ment or a cooperative exchange between two public agencies. Although conXict
between two individuals or organizations could also count as a relationship,
network institutionalism tends to presume positive relationships. Informed by a
Durkheimian perspective on social solidarity, many network studies emphasize the
socialandaVectualbases of relationships. However, it is not always necessary
to assume that networks are solidaristic. Networks may be merely patterns of
interaction or connection. For instance, two stakeholder groups may interact
frequently in the context of a policy arena or the boards of two NGOs might
share the same directors. Such relationships do not necessarily produce social
solidarity and may be rife with conXict. But they imply the possibility that these
connections are conduits, even if inadvertent, for information, ideas, or resources.
Frequent interaction in a legislative committee, for example, might be the basis for
theXow of critical information (regardless of whether the actors involved have any
sense of mutual obligation). Interdependence oVers a third way to interpret
networks. For example, one lobbyist might have information that another lobbyist
needs or two nations might have extensive trading relations. This interdependence
may motivate them to engage inexchangerelationships with each other. Successful
exchange can, in turn, generate strong norms ofmutual obligationandreciprocity
(sometimes referred to as ‘‘generalized exchange’’). The prominence of bargaining
in political relationships makes this exchange approach to networks a natural one
for political science.
Granovetter ( 1985 ) has argued that social network approaches steer a course
between oversocialized (norm determined) and undersocialized (self-interest
determined) understandings of social behavior. From this perspective, social net-
works have both a social (aVectual) and instrumental (exchange) dimension. If the
neoclassical market exchange takes places at ‘‘arms-length,’’ we should expect little
loyalty in such relationships and we should not expect them to provide the basis for
the kind of trust or reciprocity necessary to produce exchange where goods are
ill-deWned or the timeframe for exchange is poorly speciWed. It is precisely the
social character of network relationships built on loyalty and mutual obligation
that allows us to think of them as social structures. Yet, Granovetter suggests,
social actors are not mindlessly governed by these social norms. An instrumental
calculus, mediated by social norms, remains at work in most social relationships.
A relationship between two actors (dyad) is the basic unit of any network.
However, network approaches are typically interested in sets of interconnected
dyadic relationships. The termnetworktypically refers to this aggregate of inter-
connected relationships. The simplest network therefore actually requires at least
three diVerent actors—a triad. Much of network analysis is concerned with the
global properties of a network as a single social structure—that is, as an aggrega-
tion of interconnected dyads. In network analytic terms, a typical organizational
hierarchy is one kind of network. Subordinates are connected to their superordin-
ates, who are in turn connected to their superordinates, until one reaches the top of
network institutionalism 77