success of the HBI in terms of encouraging and promoting
interdisciplinary, collaborative working relationships among
neuroscientists at UCalgary. Moreover, this research will shed light on
the potential benefits of using network analysis to develop focused
themes to enhance collaborations amongst interdisciplinary groups of
researchers beyond the neurosciences.
Networks
We constructed six professional interaction networks for each time
period from the survey data: the 'advice' network; the
'conference' network; the 'grant' network; the
<co-authored' network; the 'co-supervised' network;
and the 'co-taught' network. For each analysis, we examined
three different sets of networks. First, we examined the six networks of
the 52 original HBI members reporting on activities prior to 2005.
Second, we examined the six networks of the same individuals (the 52
original HBI members) reporting on activities since 2005. Finally, we
examined the six networks of the full 81 HBI members in 2010 reporting
on activities since 2005. Although the networks of the 52 original
members reporting on activities since 2005 are in some sense
'incomplete' (because they do not include the new HBI
members), we looked at these separately in order to make statistical
comparisons with the original 52-member networks.
Descriptive Analyses
We used the network statistical software package UCINET 6.0
(Borgatti, Everett and Freeman, 2002) to conduct analyses of our network
data. To answer our first research question (How many and what type of
professional interactions existed among HBI members before the formation
of the HBI, and how many have occurred since the formation of the HBI?),
we examined the following whole network measures: density; reciprocity;
average degree; and number of isolates. Density is the proportion of
actual linkages to possible linkages among group members (Wasserman
& Faust, 1994). Density can be regarded as a measure of how
interconnected individuals in a network are, where a density value of
100% would indicate that every person in the network is directly
connected to every other person in the network (Scott, 2000).
Reciprocity is the percentage of linkages that are reciprocated, or
returned.
Isolates are individuals who are not connected to any others in a
network. The number of isolates in a network indicates the percentage of
respondents who are not involved with any other network members for a
particular activity (Waasserman & Faust, 1994). Degree or the number
of ties is the number of alters a respondent mentions.
To answer our second research question (What is the composition of
the professional interaction networks within the HBI, with regards to
individual-level attributes such as gender, department, research theme
and CIHR pillar membership) we examined and compared compositional
measures for each network, across the various individual-level
variables. We assessed the diversity of these networks with regard to
individual-level characteristics.
Statistical Analyses
To answer our third research question (What individual-level
characteristics of respondents predict professional interactions between
researchers? Have these predictors changed over time?), we performed
Quadratic Assignment Procedure (QAP) regression (Carley &
Krackhardt, 1996; Brewer & Webster, 1999; Burris, 2005). Network
data violates the assumptions of Ordinary Least Squares (OLS)
regression. The QAP regression procedure, which overcomes these
limitations, is best understood as a form of simulation (Burris, 2005).
First, OLS coefficients are calculated for the independent variables in
the regression. Next, the rows and columns of the dependent variable
matrix are randomly permuted and the OLS regression coefficients are
re-calculated. The simulation is repeated 2,000 times in UCINET 6. The
initial regression coefficients are then compared with the distribution
of all possible coefficients, and significance tests are based on these
distributions.
To answer our fourth research question (Are the appointed HBI
leaders occupying leadership positions in the professional interaction
networks?) we examined two network measures of centrality: degree
centrality and betweeness centrality. Degree centrality simply measures
the number of ties individuals have, while betweeness centrality
examines the number of times any individual lies on the shortest path
between two other individuals in the network (Freeman, 1979).
Both centrality measures can be thought of as measures of
leadership roles within a network (Katerndahl, 2012). Degree centrality
can be thought of as a comparative measure of activity, while betweeness
centrality can be thought of as a measure of comparative control or gate
keeping within the network. Gatekeepers act as important hubs within a
network, who can assist with the diffusion of knowledge and information
(Cowan and Jonard, 2004). We compared these measures for HBI leaders and
non-leaders.
Visualizations
Finally, we produced and examined a network visualization of each
of the networks. Network visualization presents information on network
structure in graphic form. This method of analyzing network data is an
important part of social network analysis because graphic
representations can reveal information that may not be statistically
obvious (Luke and Harris, 2007). In these graphs, the researchers are
represented by shapes, and the relationships are represented by lines.
In all of the graphs except those for the 'advice' network we
present reciprocal ties only. In the 'advice' network graphs,
the arrows are used to represent the direction of the interaction (the
arrow pointing away from the responding researcher (seeking advice) and
towards the researcher they nominated (from whom they sought advice)).
We do not include isolates (respondents with no ties) in the graphs, but
we note the number of isolates underneath each graph. To produce the
graphic representations of our networks, the layout uses the spring
embedded procedure in Netdraw 2.24 which graphs the nodes (the
researchers) according to their geodesic (shortest distance)
proximities. We present the most illustrative of these diagrams in the
results section.
Results
Sample
Table 1 describes the composition of the sample. Of the 52 original
members who responded to the survey, 11.5% were female. This percentage
rises to 21% of current members in 2010. HBI members come from ten
different departments. Examining the 2010 data, the largest group
(34.6%)
are from the Department of Clinical Neurosciences, followed by 17.3%
from Physiology and Pharmacology. "[he rest are fairly evenly split
between the Departments of Psychiatry, Psychology, Radiology, Cell
Biology and Anatomy, Community Health Sciences and 'Other.'
The 'Other' group is comprised of individuals from three
different departments, each of which had fewer than 2 respondents. We
combined these departments to avoid deductive disclosure.
In 2010, over 50% of respondents place themselves in the
'Biomedical Research' CIHR pillar, followed by almost 35% in
'Clinical Research.' We combined the 'Population'
and 'Health Services' pillars, as together they represented
only 8.6% of the sample. In terms of the HBI research themes, which were
formalized in 2010, 55.6% of the sample are in the Neural Systems and
Behaviour theme, with the rest evenly split (22.2% each) between Axon
Biology and Regeneration and Cerebral Circulation. HBI researchers'
offices are spread over six different buildings on the UCalgary campus.
Just over half the researchers are Full Professors, while approximately
a 25% are Associate Professors, and 23% are Assistant Professors.
Descriptive Results
We first examined the extent of professional interactions among
researchers who are members of the HBI, both before and since the
founding of the Institute. Table 2 contains the density and reciprocity
figures for each of the six professional activities across three sets of
networks. The first two columns present figures for the 52 original HBI
members, before the founding of the Institute. The middle columns
present figures for the original HBI members since the founding of the
Institute. And the final two columns present figures for all 81 HBI
members since 2005.
