Scea torrida, Miller, 2009

Miller, James S, 2009, Generic Revision Of The Dioptinae (Lepidoptera: Noctuoidea: Notodontidae) Part 2: Josiini, Bulletin of the American Museum of Natural History 2009 (321), pp. 675-1022 : 852-884

Scea torrida , new species Figures 304 , 353 ; plate 33 [EX]

DIAGNOSIS: Scea torrida and S. angustimargo are the only two Scea species exhibiting an orange longitudinal stripe along the HW anterior margin (pl. 33). They also differ from most other members of the genus in having extremely large, bulging eyes; the majority of Scea species possess relatively small eyes (e.g., fig. 346A–C, G). Scea torrida (FW length 5 17.0–18.0 mm) can be distinguished from S. angustimargo (FW length 5 14.0–16.5 mm) in its larger size and its less extensive orange FW triangle. The outer margin of the triangle in S. angustimargo reaches distally well beyond the DC, passing the fork of M 3 +CuA 1. In S. torrida , the orange triangle barely extends beyond the DC, and terminates at the fork of M 3 +CuA 1. The FW and HW ground color of S. torrida is slate gray, whereas that of S. angustimargo tends toward charcoal gray.

Male genitalia in S. torrida and S. angustimargo are extremely similar. Scea torrida differs in that St8 is constricted distally (figs. 347C, 353C), and the enlarged distal cornutus on the ventral appendix of the vesica is smaller and less robust (fig. 347E, 353E). Female genitalia in the two species differ in the configuration of the large, fluted area at the base of the CB (figs. 347D, 353D).

DESCRIPTION: Male. Forewing length 5 17.0–18.0 mm. Head: Labial palpus relatively short, curving strongly upward, its apex falling short of middle of front; Lp1 long, curved, with a dense, wedge-shaped ventral fringe; Lp2 shorter than Lp1, closely scaled; Lp3 short, bullet shaped; scales of front gray-brown, pointing dorsomedially, meeting between antennal bases to form a small tuft; clypeus and midline of front scaleless, glossy; occiput narrow, covered with short, semierect gray-brown scales; eye extremely large, bulging below ventral margin of head, gena absent; vertex covered with anteriorly pointing, slate-gray scales; antenna bipectinate, rami long; scape and dorsum of antennal shaft glossy slate gray.

Thorax: Legs evenly gray-brown; pleuron with elongate and hairlike slate-gray scales; patagium tightly covered with long, gray-brown scales; tegula gray-brown, covered with hairlike scales, these longest at margins; dorsum gray-brown to slate gray.

Forewing: (Dorsal) Ground color slate gray to gray-brown (pl. 33); an ochreous orange triangle in basal third, extending from base to immediately beyond distal margin of DC, terminating at fork of M 3 +CuA 1, its anterior margin touching SC, its posterior margin falling short of 1A+2A; veins within orange triangle, including midline of DC and anal fold, lined with gray-brown to slate-gray scales; orange area between midline of DC and anal fold with a dense covering of slate-gray to gray-brown scales. (Ventral) Ground color slightly darker than on dorsal surface; orange triangle without gray-brown veins.

Hind wing: (Dorsal) Uniformly dark slate gray (pl. 33); a faint ochreous orange stripe along anterior margin in basal half, located between Sc and radius, this stripe in some specimens obscure. (Ventral) Ground color gray-brown, slightly lighter in tone than on dorsal surface; ochreous orange stripe along anterior margin wider and much more conspicuous than on dorsal surface.

Abdomen: Dorsum and venter evenly slate gray, with a faint blue iridescence.

Terminalia (fig. 353A–C, E): Tg8 much narrower than Tg7, shorter than St8; anterior margin of Tg8 broadly concave, lateral margins slightly constricted in distal two-thirds, posterior margin with a deep, Ushaped mesal excavation; St8 approximately equal in width to St7, much longer; anterior margin of St8 forming a wide, short, truncate process, lateral margins gently convex in basal half, then constricted distally, posterior margin with a deep, quadrate mesal excavation; posterolateral angles of St8 forming broad, lobelike processes; socii/uncus complex narrow, delicate; uncus long and thin, apex slightly widened; socii narrow, laterally compressed, shorter than uncus; tegumen shorter than vinculum, much wider, abruptly expanded toward meson above junction with valva; vinculum tall, narrow in dorsal third, much wider ventrally; ventral margin of genitalia transverse, saccus small; valva large, mostly membranous; BO large, occupying two-thirds of valva; costa narrow, lightly sclerotized, sides roughly parallel, slightly narrower at base; costa forming a blunt, angular process at apex; transtillar arms narrow at bases, meeting in manica to form a wide, smooth, concave plate above aedeagus; aedeagus wide and bulbous at base, narrowed to form a pointed ventral process distally; ductus ejaculatorius simplex arising from dorsum of aedeagus near base; vesica extremely long, over four times as long as aedeagus, wider than aedeagus, with a narrow ventral appendix arising in distal third, this appendix bearing approximately 20 small, fine, spinose cornuti and a single enlarged distal cornutus.

