Biological Journal of the Linnean Society (2001), 72: 33–45. With 7 figures doi: 10.1006/bijl.2000.0484, available online at http://www.idealibrary.com on

Phenotypic plasticity in field populations of the tropical butterfly bolina (L.) ()

DARRELL J. KEMP∗

School of Tropical Biology, James Cook University, P.O. Box 6811, Cairns, Queensland 4870,

RHONDDA E. JONES

James Cook University, Townsville, Queensland 4811, Australia

Received 22 February 2000; accepted for publication 15 August 2000

Phenotypic plasticity may enable organisms to maximize their fitness in seasonally variable environments. However, in butterflies, seasonal polyphenism is often striking but functionally obscure. This paper addresses the possible adaptive significance of phenotypic variation in the tropical butterfly (L.) (Nymphalidae). Plasticity in body size and wing coloration can be elicited in this under laboratory conditions, however it is not known how this plasticity is expressed in the wild. Moreover, adult H. bolina spend the winter dry season in a reproductive diapause, which allows certain predictions regarding the occurrence of seasonal plasticity. Based on consideration of the requirements of diapausing and directly developing individuals, we predicted that if seasonal plasticity in phenotype were adaptive, then overwintering individuals should be larger and darker than their directly developing counterparts. This prediction was largely – although not entirely – fulfilled. Dry season butterflies were duller and darker than their wet season counterparts (this plasticity was superimposed on a genetic colour ), however size plasticity varied geographically. Dry season adults were consistently larger than wet season adults in the tropical north, but not in the south. We use these findings to discuss the possible adaptive significance of seasonal variation in the colour and size of this tropical butterfly.  2001 The Linnean Society of London

ADDITIONAL KEYWORDS: polyphenism – wing pattern – diapause – overwintering – body size – .