The first column of Table 2 illustrates that prior to the founding
of the Institute, the 52 neuroscientists at UCalgary were already
involved in many professional activities together. Before 2005, they
were most likely to co-supervise students, write papers together, teach
together and request advice from one another. However, they also held
grants together and organized conferences together. While the density
figures may appear low at first glance, it is important to remember that
density is recorded as a percentage of all possible ties. There are
2,652 possible ties among the 52 original members. Thus a density of
.046 for co-authorship indicates that HBI members reported 122 unique
co-authorship ties before the founding of the Institute.
The third column of Table 2 illustrates that since the founding of
the institute, the original 52 members again report that they have been
most likely to co-supervise, request advice, co-author papers and
co-teach with others. They have also held grants and organizing
conferences with the other original members.
The fifth column of Table 2 shows that since the HBI was founded,
the 81 current members report that they have been most likely to
co-supervise, seek advice and co-author with other HBI members, followed
by co-teaching, holding grants, and organizing conferences together.
There are 6,480 possible ties among the 81 current members. Thus a
density of .034 for co-authoring papers indicates that HB1 members
reported 220 unique co-authorship ties between 2005 and 2010.
Statistical comparisons can only be made for densities of networks
of the same size. UCINET 6.0 tests for statistical significance using a
bootstrap technique to compare the densities of networks with the same
members, allowing two different time points to be compared (Snijders and
Borgatti, 1999). There are statistically significant increases in
density for the original 52 members for both the grants and advice
network.
We turn now to the results for reciprocity in Table 2.
Theoretically, all of the relationships except for 'advice'
should be reciprocal. However, we find that network reciprocity (the
percent of all ties that are reciprocated) (Wasserman & Faust, 1994)
ranged from a high of 73% for 'co-authored' since 2005 down to
a low of 17% for 'conference' before 2005. There are many
potential reasons for low reciprocity. The salience of some activities
is obviously higher than others (writing papers compared to being on a
supervisory committee together, for example). Additionally, reciprocity
is generally higher for the more recent time period. However, in order
to avoid potential bias introduced by including non-reciprocal ties,
from this point forward in the paper we only examined reciprocal ties
for all networks except 'advice'. The practice of only
examining confirmed ties is considered to increase the reliability of
self-reported network data (Scott, 2000). In future analyses, we plan to
conduct analyses including the non-reciprocal ties in order to determine
if the predictors of non-reciprocal ties differ from the predictors of
the reciprocal ties.
Table 3 illustrates the average number of people researchers are
tied to for each professional relationship at each of the two time
periods, and the number of isolates in each network. We can see that on
average before 2005, the 52 neuroscientists at UCalgary reported seeking
advice from 2.8 others, co-authoring with 2.4 others, and co-supervising
with 2.3 others. Between 2005 and 2010, on average, the 81 members of
the HBI reported seeking advice from 5 other HBI members, co-authoring
with 2.8 others and co-supervising with 2.8 others.
While we cannot compare the average ties over time statistically
(because the total number of network members changes between the two
time periods), it appears that HBI members have more ties to other
neuroscientists at UCalgary since the founding of the Institute. The
average number of ties goes up for every professional activity. An
analysis of the isolate data also shows that the percent of people who
have NO ties to other neuroscientists at UCalgary has gone down across
all of the professional activities except teaching.
We next examined the composition of the professional interaction
networks over time, with regards to the individual-level characteristics
of the network members. Table 4 shows the composition of the seven
networks at the two time points in terms of gender, department, HBI
research theme, and CIHR pillar.
Table 4 shows that the advice networks have become more homogenous
by gender between the two time periods. Before 2005, the neuroscientists
reported that 58% of those from whom they sought advice were the same
gender as themselves, while after 2005 this number increased to 74%.
However, all of the other professional activities are less homogenous by
gender after the foundation of the HBI.
Overall, department does not appear to be as salient as either CIHR
pillar or HBI theme for network composition. Except for teaching and
organizing conferences before 2005 and advice and conference
organization between 2005-2010, all of the networks were composed of
more than 50% people from other departments. The importance of the CIHR
pillars for the organization of networks is clear, and went up over
time. Between 2005 and 2010, respondents reported that 68% of their
co-authors and 78% of those with whom they co-supervised were in the
same CIHR pillar as themselves.
The HBI formalized the three research themes in 2010, in an attempt
to reflect how members were organizing themselves around research areas.
Members clearly find these research themes to be salient, as they report
that 61% of the people with whom they hold grants are in the same
research theme, and 63% of their co-authors are in the same research
theme area.
QAP Regression Results
Tables 5 and 6 contain the QAP regression results for five of the
six original networks and the multiplex network for each time point. We
chose to eliminate the conference network from the remainder of the
analyses, since the network is so sparse. To create the multiplex
network, we summed across all of the networks (Koehly & Pattison,
2005). The tie value for each pair is the total number of types of
relationships between any two individuals, and can range from 0 to 5.
The multiplex network thus illustrates the extent to which actors share
multiple relationships.
The QAP regression results illustrate the simultaneous effect of
each of the independent variables on the likelihood of a tie between two
individuals. Burris (2005) argues that when interpreting QAP regression
results, the focus should be on the comparative magnitude of the
coefficients, rather than on the overall model [R.sup.2] or the level of
statistical significance for each coefficient. In Tables 5 and 6, we
report the standardized coefficients for each independent variable, and
their significance level. Our discussion will focus on the comparative
magnitude of those coefficients which are significant.
The top half of Table 5 contains the regression results for the
pre-HBI networks. There are five independent variables included in these
models: belonging to the same department, being the same gender,
belonging to the same CIHR pillar, having offices in the same building,
and having the same appointment level. Controlling for the other
independent variables, department appears to have the largest effect on
collaboration among the 52 neuroscientists at UCalgary before 2005.
Belonging to the same department was statistically significantly related
to seeking advice, co-authoring, and co-teaching, controlling for the
other independent variables.
Figure 1 shows the pre-2005 co-authorship network. Respondents are
represented by shapes, and the lines indicate reciprocal co-authorship
relations. In this figure, the respondents are shaped by departmental
membership. The clustering of certain shapes in different areas of the
graph illustrates the importance of departmental membership for
co-authorship among the 52 scientists before 2005.
[FIGURE 1 OMITTED]
Turning back to the results presented in the top half of Table 5,
the next most important independent variable for the pre-HBI networks is
being at the same appointment level, which has a positive impact on the
presence of ties for advice, co-authoring, and co-supervising, net of
the other independent variables. Interestingly, having offices in the
same building was also positively related to seeking advice and sharing
grants before 2005, net of the other variables. In fact, the only
variable that predicted shared grants in the pre-HBI time period was
office co-location.
