| Abstract: | Host‐plant data for North American and Australian butterflies were used to test the hypothesis that larval diet breadth increases with decreasing resource predictability (where the latter was estimated by host‐plant growth‐form/duration). For each region in turn we compared the diet breadths of butterflies which utilise herbaceous host‐plants with those of species having woody hosts. For North America alone we also compared the diet breadths of species having annual hosts with those utilising perennial hosts, and the diets of species having herbaceous‐annual hosts with those using woody‐perennial hosts. Studies of diet breadth may be biased by the host taxonomic level which contributes most to the diet index used. For example, the results of analyses which employ indices based on numbers of families of hosts utilised may differ from those using indices based on counts of host species or genera. To investigate this potential problem we performed cross‐species analyses where diet breadth was defined in turn as the number of host species, genera, or families eaten. We found that using different taxonomic levels did give inconsistent results. To avoid this we employed phylogenetic diet breadth indices in comparative analyses of Independent Contrasts. The former incorporate information from the whole of the host‐plant phylogeny, whilst the comparative method eliminates any confounding effects of butterfly phylogeny. The results indicated that there is a phylogenetic component to butterfly diet breadth. They also largely differed from those of the cross‐species investigations, although there were similarities (i.e. results differed between regions and varied according to whether the whole fauna or just endemics were investigated). Our results suggested that in both regions, non‐endemics which feed on herbaceous plants have wider diet breadths than non‐endemics which utilise woody hosts. However, we found no consistent evidence that the diet breadths of endemics increase with decreasing resource predictability (as estimated here). |