Female. Forewing length 5 17.0 mm. Body characters similar to male except: labial palpus shorter, porrect, and curving slightly upward to clypeus; antenna ciliate, shaft moderately wide.

Terminalia (fig. 353D): Tg7 over twice as long as Tg6, gradually narrowed distally, anterior and posterior margins simple; St7 over twice as long as St6, somewhat narrow- er, anterior margin simple, lateral margins gently convex, posterior margin with a large, U-shaped membranous area at midline; Tg8 completely membranous; AA short, narrow; PP long, extremely thing; PA small, mostly membranous, bearing elongate setae, posterior margins slightly lobate near dorsum; postvaginal area comprising a narrow, transverse band, barely sclerotized; ostium relatively wide, lightly sclerotized, dorsoventrally compressed; DB extremely long, constricted a short distance from ostium, membranous in basal half, bearing a fluted, intestinelike internal sclerite in distal half; intestinelike sclerite of DB divided and extending along both sides of CB to occupy majority of corpus, terminating at distal third; signum located distally, comprising two clusters of long, internal spines; DS arising from venter of CB in basal half.

ETYMOLOGY: This name is taken from the Latin word torridus, meaning ‘‘dry’’ or ‘‘parched’’, in reference to the dry forest where this moth lives.

DISTRIBUTION: Scea torrida is known from a single locality in southern Ecuador (fig. 304)—along the road between Loja and Catamayo (Loja Province). The site, closer to Catamayo (03 ° 58 9 51 0 S, 79 ° 21 9 20 0 W), is on the western slopes of the Cordillera de Chilla, which drains into the Río Catamayo. The Río Catamayo ultimately becomes the Río Chira, emptying into the Pacific Ocean south of Negritos, Peru. The moth fauna of this fascinating region has barely been explored.

A male specimen of S. torrida , in the collection of the Oxford University Natural History Museum, bears two labels—one handwritten, seemingly stating ‘‘N. Gran.’’; and another with a handwritten number ‘‘1008’’. The first of these may refer to Nuevo Granada, once part of Colombia, which then included most of modern Ecuador. It would be reassuring to locate additional examples of S. torrida with more definitive locality information.

DISCUSSION: Claude Lamaire and Paul Thiaucourt collected the type series of S. torrida (3³³, 1♀) on a single date in 1983. Its apparent sister species, S. angustimargo , is known mostly from southeastern South America, its type locality in Paraguay. In March 2006, Elicio Tapia, Suzanne Rab Green, and I visited the exact site where the type series of Scea torrida had been collected, along a streambed crossing the road between Loja and Catamayo. This locality is unusually arid, with coverage of short, thorny Acacia trees; other Ecuadorian sites at elevations near 2000 meters are invariably blanketed in moist cloud forest. During our visit to the habitat of S. torrida , we were unpleasantly surprised by a flash flood, and were further saddened by not capturing additional specimens of this rare moth.

HOLOTYPE: Male (pl. 33). ECUADOR: Loja: anc. Rte. Loja-Catamayo, km 29, 1900 m, 8 Feb 1983, leg. C. Lamaire & P. Thiaucourt. The type is deposited in the PTC.

PARATYPES: ECUADOR: Loja: 233, 1♀, anc. Rte. Loja-Catamayo, km 29, 1900 m, 8 Feb 1983, leg. C. Lamaire & P. Thiaucourt, ( PTC; 3 genitalia slide nos. JSM-501 , 1718 ; ♀ genitalia slide no. JSM-502 ) .

OTHER SPECIMENS EXAMINED: 13, ‘‘N. Gran.’’, ‘‘1008’’ (OUMNH).

DISSECTED: 233, 1♀.

The following have been transferred from Scea :

nudata Hering to Notascea , new genus obliquaria Warren to Notascea , new genus solaris Schaus to Lyces Walker ( Patula

Group)

vulturata Warren to Lyces Walker ( Patula

Group)

venusta Dognin to incertae sedis

DISCUSSION PHYLOGENY

The goal of this paper was to produce a revised classification for the Dioptinae , with special focus on defining monophyletic genera. The implied hypothesis of intergeneric relationships (fig. 7) is provisional, and should be revisited at some point in the future, when additional character data become available. My morphological analyses were comprehensive, and it is unlikely that reinterpretation of existing characters will offer dramatic new insights. It seems equally implausible that novel characters involving external adult anatomy will be found, unless they are discovered through SEM study of particular structures. In my view, the best hope for future research lies in adding characters from DNA and immature stage morphology (see below) to the existing data set from adult anatomy.

The Josiini has been the subject of previous work ( Miller, 1996; Miller, Brower, and DeSalle, 1997), so discrepancies between those results and the ones presented here should be addressed. The two earlier papers utilized taxon samples similar in size to the one used here, but employed more inclusive character information. Miller (1996) looked at relationships among 24 representative species in the Josiini , using 86 characters from adults, larvae and pupae. Miller, Brower, and DeSalle (1997), focusing on 21 of those same species, combined the morphological data from Miller (1996) with DNA sequence data.

