Manipulation of natal host modifies adult reproductive behaviour in the butterfly Heliconius charithonia

Abstract

Advances in understanding non-genetic inheritance have prompted broader interest in environmental effects. One way in which such effects may influence adaptation is via the transmission of acquired habitat biases. Here I explore how natal experience influences adult host orientation in the oligophagous passion vine butterfly Heliconius charithonia. As an exemplar of the ‘pupal mating' system, this species poses novelty among diurnal Lepidoptera for the extent to which male as well as female reproductive behaviours are guided by olfactory host cues. I sampled wild adult females breeding exclusively upon Passiflora incarnata, assigned their offspring to develop either upon this species or its local alternative Passiflora suberosa, and then assessed the behaviour of F₁ adults in a large rainforest enclosure. Despite the fact that juvenile performance was superior upon P. incarnata, females oviposited preferentially upon their assigned natal species. Mate-seeking males also indicated a bias for the proximity of their natal host, and there was evidence for assortative mating based upon host treatment, although these data are less robust. This study is, to my knowledge, the first to support Hopkins' hostplant principle in butterflies, and points to inducible host preferences capable of reinforcing ecological segregation and ultimately accelerating evolutionary divergence in sympatry.

1. Introduction

Unravelling how the conditions experienced throughout ontogeny shape phenotypic development presents an enduring challenge across the life sciences [1–3]. For plant-feeding insects, a captivating idea is that the hostplant experienced during juvenile development may induce assortative biases in adult host orientation [4,5]. Conditioning effects of this nature have the potential to reinforce vectors of host specialization, and ultimately maintain (if not extend) the breadth of polyphagy. Depending upon factors such as the genetic basis of host preference and the heterogeneity of hostplant environments [6], such effects could moreover influence reproductive isolation. This has broad analogy with ‘resource polymorphism', a concept considered as a driver of adaptive diversification and speciation in vertebrates [7].

The hypothesis that natal host experience may engender adult biases in holometabolous insects was first proposed and tested via Hopkins' seminal work upon mountain pine beetles [4,5]. It became thereafter cast under the general banner of ‘Hopkins’ host selection principle'. Efforts to test this principle have considered when and/or how hostplant cues might actually influence adult behaviour [8]. This work has largely focussed upon the consequence of preference induction for adult females and in the context of host choice for oviposition [9]. As shown by a singular example in the moth Spodoptera littoralis [10], natal-derived biases could also shape how males orient to hostplants in relation to mate-seeking behaviour, or influence mate choice decisions in either sex. Establishing the presence of such biases is not however clear-cut. Two salient points are that: (i) natal cues will often persist into early adulthood via chemical residues in the gut or upon the pupal cuticle, and (ii) adults will often eclose in proximity to their larval host. Both offer avenues for ‘chemical legacy', whereby biases could, in fact, become established early in adulthood [11]. Doubt has also been expressed [8] as to whether the signature of pre-imaginal conditioning could survive the extensive neural re-arrangement that occurs during metamorphosis (a point discussed in §5 vis-à-vis recent evidence to the contrary [12]). Scrutiny from this mechanistic (proximate) level has fuelled controversy as to whether Hopkins'-like effects could apply, at least in the literal sense of pre-imaginal conditioning. Revised labels such as ‘neo-Hopkins' principle' [13] emphasize a more inclusive view of how natal provenance may influence adult host association.

Deterministic effects upon adult habitat use have broad interest owing to their potential evolutionary consequences [3]. The (neo-) Hopkins hypothesis has perhaps attracted particular attention because any inducible component to adult hostplant preference can inturn bias their offspring to receive a similar effect. Should it prove adaptive [14], the presence of inducible plasticity could in this sense multiply or at least reinforce itself across generations. Interest in this hypothesis has recently expanded owing to evidence for so-called transgenerational acclimatization [15]. This effect is such that a female's natal experience primes their offspring for optimal performance under similar conditions, and invokes unconventional mechanisms of inheritance [16,17]. The ultimate significance of such possibilities will, however, depend upon how frequently and under what ecological conditions adult host induction actually occurs. Notably, almost a century of dedicated study has revealed a somewhat infrequent and taxonomically sporadic incidence of such plasticity. Whereas support has accumulated in beetles, flies and moths (reviewed in [9]), the balance of work in groups such as butterflies [14,18–23] best supports canalized innate host preference. The discrepancy between diurnal and nocturnal Lepidoptera (i.e. butterflies versus moths) is particularly salient. Both groups are near-exclusively phytophagous, yet differ profoundly in sensory biology and mating system structure [24]. The importance of olfaction for mate and host location in moths may signify a role for sensory ecology as to whether preference induction is likely to arise.