INTRODUCTION – Dingle, 1982; Jones, 1987), genetic polymorphism, and ecological generalization (Van Buskirk, McCollum Organisms that live in seasonal environments are & Werner, 1997). Alternatively, species may allow for faced with the problem that, because their living con- plasticity in their phenotype that is linked to the ditions vary from season to season, the best way to environmental variation, and hence, attempt to match make a living may vary throughout the year. In this their phenotype to their environment. sense, these organisms are faced with a ‘moving target’, The idea of phenotypic plasticity (the ability of a whereby there is not just one optimum phenotype (for single genotype to give rise to different phenotypes maximization of reproductive success), but several (or depending on the environment – Shapiro, 1976; Nylin, a continuum of) optima corresponding to different sea- 1994), is not new. Indeed, Darwin (1859) recognized sonal environments (McLeod, 1984). There are several that the environment may have ‘some direct action’ ways of dealing with this problem, including selection on phenotypes. Until recently, however, most of homogenous or favourable habitats (escape in space research has focused on those elements of phenotypic variation that are attributable to genetic factors. This may have been because environmentally-mediated ∗ Corresponding author: E-mail: [email protected] variation was viewed as non-heritable, and therefore 33 0024–4066/01/010033+13 $35.00/0  2001 The Linnean Society of London 34 D. J. KEMP and R. E. JONES unavailable for selection (Gotthard & Nylin, 1995). If plasticity in butterflies is adaptive, then there are Currently, the study of phenotypic plasticity is ex- clear reasons to expect that plasticity in field popu- panding rapidly, with significant advances being made lations of H. bolina should proceed in a specific manner. in the area of demonstrating adaptation (see reviews This is due to the breeding phenology of this species, in Gotthard & Nylin, 1995; Nylin, Gotthard & Wiklund, which includes a discrete annual period of reproductive 1996). dormancy. In Australia, H. bolina overwinter as adults One animal group that received considerable early (Jones, 1987) in sheltered gullies and creek beds in attention in the study of phenotypic plasticity is the coastal areas. There are striking differences in be- . Butterfly species show seasonal plasticity haviour and habitat selection between wet season fe- (i.e. polyphenism – Shapiro, 1976) with regard to adult males, which mate and oviposit soon after eclosion, and colour pattern (Watt, 1968, 1969; Hoffman, 1974; dry season females, which congregate at overwintering Shapiro, 1976, 1977; Douglas & Grula, 1978; Brakefield sites shortly after eclosion. Overwintering individuals & Larsen, 1984; McLeod, 1984; Rienks, 1985; Ritland, may remain behaviourally dormant at a single site for 1986; Brakefield, 1987; Kato & Sano, 1987; Brakefield up to 118 days (D. J. Kemp, unpublished data), whereas & Reitsma, 1991; Jones, 1992; Braby, 1994; Windig et ovipositing wet season females are more mobile and al., 1994) and size at maturity (James, 1987; Nylin & visually apparent (Kemp, 1998). We expect that these Svard, 1991; Braby, 1994), and these are often linked different developmental pathways would call for dif- to diapause (e.g. Yoshio & Ishii, 1994; Braby & Jones, ferent optimal phenotpes, and hence that plasticity in 1995). Interestingly however, although this group is both coloration and body size can be predicted in popular for study, the adaptive significance of seasonal certain directions (outlined specifically below). In this plasticity is often unclear (Gotthard & Nylin, 1995). paper, we aim to investigate (a) the degree to which This is especially true for some of the most striking field populations of H. bolina exhibit seasonal plas- examples of butterfly polyphenism (e.g. Araschnia le- ticity, and (b) whether the observed plasticity is con- vana, Catopsilia pomona, Polygonia c-album, sistent with predictions drawn from the adaptive octavia), which raises the important question of explanation. We therefore aim to suggest adaptive whether all of these polyphenisms are really adaptive. explanations for phenotypic plasticity in this butterfly It is possible that they relate simply to environmentally species, but not to explicitly test these explanations imposed constraints or some incidental effect that is (which would require some form of experimental similarly correlated with the generations. This ques- manipulation, such as reciprocal translocations – see tion is addressed here by examining phenotypic vari- Shapiro, 1976; Gotthard & Nylin, 1995; Kingsolver, ation in the tropical nymphalid butterfly Hypolimnas 1995, 1996). bolina (L.). The first prediction is that optimal body size should Hypolimnas bolina is a sexually dimorphic species differ between dry and wet season butterflies, with that exhibits a striking female-limited genetic colour diapausing (dry season) individuals advantaged by polymorphism (Poulton, 1924; Clarke & Sheppard, relatively larger size. This is because the ability of a 1975). Althugh research has focused exclusively on the butterfly to survive for long periods in a quiescent genetics of colour variation in this genus (e.g. Clarke state will depend on its capacity to store water and & Sheppard, 1975; Smith, 1976; Vane-Wright, Ackery other nutrients relevant to overwintering survival & Smiles, 1977), phenotypic plasticity is likely to ac- (Danks, 1987; Ohgushi, 1996), such as fat reserves count for some variation in the female phenotype (Chaplin & Wells, 1982). A larger animal has greater (Poulton, 1924; Clarke & Sheppard, 1975). Recent re- storage capacity, and also a smaller surface area:vol- search, using a split-family experimental design ume ratio, which would minimize the rate of evap- (method described by Nylin, 1994), has demonstrated orative water loss during the overwintering period. plasticity in coloration, as well as body size, in response The ability to resist desiccation is critically important to temperature and photoperiod in this species (Kemp, to the long term survival of an overwintering 2000). This result shows that plasticity in this species (Danks, 1987). Since adult male and female H. bolina is not entirely a simple effect of temperature upon insect are believed to diapause throughout their coastal range metabolic rates (refer to Atkinson, 1994). However, in tropical and sub-tropical Australia, we predict that although this experimental work indicates that plastic if body size plasticity is adaptive, then both sexes of responses are possible in H. bolina, the magnitude and all populations should exhibit plasticity in the direction seasonal alignment of plasticity expressed in the wild of larger dry season individuals. is unknown. Also, phenotypic attributes may be de- Secondly, we predict that dry season female H. bolina termined by additional environmental factors, such as should generally be darker than directly reproducing humidity and/or rainfall (Roskam & Brakefield, 1999), individuals. There are two adaptive reasons to expect which makes it important to investigate phenotypic this. The first, proposed by Brakefield & Larson (1984), variation as it occurs under field conditions. is that dry season morphs of adult-diapausing tropical PHENOTYPIC PLASTICITY IN H. BOLINA 35 butterflies should diverge from the wet season pheno- affect thermal load relevant to overheating risk (King- type in the direction of greater crypsis. If overwintering solver, 1987, and references therein), we expect that and directly reproducing butterflies select different wet-season individuals should be more advantaged by habitats, then each seasonal morph should appear relatively lighter wing markings. most cryptic in its respective habitat (see Tauber & Female H. bolina bask with their dorsal wing sur- Tauber, 1981a,b). In the case of H. bolina, over- faces exposed. In this ‘dorsal basking’ pose, the wings wintering individuals typically select dark, sheltered may contribute to body temperature through con- gullies, whilst reproductively active females operate ductance of heat gained through absorbed solar ra- close to the ground along sunlit forest edges, clearings diation. However, because of the low thermal and along exposed creek lines (Rutowski, 1992; Kemp, conductivity of butterfly wings, apparently only the 1998). Judging from beak marks on the wings of field- basal areas of dorsal wing surfaces (the inner third) caught butterflies, and observed predatory attempts may contribute significantly to heat gain in this man- (D. J. Kemp, unpublished data), females in both these ner (Kammer & Bracchi, 1973; Kingsolver, 1987, and situations are exposed to from visual hunting references therein). Similarly, only the basal areas bird predators. Thus, to maximize crypsis, we would of both wing surfaces are likely to contribute, via expect that overwintering individuals would be most conduction of absorbed heat, to overheating risk. advantaged by darker wing markings, and directly Hence, the thermoregulatory-based prediction that dry reproducing individuals by relatively lighter markings. season female H. bolina should be relatively darker than wet season butterflies pertains primarily to basal This prediction pertains specifically to ventral wing wing areas, particularly of the dorsal wing surface. surfaces, since overwintering butterflies spend long In this study we investigate variation in the wing periods (several days at a time) perching with wings coloration of females only. We excluded males because closed in shaded microenvironments. Ventral and dor- male wing coloration is potentially subject to sta- sal surfaces are exposed equally by actively ovipositing bilizing or directional selection in the context of female females (Kemp, 1998). mate choice (see Rutowski, 1992). Although the ad- The second reason to expect overwintering in- aptive predictions posed here would apply equally to dividuals to be darker in colour relates to thermo- males, sexual selection on their colour pattern may regulation. Since overwintering butterflies begin mediate against plasticity. Males of this strongly sexu- reproductive activity in the spring, when ambient tem- ally dimorphic show almost no variation (genetic or peratures are relatively low (Kemp, 1998), dark wing otherwise) in wing colour pattern (Clarke & Sheppard, markings may be thermally beneficial at these times 1975; D. J. Kemp, unpublished data). (Watt, 1968, 1969; Kingsolver, 1995, 1996). Post-dia- pause female H. bolina specifically target freshly ger- minated seedlings of the larval foodplant, MATERIAL AND METHODS nodiflora (Asteraceae), and these seedlings germinate PHENOTYPIC VARIATION IN FEMALE H. BOLINA en masse in response to spring rains (Kemp, 1998). In Four separate (supposedly genetic) colour forms have order to utilize this transient new season growth, been described for female H. bolina – nerina, eu- ovipositing females operate under relatively cool con- ploeoides, pallascens and naresi, of which detailed ditions, and intersperse periods of oviposition with descriptions and colour photographs can be found in periods of basking (Kemp, 1998). The occurrence of Clarke & Sheppard (1975). Across most of , female basking behaviour (described below) indicates that H. bolina are largely monomorphic for the form eu- females in this stuation are frequently below their ploeoides, which is believed to mimic various species thermal optimum. As such, the ability to efficiently of danaine butterflies in the genus (Vane- elevate body temperature is likely to contribute to the Wright et al., 1977). In northern Australia, New realized fecundity of these post-diapause females, as it and throughout the islands of , females does in some temperate butterflies (refer to Kingsolver, are polymorphic, with the forms nerina, pallascens and 1987, and references therein). Dark wing markings naresi present, along with intermediates between these would contribute to this process of heat gain (see forms (Clarke & Sheppard, 1975). These forms are below). By contrast, wet-season butterflies are faced believed to be non-mimetic, although Vane-Wright et with high to extreme temperatures and levels of solar al. (1977) suggest that “orange splashed nerina forms” radiation across their Australian range, and butterflies may be mimics of Danaus affinis (F.). Although discrete may risk overheating under these conditions (King- colour forms are recognized for this species, these solver & Watt, 1983; Kingsolver, 1987). Overheating, represent ‘endpoints’ along several dimensions of con- even for short periods, has been shown to reduce tinuous variation (see later). survivorship and fecundity in temperate pierid species Phenotypic colour variation in female H. bolina en- (Kingsolver & Watt, 1983). Since wing coloration can compasses two primary elements – (1) variation in the 36 D. J. KEMP and R. E. JONES