The cladogram in Miller, Brower, and DeSalle (1997) shares structural details with the one hypothesized in the current study. For example, their results foreshadowed the need to break a polyphyletic ‘‘ Josia ’’ into two groups— Josia and Lyces (fig. 283). However, the hypothesis of generic relationships as implied by that tree shows basic differences from the hypothesis here (fig. 7). Surprisingly, the two cladograms are almost completely reversed from top to bottom. In the earlier papers, J. megaera arises alone at the base of the tree as the sister group to the rest of the Josiini . Furthermore, Josia forms a basal clade immediately above J. megaera , and the other genera follow, with Lyces and Getta in the most derived positions. In contrast, according to the current hypothesis (fig. 283), J. megaera is nested within Josia , and this genus arises, along with Scea , as the most derived element of the Josiini . Getta falls near the bottom of the tree, while Lyces appears halfway up. These differences could be viewed almost as a rooting issue; the most recent cladogram is rooted near Getta , whereas the earlier ones were rooted near Josia .

It is difficult to explain these discrepancies. One possibility is that, even though the number of josiine exemplars is roughly the same in both studies, cladogram topology was influenced by the dramatically different species compositions. The exemplar list in Miller (1996) and Miller, Brower, and DeSalle (1997) was determined by availability of alcohol-preserved immature stages. In the current paper, 28 exemplars were chosen to represent adult morphological diversity across the Josiini . Two josiine genera newly described here— Proutiella and Notascea — account for important character variation within the tribe. Representatives of Proutiella and Notascea were not available to Miller (1996) or Miller, Brower, and DeSalle (1997); immature stages for both genera remain unknown. Similarly, specimens of Phintia were unavailable in the earlier analyses, but that genus is a pivotal element in the updated josiine cladogram (fig. 283).

An obvious next step toward better resolving the phylogeny of the Dioptinae would involve an expanded analysis, modeled after Miller, Brower, and DeSalle (1997). It should combine characters from adults, immatures, and DNA. Certainly, such a study would include representatives from all 11 of the josiine genera recognized in this paper (appendix 2). Rooting could effectively be achieved through inclusion of a broad species sample from the Dioptini . This plan would yield a cladogram detailing the evolution of the Josiini . As such, it would set the groundwork for precise inquiry into such issues as the history of host-plant use and the evolution of wing pattern within the group.

CLASSIFICATION

In reclassifying the Dioptinae I have adhered to the generic concepts of previous authors as closely as possible, while remaining true to a strict application of monophyly. For certain genera, dramatic changes in membership were not required. These were already monophyletic for the most part. Invariably, the species in these genera exhibit superficially obvious traits, easy to recognize without the aid of a microscope. In the best of cases, such characteristics also happen to be apomorphic. Thus, taxonomists, even those working before the turn of the 20th century, were able to make well-founded decisions regarding generic placement.

Erbessa provides an excellent example. Historically, 78 species-group names have been associated with the genus ( Bryk, 1930), and my research produced only four changes to the composition of that list. All involve taxa that properly belong in Erbessa , but had been misplaced in other dioptine genera: clite Walker is brought from Getta ; prolifera Walker is moved from Oricia ; integra C. and R. Felder is transferred from Phaeochlaena ; and longipalpata Dognin is taken from Scotura and placed as a synonym of E. leechi Prout. No species were removed from Erbessa . This relatively modest set of taxonomic changes resulted because a combination of features—greatly elongate labial palpi, ciliate male antennae, and the separation of veins M 3 and CuA 1 in the FW—can be used to diagnose Erbessa species with considerable confidence. None of these require familiarity with morphological detail.

Similarly, most Scotura species possess brilliant yellow scaling on the vertex (pls. 1, 2), and the male antennae are ciliate (fig. 11B, C). These traits together confirm genus membership, at least for the majority of taxa. In my classification, only two species names were removed from Scotura —longipalpata to Erbessa and ovisigna Prout to Pseudoricia . The erroneous placement of these taxa attests to the cursory nature of the taxonomic decisions made by previous authors; both species show a host of morphological traits unique to the genera in which they truly belong. The only taxon here added to Scotura , venata Butler , formerly placed in Oricia , is similarly instructive. This is one of the few members of Scotura in which the head is not bright yellow. Other structures of S. venata , such as its tympanum, male FW stridulatory organ and genitalia, leave no doubt regarding proper assignment to Scotura .

In contrast, my cladistic analyses demonstrate that some dioptine genera were an utter shambles. In such cases, superficial appearance and wing pattern are poor indicators of relationship. Earlier authors often seem to have thrown up their hands in frustration. However, my research demonstrates that careful morphological study, combined with cladistic analyses, can provide solutions to what initially seem to be intractable taxonomic problems.