Here I present an experimental test of host-derived adult biases in Heliconius charithonia, a member of the ‘passion vine' group of butterflies (Nymphalidae: Helinoniinae). This group has offered an important testing ground for various principles of evolutionary theory, including aposematism, Müllerian mimicry and speciation. Their intimate relationship with a single plant genus (Passiflora) has moreover seen the Heliconiinae play an instrumental role in the study of coevolution [25]. Heliconius charithonia is one of the more generalist species, with more than 20 recorded hosts across its range. The rationale for using this species comes from recent findings for how the reproductive behaviour of both sexes depends critically upon host orientation. Adult males search for hostplants in order to locate developing conspecific immatures and as part of their strategy of guarding female pupae until eclosion [26,27]. Male host orientation is mediated via the detection of phytochemical volatiles, particularly those released in response to tissue damage, and notably more so when such damage results from conspecific folivory [28]. This finely tuned olfactory ability implies the potential to seek mates developing upon particular Passiflora species. Adult females likewise rely upon detecting phytochemicals to locate and choose oviposition substrates (along with visual cues such as leaf colour and shape). As in other oligophagous heliconiine species, ovipositing females are known for innate host biases [29]. Work in several species has however also revealed that larval feeding preferences can be induced according to the host species encountered towards the end of development [30]. It remains unknown whether derived biases of this nature persist through metamorphosis to influence adult reproductive behaviour.

2. Experimental design, aims and predictions

I explore hostplant preferences using a design whereby the offspring of wild-caught H. charithonia from a single-host breeding assemblage (centred exclusively upon Passiflora incarnata) were randomly assigned to develop upon either this species or a local alternative (Passiflora suberosa). Adults were then assayed for reproductive behaviour in relation to each host within a large rainforest enclosure. Adult host orientation, therefore, offered the basis for assessing innate versus inducible preference. The presence of innate bias should see individuals preferentially associate with one particular host irrespective of their assigned natal treatment. Given the specific sampling provenance of wild females, such a bias should favour P. incarnata. Inducible bias, by contrast, predicts a positive association between natal host assignment and adult host orientation. These hypotheses are not mutually exclusive; that is, adult host preference could be shaped by combined innate and inducible effects [31]. The hypothesis test therefore emphasizes the relative magnitude of each source of bias rather than absolute presence/absence.

I also aimed to test for mate assortment based upon natal host provenance. Assortative mating of this nature could arise as an indirect consequence of biased host orientation [32], or more directly from mating preferences [33,34]. The latter opportunity exists because heliconiines discriminate among potential mates using olfactory and visual signals [35–37] that are known to vary according to natal hostplant [38].

Last, I present data on juvenile performance across each Passiflora species as a context for interpreting the potential consequences of hostplant bias in the sample population [29,30,39].

3. Material and methods

(a). Sampling site

Thirty-eight adult female H. charithonia were collected near Lakeland highlands scrub in south-central Florida, USA (27.93° N, 81.92° W). The sampling population consisted of a high-density breeding assemblage centred upon an approximately 10 m² outcrop of P. incarnata. Eggs, larvae, pupae and adults were extremely abundant at this locality. During the period of sample collection (9.00–12.00 hr), adult females were observed foraging around hostplants and laying eggs, whereas males courted females and guarded pupae. Multiple copulating pairs were seen, and adults of both sexes fed occasionally upon flowering Lantana plants. A haphazard search spanning approximately 1 km of the site indicated no evidence of P. suberosa or any other Passiflora species.

