We have demonstrated that M. sexta larvae can learn to associate odor cues with an aversive stimulus, and that this memory persists undiminished across two larval molts, as well as into adulthood. The behavior represents true associative learning, not chemical legacy, and, as far as we know, provides the first definitive demonstration that associative memory survives metamorphosis in Lepidoptera. Furthermore, the results from our differential timing of larval training are consistent with the idea that retention of memory could be due to the persistence into adulthood of intact larval synaptic connections.
Our results support those of a handful of other studies that show learning within a larval instar in Lepidoptera 
. Only the forward temporal pairing of training odor with electric shock generated aversive behaviors in larvae. Backward pairing of shock and EA did not result in an aversive response to EA in larvae; nor did exposure to EA alone or shock alone cause larvae to avoid EA, ruling out the possibility that the behavior was a result of sensitization or habituation to either stimulus. Interestingly, larval food aversion learning has not been observed in Manduca sexta 
, nor in several other lepidopteran taxa 
despite the extreme negative consequences of ingestion of the noxious or toxic food.
Like fifth instar larvae, third instar M. sexta
caterpillars could be conditioned to avoid EA, and they recalled this information 10–14 days later, after two molts, as late fifth instars. We know of only one other study that demonstrates retention of larval memory across a molt in Lepidoptera: neonate larvae of the codling moth Cydia pomonella
(Tortricidae) recall a learned aversion to noxious food across four days during the transition from first to second instar 
Over the past century, a number of investigators have experimentally evaluated the effect of larval experience on adult behavior in beetles, moths, and butterflies 
. Some of these studies have sought to assess the validity of Hopkins' Host Selection Principle 
whereas others have explicitly examined the persistence of associative memory across metamorphosis. Evidence for HHSP, whether based on chemical legacy or retention of memory across metamorphosis, is equivocal. In some lepidopteran species, adults show an increased tendency to oviposit on their larval host plant 
, while in others larval feeding experience has no effect on adult host plant preference 
. Of those studies that do show an effect of larval experience on adult behavior, none have ruled out the possibility of chemical legacy. For example, Chow et al. 
demonstrated that oviposition deterrence in the presence of a novel chemical was markedly reduced following larval consumption of the chemical. Though pupae were removed from the larval environment and rinsed prior to the adult trials, residues of the non-water-soluble chemical may still have been present in the insect haemolymph or outside the pupal case.
In our experimental design, we attempted to eliminate the possibility of chemical carryover from the larval environment by using an electric shock rather than an aversive ingested chemical as the unconditioned stimulus, and by using ephemeral exposure to a gaseous compound, EA, as the conditioned stimulus. To further ensure that chemical carryover was not a factor, we washed pupae that had experienced forward-paired shock+odor as larvae, and applied EA to pupae that developed from naïve larvae, and in neither case did our results change relative to the experimental treatments. Thus, we are confident that the observed changes in adult behavior reflect larval experience, rather than exposure of emergent adults to cues from the larval environment.
Our results demonstrate a clear effect of larval experience on adult behavior. Only the pairing of training odor with electric shock generated aversive behaviors in larvae, and this aversion was retained in the adult moths. Furthermore, our constancy calculations demonstrate that the majority of individuals in the forward-paired shock+odor treatment made the same choice as larvae and as adults, indicating that individual preferences were maintained.
Studies of a handful of other holometabolous taxa, including beetles 
, fruit flies 
, ants 
and parasitic wasps 
, have convincingly demonstrated an effect of larval experience on adult behavior that was not due to exposure of emergent adults to residual chemicals, and several have suggested, but have not explored, a neural basis for their findings. What mechanisms could account for the carryover of larval experience into adulthood in our system? Manipulation of the timing of larval conditioning may provide insight into the basis of memory retention, as regions of the MBs develop at different times, and have different fates; that is, some lobes are retained intact through metamorphosis while others are not. Our results are consistent with, but do not provide conclusive support for the survival of synaptic connections within the larval brain across metamorphosis, enabling persistence in the adult brain of memories formed during the larval stage.
