Statistics
For plants grown in competition, we averaged the data for both plants before analysis. Caterpillar weight was Ln+1 transformed to fit a normal distribution. Protease inhibitor data were square root transformed to fit a normal distribution and analyzed using a linear model. We ran a series of linear testing the interacting effects of competition, mycorrhizae and induction on each of the response variables: height, protease inhibitor activity, C/N ratio, caterpillar mass, and leaf area consumed. Each of these models were analyzed using a linear model in R using the nlme package (Pinheiro et al. 2018). We also analyzed the effects of the potential mechanisms: protease inhibitor activity and C/N ratio on caterpillar weight and leaf area consumed. For each of these analyses, we ran a full factorial model and dropped non-significant terms in a backwards stepwise fashion. Significant and non-significant results are shown in Table 1. To measure the effect of mycorrhizae on size dimorphism, the data were subset to select only plants grown in competition and we took the difference between plant A and plant B. The absolute value of the difference was log transformed to meet assumptions of normality and analyzed using a linear model with mycorrhizae as a predictor variable. Data were analyzed using R version 3.3.3.