The bottom half of Table 5 also examined the predictors of
relationships among the original 52 HBI members during the 2005-2010
time period. An additional independent variable is added to these
models, belonging to the same HBI theme.
We find that being in the same department is still positively
related to requesting advice, co-authoring and teaching, controlling for
the other independent variables. However, belonging to the same HBI
theme also becomes important for all of the relationships except
teaching. During this time period, belonging to the same theme has a
stronger effect on holding grants, co-authorship, and co-supervision
than belonging to the same department, net of the other variables.
Once again, having offices in the same building is positively
related to all the relationships. In fact, shared office location is the
most important predictor of shared grants and co-authorship for the
original 52 members since 2005, controlling for the other independent
variables.
Table 6 shows the regression results predicting relationships among
all 81 members of the HBI between 2005 and 2010. These results indicate
that belonging to the same department, belonging to the same HBI theme,
and having an office in the same building are all positively related to
ties across all of the professional activities (except teaching), net of
other variables. Shared CIHR pillar affects advice, supervising and
teaching, but not grants and papers, net of other variables.
Among the full 81 members, advice is impacted by all of the
independent variables except shared rank. It is interesting to note that
gender becomes significant in this model, suggesting that women are more
likely to seek advice from female colleagues, and men from male
colleagues, net of other independent variables. Since the founding of
the HBI, the UCalgary has implemented a gender equity policy that has
increased the number of women hired in predominantly male disciplines.
The effects of this policy are reflected in the fact that the percent of
the sample that is female rose from 11.5% in 2005 to 22% in 2010. By
2010, women had more same-gender colleagues from whom to seek advice.
However, the effect of shared gender on advice seeking remains half that
of shared department.
Figure 2 shows the co-authorship network between 2005 and 2010 with
individuals shaped by departmental membership, and Figure 3 shows the
co-authorship network between 2005 and 2010 with individuals shaped by H
BI research theme. Once again, the effect of department on co-authorship
is evident in Figure 2 by the clustering of similarly shaped squares.
Comparing Figures 2 and 3, we can see that co-authorship is even more
clearly clustered by HBI theme.
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
Finally, we compare the results for the multiplex networks across
Tables 5 and 6. For the 52 neuroscientists at UCalgary before 2005, the
strength of the tie between two researchers (or the number of different
professional activities they engage in together) is affected by shared
department, shared rank and office co-location, controlling for gender
and CIHR pillar. For these same individuals, the strength of the tie
between two researchers from 2005 to 2010 is affected by office
coqocation, shared department, shared theme, shared pillar and shared
rank, controlling for gender. For all 81 HBI members, the strength of
the tie between two researchers from 2005 and 2010 is affected by shared
department, shared theme, office co-location and shared pillar.
Leadership Analysis
The HBI has eleven members who hold leadership roles, including a
Director, a Deputy Director, an Education Director and leaders for each
of the thematic research areas. We calculated both degree and betweeness
centrality measures for the 2005-2010 data, and then compared these
measures for leaders versus non-leaders. The results of this analysis
are shown in Table 7.
Degree centrality is a measure of activity; it simply indicates the
number of ties reported by each individual. An individual with no ties
will not have a value for degree centrality. Table 7 shows that there
are significant differences between the activity levels of leaders and
non-leaders in the HBI across two of the professional activities: advice
giving and co-supervising. On average, leaders were asked for advice by
10.5 others between 2005 and 2010, while non-leaders were only asked for
advice by only 3.6 other members. Interestingly, leaders also requested
advice from more others (7.7) than non-leaders (4.5) during this time
period. Leaders co-supervised with an average of 4.5 other members,
while non-leaders reported co-supervising with 2.5 others.
Betweeness centrality measures how often an individual lies on the
shortest path between two other individuals. Betweeness centrality is
often considered a measure of bridging, or an indicator of a brokering
tie. Individuals with high betweeness can facilitate (or hinder) the
flow of information through a network. Table 7 illustrates that the HBI
leaders have significantly higher betweeness centrality than other
members in every one of the professional networks.
Figure 4 illustrates the advice-seeking network for the 81 HBI
members from 2005-2010. The eleven members who hold leadership roles are
drawn as large squares, while the rest of the members are shown as small
squares. This figure clearly shows how central the leaders are in the
network.
Discussion
We set out to evaluate how the establishment of the HBI affected
professional interactions among neuroscientists at UCalgary. We organize
our discussion as a set of reflections on our four research questions.
Findings
Our first research question asked how many and what type of
professional interactions existed among HBI members before the formation
of the HBI, and how many have occurred since the formation of the HBI.
We found that while professional interactions did occur among the 52
neuroscientists at UCalgary before the foundation of the HBI,
professional interactions have increased since the HBI was founded.
Specifically, we found a statistically significant increase in
interactions among the original 52 HBI members in terms of advice
seeking and holding grants together.
[FIGURE 4 OMITTED]
The percent of the sample who are isolates declined in each network
except the teaching network (where it rose from 48 to 49%) over the time
period. However, approximately a quarter of the members are isolates in
the 2005-2010 advice networks, approximately a third are isolates in the
co-author and co-supervise networks, and approximately half are isolates
in the grants and teaching networks. Since we only considered reciprocal
ties for the latter four networks, some isolates may have been created
by recall bias. We performed post-hoc analyses to examine the isolates,
in an attempt to determine if they shared similar characteristics. We
found that the vast majority of isolates are either recent recruits to
UCalgary or clinicians. The recent hires have not had as much time to be
involved in professional interactions with other HBI members, and the
clinicians spend significant time in their clinical roles, thus leaving
less time to be involved in research.
Among the current members, we found that the most common
professional interactions within the HBI are co-supervising and seeking
advice, followed by co-authoring, co-teaching and holding grants
together. On average, between 2005 and 2010, HBI members sought advice
from 5 other members, supervised students with 2.8 other members, and
co-authored with 2.8 other members. Seventy percent of HBI members
co-authored with at least one other member between 2005 and 2010, and
over half of HBI members held a grant with at least one other member in
the same time period, qhus it appears that the formation of the HBI has
indeed increased professional interactions among neuroscientists at
UCalgary.
Our second research question asked about the composition of the
professional interaction networks within the HBI, with regards to the
individual-level attributes of gender, department, research theme and
CIHR pillar membership. We found that the networks of grants, papers,
supervising and teaching were all more homogenous by CIHR pillar and HBI
research theme than they were by department. This result appears to
indicate that the HBI has succeeded, at least partially, in overcoming
the administrative boundaries that frequently exist between departments.
