Induced resistance to herbivores and the information content of early season attack.

Current models of induced plant defenses all assume that herbivory is predictable. Present damage must provide information about the likelihood of future attack. We tested this assumption by measuring the relationship between damage early in the season and the number of subsequent attacks by cotton leaf perforators, Bucculatrix thurberiella, to plants of wild cotton, Gossypium thurberi, at three sites in the Sonoran desert. Damage early in the season was a good predictor of the number of new mines initiated throughout the season for plants in Florida Canyon. This result was neither evidence for nor against induced resistance nor was it a consequence of induced resistance. This is because induced resistance has been found to affect survival of miners but previous damage did not affect the initiation of new mines. At two other sites, early damage was not related to future attacks. This difference in predictability of attack may be related to inducibility of plants since Florida Canyon was the only site that provided evidence of induced resistance in a previous study. We found no evidence that the usefulness of information based on early season damage decreased as the season progressed. Assessment of attacks on all shoots of the plant was more useful at predicting later damage to an assay shoot than was assessment of solely the assay shoot. Number of mines, produced only by G. thurberiella, was a better predictor of subsequent attacks by G. thurberiella than were chews and rasps which were made by many different herbivores. However, general chewing damage was a better indicator of the level of induction against G. thurberiella than was the more specific mining damage. Plants may respond more to chewing damage even though mining damage is a better predictor of future attacks.