(b). Rearing protocols

Wild-caught females were transported to the laboratory within an insulated cooler and released at roughly equal densities into three outdoor enclosures (1.8 × 1.8 × 1.8 m; Yancheng Breda, Yancheng, China) containing potted P. incarnata. The logistics of sourcing eggs was such that parentage could only be reliably tracked for perhaps five to six different mothers, so I instead prioritised genetic diversity by seeking a single pooled sample (to which all 38 females had the opportunity to contribute). Eggs were laid primarily upon new growing shoots, and collected at the end of each day by removing cuttings to plastic cups. Cups were designed such that the plant stem extended into a water reservoir separate from where foliage was contained. These were housed at 21.0 ± 1.0°C and a 14 L : 10 D photoperiod. Newly hatched larvae were randomly transferred (according to a computer-generated random series) to a cup containing a fresh cutting of either P. incarnata or P. suberosa. The transfer was done using a fine artists' paintbrush under x10 visual magnification. Larvae were then raised under locally relevant conditions of 30°C (day; 14 h) and 21°C (night; 10 h), with hostplant refreshed daily and maintained in excess. Larval mortality upon each host was negligible (less than 5%). Pupae were detached from their host substrate the day after pupation, weighed to the nearest 1 mg, then affixed to a strand of cotton string using superglue (ethyl-2-cyanoacrylate; Super Glue Corp., Ontario, USA). Pupae developed in a clean, paper-lined cup with a gauze top, and hence were isolated from their natal hostplant (refer to the electronic supplementary material, figure S1c-d). Eclosed and hardened imago were marked with an identifying number on each ventral hindwing using a white marker (‘Decocolor' extrafine 130-S, Uchida corp., Torrance, USA) prior to release the following day.

(c). Adult observations

Adults were observed in a large outdoor glasshouse (9.2 × 10.7 m area and 6.1 m height). This enclosure mimicked the local breeding habitat, with several large rainforest trees bearing epiphytes and vines, as well as understory shrubs, ferns and bryophytes (but excluding Passiflora). Airflow through moistened fibrous wall panels provided ventilation and evaporative cooling. Potted P. incarnata and P. suberosa plants were arranged at opposite sides of the glasshouse to present discrete patches of the two species. These were situated for similarity of immediate vegetative geometry and aspect, and positioned adjacent to a perimeter pathway within the glasshouse to facilitate observation.

Totals of 59 adult males (37 P. suberosa-reared) and 30 females (15 P. suberosa-reared) were released into the glasshouse. Behavioural observations were then made in discrete 10 min bouts conducted hourly between 8.00 and 17.00 over eight subsequent days. Each of the two hostplant patches were observed continuously for 5 min (in randomized order) each bout from a standard sheltered station. Short-focus binoculars were used to resolve specimen wing numbers. Strict criteria were imposed for the recording of associative events, such that females had to actually oviposit, and males had to exhibit ritualised courtship behaviours. Copulating pairs were identified where possible. Hostplants were inspected following each observation bout to remove eggs, which were scraped from shoots, tendrils and leaves, or removed by cutting off small sections of plant tissue. This was done because the presence of eggs on hostplants may serve as a visual deterrent to oviposition, and I wished to maximize the number of egg-laying observations.

As typical for butterflies, female H. charithonia land on their host and briefly pause with wings closed at the precise moment of oviposition. Such behaviour presented stationary views of their ventral hindwing and thereby aided identification. By contrast, male courtship involves aerial chasing, hovering and/or fanning behaviours, and individuals maintained vigorous flight afterwards. Identifying courting males required actively following them until they paused to rest or feed, which disrupted butterfly activity throughout the glasshouse. Priority was therefore given to observing females, with opportunities to identify courting males pursued towards the end of each observation bout. All observations were made blindly with regard to the host provenance of adult subjects.

(d). Statistical analysis

I assessed the effect of hostplant upon key juvenile performance traits (i.e. growth and development rates and pupal weight) in each sex using a general linear model approach equivalent to a two-factor analysis of variance. Each model included host treatment and sex as fixed effects, plus the host × sex interaction.

For the analysis of adult behaviour, I first calculated an index of host association (H_pref) for each study individual (sensu [40]):

$$H_{\mathrm{pref}}=\frac{(O_N-O_A)}{(O_N+O_A)}.$$ 3.1

This index expressed the relative frequency of observed oviposition (female) or courtship (male) events as a function of each individual's assigned natal host (O_N) versus the alternative species (O_A). Its domain ranged from +1.0 (complete association with the natal host) to −1.0 (complete avoidance), with zero indicating indifference. Index values were calculated per individual to eliminate pseudo replication, and then averaged across the members of each sex to derive the observed sample-wide mean values of host association for males and females.