B of seasonal and geographic variation. Butterflies were C surveyed from collections held at the Australian Na- tional Insect Collection (Canberra), the Australian Mu- seum (Sydney), the Queensland museum (Brisbane), and the Museum of Tropical Queensland (Townsville). A Smaller collections held at James Cook University (Townsville), the University of Queensland (Brisbane), the Queensland Department of Primary Industries D (Brisbane) were also used, along with butterflies con- E tained within the senior author’s private collection. F For each specimen, forewing length (from apex to insertion) was measured to the nearest 0.5 mm using Figure 1. Dorsal (left) and ventral (right) coloration of a set of plastic callipers, and the date and place of female H. bolina, indicating the dorsal colour elements capture was recorded (where supplied). Wing wear was quantitatively assessed in this study (A–F), and the subjectively classified as either ‘fresh’ or ‘worn’, using ground colour region (stippled area) of both surfaces. the extent of lost scales in addition to the number of wing chips (extent of ‘feathering’ of wing margins) as characters. These age assessments were used only as tone of the overall wing surface, and (2) variation in a basis for excluding worn individuals from as- the shape and size of discrete colour patches. Tonal sessments of colour pattern. The dorsal surface of all variation is seen on both dorsal and ventral wing female specimens was then photographed, using an surfaces, but the tone of both surfaces varies in concert Olympus OM-1n SLR camera fitted with an Achiever – individuals are either relatively ‘dark’ or ‘light’ on TZ250 electronic flash unit, using 100 ASA Kodak both wing surfaces, but never ‘dark’ on one side and Ektachrome slide film. The camera was set to an ‘light’ on the other (D. J. Kemp, unpublished data). aperture size of f/1.8 for all photographs. Similarly, tonal variation is not localized to specific regions of wing surfaces, but is seen across the entire surface of both wings. This source of variation impacts QUANTIFICATION OF FEMALE DORSAL COLORATION most obviously on the tone of the brown and rufous For all females with ‘fresh’ wing wear ratings, the tone base coloration of much of the dorsal and ventral wing of the ground colour on both dorsal and ventral surfaces surfaces (hereafter, ‘ground colour’; Fig. 1), and may of the fore- and hind-wings (stippled region on Fig. therefore bear most relevance to the outlined adaptive 1) was visually assessed as either ‘dark’ or ‘light’. predictions. By contrast, the second source of variation, Reference specimens were used to maintain con- that affecting the size and shape of discrete colour patches, proceeds primarily on the dorsal wing surfaces sistency throughout this classification. Although the (Clarke & Sheppard, 1975). This source of variation tone of ground colour varies in concert on both wing affects the patches of orange, white, and blue, on both surfaces, variation in the overall colour pattern (as fore- and hind-wings (refer to Fig. 1), and has been earlier reviewed) is limited chiefly to dorsal surfaces. used primarily to distinguish between genetic colour Six dorsal colour elements were therefore used to forms (e.g. Clarke & Sheppard, 1975; Vane-Wright et quantify this variation (A–F; Fig. 1). These were (A) al., 1977). Little is known about how these two sources the area of the orange-brown forewing patch, (B) the of variation interact, and whether either has a seasonal area of subapical white forewing bar, (C) the area of basis. suffused blue on the forewing, (D) the area of orange- brown on the hindwing, (E) the area of the white hindwing patch, and (F) the area of blue scales fringing ASSESSMENT OF PHENOTYPIC VARIATION the white hindwing patch. The approximate area of In order to investigate phenotypic plasticity within these elements was quantified by projecting slides of field populations of H. bolina, variation in colour (fe- female H. bolina (‘fresh’ wing condition) onto graph males only) and size (both sexes) was investigated paper (5 mm squares) and counting the number of by surveying individuals held in museum collections. squares covered (by greater than 50% area) by each Similar collections have been used previously to in- colour element. To standardize for wing size, slides vestigate quantitative variation in seasonal phenotype were projected from a distance chosen so that the to examine ecologically relevant questions (e.g. Nylin forewing length (from apex to insertion) of each speci- & Svard, 1991; Holloway, 1993; Roskam & Brakefield, men measured approximately 190 mm on the graph 1999). These collections are not ideal, since the in- paper. This task was completed over several days, formation is constrained by the location and inclination using the same observer, with cross-checks consistently of collectors, but they can provide strong indications made between specimens to ensure consistency in the PHENOTYPIC PLASTICITY IN H. BOLINA 37

300 NT CQ S 250 NQ

200

150

Number of specimens 100

50

0 J F M A M J J A S O N D Month

Figure 2. Monthly totals of H. bolina (sexes pooled) sampled from museum collections. S=South, CQ=Central Qld., NQ=North Qld., and NT=Northern Territory. The boundaries of these geographic regions are explained in the text. subjective assessment of colour patch boundaries. In defined with four levels – Summer (December– addition to these assessments, the dorsal coloration of February), Autumn (March–May), Winter (June– all females was classified visually with reference to August), Spring (September–November)—and the genetic colour form classifications of Clarke and GENOTYPE was included to test whether seasonal Sheppard (1975). Individuals were classed as the gen- differences were consistent across the identified gen- etic form (or the intermediate) to which they most etic forms. Any geographic variation in colour was closely resembled. assessed by performing further factorial ANOVAs on each component, using SEASON and LOCATION as independent variables (levels provided in results). STATISTICAL ANALYSES Prior to conducting ANOVAs,dependent variables were Factorial analysis of variance (ANOVA) was used to screened for normality using the Kolmogorov–Smirnov determine any sexual, seasonal, or geographic dif- test, and appropriate transformations (provided in ferences in winglength. In this analyses, SEASON was results) were used to correct for deviation from nor- categorized as either dry (June–November) or Wet mality. All ANOVAs were performed using type III (December–May), and LOCATION was categorized as sums of squares to allow for unbalanced data (Shaw ° either South (South of Bundaberg; approx. 25 S), Cent- & Mitchell-Olds, 1993), and =0.05 was chosen as the ° ral Queensland (Bundaberg–Bowen; approx. 20–25 S), critical level of significance for all tests. The relevance North Queensland (Bowen–Torres Strait; approx. of season and geographic location to variation in the ° 10–20 S) or Northern Territory (West of Karumba; tone of the ground colour was assessed using Chi- ° approx. 141 E). In order to examine differences in Squared analyses. coloration, variables describing the six elements of female dorsal coloration (A–F; defined above) were reduced to orthogonal factors using principal com- RESULTS ponents analysis. All specimens were divided into gen- etic forms on the basis of visual classifications in SAMPLED NUMBERS conjunction with obvious non-seasonal quantitative 503 male and 563 female specimens were surveyed in groupings (outlined in results). Principal components this study, with most butterflies sampled from the were then used as dependent variables in separate North Qld. and South geographic regions. When pooled ANOVAs conducted to determine seasonal differences across all localities, the numbers of males and females 2 = in wing phenotype. In these analyses, SEASON was were distributed similarly among months ( 11 13.4, 38 D. J. KEMP and R. E. JONES

