1. Theoretical models on mutualism dynamics predict that partner traits may influence the outcome of mutualistic interactions. However, most empirical data on this issue is restricted to case studies, limiting our ability to reach a more widespread comprehension of the role of partner traits on the dynamic of mutualisms.

2. We investigated how the outcome of protective mutualisms between ants and plants bearing extrafloral nectaries (EFNs) is influenced by the traits of EFNs and ants feeding on EFNs. We used a meta-analytical approach based on 35 studies investigating the effect of ant attendance on the herbivores and reproductive performance of EFN-bearing plants. We evaluated how variation in the EFN vascularization and location on plants and the ant aggressiveness can modulate the effect of ant attendance on the plants.

3. Both plant and ant traits investigated here drove the outcome of the protective mutualism for EFN-bearing plants. Plants exclusively bearing EFNs near reproductive organs benefited more from ant attendance than plants bearing EFNs on vegetative or vegetative and reproductive organs. Ants had a higher positive impact on the reproductive performance of plants bearing non-vascularized EFNs than plants bearing vascularized EFNs, although their effects on herbivores had been similar in both plant types. Regarding the ant behavior, plants often attended by more aggressive ant species had a higher reproductive performance than plants often attended by less aggressive ones.

4. Synthesis: Our results highlight that the selective pressures and evolutionary routes in ant-plant protective mutualisms may depend on the pool of traits exhibited by partner species. Although some studies have already reported some impact of species traits on the outcome of ant-plant mutualisms, this is the first time that a generalization about the role of species traits on the net balance of ant attendance was proposed. Due to this generalization, it was possible to advance our knowledge about the evolution of facultative mutualisms by showing that the role of species traits on the mutualistic outcome can vary in intricate ways due to a particular trait combination found among partners in communities where the interactions are embedded in.

We compilled information from studies providing empirical data about the ants’ impact on herbivores and/or on the performance of EFN-bearing plants. A database compiling studies that fit these requirements was provided by Leal & Peixoto (2017) in a previous meta-analysis that evaluated the role of water availability in driving the effect of ant attendance on the performance of EFN-bearing plants. Although their objective was different from ours, their database was built following a replicable and detailed protocol suitable for searches focusing on studies investigating the role of ant attendance on EFN-bearing plants. Since Leal & Peixoto’s (2017) database included studies published in English between 1970 and September 2015, we updated this database by searching for new studies published between October 2015 and December 2019. For that purpose, we searched for papers in Scopus (www.scopus.com) and ISI Web of Knowledge (www.isiknowledge.com – core collection) databases, using the following key terms: “extrafloral nectar*”, “ant-plant mutualism”, “plant AND biotic defen*e”. These keywords and search sites were the same as used by Leal & Peixoto (2017). We found 25 new studies published in English for this new search, added to the 126 studies compiled by Leal & Peixoto (2017).

After searching the literature, we screened the abstracts and methods of all papers to select the studies that fulfilled the following inclusion criteria: the studies must have (1) used plant species bearing natural EFNs as model organisms, discarding all the studies artificially manipulating EFN occurrence or extrafloral nectar availability to ants; (2) evaluated the role of ant attendance on plants by contrasting two groups of plants: one freely accessible by ant species (control group) and another group in which ants were naturally absent or experimentally excluded from plants; (3) used native plant species as a model to prevent confounding effects specific to invasive species (see Traveset & Richardson 2006); and (4) reported the sample size, the mean values, and the standard deviation or standard error for metrics related to herbivore performance or plant reproductive success in groups of plants in which ants were present or absent (see more information about these metrics below). After this screening, 38 studies remained in our database. However, we excluded from our dataset three studies that used as models two fern species: Pteridium aquilinum and Polypodium plebeium (Koptur et al 1998, Jones & Paine 2012, Rashbrook et al. 1992; see Table S2).  Then, we created the dataset presented here, including 132 effect sizes from 35 studies undertaken in 31 different locations and using 42 EFN-bearing plant species belonging to 20 families of angiosperms as model species.

All analysis were performed using R environment. All packages required for the analysis are detailed in the script file.