Supplemental Table 1: Raw Data: Egg Heights and Clutch Sizes of E. Editha in Rabbit Patch-Pair

Supplemental Table 1: Raw data: Egg heights and clutch sizes of E. editha in Rabbit patch-pair

Egg Heights at Rabbit clearing on Collinsia in 1991 (cm). 2.4; 1.8; 1.1; 0.9; 0.9; 0.7; 0.6; 0.5; 0.4; 0.3; 0.3; 0.3; 0.2; 0; 0; 0; 0; 0.(Previously unpublished)

Egg Heights in Rabbit open-forest on Pedicularis in 1993 and 2002: 3.0; 1.1; 0.8; 0.8; 0.8; 0.7; 0.6; 0.5; 0.4; 0.4; 0.3; 0.3; 0.2; 0.2; 0.2; 0.1; 0; 0; 0. (more data in Singer and McBride 2010)

Clutch sizes at Rabbit in field, estimated to nearest ten when hard to count eggs wthout breaking up the clutch. Potentially reduced by predation from starting clutch size.

Clutch sizes in clearing on Collinsia in 1982: 140; 130; 120; 110; 110; 110; 103; 90; 85; 75; 75; 70; 62; 60; 60; 60; 55; 54; 53; 50; 50; 48; 47; 46; 45; 45; 45; 44; 40; 39; 36; 35; 34; 34; 33; 33; 30; 22; 22; 21; 18; 17; 15; 15; 14; 14; 11; 8; 7; 7

Clutch sizes in open-forest on Pedicularis in1982: 120; 120; 112; 100; 95; 95; 86; 85; 85; 85; 80; 75; 74; 68; 65; 60; 60; 60; 60; 57; 55; 55; 54; 53; 52; 52; 52; 50; 50; 50; 47; 47; 45; 45; 41; 41; 40; 40; 40; 40; 38; 37; 35; 35; 35; 35; 35; 33; 33; 32; 31; 30; 29; 27; 25; 25; 21; 21; 20; 20; 20; 20; 20; 19; 19; 18; 16; 14; 13; 13; 12; 12; 11; 11; 11; 8; 5; 3; 3.

Supplemental Table 2: Qualititative description of host use, host adaptation and host-specific survival of E. editha in adjacent habitat patches at Rabbit Meadow, 1979-84. For completeness, table lists minor hosts censused by Singer (1983) but not discussed in this paper. Details in Singer (1983) Moore (1989), Singer and Thomas (1996), Thomas et al. (1996), Boughton (1999).

Appendix 1: Butterfly amenability to studies of dispersal and local adaptation.

The size, manipulability and discrete population structure of many butterflies render them amenable to studies of dispersal (Ford and Ford 1930; Ehrlich 1961, Hanski 2011). For example, a small team of researchers took less than a month to perform mark-release-recapture of Euphydryas aurinia across 1,500km2 in the Czech Republic, incorporating all known populations in the country. They wrote individual numbers on the wings of 9,118 individuals, recaptured 2,911 and recorded movements within and among populations (Zimmerman et al. 2011).

Unlike fruit flies, butterflies can be followed as they search for resources. Their patterns of flight, alighting and resource discovery can be observed and recorded, revealing the identities of individual plants accepted and rejected for oviposition. With current technology it is feasible to attach radio transponders to individuals and track them with harmonic radar (Cant et al. 2005).

That butterflies indulge in local adaptation has become well-known with respect to mimicry (Jiggins et al. 2001) and, if the Peppered Moth is momentarily nominated as an honorary butterfly, camouflage (refs in Cook and Saccheri 2013). Therefore, we also expect to find local adaptation to physical characteristics and spatial distributions of habitats and the hosts that they contain. These expectations are fulfilled. Butterflies in one population of M.cinxia clung to an isolated, windy island by “improving their grip,” evolving stronger claws compared to conspecifics from a connected landscape (Duplouy and Hanski 2013). In response to a similar problem, a population of Lycaeides on a windy mountain evolved glue-free eggs that fell off the host and remained in the habitat when the hosts were blown off the mountain in winter (Fordyce and Nice 2003).

Local adaptation to selection on butterfly dispersal generates complex suites of behavioral, genetic and metabolic traits (Hanski 2011; Marden et al. 2013) whose correlations with known population history are well-replicated in a metapopulation context (Hanski 2011). Dispersal evolution has been informatively dynamic at range margins expanding under warming climate (Thomas et al. 2001, Buckley and Bridle 2014) and in anthropogenic landscapes (Ockinger and van Dyck 2012; Stevens et al. 2012). In parallel to these responses to selection on dispersal, local adaptation of butterflies to their hosts also involves complex suites of rapidly-evolving life-history and behavioral traits (Singer et al. 1993, Singer and McBride 2010; Bennett et al. 2014).

Cant ET, Smith AD, Reynolds DR, Osborne JL (2005) Tracking butterfly flight paths across the landscape with harmonic radar. Proc Roy Soc B 272:785-790

De Meeus T, Hochberg ME, Renaud F (1995) Maintenance of two genetic entities by habitat selection. EvolEcol 9:131-138.

Duplouy A, Hanski I (2013) Butterfly survival on an isolated island by improved grip. Biol Lett 9:20130020

Ford HD, Ford, EB (1930) Fluctuation in numbers and its effect on variation in Melitaea aurinia (Rottembourg, 1775) (Lepidoptera: Nymphalidae). Trans Roy Ent Soc Lond 78:345-351.