Table 1. Factorial ANOVA of forewing length, with sex (2 levels), season (2 levels) and location (4 levels) as independent factors. Significant effects (P<0.05) are in bold

Effect MS df FPvalue

Sex 1776.2 1 93.2 <0.001 Season 261.3 1 13.7 <0.001 Location 40.0 3 2.1 0.098 Sex×season 5.7 1 0.3 0.584 Sex×location 6.0 3 0.3 0.813 Season×location 113.2 3 5.9 <0.001 Sex×season×location 8.9 3 0.5 0.705 Error 19.1 1020

P=0.268), indicating that sexes were sampled equally the north accounted for the seasonal size plasticity not at all times. The total number of specimens caught recorded from southern populations. each month in each geographic region (males and Although there was significant sexual size di- females pooled) is depicted by Figure 2. Captures were morphism (SEX effect; Table 1), the lack of significant high from January to May, and low from June to interaction between SEX and SEASON indicated that December, with these two periods separated by rel- both males and females showed similar seasonal re- atively sudden changes in monthly capture totals. gimes of variation. Hence, females were larger than The distribution of monthly captures was significantly males to the same extent in both directly developing different between the North Qld. and South geographic and diapausing generations. 2 = regions ( 11 63.9, P<0.0001), which appeared to be due to a slightly shorter period of high captures in the FEMALE DORSAL COLORATION south (beginning later). Initial visual classifications, made according to Clarke & Sheppard’s (1975) genetic forms, placed most in- SIZE PLASTICITY dividuals as either nerina (n=417) or euploeoides- Mean forewing length of all surveyed individuals nerina (n=119). Other genetic forms, including pure varied significantly with SEX and SEASON (Table 1), euploeoides (n=6), naresi (n=8) and pallascens (n= and the effect of these two variables (with season 9), occurred rarely in collections. Principal components divided monthly) is illustrated by Figure 3. As pre- analysis indicated that of the six measured dorsal dicted by the adaptive size hypothesis, dry season colour elements, the areas of like colours (orange, white butterflies were on average considerably larger than and blue, respectively) on the fore- and hind-wings their wet season counterparts. However, this seasonal were significantly correlated. The original six colour variation was not consistent across all geographic variables were therefore reduced to three, and with populations. Although mean winglength did not differ the component axes varimax rotated, these cor- significantly between populations (there was no effect responded to areas of orange (PC1), white (PC2) and on winglength due to LOCATION alone), there was blue (PC3) on both dorsal wing surfaces (Table 2). These significant interaction between LOCATION and SEA- three components explained 87.1% of the variance SON (Table 1). As shown by Figure 4, the degree of described by the original six. seasonal plasticity in winglength decreased uniformly Figure 5 provides a plot of each sampled specimen from the most northerly populations to the most south- in three dimensional ‘phenotype space’, with each axis erly, with southerly populations (South and Central of the graph representing the value of each respective Qld.) exhibiting little seasonal plasticity in size. colour component, and individuals represented by their This interaction effect was further investigated by visual colour classification symbol. Individuals se- conducting separate ANOVAs for dry and wet season gregate into two appreciably different forms only on the butterflies. Although mean winglength of wet season white (PC2) axis, and this axis differentiates primarily individuals differed significantly between the geo- between the forms classified visually as nerina and = graphic populations (F3,823 16.1, P<0.001), the size of euploeoides-nerina. All individuals were subsequently = = dry season individuals did not (F3,197 2.6, P 0.053). divided into two groupings on the basis of differences Hence, the observed geographic difference in size plas- along this component axis (nerina=PC2 <−0.5, eu- ticity related solely to the size at maturity of wet ploeoides-nerina=PC2>−0.5). The relative frequency season individuals. Smaller wet season butterflies in of these two groups did not differ significantly between PHENOTYPIC PLASTICITY IN H. BOLINA 39

52 18 24 154 22 23 23 48 89 5 19

22 82 69 44 23 17 47 155 20 45 10 10 40 86

Forewing length (mm) Forewing 54 59 15 36

32 Jan FebMar Apr May Jun Jul Aug Sep Oct Nov Dec Month

Figure 3. Annual variation in forewing length of male (Φ) and female (Ε) H. bolina specimens. Error bars indicate ±1 standard error of each sample mean, and sample sizes are given above each data point. The fitted lines are weighted least squares approximations fitted by the STATISTICA computer program.

50 2) and naresi (n=5) were excluded from analyses of seasonal variation. 46

SEASONAL COLOUR VARIATION 42 Seasonal variation in H. bolina loosely supported pre- 38 dictions drawn from the proposed adaptive hypotheses (crypsis and thermoregulation), with wet season but- 34 terflies lighter and brighter on both wing surfaces than Forewing length (mm) Forewing their dry season counterparts. This seasonal variation 30 impinged upon both sources of variation in this species South Central Q. North Q. NT (ground colour tone and dorsal markings). Firstly, Geographic region the tone of the dorsal and ventral ground colour was 2 = seasonally variable ( 3 48.1, P<0.0001), with lighter Figure 4. Interaction plot showing the mean forewing butterflies disproportionately common in summer size of H. bolina in each season across the different (Dec.–Feb.; the ‘height’ of the wet season; 81.4% ‘light’ Φ= Ε= geographic areas. Male points ( dry season males, butterflies) compared with the three other seasons wet season males) are joined by dotted lines; female points (range of 24.3–32.2% ‘light’ butterflies). Secondly, two- (Β=dry season females, Χ=wet season females) by solid way ANOVAs performed on PC1 (log transformed) and lines. PC3 (squared), using SEASON (4 levels) and GENO- TYPE (2 levels: nerina, euploeoides-nerina)asin- dependent factors revealed significant seasonal 2 = = seasons (Summer, Autumn, Winter, Spring; 3 3.53, differences in both orange (PC1; F3,214 9.44, P<0.0001) = = = P 0.317) or geographic locations (South, Central Qld., and blue components (PC3; F3,214 6.43, P 0.0003). 2 = = North Qld., Northern Territory; 3 3.65, P 0.302). Wet season butterflies had more extensive areas of On this basis, these were taken to represent two dif- both orange and blue than their dry season counter- ferent genotypes (as suggested by the classifications parts (Fig. 6). This plasticity was uniform across gen- = = of Clarke & Sheppard, 1975), and seasonal differences etic forms (no effect due to GENOTYPE [F3,214 0.7, P in wing phenotype were examined separately for each 0.553] and no significant interaction between SEASON = = (proposed) genotype. Owing to their relatively low and GENOTYPE [F3,214 1.1, P 0.350]). Due to its occurrence, individuals classified as pallascens (n= bimodality, seasonal variation in PC2 was assessed 40 D. J. KEMP and R. E. JONES

