Benefits of a mechanistic framework for the ERH
We describe three ways that our proposed framework can increase understanding of the ERH, including predictive understanding.
Explain inconsistencies within and between studies
Context increases variation in evidence for the ERH, both within and between studies. If context is appropriately reported, then this variation has the potential to be informative, as the above reanalysis example shows (Fig. 6). As another example, by explicitly considering the local environment and resource availability of populations ofClidemia hirta , Dewalt et al. (2004) found evidence that populations in its invaded Hawaiian range had experienced enemy release, but only when compared with understory sites in its home range. This was likely because open sites in its home range provided enough resources (light) for C. hirta to compensate for enemy damage (DeWaltet al. 2004). Due to this variation, Jeschke & Heger (2018) list this study’s evidence for the ERH as “undecided”, as the current conception of the ERH is not nuanced enough to capture these sorts of dynamics. In contrast, our framework emphasises that reporting key ecological contexts is vital for contextualising the results of individual studies and provides a checklist of these contexts. The Dewalt et al. (2004) example highlights apparent inconsistencies within the same study system but shows that these inconsistencies and ‘undecided’ nature of the results are explicable when ecological context (resource availability) is considered (Box 2; Fig. 4). Not all studies can be carried out on large scales encompassing multiple contexts (or broad ranges of a single context), nor is it easy to study all three factors that underpin the ERH (Fig. 1b). However, the factors and contexts pertinent to each study must be considered and reported as explicitly as possible (Catford et al. 2022). The near-absence of several contexts from ERH meta-analyses to date (i.e. time since introduction, asynchronicities, number of introductions; Fig. 6) would at least partly reflect lack of information in the primary literature.
Reduce redundancy in invasion hypotheses
At least eight enemy-related hypotheses have been proposed in invasion ecology (Table 2; see also Catford et al. 2009). Our framework effectively integrates these hypotheses by showing that they share the same set of factors and contexts, and represent variations of the same underlying concepts (Jeschke 2014; Enders et al. 2018) (Table 2). For example, ‘enemy release’ and ‘enemy reduction’ functionally represent the same process, just to a different degree. If the set of factors or conditions underlying different hypotheses are not integrated, we are left with a “heady mix of acronyms” (Evans 2008) and a suite of loosely related hypotheses with little way of predicting when one is more likely than another. Highlighting possible conditions that lead to each hypothesis (Table 2) should facilitate better predictions of when, where and how exotics may benefit from the ERH.
Identify gaps in meta-analyses and meta-syntheses and reduce unexplained variation
The power of meta-analyses and -syntheses can be improved by explicitly incorporating the factors and contexts of the ERH. A meta-analysis’ power is hampered when spurious variation is introduced, leading to inconsistencies between different meta-analyses (Table S1). A key goal of meta-analyses is to attribute this variation to meaningful predictors (Spake et al. 2022). To comprehensively test the ERH, we suggest that a meta-analysis of the ERH should consider evidence of one metric for at least two of the three factors (Table 1b) and report them as discrete effect sizes. Meta-analyses that comprehensively test one factor are still valuable, but they should be reported as partialexaminations of the ERH, and should not combine results from studies that focus on different factors or metrics into a single effect size. When information is available, contexts should be included as moderator variables to better reveal generalities (i.e., parameterising Fig. 3) and reduce ‘unexplained’ variance (Fig. 6), both within and between meta-analyses. We refer readers to a rich literature that provides specific guidance on effective strategies and approaches to account for within- and between-study variance and for incorporating context into studies, including meta-analysis and meta-synthesis (e.g. Koricheva & Gurevitch 2014; Doncaster & Spake 2018; Catford et al. 2022; Spake et al. 2022, 2023).