The HBI themes, which have emerged naturally among the members since the
founding of the Institute, appear to provide salient groupings within
which researchers can cluster.
Our third research question addressed the individual-level
characteristics of respondents that predict professional interaction
between researchers. We first examined the predictors of professional
relationships at two time periods for the original 52 researchers. We
found that although shared department was important for predicting
relationships at both time periods, its importance had declined by the
2005-2010 time period. Shared theme became more important than
department for holding grants together, co-authoring, and
co-supervising. Once again, this result suggests that the HBI themes are
proving relevant for researchers in the Institute.
We next examined the predictors of relationships among the full 81
members between 2005 and 2010. We found that shared HBI theme predicted
all professional relationships among this group. Also important were
shared CIHR pillar and shared department. Shared gender also mattered
for advice seeking in this group.
An interesting finding in the 2005-2010 data is that office
co-location also predicts almost all the professional relationships
(except co-teaching among the full 81 members). For relationships among
the 52 original members before 2005, office co-location only affected
advice seeking and sharing grants. It is important to note that since
offices at the HBI are not assigned by department (and departmental
co-membership is controlled for in these models), the effect of office
co-location is not a disciplinary effect. These findings mirror recent
findings in the literature on 'networked organizations' that
suggest that even within organizations with an international reach,
where workers use communication technology to collaborate across
geographic distance, communication with co-located co-workers is
considered most important (Olson and Olson, 2010; Quan-Haase and
Wellman, 2004). A university or institute may want to consider
capitalizing on such findings by relocating members so that all of their
offices are closer together, perhaps in a dedicated building.
Collaboration may increase even further if all members' offices
were all in the same location.
Our final research question asked whether the appointed HBI leaders
occupy leadership positions in the professional interaction networks
between 2005 and 2010. Using measures of degree centrality, we found
that leaders are more active, and therefore more central in both the
advice and the supervision networks. Using betweeness centrality, we
found that the leaders are more central in each of the professional
relationship networks. The HBI leaders are clearly serving as bridges in
the Institute, linking members to one another through the various
professional relationships. Recently, bibliometric and organizational
analyses have begun using measures of betweeness centrality (for
journals, for authors, and for organizations) to measure the degree of
interdisciplinarity of a journal, an individual author, or an
organization (Cassi, Corrocher, Malerba, and Vonortas, 2008; Wagner et
al., 2011). We argue that in the HBI context, the high betweeness
centrality of the leaders indicates that they are promoting the
interdisciplinary goals of the Institute.
Limitations
We recognize that our data and analyses are limited by several
factors. First, and most importantly, we are presenting results from a
single case study. We do not have data from a comparable group of
scientists (either another group of researchers at UCalgary or
neuroscientists at another institution). The increases in professional
interactions that we observe over time among the HBI members may simply
reflect a historical trend of increased interdisciplinary research
(Rimer and Abrams, 2012), and may have occurred even without the
establishment of the HBI.
Second, although our response rates are high (81% for the pre-HBI
network and 84% for the post-HBI network), we did not receive survey
data from all the members of the HBI. In any network survey, missing
data issues are compounded by the fact that those who do not take the
survey are not only eliminated as respondents, but they are also
eliminated as potential ties. We do not know the extent to which those
who did not respond differ from our respondents (for example, in
engagement with others), and thus cannot comment on the
representativeness of our sample.
Third, we may be underestimating the number of pro fessional
interactions among these researchers by only including reciprocal ties
in our analyses. However, we have more confidence in the self-report
data by using only the confirmed ties. Additionally, it is important to
note that we do not have any information on professional interactions
with researchers outside of UCalgary for either time period. We stress
that we recognize that sometimes the most productive collaborative
relationships occur cross-institution, and, as yet, we have no data on
such relationships.
Finally, we must address the many ways in which time may affect our
findings. First, the pre-2005 data may suffer from recall issues.
Several respondents reported to us that it was sometimes hard for them
to remember relationships they had before 2005 (especially those that
did not result in a tangible outcome such as a publication). Second, we
have no information on people who were members of the HBI when it was
founded in 2004 but left the university before our survey in 2010.
Third, the neuroscientists who joined the UCalgary since 2004 joined
knowing that they would become HBI members. Thus, the newer Institute
members may be pre-disposed to have more professional interactions with
other members of the HBI than those who automatically became members in
2004.
Conclusion
Despite the limitations noted above, this study serves as an
example of the use of social network analysis to evaluate the effects of
the establishment of a multidisciplinary research institute on
scientists' working relationships in a Canadian University setting.
We believe our analyses demonstrate that the establishment of the HBI
has fostered, encouraged and increased interdisciplinary professional
interactions among neuroscientists at UCalgary.
The HBI is emerging as a Canadian centre for neuroscience research
and has received substantial funding since its inception. By all
traditional measures (publication output, grant funding, and
international reputation), the H BI is a success (Haustein et al.,
2013). Social network analysis enables us to assess the additional
impact of institute membership on scientific relationships among
members.
Reflecting back on the literature on successful collaborative
research institutes, four factors were seen as most important:
communication, infrastructure, shared vision, and leadership. The
HBI's success rests on the integration of these four factors.
Communication among the members is facilitated through regular meetings.
The institute has recently received another large donation, further
strengthening its already strong infrastructure and funding environment.
Shared vision appears to be working through the three thematic research
areas, which are drawing scientists together across disciplines and
departments. And the leaders appear to be playing a strong leadership
role, bringing together researchers across all the professional
relationships.
This analysis has enabled us to quantify the success of the HBI in
more than just financial terms or the co-authorship of publications. The
Institute has been successful in promoting professional interactions
among its members. By allowing the natural emergence of research themes,
the Institute leadership built on existing interactions and
relationships among scientists and encouraged the further formation of
professional ties across disciplines.
Caption: Figure 1. Co-authorship network, Pre-2005, by Department
(N = 52)
Caption: Figure 2. Co-authorship network, 2005-2010, by Department
(N = 81)
Caption: Figure 3. Co-authorship network, 2005-2010, by HBI
research theme (N = 81)
Caption: Figure 4. Advice Network, 2005-2010 (N = 81)
Appendix--Network Survey Questions
Part One (Sent to all current (2010) HBI members).
The following six questions ask about your professional
relationships with other full members of the HBI during the time period
from January 2005 to the present.
Please note that we have provided 20 boxes for names under each
question. You may fill as many or as few boxes as necessary under each
question.
