Species interactions often occur along a gradient, with much of the research pertaining to gradient extremes of mutual benefit or antagonism. Ant-seed interactions take several forms, including dispersal, predation (or granivory), and parasitism, whereby ants consume seed appendages without dispersal of seeds. These interactions might be dependent on the identity and corresponding traits of the participating plant and ant. We hypothesized that these interaction outcomes could be predicted by ant and plant traits and habitat, with outcomes falling along a gradient of cost and benefit to the plant. To test this hypothesis, we conducted a global literature review and classified over 6000 pairs of ant-seed interactions from 753 studies across six continents. For all appropriate references, the corresponding plant and seed characteristics were obtained for family, growth habit, life-form, dispersal mechanism, seed length, seed shape, and seed surface texture. The data were analyzed using seed and ant trait data as proxies for identity using logistic regressions with seed fate as the independent variable. Multiple correspondence analysis was used to elucidate the interactions of plant family and ant genus in relation to each other and seed fate and if those interactions are in accordance with our predictions.
Results/Conclusions
Linear models showed that seed and ant size, habitat, and dispersal syndrome were the most consistent predictors. Predation was less likely than parasitism and seed dispersal among myrmecochorous plants. A classification tree of the predicted outcomes from linear models revealed that dispersal and predation formed distinct categories based on habitat, ant size, and dispersal mode, with parasitism outcomes forming a distinct subgroup of predation based on seed size and shape. Multiple correspondence analysis indicated some combinations of ant genera and plant families were strongly associated with particular outcomes, whereas other ant-seed combinations were much more variable. Instead of the single axis association we predicted, the outcome appears to correspond to two axes –from dispersal to predation and from non-cheating to cheating. Taken together, these results demonstrate that ant and plant traits are important overall predictors of potential seed fates in different habitat types. More data on ant diet preferences and final seed fate (i.e. germination or death) would further clarify these results, but these data could indicate trends for future interaction predictions and applying this knowledge for both conservation and control purposes.