Perhaps nowhere was such confusion more apparent than in the earlier classification of Tithraustes . Figure 354 shows the resulting positions of the 13 species, formerly placed in Tithraustes (table 3), employed as exemplars in the cladistic analyses. Prior to this paper, there were 42 species in the genus ( Bryk, 1930). I here transfer 35 of them elsewhere. One taxon, albilinea Schaus , goes to the Arctiidae . Two species are placed as incertae sedis. The majority—18 species—is spread among the six species groups of the new genus Nebulosa . Chrysoglossa and Dioptis receive six and five species respectively, while the remainder goes to Polypoetes (2 species) or Dolophrosyne (1 species). Tithraustes , now a mere shadow of its former self with only 10 included species, two of which are described as new, is finally monophyletic. This conundrum resulted from the fact that previous authors characterized Tithraustes by plesiomorphic traits.

Josia provides another case in point, where major changes from the previous classifications were required. In my arrangement, the genus includes 21 species arrayed in four species groups. Previous authors recognized a much larger Josia . For example, Bryk (1930) listed 67 species. A thorough morphological analysis demonstrates, first of all, the existence of a tight monophyletic group containing Josia ligula Hübner , the type species of the genus. That clade becomes the nowconfined genus Josia . The 47 remaining species are dispersed to five different genera in the Josiini . The main beneficiary is Lyces Walker , reinstated here to receive 27 of the species names from Josia of previous authors. Lyces is further divided into three species groups. The cladogram produced in the current work only serves to strengthen the theory put forth earlier ( Miller, 1996; Miller, Brower, and DeSalle, 1997)—longitudinal wing stripes are not homologous in Lyces and Josia , but instead arose by convergence. The other major recipients of former Josia species are Ephialtias (6 species) and the new genus Proutiella (11 species). Finally, Phintia Walker is elevated to generic status to receive two names, while Phryganidia , far removed in the Dioptini , gets one ( Josia brevifascia Prout , a new junior synonym of Phryganidia naxa Druce ). Wing pattern, essentially the only character system employed by the early taxonomists to construct a Josia classification, obviously fails as an indicator of relationship. On the other hand, a more sophisticated analysis of wing patterns in the Josiini , perhaps refined through developmental study of the pattern elements themselves, might be extremely informative.

Stenoplastis C. and R. Felder serves as the final example where a genus required dramatic changes. Previous authors characterized Stenoplastis using a single trait—male antenna lacking pectinations. At last count, the genus contained 22 species ( Bryk, 1930). Here, 19 of those are removed, leaving Stenoplastis with only four included species (appendix 2), one of which is newly described. The remainder has been transferred to Polypoetes (12 species), Argentala (3 species), Scoturopsis (2 species), and Momonipta (1 species). Figure 354 shows the phylogenetic placement of the six exemplar species from Stenoplastis of earlier authors. In this paper, ‘‘male antennal pectinations absent’’ shows extreme levels of homoplasy. Pectinations are absent in 10 separate clades within the Dioptini .

I feel confident in stating that each of the genera defined in this paper is monophyletic. Membership in a few may require minor changes, but overall, this research has produced a stable generic classification for the Dioptinae . Less confidence should be afford- ed the various species groups proposed here. For these, wherever observable morphological variation could be found, an attempt was made to reflect this by subdividing the genus into species groups. However, many of these groups may fall, or their composition may change as they are subjected to further scrutiny. Nevertheless, the categories provide hypotheses for future testing.

Polypoetes , the largest and most complex dioptine genus, is particularly problematic, as are Dioptis , Nebulosa , and Lyces . Since each of these is morphologically diverse, they will probably yield to a species-level approach based on adult structure. Other genera will be more challenging. Erbessa , though remarkable in its wing-pattern variation, is otherwise morphologically homogenous. The genus may ultimately prove tractable only through an analysis based on characters from DNA. Finally, there are genera, such as Xenomigia , where less than 20% of the species are as yet described. These will require an attack on all fronts—through additional collecting, basic descriptive work, morphological study, and DNA analyses. For certain genera, subgroups may be best reflected in the morphology of immature stages.

BIODIVERSITY

Completion of this paper places the Dioptinae in a rare position. Taxonomically, they become one of the best-documented tropical moth clades. Estimates suggest that 90–95% of butterfly species have been described ( Robbins and Opler, 1997). Certain parts of the neotropical fauna, especially in the Hesperiidae and Lycaenidae , still require significant amounts of basic descriptive work. Perhaps the most significant remaining challenge is the nymphalid subfamily Satyrinae , where approximately 25% of the fauna remains undescribed ( Lamas et al., 2004). This begs the following questions: How does the situation for Dioptinae compare with that in butterflies? How many dioptine species remain to be discovered and described, and what estimate could we put on the size of the world fauna?

Table 7 shows the number of previously described dioptine species, the number newly described in this paper, and the number for which descriptions remain. The third column is particularly interesting. It represents an informed estimate of the number of undescribed species observed among the museum material studied for this paper. Importantly, each of those was verified by genital comparison with described material. If descriptions for all the species in column 3 were to be completed, certain genera, such as Nebulosa , Xenomigia , and Polypoetes , would increase significantly in size. My research thus revealed 173 unnamed species (64 newly described, 109 remaining to be described). Forty-seven species of previous authors were newly synonymized, while 31 were revived from synonymy. The net increase for the Dioptinae thus stands at 157 species. Overall, based on what is currently known the dioptine fauna would total 563 species.