Table 2. Linear correlation between extracted principal com- ponents (varimax rotated) and measured colour elements. High- lighted component ladingss are >0.70, italicized loadings are significant at P<0.05, and n=317 for all comparisons

Extracted components Colour element PC1 PC2 PC3

Forewing orange 0.885 0.302 0.042 Hindwing orange 0.886 −0.099 0.235 Forewing white 0.016 0.949 0.081 Hindwing white 0.132 0.946 0.073 Forewing blue −0.006 −0.236 −0.898 Hindwing blue −0.367 0.085 −0.818 Explained variance 1.72 1.96 1.55 Proportion of total (%) 28.7 32.7 25.7

1.2 3

2 0.8

PC3 1 0.4

– 0 Blue component 0 –1

–2 –0.4

–3 Principle component value –0.8 –4 Win Aut Sum Spr –5 Season

6 Figure 6. The nature of seasonal plasticity in areas of PC1 5 Φ Μ 4 orange (PC1; and solid line) and blue (PC3; and – Orange component3 dotted line) markings on the female dorsal wing surface. 2 The fitted lines are weighted least squares approximations 1 fitted by the STATISTICATM computer program. 0 3 2 –1 1 0 –1 –2 PC2–White component GEOGRAPHIC COLOUR VARIATION In order to investigate geographic variation, two-way Figure 5. Plot of all surveyed female H. bolina in three- factorial ANOVAs were carried out on the transformed dimensional ‘phenotype space’. Individuals were visually (as above) values of PC1 and PC3, using SEASON (4 classified as the genetic colour forms nerina (Χ), eu- levels: Summer, Autumn, Winter, Spring) and LOCA- ploeoides-nerina (Α), naresi (Φ) and pallascens (+) prior to quantitative analyses. TION (2 levels: South [South and Central Qld.] and North [North Qld. and Northern Territory]) as in- dependent variables. Although seasonal effects had already been established, SEASON was included in the separately for each genotype (nerina and euploeoides- analysis to test for any interaction with LOCATION. nerina). One-way ANOVAs carried out on PC2 (un- These analyses indicated clear seasonal differences (as transformed), using SEASON (4 levels as above) as the already reported), but no effects due to LOCATION = = independent variable, indicated no significant seasonal (F1,288 2.3, P 0.131) and no significant interaction = = = = differences in either nerina (F3,159 0.70, P 0.553) or between SEASON and LOCATION (F3,288 2.1, P = = euploeoides-nerina (F3,55 1.8, P 0.158). Hence, even 0.100). Similar analyses were performed for the values within proposed genotypes, the area of white on the of PC2 separately for each genotype. Again no sig- wings of female H. bolina did not vary across the year. nificant effects were reported for either LOCATION PHENOTYPIC PLASTICITY IN H. BOLINA 41