NOTE: For each box, respondents were provided with a drop down menu
that listed the names of all of the current (2010) members of the HBI,
in alphabetical order by last name.
1. With whom have you held grants between January 2005 and the
present?
2. With whom have you co-authored papers between January 2005 and
the present?
3. With whom have you taught in a course (graduate or
undergraduate) between January 2005 and the present?
4. With whom have you co-supervised students, fellows or trainees,
or served on student supervisory committees between January 2005 and the
present?
5. With whom have you co-organized conferences / symposia between
January 2005 and the present?
6. From whom have you sought professional advice or mentorship
between January 2005 and the present?
Part Two (Sent to all original (2005) HBI members).
The following six questions ask about your professional
relationships with other full members of the HBI prior to the formation
of the HBI. Please report on any relationships you had with your
colleagues BEFORE January 2005.
Please note that only your colleagues who also became members of
the HBI in 2004 are listed for these six questions.
Please note that we have provided 20 boxes for names under each
question. You may fill as many or as few boxes as necessary under each
question.
NOTE: For each box, respondents were provided with a drop down menu
that listed the names of all of the current (2010) members of the HBI,
in alphabetical order by last name.
1. With whom had you held grants prior to January 2005?
2. With whom had you co-authored papers prior to January 2005?
3. With whom had you taught in a course (graduate or undergraduate)
prior to January 2005?
4. With whom had you co-supervised students, fellows or trainees,
or served on student supervisory committees prior to January 2005?
5. With whom had you co-organized conferences / symposia prior to
January 2005?
6. From whom had you sought professional advice or mentorship prior
to January 2005?
References
Aboelela, S. W., Merrill, J. A., Carley, K. M., Larson, E., 2007.
Social network analysis to evaluate an interdisciplinary research
center. Journal of Research Administration 38(1), 61-78.
http://go.galegroup.com/ps/i.do?id=GALE%
7CA174325643&v=2.1&u=ucalgary
&it=r&p=AONE&sw-w&asid=5bf330ealc251431e5d81c871b4f8721
Adams, J.D., Black, G.C., Clemmons, J.R., Stephan, EE., 2005.
Scientific teams and institutional collaborations: Evidence from US
universities, 1981-1999. Research Policy 34, 259-285.
http://dx.doi.org/10.1016/j.respol.2005.01.014
Bammer, G., 2008. Enhancing research collaboration: Three key
management challenges. Research Policy 37, 875-887.
http://dx.doi.org/10.1016%2Fj.respol.2008.03.004
Barabasi, A.L., Jeong, H., Neda, Z., Ravasz, E., Schubert, A.,
Vicsek, T., 2002. Evolution of the social network of scientific
collaborations. Physica A311(3-4), 590-614. http://
dx.doi.org/10.1016%2FS0378-4371%2802%2900736-7
Birnholtz, J. P., 2007. When do researchers collaborate: Toward a
model of collaboration propensity. Journal of the American Society for
information Science and Technology 58, 2226-2239.
http://dx.doi.org/10.1002%2Fasi.20684
Boardman, EC., Corley, E.A., 2008. University research centers and
the composition of research collaborations. Research Policy 37, 900-913.
http://dx.doi.org/10.1016%2Fj. respol.2008.01.012
Borgatti, S. P., Everett, M. G., Freeman, L. C., 2002. UCINET for
Windows: Software for Social Network Analysis. Harvard, MA: Analytic
Technologies.
Braun, T., Glanzel, W., Schubert, A., 2001. Publication and
cooperation patterns of the authors of neuroscience journals.
Scientometrics 51,499-510.
Brewer, D.D., Webster, C.M., 1999. Forgetting of friends and its
effects on measuring friendship networks. Social Networks 21, 361-373.
http://dx.doi.org/10.1016%2 FS0378-8733%2899%2900018-0
Burris, V., 2005. Interlocking directorates and political cohesion
among corporate elites. American Journal of Sociology 111,249-283.
http://dx.doi.org/10.1086%2F428817
Carley, K.M., Krackhardt, D., 1996. Cognitive inconsistencies and
non-symmetric friendship. Social Networks 18, 1-27.
http://dx.doi.org/10.1016% 2F0378-8733%2895%2900252-9
Cassi, L., Corrocher, N., Malerba, E, Vonortas, N. 2008. Research
networks as infrastructure for knowledge diffusion in European regions.
Economic Innovations and New Technologies, 17, 663-676.
http://dx.doi.org/10.1080%2F 10438590701785603
The Canadian Institutes of Health Research (CIHR) National
Consultation Summary Report. July 2009. Retrieved from
http://www.cihr-irsc.gc.ca/e/39976.html
Chompalov, I., Genuth, J., Shrum, W., 2002. The organization of
scientific collaborations. Research Policy 31, 749-767.
http://dx.doi.org/10.1016%2 FS0048-7333%2801%2900 145-7
Chompalov, I, & Shrum, W., 1999. Institutional collaboration in
science: A typology of technological practice. Science, Technology,
& Human Vaktes 24(3), 338-372. http://
dx.doi.org/10.1177%2F016224399902400302
Cowan, R., Jonard, N., 2004. Network structure and the diffusion of
knowledge. Journal of Economic Dynamics and Control, 28, 1557-1575.
http://dx.doi.org/10.1016%2Fj. jedc.2003.04.002
Cummings, J. N., Keisler, S., 2005. Collaborative research across
disciplinary and organizational boundaries. Social Studies of Science
35, 703-722. http://dx.doi. org/10.1177%2F0306312705055535
Dimitrova, D., Koku, E., 2010. Managing collaborative research
networks: The dual life of a virtual community of practice.
International Journal of Virtual Communities and Social Networking 2(4),
1-22. http://dx.doi.org/10.4018%2Fjvcsn.2010100101
Doty, R.W., 1987. "Neuroscience." Chapter 18 in History
of the American Physiological Sociegy: The First Century 1887- 1987.
Edited by J.R. Brobeck, O.E. Reynolds, and T.A. Appel. Pp. 427-434.
Emirbayer, M., 1997. Manifesto for a relational sociology. American
Journal of Sociology 23, 281-317. http://dx.doi.org/10.1086%2F231209
Freeman, L.C., 1979. Centrality in social networks I: concept
clarification. Social Networks 1, 215-39.
Friedkin, N. E., 1982. Information flow through strong and weak
ties in interorganizational social networks. Social Networks 3, 273-285.
http://dx.doi.org/10.1016%2F03788733%2882%2990003-X
Genuth, J., Chompalov, I., Shrum, W., 2000. How experiments begin:
The formation of scientific collaborations. Minerva 38, 311-348.