These numbers contrast with the findings of Scoble et al. (1995), who documented taxonomic changes in a survey of eight different revisions for neotropical Geometridae . The authors found that levels of synonymy (32%) were higher than the percentage of new species described (21%). Thus, at least for geometrids, revisionary work produced a net decrease in estimates of species richness. Before this paper, 406 species of Dioptinae were recognized ( Bryk, 1930) with a total of 567 species-group names. The rate of new synonymy (n 5 47) was 12%, a number similar to that found by Holloway (1994), where 13% of ennomine geometrid species from Borneo were placed in synonymy. However, when species revived from synonymy (n 5 31) are taken into account, the net change in preexisting names for Dioptinae is a decrease of only 3%. When the 173 unnamed species are considered, species richness for the Dioptinae is increased by 44%. Compared to tropical insect groups other than Geometridae , this number is, if anything, low. For example, in a sample of African Hemiptera, Hodkinson and Casson (1991) collected 1690 species, of which 62.5% were undescribed. In certain extremely important, but relatively obscure Lepidoptera groups, a remarkably small percentage of species is known. Recent estimates suggest that only 25% of the species diversity for Gelechioidea has been described ( Hodges, 1999). Ultimately that superfamily could total 65,000 species ( Bucheli and Wenzel, 2005), a number rivaling the Noctuoidea.

The most difficult estimate to provide is the number of Dioptinae that remain completely undiscovered. For conspicuous, thoroughly studied lepidopteran groups, such as butterflies, Saturniidae , and Sphingidae , these numbers would be relatively low. Dioptines, on the other hand, may conform more closely to taxonomically neglected groups, such as micro Hymenoptera and Diptera . Meier and Dikow (2004) used the tropical predatory fly genus Euscelidia ( Diptera : Asilidae ) to examine what percentage of that fauna remained undiscovered. Their calculations suggested that, after a revision of the group had been completed ( Dikow, 2003), 36%–41% of the species were still uncollected. Applying the estimate of Meier and Dikow (2004) to the Dioptinae , we could expect the fauna to ultimately total between

TABLE 7 Estimated species diversity for dioptine genera (arranged according to the checklist for the Dioptinae , appendix 2)

750–800 species, roughly double its size before modern revisionary work was undertaken. This number is in line with Dolphin and Quicke (2001), who predicted an increase of 100% to 200% undescribed species for the world Braconidae . At the extreme end of this spectrum are spittlebug flies, the genus Cladochaeta ( Diptera : Drosophilidae ). After a revision in which 105 new species were described, adding to a preexisting fauna of only 14 taxa, Grimaldi and Nguyen (1999) estimated that 85% of Cladochaeta species remained undescribed.

Two factors conspire to make the numbers in column 3 of table 7 low. First, the list of Dioptinae remaining to be described includes only those observed either in the current AMNH holdings, or on loan from one of the 35 collections sampled for this research. When additional collections are studied, more undescribed species will invariably be found. Secondly, focused collecting of Dioptinae in neglected regions of the Neotropics will vastly increase the number of new taxa discovered. For any insect group, museum material is far from a fair biogeographical representation of the world fauna ( Gaston et al., 1995; Scoble et al., 1995). For a variety of reasons, such as the availability of supportive infrastructure for local and visiting scientists, countries such as Ecuador and Costa Rica have been inordinately sampled in recent years. This then becomes reflected in the provenance of loan material. Those two countries are overrepresented compared to Colombia, Nicaragua, and Bolivia. Furthermore, the Dioptinae provide a case where comprehensive sampling leads to new species discovery.

This is illustrated by examining the Dioptinae of Ecuador in more detail. Before this study, 94 species were known to occur there ( Prout, 1918; Hering, 1925; Bryk, 1930). Twenty-four Ecuadorian taxa are newly described here, and of the 115 species in collections remaining to be described (table 7), 43 are from Ecuador. When the newly described and remaining undescribed species are totaled, dioptine diversity for Ecuador goes up by 67 species, a 70% increase from what was known prior to this research. Furthermore, it is false to assume that these numbers have reached their limits. New Ecuadorian species are constantly being discovered. On a recent visit to Ecuador (March 2006), I captured several undescribed Dioptinae that had never appeared in collections.

Endemism appears to be quite high for Ecuador (table 8). Overall, however, such estimates for the Dioptinae are crude. For example, based on what is currently known, 44% of the dioptine fauna of Colombia has not been recorded elsewhere (table 8). This obscures the fact that the vast majority of Colombian Dioptinae available for study in the world’s museums was essentially collected between the years 1900 and 1910 by a single individual—Anton Fassl. As amazing as Fassl’s accomplishments are, his samples in no way reflect the actual geographical distribution of most of these species. Many Colombian dioptines are known from a single specimen or from a single series collected at one locality. Unlike the situation that currently obtains for butterflies and some other heavily sampled insect groups, such as tiger beetles ( Cassola and Pearson, 2000), many years of widespread collecting throughout the Neotropics will be required before levels of endemism can be understood for the Dioptinae .