I then used a randomized simulation procedure [41] to define the null distribution of H_pref and to derive p-values for observed host association in each sex. Null distributions were simulated via 50 000 iterations of n = i individuals, each with a 0.5 probability of associating with either host across each of n = k oviposition (female) or courtship (male) events. Values of i and k were precisely as in the empirical sample, and H_pref was likewise calculated per individual and averaged for each sex in each iteration. p-values were obtained from the null distribution to test the prediction of H_pref (obs) > 0.0; that is, for individuals to orient non-randomly in favour of their natal host. The same procedure was also used to account for pseudo replication in the analysis of host-based mating patterns, which was necessary because some males mated multiple times. An index of mate assortment (M_pref) was calculated for individual males by applying equation (3.1) to data on the hostplant provenance of their mating partners. Predictions for H_pref and M_pref were formulated a priori to define one-tailed hypotheses, and are tested as such. As true to the logical basis of disproof (modus tollendo tollens) [42], the null hypothesis therefore is that adults are not more likely to orient or mate assortatively for natal host treatment.

I supplemented this approach with a binomial generalized linear model analysis of host orientation across all individuals and events. This analysis was not designed to evaluate positive host assortment per se, but to test whether the probability of orienting towards the natal host differed according to sex (male versus female), host species (i.e. P. incarnata versus P. suberosa) or time of day. Individual (nested within sex) was included as a random factor. Fit to the binomial distribution was modelled using a log-link function with terms assessed via log-likelihood (G) statistics.

4. Results

(a). Does juvenile performance vary across hosts?

Of the two studied hosts, P. incarnata clearly supported superior juvenile performance. Larvae assigned to this species developed faster, pupated at heavier weights, and thereby achieved higher growth rates than those assigned to P. suberosa (table 1 and figure 1). Despite their greater average weight, P. incarnata–reared individuals also completed pupal development faster than their P. suberosa–reared counterparts. Female pupae were heavier than males irrespective of larval host, but pupal development rates were equivalent across the sexes.

Table 1.

Results for the linear model analysis of key juvenile performance traits in H. charithonia as a function of sex and assigned natal host treatment (i.e. P. incarnata versus P. suberosa). Significant effects in bold are shown in figure 1.

source of variation	developmental rates		pupal weight	larval growth rate
	larval	pupal
intercepta	F1,138 = 26 405	F1,137 = 8198	F1,135 = 7283	F1,135 = 24 286
assigned host	F1,138 = 21.5 p < 0.001	F1,137 = 3.96 p < 0.05	F1,135 = 15.2 p < 0.001	F1,135 = 31.5 p < 0.001
sex	F1,138 = 1.03 p = 0.312	F1,137 = 2.17 p = 0.143	F1,135 = 11.0 p < 0.005	F1,135 = 0.00166 p = 0.968
host × sex	F1,138 = 0.0184 p = 0.892	F1,137 = 1.76 p = 0.187	F1,135 = 2.40 p = 0.124	F1,135 = 0.103 p = 0.748

^aAll model intercepts were significant at p < 0.001.

Figure 1.

Key juvenile performance measures for Heliconius charithonia that were randomly assigned to develop upon either Passiflora incarnata (the immediate breeding host of wild-caught dams) or P. suberosa (an alternative local host). Larvae assigned to P. incarnata completed development faster (a) and gained weight more rapidly (b) than their P. suberosa-reared counterparts, and consequently pupated as heavier individuals (c). Despite their greater weight (i.e. size), however, individuals in this treatment completed pupal metamorphosis significantly faster (d). All findings are consistent with a superior ability to acquire and/or assimilate resources upon P. incarnata as a host. Means are indicated with 95% confidence intervals and the corresponding statistics are given in table 1.

(b). Do F₁ adults associate with particular hosts?

Glasshouse observations yielded 42 oviposition events involving 13 unique females, and seven courtship events involving five unique males. In support of the inducible bias hypothesis, these events were non-randomly associated with each hostplant patch in favour of each individual's assigned natal species (figure 2b,c). The index of host association was positive, and revealed by simulation to deviate significantly from zero in both sexes (females: null distribution approx. N_0.00,0.225, H_pref = +0.467, P_1-tailed = 0.0190; males: approx. N_0.00,0.416, H_pref = +0.867, P_1-tailed = 0.0186; electronic supplementary material, figure S2). The generalized linear model indicated no evidence for variation in the strength of host assortment among sexes (G₁ = 0.475, p = 0.490), host species (G₁ = 0.058, p = 0.809), study individuals (G₁₅ = 19.9, p = 0.176), or different times of day (G₁ = 0.09, p = 0.768).

Figure 2.