Geiger, R.L., 1990. Organized research units: Their role in the
development of university research. The Journal of Higher Education 61
(1), 1-19. http://dx.doi. org/10.2307%2F 1982031
Godley, J., Barron, G., Sharma, A., 2011. Using social network
analysis to assess collaboration in health research. The Journal of
Healthcare, Science and the Humanities 1(2), 99-116.
http://issuu.com/navynmedicine/docs/jhsh_vol_1_no_1
Hackett, E. J., 2005. Special guest edited issue on Scientific
Collaboration. Social Studies of Science 35, 667-671.
Hagen, N.A., Stiles, C.R., Biondo, ED., Cummings, G.G., Fainsinger,
R.L., Moulin, D.E., Pereira, J.L., Spice, R., 2011. Establishing a
multicentre clinical research network: lessons learned. Current Oncology
18(5), e243-e249, http://dx.doi.org/10.3747%2Fco.v 18i 5.814
Haines, V.A., Godley, J., Hawe, P., 2011. Understanding
interdisciplinary collaborations as social networks, american Journal of
Community Psychology 47, 1-11. http://dx.doi. org/10.1007%2Fs
10464-010-9374-1
Harris, M., 2010. Interdisciplinary strategy and collaboration: a
case study of American research universities. Journal of Research
Administration 41(1), 22-39. http://go.galegroup.
com/ps/i.do?id=GALE%7CA223281015&v=2.1&u=ucalgary&it=r&p= AONE&sw=w&asid=9654e61a0135bea38678aa009a4e1f39
Haustein, S., Cote, G., Beaudet A., 2013. State of Knowledge
Production in Neuroscience in Alberta: A Bibliometric Assessment.
Montreal: Science-Metrix. [Unpublished Report]
Huutoniemi, K., Klein, J.T., Bruun, H., Hukkinen, J., 2010.
Analyzing interdisciplinarity: Typology and indicators. Research Policy
39, 79-88. http://dx.doi.orgl/10.1016%2Fj. respol.2009.09.011
Jha, Y., Welch, E.W., 2010. Relational mechanisms governing
multifaceted collaborative behavior of academic scientists in six fields
of science and engineering. Research Policy 39, 1174-1184.
http://dx.doi.org/10.1016%2Fj.respol.2010.06.003
Johnson, J.C., Christian, R.R., Brunt, J.W., Hickman, C.R., Waide,
R.B., 2010. Evolution of collaboration within the US Long Term
Ecological Research Network. BioScience 60(1), 931-940.
http://dx.doi.org/10.1525%2Fbio.2010.60.11.9
Katerndahl, D., 2012. Evolution of the research collaboration
network in a productive department. Journal of Evaluation in Clinical
Practice 18, 195-201. http://dx.doi.
org/10.1111%2Fj.1365-2753.2011.01791.x
Katz, J.S., Martin, B.R., 1997. What is research collaboration?
Research Policy 26,1-18.
http://dx.doi.org/10.1016%2FS0048-7333%2896%2900917-1
Koehly, L.M., Pattison, P., 2005. Random graph models for social
networks: Multiple relations or multiple raters. In: Carrington PJ,
Scott J, Wasserman S (eds). Models and Methods in Social Network
Analysis. (pp. 162-191). New York: Cambridge University Press.
Krackhardt, D., Porter, L., 1986. The snowball effect: Turnover
embedded in communication networks. Journal of Applied Psychology 71,
50-55. http://dx.doi.org/10.1037%2F% 2F0021-9010.71.1.50
Levitt, J.M., Thelwall, M., 2008. Is multidisciplinary research
more highly cited? A macrolevel study. The Journal of the American
Society for Information Science and Technology 59(12), 1973-1984.
http://dx.doi.org/10.1002%2Fasi.20914
Luke, D.A., Harris, J.K., 2007. Network analysis in public health:
history, methods and applications. Annual Review of Public Health 28,
69-73. http://dx.doi.org/10.1146%2Fannurev.publhealth.28.021406.144132
Marsden, P. V., 1990. Network data and measurement. Annual Review
of Sociology 16, 435-463. http://dx.doi.org/10.1146%2Fannurev.soc.
16.1.435
Mattson, P., Laget, E, Vindefjard, A.N., Sundberg, C.J., 2010. What
do European research collaboration networks in life sciences look like?
Research Evaluation 19(5), 373-384.
http://dx.doi.org/10.3152%2F095820210X12809191250924
Meyer, R.E., Fabor, T.E, Hesselbrock, V., 1988. An alcohol research
center in concept and practice: Interdisciplinary collaboration at the
UConn ARC. British Journal of Addiction 83, 245-252.
http://dx.doi.org/10.1111%2Fj.1360-0443.1988.tb00464.x
Olson, G., Olson, J., 2010. Mitigating the effects of distance on
collaborative intellectual work. Economics of Innovation and New
Technology 12(1), 27-42. http://dx.doi. org/10.1080%2F10438590303117
Ponomariov, B.L., Boardman, EC., 2010. Influencing scientists'
collaboration and productivity patterns through new institutions:
University research centers and scientific and technical human capital.
Research Policy 39, 613-624. http://dx.doi.
org/10.1016%2Fj.respol.2010.02.013
Quan-Haase, A., Wellman, B., 2004. Local virtuality in a high-tech
networks organization. Analyse und Kritik 28, 1-17.
Quinlan, K.M., Kane, M., Trochim, W.M.K., 2008. Evaluation of large
research initiatives: Outcomes, challenges and methodological
considerations. In C.L.S. Coryn and M. Scriven (Eds.), Reforming the
evaluation of research. New Directions for Evaluation, 118, 61-72.
http://dx.doi.org/10.1002%2Fev.261
Rimer, B.K., Abrams, D.B., 2012. Present and future horizons for
transdisciplinary research. American Journal of Preventive Medicine, 42,
200-201. http://dx.doi.org/10.1016%2Fj.amepre.2011.10.010
Rosenfield, P. L., 1992. The potential of transdisciplinary
research for sustaining and extending the linkages between the health
and social sciences. Social Science & Medicine 35, 1343-1357.
http://dx.doi.org/10.1016%2F0277-9536%2892%2990038-R
Schensul, J. J., Robison, J., Reyes, C., Radda, K., Gaztambide, S.,
Disch, W., 2006. Building interdisciplinary/intersectoral research
partnerships for community-based mental health research with older
minority adults. American Journal of Community Psychology 38, 79-93.
http://dx.doi.org/10.1007%2Fs 10464-006-9059-y
Scott, J., 2000. Social Network Analysis: a handbook (second
edition). London: Sage.
