Our results demonstrate three essential components of self-medication predicted by adaptive plasticity theory: 1) self-medication behavior improves fitness of animals infected by parasites; 2) self-medication behavior decreases fitness in uninfected animals; and 3) infection induces self-medication behavior.
Predictions 1 and 2 are supported by the survival and resistance experiment and, to a lesser extent, the feeding choice experiment. The expected fitness trade-off of PA ingestion in the presence and absence of parasitism was most clearly seen in the survival and resistance experiment (). In the feeding choice experiment, increased PA ingestion was likewise associated with increased survival in caterpillars that received 2 eggs, whereas the opposite relationship occurred for unparasitized caterpillars. Previous work on this system suggested the existence of this trade-off 
, but the present study shows it directly for the first time. Importantly, these results unambiguously identify PA as an agent of anti-parasitoid resistance for G. incorrupta
. Although contrary cases exist [e.g., 26]
, this work complements several previous studies of various caterpillar species showing that the host's ingestion of plant secondary compounds can retard growth and development of its parasitoids [reviewed in 27]
; [see also 29]
. However, very few studies have shown that such anti-parasitoid effects translate into resistance benefitting host survival, as shown here. The physiological mechanism by which dietary PA confers resistance against parasitoids is not yet known.
Together the feeding choice and no-choice experiments lend support to prediction 3, that parasitism induces self-medication. However, the evidence from the feeding choice experiment is relatively weak, being complicated by extensive variation in PA feeding responses among individuals within parasitism treatments. It is presently unclear why some individual caterpillars exhibited increased PA feeding in response to parasitism while others did not. The no-choice feeding experiment provides more straightforward evidence in support of prediction 3. The presence of infection by parasitoids was clearly associated with increased PA ingestion by caterpillars. We believe the two feeding experiments differed in the variability of PA feeding response in part because of methodological differences in how we scored parasitism in relation to host feeding behavior. In the no-choice experiment, the dissection and scoring of parasitism soon after the feeding assay gave a relatively accurate measure of the effects of parasitoids during the PA feeding assay. By contrast, the caterpillars in the choice experiment received a controlled number of fly eggs in the larval stage before the PA feeding assay, and their parasitoid loads during the feeding assay were unknown. We suspect that many early-stage parasitoids were destroyed prior to the feeding assay by the host encapsulation response, as G. incorrupta
appears to have an unusually strong encapsulation response 
. This would have introduced considerable, uncontrolled variation in the parasitoid loads experienced by individual caterpillars during the feeding choice assay.
Taken together, the feeding choice and no-choice experiments show that parasitized caterpillars forage differently than unparasitized caterpillars. One aspect of this foraging difference is an adaptive increase in PA ingestion by caterpillars facing a high threat of mortality from parasitism. Whether the threat of mortality reflects a parasitoid dose-dependent effect (i.e., the number of parasitoid larvae in a host), variation in the developmental stage of individual parasitoids (i.e., early vs. late instars of parasitoid larvae in host), or both is not clear from these experiments.
General observations suggest it is likely that other plant-feeding insect species engage in self-medication because of the ubiquity of dietary chemical defenses 
, and the substantial frequency of parasitism 
among herbivorous insects. Moreover, many herbivorous insects exhibit various forms of adaptive plasticity 
. Even herbivores with specialized diets might alter their intake of plant tissue types of varying defensive value in response to parasitism or disease. There exists one other published account of possible self-medication by an herbivorous insect. Parasitized Platyprepia virginalis
caterpillars (Arctiidae) increased their likelihood of survival by feeding on poison hemlock plants, and parasitized caterpillars preferred poison hemlock over bush lupine, unlike unparasitized caterpillars 
. Interestingly, this case involved tolerance to (rather than resistance against) parasitoids by host caterpillars, as both host and parasitoid survived in numerous instances. This study is an ambiguous case of self-medication because it is unclear to what extent the results might be due to the parasitoid adaptively manipulating host behavior, as both parasitoid and host benefit from the change in host behavior. Other foraging behaviors in insects have been shown to function as defenses against parasites (e.g., resin-collecting by ants 
), but none of these other examples shows an adaptive change in behavior in response to infection by parasites.
