We study how knowledge is produced at the intersection of science, environmental policy and public engagement. Based on analysis of monarch butterfly conservation, we critically evaluate models of knowledge production. The monarch butterfly and its migration have engaged science and enchanted people for over a century, and current threats to monarchs catalyze debates and actions. This paper traces the historical development of knowledge regarding (i) long-term monarch population trends, (ii) the monarch's dependence on a particular food plant, the milk-weed, and (iii) the monarch as a pollinator. Our analysis indicates that knowledge production and science-policy-practice interfaces cannot be satisfactorily understood through reference to the classical linear model and more recent conceptions of relationally produced knowledge (i.e. co-production). We identify powerful and sometimes contradictory knowledge claims that emerge from unmediated interactions among scientists, advocates, policy makers and diverse publics. The emergent model complements existing models of knowledge production, thereby expanding the conceptual foundation available for making sense of science-policy-practice interfaces.In the past 30 years or so, an interactive, relational conception of knowledge production and social change has emerged. This multi-polar model highlights tensions and complementarities among scientists and variously situated local actors (i.e. practitioners, engaged citizens) (Rosenberg, 1982;Yearley, 2005). Applied to policy decisions, the stakes, uncertainties and fundamental importance of the particular values of individuals and social groups that characterize contemporary debates about the environment present a situation in which knowledge production should, it is argued, increasingly be understood and advanced through dialog among differently situated actors (Funtowicz and Ravetz, 1993; Hajer and Wagenaar, 2003). Acknowledgement of different knowledge practices and the importance of 'co-production' of knowledge represent an explicitly relational conception of and prescription for science, policy and development (Jasanoff, 2004;Palmer, 2012;Weber and Khademla, 2008). This commitment to interaction stands in contrast to the under-socialized, decontextualized conception of knowledge production at the heart of the linear model.Contemporary treatment of science-policy-practice interfaces recognizes linear and relational aspects of knowledge production as useful heuristics (e.g. van Kerkhoff and Lebel, 2006). In problematizing these models as ideal types and specifying their shortcomings, empirical studies highlight interdependence of linear and relational processes, as well as the significance of context and interests in shaping expressions of knowledge. For example, Cash et al. (2003) emphasize how credibility, relevance and legitimacy of knowledge in particular circumstances mediate how scientific knowledge comes to be employed for problem solving in policy domains. Sarkki et al. (2017) highlight how pragmatismi.e. the p...