Not all bridges are equal: brokerage roles and innovation in research networks
One of the most robust findings in the sociology of science is that researchers who bridge otherwise disconnected groups tend to be more innovative. The idea is intuitive: spanning structural holes gives you access to information and knowledge combinations that your peers — embedded in more homogeneous networks — never encounter.
But this framing hides a question that turns out to matter quite a lot: not all bridging positions are the same. A researcher who connects two groups they both belong to is doing something quite different from one who connects two groups on behalf of a third party. Both are brokers. But they face different constraints and different opportunities.
The brokerage taxonomy
Sociologist David Krackhardt and others have identified several distinct brokerage roles depending on who the broker connects and whether the groups involved share institutional membership:
- Coordinator: broker and both alters are from the same group
- Itinerant: broker and one alter from the same group, the other from outside
- Gatekeeper: information flows into the broker’s group from outside
- Representative: information flows out of the broker’s group
- Liaison: broker connects two groups neither of which they belong to
Each of these structures implies a different combination of access to novel knowledge and integration costs. Coordinating within a shared group is easy — everyone speaks the same language — but offers little in terms of novel recombination. Liaisons have access to diverse, non-overlapping knowledge, but the coordination costs are steep.
Balanced vs. unbalanced triads
In a paper I co-authored with Pablo D’Este (Journal of Product Innovation Management, 2022), we argued that what matters is not just whether you broker, but how balanced the triad you occupy is.
We call a triad balanced when the broker has some institutional overlap with each alter — enough shared ground to coordinate, but enough distance to encounter genuine knowledge diversity. In Krackhardt’s taxonomy, gatekeeper and itinerant roles fit this description. The liaison role, by contrast, is maximally diverse but also maximally costly to coordinate.
Using data from over 1,000 biomedical scientists, we found that balanced brokerage roles (gatekeeper and itinerant) were consistently associated with higher individual innovativeness — measured by involvement in patent applications, clinical trials, and commercialisation activities. The liaison role, despite offering the greatest structural access to diverse knowledge, showed weaker effects.
The interpretation: being a bridge is not enough. The bridge has to be one you can actually walk.
What this means for how scientists build networks
If you are a researcher deliberately thinking about your collaborative network — and there are good reasons to do so — these findings suggest a few things:
Seek diversity, but not at the expense of coordination capacity. Crossing institutional and disciplinary boundaries is valuable. But if the distance is so large that you cannot find common ground, the knowledge stays inaccessible.
Your institutional context shapes what works. We also found that the innovation benefits of balanced brokerage are especially pronounced for scientists embedded in settings distant from the application context — basic researchers who are more constrained by institutional norms and need the social resources that brokerage provides to overcome those constraints.
Network position is not just a structural accident. In later work (Research Policy, 2021), we explored how tertius iungens orientation — the active disposition to introduce and connect others — interacts with structural network properties. The finding: strategic behaviour matters. Two researchers in identical structural positions can get very different innovation outcomes depending on whether they actively mobilise the potential their position affords.
A note on measurement
One reason these distinctions matter less in practice than they should is that most analyses aggregate all brokerage into a single structural holes measure. The Burt (1992) constraint index is powerful, but it tells you how open a researcher’s ego-network is, not what kind of bridges they occupy.
Disaggregating by brokerage role requires knowing something about the group membership of each ego–alter pair, which means collecting richer data. Biomedical science is a good context for this because institutional affiliation maps reasonably well onto the institutional logic distinction (science vs. care vs. industry) that matters for medical innovation.
If you are working on related questions — network analysis in science, institutional logics, or individual innovation behaviour — I would be glad to hear from you. Drop me a line.