= = 35 (nerina F1,150 0.34, P 0.561 and euploeoides-nerina = = F1,49 1.44, P 0.236) or the interaction between = = LOCATION and SEASON (nerina F3,150 0.51, P = = 30 0.676 and euploeoides-nerina F3,49 0.17, P 0.916). Also, there was no difference in the ratio of individuals with dark to light ground colour across the geographic 2 = = 25 zones ( 3 4.1, P 0.251). Temperature (Deg.C) Temperature INTERACTION BETWEEN SOURCES OF SEASONAL COLOUR 20 VARIATION D J F M A M J J A S O N In order to determine how the different sources of seasonal colour variation interact, two-way ANOVAs 400 were conducted on each dorsal colour principal com- ponent, using TONE (2 levels: light, dark) and GENO- 300 TYPE (2 levels: nerina, euploeoides-nerina)as independent factors. These analyses revealed sig- 200 nificant effects on PC1 (log-transformed) and PC3 Rainfall (mm) 100 (squared) due to TONE (F1,201>22.8, P<0.0001), with no interaction between TONE and GENOTYPE 0 (F1,201<3.1, P>0.080). No effect on PC2 (untransformed) = D J F M A M J J A S O N was detected due to TONE (F1,201 0.7, P>0.404). These Month results indicate that the tone of the ground colour (recall that this is expressed on both dorsal and ventral Figure 7. Seasonal variation in average daily maximum wing surfaces) varies in concert with the extent of temperature (top) and mean monthly rainfall (bottom) at orange and blue, but not white, on the dorsal wing Darwin – 12°42′S, 130°88′E(Φ; representative of North- surfaces, and that this relationship is consistent across ern Territory), Cairns – 16°89′S, 145°76′E(Β; rep- genotypes. resentative of North Qld.), Rockhampton – 23°38′S, 150°48′E(Ε; representative of Central Qld.), and Bris- bane – 27°39′S, 153°12′E(Μ, representative of South). DISCUSSION The region classified as the ‘dry season’ in this study is shaded. Data are taken from 47–55 year archives of the SIZE PLASTICITY Australian Bureau of Meteorology. In line with many other adult-diapausing butterfly species (e.g. Bicyclus spp. – Brakefield & Reitsma, 1991; Eurema spp. – Jones, 1992; Mycalesis spp. – populations, with butterflies in the south equivalent Braby, 1994; Polygonia c-album – Nylin, 1994; Karls- in size to overwintering northern butterflies through- son & Wickman, 1989), tropical populations of H. bol- out the entire year (Fig. 4). This result was not ori- ina exhibited clear seasonal size plasticity in the ginally predicted, since seasonal variation in direction of larger overwintering adults. Under ex- temperature and photoperiod increases to the south perimental rearing conditions, plasticity in body size (refer to Fig. 7), and therefore, the extent of size of H. bolina can be elicited by variation in both tem- plasticity would also be expected to increase in south- perature and photoperiod, with cooler temperatures ern populations (based upon the experimental data and shorter photoperiods both leading to larger adults presented by Kemp, 2000). (Kemp, 2000). Since individual H. bolina respond to The adaptive plasticity hypothesis predicts that, in photoperiod as well as temperature, the plastic re- order to maximize survival during diapause, over- sponse is not merely a direct physiological effect of wintering should grow to a larger size than temperature on larval development (refer to Atkinson, directly developing individuals (Braby & Jones, 1995). 1994). Furthermore, the magnitude of the seasonal This prediction was fulfilled by the seasonally plastic size variation observed in field specimens is com- northern populations of H. bolina, but not by southern parable with the range of sizes elicited by the rearing populations. So the question is: why do butterflies conditions of Kemp (2000). This experimental in- in the sub-tropical south not exhibit the predicted formation therefore provides the proximate basis for seasonal size variation? One possibility is that H. the size plasticity observed here, because on average, bolina is univoltine at southerly latitudes, and there- (the larger) dry season butterflies would develop under fore all individuals develop under the same set of cooler and shorter days than (the smaller) wet season environmental conditions. However, judging by the butterflies. However, this was only true of northern length of the breeding season (given by the five month 42 D. J. KEMP and R. E. JONES period of high captures in Fig. 2), this species has H. bolina. The non-seasonal variation corresponded to ample time to complete three to four generations in visual classifications made according to the genetic the south (D. J. Kemp, unpublished data). Another classification of Clarke & Sheppard (1975), and two potential explanation based on the adaptive hypothesis forms dominated museum material: nerina and eu- is that butterflies do not overwinter as adults at south- ploeoides-nerina. These forms differed chiefly in the erly latitudes. Under this scenario, southern butterflies extent of white scales on the hindwing patch and should mature at roughly the size of directly developing subapical forewing bar, which may correspond to dif- wet season butterflies (if this is indeed the optimum ferences in the allele frequencies at the first, ‘dark’ size for this developmental pathway), however this locus of Clarke & Sheppard’s (1975) two-locus genetic clearly is not the case. It therefore appears that the model. The seasonal variation corresponded to dif- adaptive requirements of diapause are not the only, or ferences in a correlated suite of colour elements the best, explanation for seasonal size variation in H. (ground colour tone and orange and blue dorsal mark- bolina. ings), and was much more subtle than these (proposed) Seasonal size variation in field populations of H. genetic differences. Overall, wet season individuals bolina may represent a mix of adaptive and non- tended to be brighter and lighter and dry season in- adaptive mechanisms. For instance, overwintering in- dividuals darker, and this plasticity was consistent dividuals of all populations may aim for larger body across geographic range and across genotypes. Sea- sizes, as adaptively predicted, but they may be unable sonal variation in the dorsal phenotype appeared to to reach this size at southerly latitudes. Since dry relate primarily to the tone of the ground colour, which season adults (those caught in June–August) are the when dark, obscured the suffused blue and orange product of late wet season larvae, foodplant quality markings. This tonal change in ground colour was may become limiting to larval growth at this time expressed equally on the ventral wing surfaces, al- (Jones, 1992). This would lead to the plasticity observed though colour pattern elements on this surface were here if late wet season foodplant quality becomes lim- not specifically quantified. iting only in the south, which is possible since this On a general level, the observed plasticity supported species uses different hostplant species at different the adaptive predictions that dry season (over- latitudes (Common & Waterhouse, 1981). Equally, it wintering) butterflies should be darker than their wet could be proposed that tropical wet season individuals season counterparts. However, plasticity was not seen simply run out of available larval foodplant, and that to act exclusively upon the wing areas of specific rel- this is responsible for pure nutritionally-driven size evance to each adaptive hypothesis (thermoregulation variation not seen at southerly latitudes (although or crypsis). For instance, only the basal areas of the plasticity under this scenario obviously arises from a dorsal wing surface should function in thermo- constraint, the ablity to pupate at small size in re- regulation (via dorsal basking—Kammer & Bracchi, sponse to seasonal resource limitation may itself be 1973; Kingsolver, 1987, and references therein), considered adaptive – Nylin, 1994). However, we con- whereas only the ventral surface should provide for sider this possibility unlikely, since larval foodplant crypsis throughout the overwintering period in this quality should be highest at the beginning of the species (see Introduction). By contrast, plasticity in breeding season. In tropical north Queensland, early surveyed individuals involved a correlated suite of wet season juvenile H. bolina feed on freshly ger- wing colour elements, impacting on the greater area minated seedlings of the annual plant S. nodiflora of both wing surfaces. The non-specific nature of this (Kemp, 1998), and this supports better growth and plasticity does not rule against possible adaptive func- survival than late season mature growth (D. J. Kemp, tion, because the tone of specific wing areas did vary unpublished data). Also, although seasonal variation as predicted, but it does make it difficult to evaluate in rainfall increases to the north (refer to Fig. 7), the which specific hypothesis (if not both) may be most finding that wet season individuals are smaller in the functionally relevant. tropics is counter-intuitive with respect to the idea Seasonal colour plasticity has been described in sev- that drought stress leads to reduced foodplant quality eral insect groups, and is often linked with the in- and hence, smaller adult size. In these regions, sea- cidence of reproductive diapause (see Danks, 1987 for sonal resource-based constraints would be most likely a review). In Eristalis hoverflies, individuals exhibit to result in smaller dry season butterflies (Jones, 1992) plasticity with respect to the coloration of the abdomen, rather than smaller wet season individuals as observed which has been suggested to serve a thermoregulatory here. function (Holloway, 1993, and references therein). This adaptive hypothesis has also been proposed to explain COLOUR PLASTICITY the seasonal plasticity in thoracic coloration in Dro- This study identified clear seasonal and non-seasonal sophila (Capy, David & Robertson, 1988). In butterflies, variation in the dorsal coloration of field-caught female colour plasticity impinges mainly upon the pigmented PHENOTYPIC PLASTICITY IN H. BOLINA 43 wing surfaces. Such variation has been reported for plasticity in life history traits, such as body size, should many tropical butterfly genera, including Bicyclus not be particularly clear cut. This is because size at (Brakefield & Reitsma, 1991; Windig et al., 1994; Ro- maturity, developmental time, and growth rates are skam & Brakefield, 1999), Catopsilia (Rienks, 1985), related in complex ways (Abrams et al., 1996) and are Eurema (Jones, 1992), Melanitis (Brakefield, 1987; potentially influenced by many factors (Gotthard & Yoshio & Ishii, 1994), Mycalesis (Brakefield & Larsen, Nylin, 1995). On this basis, data presented in this study 1984; Braby, 1994), and Precis (McLeod, 1984). Polyp- appear consistent with the hypotheses that phenotypic henisms in the satyrine genera (Bicyclus, Melanitis plasticity in tropical butterflies contains adaptive ele- and Mycalesis) act on either eyespot apparency and/ ments. However, in H. bolina and other tropical species or ventral ground colour, and have been interpreted (e.g. Catopsilia pomona, ), the proposed in light of seasonally alternating requirements for adaptive function of specific phenotypes across specific either cryptic or deflective wing patterns (Brakefield & seasonal environments remains to be demonstrated. Larsen, 1984; Brakefield, 1987; Brakefield & Reitsma, Experimental approaches, such as reciprocal trans- 1991; Braby, 1994). Members of the tropical pierid locations (see Shapiro, 1976; Gotthard & Nylin, 1995; genera Catopsilia and Eurema vary primarily in vent- Kingsolver, 1995, 1996), may prove most beneficial in ral wing colour and tone, however, the function of this unequivocally demonstrating the degree of adaptation plasticity is more obscure (Rienks, 1985), and may act of this currently puzzling source of phenotypic vari- for both crypsis and thermoregulation (Jones, 1992). ation. Interestingly, many examples of plasticity in but- terflies are reported to impinge only (or primarily) ACKNOWLEDGEMENTS upon ventral wing surfaces, which indicates that col- our plasticity may be expressed on one wing surface We acknowledge the assistance of Max Moulds (Aus- independently of the other. However, where variation tralian Museum), Michael Braby (Australian National in the tone of wing ground colour occurs, such as that Insect Collection), Geoff Monteith (Queensland Mu- displayed by female H. bolina, it generally proceeds seum), Chris Burwell (Queensland Museum) and John over the entire wing surface (see Douglas & Grula, Arnold (Museum of Tropical Queensland). So¨ren Nylin 1978; McLeod, 1984; Rienks, 1985; Jones, 1992). Thus, and Christer Wiklund also provided particularly help- it appears possible that mechanistic or developmental ful feedback on an earlier version of this manuscript. constraints may act against tonal variation of loc- This research was supported by an Australian Post- alized wing areas, such as specifically predicted by graduate Research Award to D.J.K. and James Cook the adaptive hypotheses here (although see Watt, University internal funding. 1969; Shapiro, 1977). The obvious non-adaptive pos- sibility is that selection on wing phenotype is not REFERENCES very strong, and that tonal variation occurs solely in response to seasonal variation in some environmental Abrams PA, Leimar O, Nylin S, Wiklund C. 1996. The factor or factors. Clearly, some degree of ex- effect of flexible growth rates on optimal sizes and de- perimentation is required to distinguish un- velopment times in a seasonal environment. The American equivocally between the possible adaptive and Naturalist 147: 381–395. incidental explanations for the occurrence of seasonal Atkinson D. 1994. Temperature and organism size – a colour plasticity in H. bolina. biological law for ectotherms? Advances in Ecological Re- search 25: 1–58. Braby MF. 1994. Phenotypic variation in adult Mycalesis CONCLUSION Hubner (Lepidoptera: Nymphalidae: Satyrinae) from the Although phenotypic plasticity in butterflies has been Australian wet-dry tropics. Journal of the Australian En- tomological Society 33: 327–336. extensively studied, there remain examples of striking Braby MF, Jones RE. 1995. Reproductive patterns and plasticity that are functionally obscure (Gotthard & resource allocation in tropical butterflies: influence of adult Nylin, 1995). This study represents one case where diet and seasonal phenotype on fecundity, longevity and there were clear adaptive grounds for expecting plas- egg size. Oikos 72: 189–204. ticity of a given nature in a species for which plasticity Brakefield PM. 1987. Tropical dry and wet season poly- had never been investigated, and was not immediately phenism in the butterfly Melanitis leda (Satyrinae): pheno- obvious. Colour variation in field populations of H. typic plasticity and climatic correlates. Biological Journal bolina agreed, in a general sense, with adaptive pre- of the Linnean Society 31: 175–191. dictions, whilst variation in body size proceeded as Brakefield PM, Larsen TB. 1984. The evolutionary sig- expected only in the tropical north. Although pre- nificance of dry and wet season forms in some tropical dictions regarding size plasticity were not entirely butterflies. Biological Journal of the Linnean Society 22: fulfilled, there are reasons to expect that seasonal 1–12. 44 D. J. KEMP and R. E. JONES

