Interdisciplinarity is essential to tackle complex scientific challenges. However, not all researchers are positioned equally to realize the potential of interdisciplinary collaborations. Network structure, the configuration of connections among researchers, is an important indicator of researchers' access to unique information, resources, and career opportunities.

Team science scholarship has shown that, beyond access to boundary-spanning relations, interdisciplinary teams must enact specific communication behaviors to realize the collaborative potential afforded
by spanning networks. Yet, institutional structures do not always recognize or reward this labor. Interdisciplinary research institutes are an organizational form established to enable scientists to build innovative, collaborative research. But, whose collaborations benefit the most from these resources? Understanding when and how research institute resources facilitate brokerage is essential to building evidence-based policy recommendations called for within a recent NASEM consensus report on Team Science.

We analyze longitudinal collaboration networks in a large interdisciplinary research institute we call the "Nature Institute" (a pseudonym) to examine how departmental conventionality (a measure of whether researchers joining an institute come from departments that are commonly or uncommonly represented among the institute's membership) shapes the development of boundary spanning relationships in researchers' careers. Based on prior literature, we expected that department conventionality might produce two differing (and competing) effects on how researchers may utilize institute resources: first, an institute's resources may best support work required of researchers from conventional departments and, as such, conventional researchers may be able to leverage those resources more effectively to develop their collaboration structures than those from rare departments. At the same time, researchers from rarer departments may see greater benefits to their collaboration networks by joining an institute full of individuals who have radically differing expertise from their own.

Using longitudinal publication data from 256 researchers over 17 years, we estimate whether researchers' collaboration networks become more redundant (high constraint, based on Burt's definition) or more bridging (low constraint) after joining one such institute. Generalized linear mixed effects models reveal that institute members experience a significant decline in network constraint over time compared to a control group of comparable scientists who did not become members of the institute. Moreover, among members, those from less conventional departments experienced a significantly lower constraint after joining than those from conventional departments. Non-members from a matched-case control group display similar changes in constraint regardless of their home departments. A significant three-way interaction shows that departmental conventionality
conditions the effect of membership: in less conventional departments, institute membership amplifies the decrease in network constraint, providing greater access to diverse and non-redundant co-authorship connections.

These findings hold direct implications for team science policy and practice. If our findings hold, they suggest to potentially counter-intuitive conclusion that team science institutes may have the greatest impact on research networks when members join from rarer departments, which creates actionable implications for policies ensuring that these researchers are supported for their brokerage roles in interdisciplinary collaboration. We will close by discussing these practical and policy implications in depth.