Figure 2. Estimated interaction strengths from S-mapping analysis, based on pooled data including all experimental treatments. Points show results for individual time steps, whereas lines show mean ± one standard deviation. Recall that partial derivatives can be interpreted in the same way as the slope of a linear regression – i.e. ∂y /∂x describes the slope of a regression of y as a function of x . (A) Estimated effect of herbivory on shoot growth, vs. estimated effect of climate on shoot growth for a time lag of -1. Note that cases with more positive effects of climate are associated with more negative effects of insect herbivory. (B)Estimated effect of climate on shoot growth, vs. time-lagged climate (GDD). Note that warmer years are associated with a more positive effect of climate on shoot length. (C) Estimated effect of insect herbivory on shoot growth, vs. time-lagged climate (GDD). Note that in warmer years, negative effect of insect herbivory is intensified.