(a) The F₁ progeny of wild female H. charithonia caught at a P. incarnata-based breeding site were assigned to develop upon either this host (lightly shaded blue) or P. suberosa (lightly shaded beige), then released into a large rainforest enclosure as adults. Consistent colour-coded shading is used to indicate natal host provenance in lower panels. (b) Frequencies of adult host association events (indicated for each host as per the leaf symbols indicated in panel (a) summed for each sex and according to natal host treatment. Association was defined for females by oviposition, and for males by host-based mate-searching behaviour. (c) Mean (±1 s.d.) host preference of adult males and females in each natal host treatment. The y-axis indicates strength of individual preference for P. incarnata (arbitrarily here greater than 0.0) as opposed to P. suberosa (less than 0.0), with zero indicating indifference. (d) Observed in-copular pairs, arranged according to the host provenance of each pair member. Males no. 5 and no. 39 are notable for achieving multiple copulations, which are indicated in sequential order from left to right.

(c). Is mating assortative according to natal host?

A total of 12 copulations were observed (figure 2d). These involved 12 different females, but only nine different males because individuals M5 and M39 achieved two and three copulations respectively. There was no evidence that natal host influenced individual mating status per se (goodness of fit for mated versus virgin individuals from P. suberosa versus P. incarnata – males: G₁ = 0.167, p = 0.683; females: G₁ = 0.0, P = 1.0). Adults from each host, therefore, mated at expected rates given their frequency in the enclosure. Eight (67%) of the 12 observed copulations however proved assortative for the natal host identity of pair members. Notably, three of the four non-host-assortative cases occurred as superfluous copulations by males M5 and M39, meaning that eight of the initial (or sole) copulations achieved by the nine different males were assortative for natal host (figure 2d). The overall index of mate assortment was significantly positive (M_pref = +0.519, null distribution approx. N_0.00,0.31, P_1-tailed = 0.0476; electronic supplementary material, figure S3), which supports the presence of host-assortative mating.

Mating did not otherwise appear assortative for body size (log pupal mass: Pearson's r = −0.207, n = 9, p = 0.594), and copulating individuals did not differ significantly in size from the remainder of their cohort (females: F_1,28 = 1.16, p = 0.291; males: F_1,57 = 0.245, p = 0.622).

5. Discussion

The idea that natal experience could shape how adult insects prioritize alternative hosts has captivated ecologists since Hopkins' seminal work a century ago [8]. Despite mounting support in select holometabolous insect taxa [9], the weight of evidence in other groups points to innate and/or canalised host preference. This study presents, to my knowledge, the first clear evidence for inducible adult host biases in butterflies. Critical support is given by the host-orientation behaviours observed among F₁ individuals of common wild provenance that were raised identically aside from their assignment to alternate Passiflora species. Wild females (the dams of test subjects) were sampled specifically for their natural breeding association with P. incarnata, a popular local host which was subsequently found to furnish superior offspring development (figure 1). Despite circumstantial reasons to therefore expect a preference for this host [18], a strong bias towards P. suberosa was induced after a single generation of natal exposure. Interestingly, this finding was supplemented by an indication of natal host-based mate assortment, a phenomenon known in nocturnal lepidopteran taxa to arise from experiential conditioning for host-derived cues [8,9]. I expand upon these results according to their relative strength of support and in light of the broader evolutionary significance of host-induced plasticity.

A notable result of this study concerns the statistical evidence for host-induced behaviour in both sexes; that is, biases expressed not only via female oviposition choice but also via the host-proximity of courting males (figure 2b,c). Bisexually aligned plasticity of this nature is rarely demonstrated, but has likely consequences for processes such as localized adaptation, ecological divergence and speciation [43]. In reality, however, the present sample of male observations is probably too limited to conclusively establish this effect. It is also possible that direct olfactory [26] or visual [38] cues from females operating near each host elicited more frequent or prolonged courtship attention from males, and hence that male host proximity was in fact driven by female attractiveness (see further below). For these reasons, the observed host treatment effect upon male behaviour is best considered conservatively as a trend, albeit one that aligns with the more robust effect in females. Bearing this in mind, the indication of host-assortative courtship has notable congruence with the limited available evidence for positive mate assortment (figure 2d). Such conformity among different features of the results adds some credence to each finding individually. Host-assortative mating behaviour is known in moths, where it has been studied according to two main, non-exclusive hypotheses: (i) that host orientation biases common to each sex predispose mate assortment owing to shared habitat proximity (sensu [32]), and/or, (ii) the presence of host-induced associative mate preferences [33,34]. Neither hypothesis is presently discountable for H. charithonia. Other potential drivers of mate assortment are however either rejected by the data (e.g. body size) or discounted by the design (e.g. age and host proximity upon eclosion). The results of this study pose relationships among natal host experience, male host orientation and mate-seeking behaviour, and mating preferences as key areas for future research in heliconiines.