We found that adults that developed from larvae trained at fifth instar recalled their larval experience, whereas those that were trained at third instar did not. In Drosophila
, the only holometabolous insect for which individual MB neurons have been tracked through metamorphosis, development of the tri-lobed MB occurs in a sequential fashion, with the γ lobe forming embryonically, the α′/β′ lobe developing just prior to pupation, during mid-third instar, and neurogenesis of the α/β lobe initiating at the onset of pupation 
. During pupation, γ lobe neurons are pruned to the main process prior to production of adult-specific projections, while α′/β′ neurons maintain intact projections throughout metamorphosis 
. Since M. sexta
progress through five instars prior to pupation while Drosophila
progress through only three, it is likely that third instar training in M. sexta
occurs before α′/β′ neurogenesis. If M. sexta
MB development is analogous to that of Drosophila
, then our findings are consistent with the idea that the memory resulting from third instar training depends upon the embryonically-formed γ lobe, which is intact at fifth instar and so could enable recall at that stage, but is lost in adults subsequent to γ lobe pruning during pupation. The memory resulting from fifth instar training, however, could be retained in the later-forming α'/β' lobe, which remains intact throughout pupation and could therefore allow recall at the adult stage. As such, it would be interesting to examine the effects of α′/β′ ablation on adult memory of larvae trained at late instars.
Many studies of insect learning use appetitive as opposed to aversive training to mimic positive feeding experiences that occur in nature. Honeybees, butterflies and moths, for example, have been shown to associate both colors and odors with food rewards 
. However, insects also learn aversive cues in a variety of ecological contexts. For example, mantids rapidly learn to avoid noxious and aposematically colored milkweed bugs 
and Manduca sexta
larvae become sensitized to repeated pinching (analogous to bird attacks), showing increased defensive behavior in response to recurring assailment 
. Thus, although the current study uses an artificial electrical shock as the aversive stimulus, this type of conditioning is consistent with aversive experiences in nature.
Duration of associative memory in insects varies considerably, from minutes to months, depending on identity, age, and gender of test organism, strength of rewarding or aversive stimulus, number of training repetitions, and assay type 
. These variables notwithstanding, memory of aversive conditioning often lasts longer than that of appetitive conditioning 
. In the current study, avoidance of EA by M. sexta
subjected to forward-paired shock+odor was almost identical before and after the 4–5 week pupal period (78% of larvae and 77% of adults avoided EA), demonstrating a long-lasting and stable memory. A similarly long-lasting aversive memory is seen in the hemimetabolous cricket, Gryllus bimaculatus
, which retained an association between an odor and salt water for up to 10 weeks 
The present study has both ecological and evolutionary implications. as retention of memory through metamorphosis could influence host choice by polyphagous insects, and could further lead to the formation of host races or even to eventual sympatric speciation 
. While some studies of this phenomenon suggest chemical legacy as the process by which HHSP occurs, our data also implicate retention of memory, although both could lead to the same result 
. In addition, the mechanism for HHSP could vary between taxa, as is observed in lepidopteran host plant induction 
Carryover of larval experience into adulthood could have important consequences not just for insects in nature, but also for laboratories studying adult insects. Larval “chemical legacies” have already been shown to generate spontaneous odor attraction in adults 
. Furthermore, evidence suggests that larval artificial diets can impact aspects of adult physiology, such as color vision 
. These observations, in conjunction with the survival of memory across metamorphosis demonstrated in the present study, argue for the standardization of rearing conditions and protocols between labs. Variation in factors as seemingly irrelevant as larval environment, diet, or cage color could lead to unexpected effects on adult behavior, which could then contribute to significant variation in observations between labs, or the inability to replicate results if animals are obtained from different rearing facilities.
Our behavioral results are exciting not only because they provoke new avenues of research into the fate of sensory neurons during pupation, but also because they challenge a broadly-held popular view of lepidopteran metamorphosis: that the caterpillar is essentially broken down entirely, and its components reorganized into a butterfly or moth. Further studies of neuronal fate in holometablous organisms will yield greater insight into the process of complete metamorphosis and move us closer to an integrated understanding of organisms, providing links between complex cognitive behaviors and the molecules and developmental processes that give rise to them.