Self-medication by G. incorrupta
is distinct from well-understood cases of self-medication in vertebrates by showing a quantitative rather than qualitative change in behavior. That is, parasitism can cause an increase in PA-pharmacophagy, a routine behavior for unparasitized caterpillars. Sick chimpanzees, by contrast, do not typically engage in leaf-swallowing or another specific self-medicative behavior, bitter pith-chewing, in the absence of stress caused by parasites 
. Self-medication based on a quantitative behavioral change, as seen for G. incorrupta
, does not easily distinguish itself from routine foraging behavior in observations of wild animals 
. Consequently, other existing cases of self-medication might be easily overlooked, with behavioral extremes attributed to random variation even for closely observed animals.
We argue that self-medication by G. incorrupta
is functionally, if not mechanistically, congruent with cases of self-medication by vertebrates. In the vertebrate literature, self-medication has been given the name zoopharmacognosy 
. The original definition of zoopharmacognosy is “the process by which wild animals select and use specific plants with medicinal properties for the treatment and prevention of disease” 
, broadly encompassing a variety of possible mechanisms such as adaptive plasticity (self-medication as defined here) and prophylaxis (which includes pharmacophagy per se). The role of associative learning in self-medication is a further important mechanistic distinction, as some authors have assumed that associative learning is an essential component of self-medication 
. Clear experimental proof of self-medication via individual learning was recently demonstrated in domesticated sheep, which learned to ingest particular chemicals that countered toxicity from experimentally applied dietary toxins 
. Self-medication via social learning is exemplified by wild chimpanzees, which can learn from other individuals' leaf-swallowing behavior to alleviate infection by intestinal nematodes 
. Among insects such as caterpillars, however, self-medication behavior need not be learned. A previous study of G. incorrupta
and the related caterpillar Estigmene acrea
showed that the phagostimulatory taste responses to PA differed between parasitized and unparasitized caterpillars 
. The gustatory cells of parasitized caterpillars fired action potentials more rapidly than those of unparasitized caterpillars in response to PA, but did not differ in their response to sucrose (a non-medicative feeding stimulant). This specific change in gustation in parasitized caterpillars implies that self-medication in G. incorrupta
is mediated through plasticity in the peripheral nervous system, without the necessity of associative learning. We hypothesize that parasitized caterpillars can immunologically recognize the presence of internal parasites, and chemically signal the taste system, thus adaptively altering their taste and feeding responses to PA in isolation (as shown here) or in the context of natural host plants. The plausibility of these mechanisms rests on extensive evidence that caterpillars and other insects can immunologically recognize the presence of internal parasites 
, and that chemical feedbacks from the blood to the insect taste system can adaptively alter taste and feeding responses to macronutrients 
A functional alliance of self-medication with other forms of adaptive plasticity reinforces the potential importance of self-medication in the ecology, evolution, and conservation of species interactions. A surge of recent publications by many different authors emphasizes the profound consequences of adaptively plastic responses of individuals for understanding population, community, and evolutionary dynamics 
. As individuals adaptively alter their behavior and phenotypic traits in response to their ecological circumstances, they can and do alter the demographic outcomes of trophic, competitive, and mutualistic interactions among species 
. The ecological, evolutionary, and conservation consequences of self-medication are virtually unstudied, despite increasing environmental stresses faced by some of the species, such as great apes, known to self-medicate 
In conclusion, our demonstration of self-medication through a shift in the extent of pharmacophagy by G. incorrupta caterpillars points to the possibility that more animal taxa than previously believed self-medicate and that known behavioral and physiological mechanisms can mediate self-medication even without associative learning. Our support for self-medication by G. incorrupta as a form of adaptive plasticity places the science of self-medication by non-human animals in a theoretical context with broad but relatively unstudied implications for ecology, evolution, and conservation of species interactions.