An important feature of the experimental design involved selectively sampling wild females from a localized P. incarnata-based breeding site. This was done in order to assess whether F₁ adults prefer this species irrespective of natal host assignment, which would be expected if their natural incumbency at the sampling site reflected an innate bias. The data cannot strictly discount this possibility (or its premise), but they do indicate a strong inducible component to adult host orientation, at least in daughters. It remains possible, for example, that test subjects in fact harboured an innate preference that was over-ridden by natal host experience [10,44]. In any event, the strong observed plasticity runs counter to the existing consensus for canalised adult host preference in diurnal Lepidoptera [14,20,21]. Some precedent is given by the occurrence of host induction in moths [9] and by recent findings in the butterfly Pieris rapae [15]. In the latter case, ovipositing females were found to favour hostplant foliage matching the nitrogen content that they themselves experienced, with their offspring moreover pre-disposed for higher performance in alignment with this bias [15]. Such findings support ‘transgenerational acclimatization', and invoke inducible variation in gene expression [45] that may shape hostplant interactions not only within an individual's lifespan but across subsequent generations.

This investigation was motivated partly by knowledge of inducible larval feeding preferences in heliconiines [30]. Interestingly, the sensitive period for induction occurs late in larval development, perhaps in the pre-pupal instar [30]. This is relevant in light of what is presently known about lepidopteran post-metamorphic memory retention. The most definitive work [12] has shown that associative odour-based conditioning can indeed ‘survive' metamorphosis, but largely (if not exclusively) when established in the final instar. Studies of the ontogeny of neurogenesis in holometabolous insects have inturn identified a candidate region of the brain—the α′/β′ mushroom body lobe—which develops in the pre-pupal instar and retains neural integrity into adulthood [46]. A proximate basis of this nature would account for the formation of host biases in late-stage heliconiine larvae [30] that carry over to the adult phenotype; that is, pre-imaginal conditioning as envisaged by Hopkins [4,5]. While this could account for the patterns of host association observed here, it remains, of course, possible that plant residues in the gut or upon the pupal cuticle caused or co-contributed to adult preference [13]. The present approach can, however, rule out any influence of hostplant proximity upon eclosion because all adults emerged under laboratory conditions in isolation of their host (see the electronic supplementary material, figure S1c-d) and were naive to the glasshouse environment.

Finally, given that lepidopterans face a trade-off between larval resource acquisition and feeding duration [47,48], the higher growth rates observed on P. incarnata indicate greater suitability of this hostplant for the sample population. A faster rate of pupal development (despite larger size), is likewise consistent with the ability to more efficiently assimilate and/or appropriate host-derived resources. Both lines of evidence suggest that—all else being equal—a bias towards P. incarnata would prove adaptive in the wild. As discussed by Janz et al. [14], however, when ‘all else' does, in fact, prove equal will depend upon myriad features of (at least) host ecology, breeding phenology and population demography [29]. If P. incarnata is locally ephemeral in abundance or foliar quality, for example, such a bias may prove maladaptive by delaying or misguiding reproductive behaviours in either sex. Individuals that instead reproduce in association with P. suberosa (and potentially other hosts) under such conditions would produce offspring that possess modified biases and potentially realize higher inclusive fitness as a result. Possible scenarios of this nature are non-trivial considering the critical role that hostplant occurrence plays in determining the adult ‘home range' and overall recruitment of H. charithonia populations elsewhere in Florida [49]. Inducible biases may, therefore, contribute a degree of population-level lability that in turn facilitates more accurate tracking of environmental resource availability. The consequences of such associative plasticity for long-standing hypotheses in coevolution [25], phenotypic variation [1,38], sympatric speciation [35] and adaptive divergence (e.g. Baldwin's ‘organic origin’ hypothesis [50]) offer prominent avenues for future investigation.

Data accessibility

The datasets supporting this article have been uploaded as part of the electronic supplementary material.

Competing interests

I declare I have no competing interests.

Funding

This research was supported jointly by the Australian Research Council Discovery grants DP140104107 and Macquarie University internal funding.