Brakefield PM, Reitsma N. 1991. Phenotypic plasticity, Kingsolver JG. 1987. Evolution and coadaptation of thermo- seasonal climate and the population biology of Bicyclus regulatory behaviour and wing pigmentation pattern in butterflies (Satyridae) in Malawi. Ecological Entomology Pierid butterflies. Evolution 41: 472–490. 16: 291–303. Kingsolver JG. 1995. Fitness consequences of seasonal Capy P, David JR, Robertson A. 1988. Thoracic trident polyphenism in western white butterflies. Evolution 49: pigmentation in natural populations of Drosophila simu- 942–954. lans: a comparison with D. melanogaster. Heredity 61: Kingsolver JG. 1996. Experimental manipulation of wing 263–268. pigment pattern and survival in western white butterflies. Chaplin SB, Wells PH. 1982. Energy reserves and metabolic The American Naturalist 147: 296–306. expenditures of monarch butterflies overwintering in Kingsolver JG, Watt WB. 1983. Thermoregulatory strat- southern California. Ecological Entomology 7: 249–256. egies in Colias butterflies: Thermal stress and the limits Clarke C, Sheppard PM. 1975. The genetics of the mimetic to adaptation in temporally varying environments. The butterfly Hypolimnas bolina. Philosophical Transactions American Naturalist 121: 32–55. of the Royal Society of London (B) 272: 229–265. McLeod L. 1984. Seasonal polyphenism in African Precis Common IFB, Waterhouse DF. 1981. Butterflies of Aus- Butterflies. In: Vane-Wright RI, Ackery PR, eds. The Bio- tralia. Sydney: Angus and Robertson. logy of Butterflies. Symposium of the Royal Entomological Danks HV. 1987. Insect dormancy: an ecological perspective. Society, 11. London: Academic Press, 313–315. Ottawa: Biological Survey of Canada. Nylin S. 1994. Seasonal plasticity and life-cycle adaptations Darwin C. 1859. The origin of species. London: Murray. in butterflies. In: Danks HV, ed. Insect Life-cycle Poly- Dingle H. 1982. Function of migration in the seasonal syn- morphism. Dordrecht: Kluwer Academic Publishers, 41–67. chronization of insects. Entomologia Experimentalis et Ap- Nylin S, Svard L. 1991. Latitudinal patterns in the size of plicata 31: 36–48. European butterflies. Holarctic Ecology 14: 192–202. Douglas MM, Grula JW. 1978. Thermoregulatory ad- Nylin S, Gotthard K, Wiklund C. 1996. Reaction norms aptations allowing ecological range expansion by the pierid for age and size at maturity in Lasiommata butterflies: butterfly, Nathalis iole Boisduval. Evolution 32: 776–783. predictions and tests. Evolution 50: 1351–1358. Gotthard K, Nylin S. 1995. Adaptive plasticity and plas- Ohgushi T. 1996. Consequences of adult size for survival ticity as an adaptation: a selective review of plasticity in and reproductive performance in a herbivorous ladybird animal morphology and life history. Oikos 74: 3–17. beetle. Ecological Entomology 21: 47–55. Hoffman RJ. 1974. Environmental control of seasonal vari- Poulton EB. 1924. in the butterflies of Fiji con- ation in the butterfly Colias eurytheme: effects of photo- sidered in relation to the Euploeine and Danaine invasions period and temperature on pteridine pigmentation. Journal of Polynesia and to the female forms of Hypolimnas bolina of Insect Physiology 20: 1913–1924. L. in the Pacific. Transactions of the Entomological Society Holloway GJ. 1993. Phenotypic variation in colour pattern of London 564–691. and seasonal plasticity in Eristalis hoverflies (Diptera: Rienks JH. 1985. Phenotypic response to photoperiod and Syrphidae). Ecological Entomology 18: 209–217. temperature in a tropical Pierid butterfly. Australian James DG. 1987. Effects of temperature and photoperiod on Journal of Zoology 33: 837–847. the development of Vanessa kershawi McCoy and Ritland B. 1986. The effect of temperature on expression of villida Godart (Lepidoptera: Nymphalidae). Journal of the the dark phenotype in female Papilio glaucus (Pa- Australian Entomological Society 26: 289–292. pilionidae). Journal of Research on the Lepidoptera 25: Jones RE. 1987. Reproductive strategies for the seasonal 179–187. tropics. Insect Science and its application 8: 515–521. Roskam JC, Brakefield PM. 1999. Seasonal polyphenism Jones RE. 1992. Phenotypic variation in Australian Eurema species. Australian Journal of Zoology 40: 371–383. in Bicyclus (Lepidoptera: Satyridae) butterflies: different Kammer AE, Bracchi J. 1973. Role of the wings in the climates need different cues. Biological Journal of the absorption of radiant energy by a butterfly. Comparative Linnean Society 66: 345–356. Biochemical Physiology 45: 1057–1063. Rutowski RL. 1992. Male mate-locating behaviour in the Karlsson B, Wickman P-O. 1989. The cost of prolonged common eggfly, Hypolimnas bolina (Nymphalidae). Journal life: an experiment on a nymphalid butterfly. Functional of the Lepidopterists’ Society 46: 24–38. Ecology 3: 399–405. Shapiro AM. 1976. Seasonal polyphenism. Evolutionary Kato Y, Sano M. 1987. Role of photoperiod and temperature Biology 9: 259–333. in seasonal morph determination of the butterfly Eurema Shapiro AM. 1977. Phenotypic induction in Pieris napi L.: hecabe. Physiological Entomology 12: 417–423. role of temperature and photoperiod in a coastal California Kemp DJ. 1998. Oviposition behaviour of post-diapause population. Ecological Entomology 2: 217–224. Hypolimnas bolina (L.) (Lepidoptera: Nymphalidae) in Shaw RG, Mitchell-Olds T. 1993. ANOVA for unbalanced tropical Australia. Australian Journal of Zoology 46: 451– data: An overview. Ecology 74: 1638–1645. 459. Smith DAS. 1976. Phenotypic diversity, mimicry and natural Kemp DJ. 2000. Investigating the basis of life history plas- selection in the African butterfly Hypolimnas misippus ticity in Hypolimnas bolina (L.) (Lepidoptera: Nymph- L. (Lepidoptera: Nymphalidae). Biological Journal of the alidae). Australian Journal of Zoology 48: 67–78. Linnean Society 8: 183–204. PHENOTYPIC PLASTICITY IN H. BOLINA 45

