In this paper, we illustrate how multivariate exponential random graph models can be applied to draw inferences about the interplay of positive and negative ties, in the context of an environmental governance network. Structural balance theory (Cartwright and Harary, 1956) is the classic social network treatment of positive and negative ties but presupposes that the two types of tie are mutually exclusive. Yet ambivalent relationships (ties that are both positive and negative) are considered in other disciplines. Bivariate exponential random graph models (ERGMs) for two types of directed tie can be used to infer the presence of social processes involving entrainment (i.e. the two types tend to occur together) or exchange (i.e. the two types tend to be reciprocated). Entrainment of positive and negative ties can take on particular theoretical meanings in different research domains. Drawing on an influential theory of the structure of effective network governance (Jones et al, 1997), we argue that for environmental governance systems entrainment and/or exchange of positive and negative ties may indicate a failure in shared macro-culture among the actors. As an empirical example, we use an ERGM to analyse a network governance system of the Swan River, Western Australia, since changed by legislation. The analysis shows that the preconditions for effective governance proposed by Jones et al (relational and structural embeddedness, shared macroculture) were seemingly absent. Our conclusions indicate the importance of understanding both positive and negative ties in the study of social structure and effective action.