The 70% increase in dioptine species for Ecuador resulted from moderately comprehensive collecting, followed by thorough alpha taxonomy. Such increases can be expected for other Andean countries, where similar effort has not yet been applied. A total of 133 Dioptinae occur in Peru (table 8), making that country home to approximately 30% of the world fauna. My own field experience in Peru is relatively limited. Two expeditions were undertaken during the course of this research—the first to the Tambopata Reserve in Amazonian Peru, and the second to cloud forests east of Cuzco. In general, collecting was difficult and relatively low numbers of dioptine specimens were captured. Nevertheless, the two trips together produced 20 undescribed species of Dioptinae . Only eight of those are described in the current work. Because of their topographical complexity and the extreme paucity of existing samples, percentage increases for Peru, Bolivia, Colombia, and Venezuela could easily exceed those for Ecuador.

Central America shows levels of faunal increase somewhat lower than the Andean countries. Costa Rica provides an estimator. Over a period of 25 years, researchers at the Instituto Nacional de Biodiversidad (INBio) have sampled the country’s moth fauna across a wide spectrum of sites, providing relatively thorough coverage of Costa Rica’s geography. In preparation for this paper, I studied all of that institution’s accumulated dioptine holdings, and also examined the extensive Costa Rican material on loan from other collections. The dioptine fauna of Costa Rica currently stands at 54 species (table 8), the highest in Central America, with 22% being endemic. The situation in Costa Rica is analogous to that in Ecuador; the relatively large number of species record- ed there can in part be attributed to disproportionately thorough sampling compared to what has taken place in other Central American countries. Prior to my research, 43 species were known from Costa Rica. Eleven are newly described, while of the 115 dioptines remaining to be described, 10 occur there. My research thus uncovered 21 undescribed Costa Rican Dioptinae , producing a biodiversity increase of 49%.

Providing an estimate of undiscovered species for the Amazonian area is particularly difficult. This region, the bulk of which is encompassed by Brazil, also includes the Guyana Shield and southern Venezuela, as well as the eastern portions of Peru, Ecuador, and Colombia. A detailed count of the described Dioptinae endemic to Amazonian South America, as documented by the current work, provides a total of 128 species (table 9). Certain genera, such as Scotura , Erbessa , and Dioptis , are overwhelmingly Amazonian (table 9). Importantly, the alpha level taxonomy of each of these genera is in chaos. It is nearly impossible to estimate how high their species numbers could ultimately go. Dioptis might easily double in size.

Sampling for the region is extremely inadequate. In this paper I describe only six new Amazonian species—from Guyana, French Guiana, Brazil, and Peru. The material on which these descriptions are based was collected fairly recently, usually within the past 20 years. However, modern collections of day-flying moths from the Amazon are scarce. The vast majority of specimens in museum collections were collected over 100 years ago. For example, H.W. Bates, during the trips in the 1850s that inspired his famous writings ( Bates, 1862), captured much of the Amazonian dioptine material housed in the BMNH. French Guiana is the only Amazonian country for which fairly thorough collections of Dioptinae have been made. My studies of that fauna, based on material at the MNHN in Paris, on the holdings of several important private Lepidoptera collections, and on my own fieldwork, suggest that we have barely scratched the surface in terms of understanding the country’s dioptine biodiversity. Remarkable taxa, such as the French Guiana endemic Phavaraea rectangularis (pl. 28) described by Toulgoët and Navatte (1997), are recent discoveries. Species such as this are often extremely rare. After more than 20 years of intensive collecting by Bernard Hermier and his compatriots, P. rectangularis is known from only three specimens. With this accumulation of albeit anecdotal evidence at hand, it seems impossible to believe that comprehensive sampling of the Amazonian fauna would not produce between 50 and 100 undescribed Dioptinae .

In summary, while it is difficult to estimate the number of species that future collecting will yield, this publication provides compelling evidence for the phenomenal levels of unexplored and undocumented biodiversity that exist throughout the Neotropics. Montane habitats will probably yield the greatest number of undescribed species. Willmott et al. (2001) showed that, even in a butterfly group such as Hypanartia ( Nymphalidae ), long regarded as taxonomically static, many new montane species remain to be discovered. These environments are under extreme threat (e.g., see Aldrich et al., 1997).

TABLE 8 Dioptine species richness Records based on specimen label data from the material examined for this study (see table 2).

CHARACTER EVOLUTION

Almost every morphological character utilized in this study shows homoplasy within the Dioptinae . Some of this can perhaps be attributed to rudimentary interpretations of what are in fact composite, or nonhomologous, traits. In the future, as our understanding of particular structures becomes better refined, characters can be rescored and a certain amount of homoplasy will thus be eliminated. On the other hand, much of the homoplasy observed in this study applies to structures for which homology assessments seem certain. Homoplasy occurs in traits unique to the Dioptinae , some seemingly fundamental to the group, as well as in characters broadly distributed across the Notodontidae . A few poignant examples will illustrate these phenomena.

