Mechanistic and scale‐specific analyses advance the preference–performance hypothesis

Abstract Oviposition choice and larval performance are essential determinants of fitness in plant–herbivore interactions. The long‐standing preference–performance hypothesis (PPH) predicts that parents should oviposit on plants that maximize offspring performance; an abundance of work supports the PPH, but many exceptions limit its generality. These exceptions may stem from applying the PPH across diverse contexts without adequately considering mechanistic processes and scale‐specific factors. Using the specialist seed predator moth Schinia florida and the common evening primrose Oenothera biennis , we investigated the mechanistic drivers of preference and performance and how these dynamics shift across scales, between versus within plants. We measured a suite of 126 plant traits, most of which were plant specialized metabolites captured by ultrahigh‐performance liquid chromatography and mass spectrometry and applied a novel machine learning approach to identify traits that predict both preference and performance. Between plants, we found no relationship between maternal oviposition preference and offspring performance. Nevertheless, nine overlapping plant traits independently predicted both preference and performance, with all traits positively correlated with preference showing a negative association with performance. This negative trait‐based preference–performance link may result from trade‐offs of moths using plant defenses as host‐finding cues. Overall, weak correlations between the nine predictive traits appear to limit a link between preference and performance between plants. Within plants, moths disproportionately oviposited on flower buds over fruits and concordantly showed higher relative performance on buds. At this scale, only one overlapping trait predicted both preference and performance, and had positive associations with preference and performance. Thus, our results show that (1) a specialist herbivore faces trait‐based trade‐offs between host selection among plants and larval performance, while making more adaptive decisions at finer scales within plants, and (2) mechanistic and cross‐scale approaches offer insight into when adult preference predicts larval performance.