Trade-offs in performance on different hosts are thought to promote th
e evolution of host specificity by blocking host shifts. Yet, in contr
ast, most experiments using phytophagous insects have shown performanc
e on alternative hosts to be uncorrelated or positively correlated. Re
cent quantitative genetic models based on mutation-selection balance i
ndicate that underlying constraints on the simultaneous maximization o
f different components of fitness may not always generate negative gen
etic correlations. We suggest an alternative or additional explanation
for the lack of observed negative genetic correlations. If performanc
e is polygenically controlled and some performance loci possess only a
ntagonistically pleiotropic alleles, then the expression of trade-offs
in performance will vary over time in populations. Consequently, a tr
ade-off will be seen only in populations that have adapted to two host
s and are at or close to genetic equilibrium. Therefore, studies testi
ng performance on a novel as compared with a normal host will generall
y yield non-negative genetic correlations between performance on the t
wo hosts. The results of published studies are consistent with the pre
dictions of this hypothesis.