CALTROP CORNUTI: The male genitalia of Notodontidae possess certain features unusual to the family. For example, while almost all Lepidoptera bear cornuti on the vesica (Kristensen, 2003), star-shaped ones that dehisce during copulation and remain in the female corpus bursae (e.g., figs. 41D, 41F, 44D, 44F, 54C, 55D), are found only in Notodontidae ( Forbes, 1948; Miller, 1991) and Sphingidae ( Rothschild and Jordan, 1903). The function of these strange structures, termed deciduous caltrop cornuti throughout this paper, is open to speculation. Of the nine notodontid subfamilies currently recognized, caltrop cornuti are found in seven ( Kitching and Rawlins, 1999). Howev- er, within each of these seven subfamilies, the trait shows a homoplastic distribution ( Holloway, 1983; Miller, 1991). The Dioptinae are no exception.

Caltrop cornuti occur widely in the Nystaleinae ( Weller, 1992) , the sister subfamily to the Dioptinae . In the analyses described here, optimization onto the dioptine cladogram suggests that their presence is plesiomorphic for the subfamily. Within the Dioptinae , caltrop cornuti are found in 123 species (28% of the group) belonging to 10 genera. However, these structures show an unusual phylogenetic distribution. First, caltrop cor-

TABLE 9 Number of Dioptinae recorded from the Amazon region, including the Guyana Shield and Upper Amazon Basin nuti occur in basal genera of the Dioptini , such as Clade 1 (fig. 7)—which includes Cleptophasia (1 sp.), Scotura (24 spp.), Eremonidia (1 sp.), Oricia (4 spp.) and Erbessa (60 spp.)—as well as in Xenorma (9 spp.) at the base of Clade 2. This accounts for the vast majority of taxa that possess them. Above the level of Xenorma (Clade 3; fig. 7), these structures were apparently lost. However, they then reappear in isolated, derived positions on the cladogram.

For example, caltrop cornuti occur in Brachyglene (fig. 147E) and Chrysoglossa (e.g., fig. 154D, F), two genera that fall close together on the dioptine cladogram (Clade 9; fig. 7). However, the sister groups of these genera do not possess them. Character optimization suggests two independent derivations. Another large lineage in the Dioptini (Clade 6; fig. 7), comprising species with elongate labial palpi and bipectinate antennae, includes Phaeochlaena (7 species), Pikroprion (1 species), Argentala (6 species) and the largest genus in the Dioptinae , Polypoetes (63 species). This clade is characterized by absence of caltrop cornuti. Polypoetes is further subdivided into five species groups. Oddly, within one of those, the Rufipuncta Group, a small subclade of six species occurs in which the vesica bears tiny, deciduous caltrop cornuti (fig. 120E). Perhaps the most surprising reoccurrence is in the Josiini . Here, such cornuti are absent throughout the tribe. However, high up on the josiine cladogram (fig. 7), they occur in a small clade containing four species, here recognized as the new genus Notascea (figs. 334C, 334E, 335C). Thus, while presence of caltrop cornuti is perhaps plesiomorphic for the Dioptinae , they were lost and subsequently regained no fewer than five separate times within the subfamily.

LABIAL PALPI: The second example of homoplasy within the Dioptinae involves a trait not found elsewhere in the Notodontidae . Earlier authors on the group largely ignored the striking diversity of labial palpus morphology found in dioptines. In the current study, on the other hand, labial palpus structure provided some of the most useful features for phylogenetic analysis, and the palpi are crucial for genus identification. Morphological variation of the labial palpi is summarized by 11 characters and 39 characters states (Characters 1–11, appendix 1). In their most elaborate development, Lp2 is much longer than the other segments and the palpi arch over the front, often extending posteriorly well beyond the antennal bases (e.g., fig. 35A–E). Such palpi are typically held elbowlike against the front (figs. 31E, 36A). My phylogenetic analyses suggest that elbow-shaped labial palpi have evolved three separate times in the subfamily: once in Oricia (4 spp.) + Erbessa (60 spp.); secondly in a clade containing 77 species in four genera— Phaeochlaena , Pikroprion , Argentala , and Polypoetes ; and finally, in the Josiini , where such palpi appear as a synapomorphy for the small genus Phintia (fig. 328A, D).

This example shows fundamental differences from the one involving deciduous cornuti. For the cornuti, each of the separate derivations shows essentially identical morphology. In these three occurrences of elongate labial palpi, on the other hand, subtle structural differences can be seen when the palpi of Erbessa , members of the Phaeochlaena -clade, and Phintia are compared. A more finely tuned morphological analysis than the one provided here might, in fact, reveal alternative ways of scoring ‘‘elongate labial palpi’’ to reflect nonhomologous aspects of the three variants—a classic case of reciprocal illumination (sensu Hennig, 1966).

SOUND PRODUCTION AND HEARING: As the final example of homoplasy, I again showcase a modification not found elsewhere in the Notodontidae or elsewhere in the Noctuoidea, for that matter. Males of many Dioptinae possess a sound-producing organ, located beyond the distal margin of the FW discal cell. This structure is variously developed within the subfamily (Miller, 1989; Rawlins and Miller, 2008). Often it is highly complex (e.g., figs. 137F, 255F). Morphological variation in the FW stridulatory organ is described in this paper using three characters (Characters 77–79, appendix 1), involving length of the discal cell, the condition of veins M 1 and M 2, and the state of the wing surface surrounding the organ itself.

