Introduction
The idea that predators might influence prey non-consumptively by eliciting trait changes has a long history. Indeed, Darwin (1839) hypothesized that prey escape responses cost time and energy to maintain and, consequently, should attenuate in the absence of predators. Today, these predator-induced trait changes, or non-consumptive effects (NCEs), have a strong conceptual basis (Charnov et al . 1976; Lima & Dill 1990; Lima 1998) and may rival or even exceed direct predation in terms of their impacts on prey populations and ecosystems (Kotler & Holt 1989; Peacor & Werner 2001; Schmitz et al . 2004; Preisseret al . 2007). Once the purview of laboratory and short-term field experiments involving small-bodied taxa (Kotler 1984; Preisser et al . 2005; Weissburg et al . 2014), NCEs and their broader consequences are increasingly being explored in large vertebrate systems (e.g., Willems & Hill 2009; Burkholder et al . 2013; Basilleet al . 2015; Moll et al . 2016; Le Roux et al . 2018; Smith et al . 2019; Valeix et al . 2019). This expansion has shed new light on how NCEs manifest in communities of larger-bodied species. Yet, it has also revealed the challenges associated with their prediction in the field, underscoring the need for standardized methodology for evaluating these phenomena across species and environmental contexts (Ford & Goheen 2015; Prugh et al . 2019) and conceptual clarity (Peacor et al . 2013; Gaynor et al . 2019) to guide research.
A growing literature suggests that contingency in NCEs hinges on key properties of the organisms involved as well as the environments in which they interact (e.g., Preisser et al . 2007; Heithauset al . 2009; Creel 2011; Schmitz & Trussell 2016). Accordingly, there have been several recent calls for these properties to be characterized and leveraged to improve our understanding of the nature and consequences of NCEs within ecological communities (e.g., Cresswell 2008; Creel 2011; Peacor et al . 2013; Moll et al . 2017). Here, we address this need by: (i) conceptualizing the multi-stage process by which predators may trigger direct and indirect NCEs; (ii) reviewing key drivers of context dependence in NCEs; and (iii) synthesizing these drivers into a general framework for predicting both the nature and strength of direct NCEs. We then (iv) conclude with a prospectus for future work. Our review spans aquatic and terrestrial ecosystems, addresses invertebrates and vertebrates, and focuses on a prevalent form of prey trait plasticity that is often implicated in the transmission of NCEs, anti-predator behaviors . We emphasize, however, that many of the sources of context dependencies that we address likely also apply to other forms of predator-induced trait modification (e.g., prey development, morphology, and physiology).