River restoration projects in gravel-bed rivers are becoming increasingly sophisticated and complex as river managers and scientists attempt to deliver the goals of catchment-scale ecosystem restoration. With increased sophistication, come the dual challenges of recognizing and responding to the uncertainty inherent in the restoration process. Uncertainty is rarely explicitly recognised in current restoration projects and, where it is, the scope and definition are limited. In this paper we argue that uncertainty is a fundamental element of river restoration and that the sources of uncertainty are varied. A typology for understanding and communicating uncertainty in terms of these sources is presented. One of the myths surrounding uncertainty is the notion that being uncertain is the same as not knowing anything. In fact, when uncertainty is expressed as a statement of plausible outcome and/or significance, expressing uncertainty is a very informative statement of knowledge. The significance of uncertainty is explored conceptually and quantified for two contrasting examples from two gravel-bed river restoration projects. Respectively, these demonstrate that uncertainty in the conceptual model applied to a restoration project can have significant impacts on the restoration process and that unreliability uncertainties can affect the design of bankfull channel dimensions. The paper concludes with a discussion of the approaches to incorporating uncertainty in river restoration projects, and argues for one that embraces uncertainty. We present an approach for embracing geomorphic uncertainty in physical habitat restoration, that uses coupled habitat and landscape evolution models to define the plausible outcomes for a given restoration project.