Smith-Lovin, L., 1999. Core concepts and common ground: The
relational basis of our discipline. Social Forces 78, 1-23.
http://dx.doi.org/10.2307%2F3005788
Snijders, T. A. B., Borgatti, S. P., 1999. Non parametric standard
errors and test for network statistics. Connections 22(2), 1-11.
http://www.stats.ox.ac.uk/-snijders/Snijders_Borgatti.pdf
Sun, S., Manson, S.M., 20l 1. Social network analysis of the
academic GIScience community. The Professional Geographer 63(1), 18-33.
http://dx.doi.org/10.1080%2F00330124.2010.533560
Van Rijnsoever, EJ., Hessles, L.K., 2011. Factors associated with
disciplinary and interdisciplinary research collaboration. Research
Policy 40, 463-472. http://dx.doi.org/10.1016%2Fj.respol.2010.11.001
Wagner, C. S., 2005. Six case studies of international
collaboration in science. Scientometrics 62, 3-26.
http://dx.doi.org/10.1007%2Fs 11192-005-0001-0
Wagner, C. S. & Leydesdorff, L., 2005. Network structure,
self-organization, and the growth of international collaboration in
science. Research Policy 34, 1608-1618. http://dx.doi.org/10.1016%2Fj,
respol.2005.08.002
Wagner, C. S., Roessner, J.D., Bobb, K., Klein, J.T., Boyack, K.W.,
Keyton, J., Rafols, I., Borner, K. 201 I. Approaches to understanding
and measuring interdisciplinary scientific research (IDR): A review of
the literature. Journal of Infometrics 165, 14-26.
http://dx.doi.org/10.1016%2Fj.joi. 2010.06.004
Wasserman, S., Faust, K., 1994. Social Network Analysis: Methods
and Applications. New York: Cambridge University Press.
Yang, E.H., Park, H.W., Heo, J., 2010. A network analysis of
interdisciplinary research relationships: the Korean government's
R&D grant program. Scientometrics 83, 77-92.
http://dx.doi.org/10.1007%2Fs11192-010-0157-0
Author's Note
Keith A. Sharkey is the recipient of a Killam Professorship and is
the Crohn's and Colitis Foundation of Canada Chair in Inflammatory
Bowel Disease Research. Samuel Weiss is an Alberta Innovates-Health
Solutions Medical Scientist. Correspondence concerning this article
should be addressed to Jenny Godley, PhD, Dept. of Sociology, University
of Calgary, 2500 University Dr NW, Calgary, AB T2N 1N4, Canada, Email:
jgodley@ucalgary.ca
Jenny Godley, PhD
Dept. of Sociology
Faculty of Arts
University of Calgary
2500 University Dr NW
Calgary, AB T2N 1N4 Canada
Email: jgodley@ucalgary.ca
Tel: (403) 220-7566
Fax: (403) 282-9298
Keith A. Sharkey, PhD
Hotchkiss Brain Institute
Dept. of Physiology and Pharmacology
University of Calgary
Email: ksharkey@ucalgary.ca
Samuel Weiss, PhD
Hotchkiss Brain Institute
Dept. of Cell Biology & Anatomy
University of Calgary
3330 Hospital Dr. NW
Calgary, AB T2N 4N 1 Canada
Email: weiss@ucalgary.ca
Jenny Godley
University of Calgary
Keith A. Sharkey
University of Calgary
Samuel Weiss
University of Calgary
Table 1. Sample Composition
Pre-HBI Post-HBI
(N = 52) (N = 81)
F % F %
Gender
Male 46 88.5 64 79
Female 6 11.5 17 21
CIHR Pillar
Biomedical 35 67.3 46 56.8
Clinical 14 26.9 28 34.6
Population / Health 3 5.8 7 8.6
Services
Theme
Axon Biology & 10 19.2 18 22.2
Regeneration
Cerebral Circulation 12 23.1 18 22.2
Neural Systems & 30 57.7 45 55.6
Behaviour
Department
Psychiatry 2 3.8 6 7.4
Psychology 3 5.8 8 9.9
Physiology & 13 25 14 17.3
Pharmacology
Clinical 19 36.5 28 34.6
Neurosciences
Radiology 5 9.6 6 7.4
Cell Biology & 6 11.5 8 9.9
Anatomy
Community Health 2 3.8 4 4.9
Sciences
Other 2 3.8 7 8.6
Office Location
Main 4 7.7 10 12.3
FHH 16 30.8 25 30.9
HM 7 13.5 11 13.6
HRIC 10 19.2 12 14.8
HSC 12 23.1 16 19.8
TRW 3 5.8 7 8.6
Rank
Full 33 63.5 42 51.9
Associate 13 25.0 20 24.7
Assistant 6 11.5 19 23.4
Year Joined
2004 52 100 52 64.2
2005 5 6.2
2006 3 3.7
2007 6 7.4
2008 9 11.1
2009 6 7.4
Table 2. Density and Reciprocity of Networks Over Time
Pre-HBI (N = 52) Post-HBI (original 52)
Density Reciprocity Density Reciprocity
Advice .040 * .10 .072 * .20
Conferences .009 .17 .012 .42
Grants .019 * .61 .032 * .55
Papers .046 .68 .051 .71
Supervise .059 .42 .076 .51
Teach .044 .40 .050 .35
Post-HBI (N = 81)
Density Reciprocity
Advice .047 .16
Conferences .009 .34
Grants .024 .62
Papers .034 .73
Supervise .050 .50
Teach .033 .42
Note: * Densities are significantly different at p<.05
Table 3. Ties Over Time
Number of Ties Isolates
Average Std.Dev. Maximum Number %
Pre-HBI (N=52)
Advice (In) 3.62 3.34 14 23 44
Advice (Out) 2.84 1.57 7 15 29
Conference 1.00 0.00 1 48 92
Grants 1.36 .658 3 30 58
Papers 2.40 1.67 8 17 33
Supervise 2.28 1.25 6 23 44
Teach 1.70 1.10 5 25 48
Post-HBI (N=81)
Advice (In) 4.84 5.31 27 19 23
Advice (Out) 5.00 4.08 18 21 26
Conference 1.18 0.39 2 64 79
Grants 2.33 1.73 7 39 48
Papers 2.77 2.00 11 24 30
Supervise 2.83 1.99 10 23 28
Teach 2.20 1.71 8 40 49
Note: All networks contain only reciprocal ties except far the advice
network. Advice (In) indicates reports that others sought advice
from the member. Advice (Out) indicates that the member reports
seeking advice from others.
