Fig. 2.

Simulated pairs of populations with stronger interactions have weaker use–resistance associations. (A) Each filled circle represents a simulated population. When spillover is weak (i.e., d₀, measured in arbitrary length units, is small), populations fall along an idealized use–resistance association (dashed black line). As spillover is increased, the cross-sectional use–resistance association becomes weaker and has greater variance. Gray lines show the simple linear regression best fit. (B) Each open circle represents a pair of the populations shown in the corresponding facet in A. For amounts of spillover, all pairs have the same use–resistance association Δρ/Δτ (i.e., 1, the slope of the dashed black line in A). For stronger spillover, there is substantial variance in the associations. For nonzero spillover (d₀ > 10⁻⁶), stronger-interacting pairs have weaker use–resistance associations (Spearman’s ρ; SI Appendix, Table S2). The red line shows the robust regression best fit. The vertical axis is truncated for visual clarity.