FIG. A1. Logistic regression analysis revealed a
significant interaction between source population and herbivory on survival
probability ( < 11.05, df < 3, P < 0.01), and analyses of each source population independently
indicated that herbivory tended to reduce the survival probability of the two
serpentine populations (S1 and S3, solid lines) more than the two
non-serpentine populations (NS1 and NS2, dashed lines) (logistic regression
coefficients: S1: -0.035, = 54.86, P < 0.0001; S3: -0.020, = 16.94, P < 0.0001; NS1: -0.009, = 2.13, P = 0.14; NS2: -0.019, = 11.58, P = 0.0007). The logistic
regression coefficient can be interpreted as a measure of tolerance to
herbivory, where more shallow slopes indicate populations that are more
tolerant of herbivory and steeper slopes indicated less tolerant
populations. While the serpentine
populations show higher survival probabilities than non-serpentine populations
at low levels of herbivory, their fitness declines more rapidly with increasing
herbivory. |