Table 4. Network Composition Over Time
% Same % Same % Same Pillar % Same Theme
Gender Dept. (CIHR) (HBI)
Pre-HBI (N=52)
Advice (In) 45 23 37 --
Advice (Out) 58 30 50 --
Conference 100 50 100 --
Grants 93 33 55 --
Papers 86 42 68 --
Supervise 77 31 76 --
Teach 77 56 82 --
Post-HBI (N=81)
Advice (In) 72 50 70 54
Advice (Out) 74 50 67 61
Conference 88 53 74 77
Grants 67 41 57 61
Papers 70 40 68 63
Supervise 76 41 78 74
Teach 64 45 87 66
Notes: All networks contain only reciprocal ties, except for the
advice networks. HBI themes were not formalized until 2010, thus use
do not calculate percent same theme for the earlier networks.
Table 5. QAP regression results, PRE-2005 and 2005-2010 Original
Members (N = 52)
Original Members (N = 52)
PRE-2005 Advice Grants Papers Supervise
Same department .067 ** 0.023 .077 ** 0.023
(0.005) (0.201) (0.010) (0.213)
Same gender 0.013 0.040 0.047 -0.006
(0.376) (0.080) (0.106) (0.405)
Same pillar 0.035 -0.019 0.007 .082 **
(0.110) (0.265) (0.404) (0.007)
Same office .049 * .057 * 0.037 -0.007
(0.019) (0.031) (0.113) (0.412)
Same rank .040 * 0.021 .058 * .052 *
(0.029) (0.237) (0.030) (0.038)
[R.sup.2] (adjusted) 0.012 0.005 0.014 0.009
PRE-2005 Advice Grants Papers Supervise
Same department .109 ** 0.052 .057 * .090 **
(0.001) (0.059) (0.039) (0.003)
Same gender 0.024 0.018 0.034 0.008
(0.292) (0.318) (0.167) (0.413)
Same pillar .089 ** -0.031 0.010 .088 **
(0.005) (0.152) (0.371) (0.002)
Same theme .075 ** .055 * .074 ** .098 **
(0.006) (0.032) (0.006) (0.001)
Same office .097 ** .077 ** .106 ** .064 *
(0.001) (0.010) (2) (0.019)
Same rank .065 * .047 * 0.040 0.002
(0.028) (0.023) (0.087) (0.467)
[R.sup.2] (adjusted) 0.049 0.018 0.024 0.036
Original Members (N = 52)
PRE-2005 Teach Multiplex
Same department .088 ** .088 **
(0.006) (0.001)
Same gender 0.004 0.031
(0.503) (0.225)
Same pillar .064 * 0.053
(0.025) (.069
Same office 0.009 .061 *
(0.368) (0.022)
Same rank 0.019 .063 *
(0.249) (0.013)
[R.sup.2] (adjusted) 0.013 0.024
PRE-2005 Teach Multiplex
Same department .086 ** .128 **
(0.004) (0.001)
Same gender -0.036 0.020
(0.143) (0.283)
Same pillar .080 ** .084 **
(0.008) (0.009)
Same theme 0.037 .111 **
(0.106) (0.001)
Same office .075 ** .134 **
(0.008) (0.001)
Same rank 0.017 .067 *
(0.278) (0.016)
[R.sup.2] (adjusted) 0.027 0.072
Notes:
* significant at the 0.05 level
** significant at the.001 level
Standardized coefficients and proportion significance in parentheses.
All the networks are reciprocal ties only, except for Advice, which
includes non-reciprocal ties.
All the networks are dichotomized, except for the multiplex network,
which is a valued network (0-5).
For the multiplex network, the coefficients are predicting the
strength of the tie, rather than just the existence of a tie.
Table 6. QAP regression results--2005-2010
All Members (N = 81)
Advice Grants Papers Supervise
Same department .100 ** .050 * .056 * .061 **
(.001) (.018) (.011) (.004)
Same gender .045 * -.003 .009 .032
(.036) (.466) (.328) (.058)
Same pillar .080 ** -.015 .020 .096 **
(.001) (.262) (.160) (.001)
Same theme .075 ** .045 * .066 ** .095 **
(.001) (.017) (.002) (.001)
Same office .075 ** .071 ** .085 ** .046 *
(.001) (.001) (.001) (.013)
Same rank .022 .042 * .034 * .008
(.118) (.027) (.050) (.328)
[R.sup.2] (adjusted) .040 .012 .020 .031
All Members (N = 81)
Teach Multiplex
Same department .044 * .116 **
(.023) (.001)
Same gender -0.028 .028
(.111) (.134)
Same pillar .099 ** .091 **
(.001) (.001)
Same theme .043 * .112 **
(.022) (.001)
Same office 0.034 .106 **
(.052) (.001)
Same rank -.012 .031
(.293) (.057)
[R.sup.2] (adjusted) .019 .063
Notes:
* significant at the 0.05 level
** significant at the .001 level
Standardized coefficients and proportion significance in parentheses.
All the networks are reciprocal ties only, except for Advice, which
includes non-reciprocal ties.
All the networks are dichotomized, except for the multiplex network,
which is a valued network (0-5).
For the multiplex network, the coefficients are predicting the
strength of the tie, rather than just the existence of a tie.
Table 7. Comparing Degree and Betweeness Centrality Measures for HBI
Leaders and Members
Degree Centrality
Members Leaders
Mean (S.E.) N Mean (S.E.) N
Advice (In) 3.63 * 51 10.45 * 11
(.499) (2.46)
Advice (Out) 4.46 * 50 7.70 * 10
(.540) (4.47)
Grants 2.06 32 3.20 10
(.265) (.696)
Papers 2.60 47 3.60 10
(.294) (.562)
Supervise 2.45 * 47 4.45 * 11
(.204) (.957)
Teach 2.31 32 1.78 9
(.325) (.364)
Betweeness Centrality
Members Leaders
Mean (S.E.) Mean (S.E.)
N = 70 N = 11
Advice (In) .420 * 4.38 *
(.128) (1.92)
Advice (Out) .420 * 4.38 *
(.128) (1.92)
Grants .506 * 2.52 *
(.134) (.891)
Papers 1.68 * 4.27 *
(.349) (1.36)
Supervise 1.34 * 4.40 *
(.231) (1.36)
Teach 1.29 * 3.22 *
(.335) (1.16)
* difference between members and leaders significant at p<.05