Fordyce, JA, Nice CC (2003) Variation in butterfly egg adhesion: adaptation tolocal host plant senescence characteristics? Ecol Lett 6:23-27

Gripenberg S, Mayhew PJ, Parnell M, Roslin T (2010) A meta-analysis of preference-performance relationships in phytophagous insects. Ecology Letters 13:383-39

Zimmerman K, Fric Z, Jiskra P et al (2011) Mark-recapture on large spatial scale reveals long distance dispersal in the Marsh Fritillary, Euphydryas aurinia. Ecol. Entomol. 36:499-510.

Appendix 2: Discrimination within and among host species, and effects of host quality and density on butterfly diet.

Example 1: Strong effects of host species on fitness of E. editha were found when fates of naturally-laid eggs were followed (Moore 1989) and when eggs were experimentally placed on different plants in the field (Singer et al 1994). Similar effects appeared in experiments in which growth and survival rates of captive larvae differed between host species by more than an order of magnitude (Rausher 1982, Singer and McBride 2010).

We classified as “potential hosts” those species that served as principal hosts of E. editha at some sites. We then asked how many of these potential hosts were available to each population and how many were actually used (Singer and Wee 2005). We censused 57 populations across an area of 250 x 1200km and found that 37 of the sites contained more than one potential host, but 43 of the 57 populations were monophagous. The insects typically failed to use the full range of hosts available to them: the majority of populations contained abundant hosts that were used elsewhere but not used at the focal site because they were not locally preferred (Singer and Wee 2005). These preference differences were important mechanistic causes of inter-site variation in diet: the butterflies had evolved post-alighting chemical preferences for different host genera at different sites and any one of five genera could be the most preferred (Singer 1971, Singer and Parmesan 1993, Singer et al. 1994; Singer and McBride 2012).

In the same survey of 57 populations the most abundant potential host was preferred at 19 sites and the least abundant at 12. At the other sites there was either a tie or no choice. The trend for the most abundant host to be preferred was suggestive but fell short of significance (p=0.07 by binomial test with expected 50:50 probability) (Singer and Wee 2005).

In 10 of the 57 populations we asked butterflies to rank the actual and potential hosts in their habitat in order of preference. We then ranked the same plants in order of their suitability for offspring by manipulating butterflies to lay eggs on them in the field and recording offspring survival. Of the ten populations, two were in the throes of rapid diet evolution at the time of the study (Singer et al. 1993). At these two sites a recently-acquired host was the most suitable but least preferred; in the other 8 cases the rank order of adult preference was exactly concordant with the rank order of offspring performance (Singer et al 1994). Since each site contained 3 or 4 hosts, this result represents a highly significant trend for concordance between preference and performance within the set of 8 “stable” populations. This result suggests that the quality of hosts had more influence on evolution of preference than their abundance. The same conclusion was drawn from a recent, separate study in the Western Sierra Nevada (California) in which E. editha used Pedicularis at some sites and Collinsia at others in a geographic mosaic, despite the presence of both hosts at all sites. Plant censuses across this set of sites showed no trend for variation of insect diet to be associated with variation of host abundance; instead it was clearly associated with host quality (Singer and McBride 2012).

Example 2: The Glanville Fritillary, M. cinxia, used just two hosts in the Åland islands: Veronica spicata and Plantago lanceolata. Some habitat patches contained only Plantago, many patches contained both hosts and a few contained only Veronica. In most patches where both hosts occurred, both were used by ovipositing butterflies. Surveys of survival from naturally-laid egg clutches were conducted in each of 7 years across 1,923 populations (van Nouhuys et al 2003). Mean survival per clutch was 20 on Plantago and 19 on Veronica, but even this small difference disappeared when analysis controlled for differences among habitats that might not necessarily be host-associated. Laboratory experiments showed very large differences in the ability to support larvaeamong individual plants within host populations, but no overall difference between plant species (van Nouhuys et al. 2003).

In sum, despite a great deal of research effort over many years, no evidence was found of any overall effect of host genus on fitness of M. cinxia: when plants were categorized taxonomically there was no detectable selection on host use. However, there WAS substantial heritable variation of post-alighting oviposition preference for Plantago versus Veronica, which emerged clearly when M cinxia from different parts of Åland were raised in Texas on a common host and preference-tested on the same plants (Kuussaari et al. 2000). This difference persisted in the laboratory for several generations (Singer and Lee 2000); however, differences in maternal preference were not associated with variation in offspring performance; there was no preference-performance correlation (van Nouhuys et al. 2003).

If these preference differences arose in the context of the equality of fitness on the two hosts that was observed, then, by elimination, we expect them to be responses to host density rather than host quality, resulting simply from the fact that preference for the more abudant host saves time by shortening oviposition searches. This expectation was fulfilled: tested post-alighting preferences varied across the metapopulation in parallel with variation in the relative abundance of the two hsosts (Kuussaari et al. 2000).

Discrimination within host species: as motivation increases during a search, the fraction of plants that would be accepted (if encountered) increases within each host species. For some butterflies, all members of the most-preferred host population are preferred over all members of the second-ranked species, but frequently there is overlap of acceptability between host species and insects may respond more strongly to chemical traits that vary within host populations rather than between different host species (Singer and Lee 2000). Regardless of this complication, butterflies must suffer costs when they fail to encounter their most-preferred hosts and are forced to continue to search until they accept lower-ranked hosts

The E. editha population at Rabbit Meadow contained some individuals that were specialists with respect to chemical variation within the host Pedicularis population and others that were generalists. Further, there was a preference/performance correlation in the strict sense of the term: specialists produced offspring that survived better on plants accepted by specialists than on rejected plants, while offspring of generalists survived equally well on plants accepted and ejected by specialists. The difference between the two classes of offspring in relative performance on the two plant classes was significant (Ng 1988).