Path models and explanatory hypotheses.

Black arrows indicate causal paths between variables

Treatments of intraspecific richness, interspecific richness, and water addition used in each of the five study sites.

White tiles represent treatment combinations which were not tested in a given site

Hierarchical Bayesian model parameter estimates for the effects of water availability, as well as intraspecific and interspecific Piper richness on herbivory, herbivorous insect richness, and Piper survivorship across sites

Violins represent the cross-site posterior parameter distribution for each relationship in site level hierarchical Bayesian models. Black lines represent the median posterior estimate and white bars represent 95% credible intervals. Percentages above violins indicate the probability of an effect being positive or negative (as indicated by a negative probability) in response to an increase of the independent variable. Distributions for water addition compare watered and control plots; distributions for interspecific richness compare Piper species richness standardized as the proportion of the maximum richness used at a site; distributions for intraspecific richness compare low and high intraspecific richness treatments; distributions for insect richness compare responses per insect taxon present on an individual leaf

Direct and indirect effects of plant diversity and water availability on insect herbivores at five study sites.

Bayesian structural equation models comparing effects of different drivers of herbivorous insect richness and herbivory at five sites. Standardized path coefficients are means of the posterior distribution for the effects estimated at each causal path. Positive relationships are indicated in blue with triangular heads, and negative relationships are indicated in red with circular heads. Black arrows indicate path coefficients of zero magnitude. Dot plots summarize the standardized mean of the posterior distribution for each causal path with 95% and 80% credible intervals. Asterisks indicate causal paths where the probability of an effect being positive or negative is >95%

Interactions between intraspecific or interspecific richness and water availability on herbivory, herbivorous insect richness, and Piper survival

Bars indicate mean response and standard error of the mean. Percentages above each sub-figure indicate the probability that the two slopes are different as calculated using site level hierarchical Bayesian models. Due to high mortality in Peru, interactions between water and intraspecific richness could not be compared for any responses except mortality. Single and two species richness plots have been combined for visualization purposes only