Tauber CA, Tauber MJ. 1981a. Insect seasonal cycles: pigment in relation to thermoregulation. Evolution 22: genetics and evolution. Annual Review of Ecology and 437–458. Systematics 12: 281–308. Watt WB. 1969. Adaptive significance of pigment poly- Tauber MJ, Tauber CA. 1981b. Seasonal responses and morphisms in Colias butterflies. II. Thermoregulation and their geographic variation in Chrysopa downesi: eco- photoperiodically controlled melanin variation in Colias physiological and evolutionary considerations. Canadian eurytheme. Proceedings of the National Academy of Sciences Journal of Zoology 59: 370–376. 63: 767–774. Van Buskirk J, McCollum A, Werner EE. 1997. Natural Windig JJ, Brakefield PM, Reitsma N, Wilson JGM. selection for environmentally induced phenotypes in tad- 1994. Seasonal polyphenism in the wild: survey of wing poles. Evolution 51: 1983–1992. patterns in five species of Bicyclus butterflies in Malawi. Vane-Wright RI, Ackery PR, Smiles RL. 1977. The poly- Ecological Entomology 19: 285–298. morphism, mimicry, and host plant relationships of Hy- Yoshio M, Ishii M. 1994. Photoperiodic determination of polimnas butterflies. Biological Journal of the Linnean seasonal morphs associated with reproductive diapause Society 9: 285–297. in Melanitis phedima oitensis (Lepidoptera, Satyridae). Watt WB. 1968. Adaptive significance of pigment poly- Japanese Journal of Entomology 62: 40. morphisms in Colias butterflies. I. Variation of melanin