A survey of the Dioptinae shows that the male FW stridulatory organ, in its various forms, occurs in 180 of the described species, approximately 40% of the subfamily. Based on my cladistic analyses, the structure evolved two times—once in Clade 1 (fig. 7), and again in a diverse group comprising 18 genera (Clade 8). Interestingly, within this large clade, my phylogenetic hypothesis suggests that the organ was subsequently lost twice—once in Dioptis and once in a clade comprising Dolophrosyne + Scoturopsis .

Loss of the FW stridulatory organ within Dioptis is particularly interesting. In basal elements of the genus, the Chloris and Butes groups, the structure is well developed (fig. 190D). Here, the FW discal cell is short, veins M 1 and M 2 are swollen, and the surface of the fascia is corrugated. During the evolution of Dioptis , the stridulatory organ was lost so that these features are absent in more derived species, the bulk of the genus. The only apparent sign that this structure occurred in their common ancestor is the presence of a short FW discal cell (fig. 190E, F). Because the taxa now in the Chloris and Butes groups posses a stridulatory organ, Prout (1918) and Hering (1925) incorrectly assigned them to Tithraustes rather than Dioptis , thus obscuring a fascinating example of character transformation.

Perhaps the most intriguing aspect of this story involves the concordant loss within Dioptis of a structure fundamental to the Noctuoidea—the metathoracic tympanum. A tympanum is present in the Chloris and Butes groups, but the structure becomes progressively smaller in more derived members of Dioptis . In the Vitrifera , Fatima , and Cyma groups, there is a tiny depression where the tympanum would be (figs. 187I, 189B, 189C), but the membrane is absent altogether. These observations suggest that a basic biological change took place during the course of Dioptis evolution, perhaps involving an aspect of interspecific communication. Whereas most members of the Dioptinae possess ears and some means for producing sound, both functions were lost in Dioptis .

As a fascinating historical aside, Forbes (1916) regarded the metathoracic hearing organ as the most crucial character for understanding the evolution of the Noctuoidea. Forbes chose a sample of species to build the first phylogeny for group, basing his hypothesis entirely on structure of the tympanum. Dioptis , perhaps an obvious choice to represent what at that time was the family ‘‘Dioptidae’’, turned out to be a misleading one. Forbes examined Dioptis , observed that no tympanum is present, and on that basis placed the Dioptidae at the base of the noctuoid phylogeny. He later recanted ( Forbes, 1922), noting that most Dioptidae do in fact possess a tympanum. He went on to urge further study of the group’s phylogenetic position. Richards, a graduate student of Forbes, took up the call, and his amazing efforts ( Richards, 1932) greatly refined our understanding of the noctuoid tympanum. Although Richards produced a more elaborate phylogeny for the Noctuoidea, he was unable to make progress on the position of the Dioptidae .

Though my research highlights often startling examples of homoplasy in the Dioptinae , such results in no way detract from the utility of morphology in phylogenetic reconstruction. Instead, morphological characters have offered a window into dioptine evolution, showcasing the group’s remarkable capacity for adaptation.

WING PATTERN

For a group of its size, the Dioptinae presents one of the most diverse wing-pattern arrays to be found in the Lepidoptera ( Seitz, 1925; Hering, 1925). This amazing complexity bedeviled early workers. As a result, 19thcentury taxonomists described dioptines in a wide range of moth families, even including the Psychidae and Pyralidae . Most commonly they were assigned either to the Arctiidae ( Pericopinae and Ctenuchinae ), or to the geometrid subfamily Sterrhinae , which contains diurnal genera—such as Cyllopoda — exhibiting boldly patterned black and yellow wings. It was not until Prout (1918) and Hering (1925) more thoroughly summarized characters for recognizing dioptines, involving wing venation and palpus structure, that the fog began to lift.

Surveying wing patterns across the Dioptinae , it seems difficult to imagine general evolutionary trends at work. On the contrary, it is almost as if, at some point in time, the group became free from evolutionary constraint. According to Nijhout (1991, 2003), there are two basic modes of wing-pattern formation in Lepidoptera : In the first, exemplified by the Nymphalid ground plan ( Schwanwitsch, 1924; Nijhout, 1991; Willmott, 2003), each pattern element retains its distinctive character and morphology, producing a finely detailed, often camouflage, appearance. In the second, boundaries between component pattern elements are difficult to discern, resulting in bold, usually aposematic patterns with little detail. Both modes seem to operate within the Dioptinae . For example, one could envision the cryptic, often intricate patterns of Chrysoglossa (pl. 15) and Xenomigia (pls. 22, 23) conforming to a system analogous to the Nymphalid Ground Plan. In contrast, the wings of Phaeochlaena (pls. 9, 10) and the Josiini (pls. 26–35) exhibit simple, aposematic patterns like those described in mode two. However, the majority of dioptine species do not easily fit into either of these categories. Large genera such as Scotura , Polypoetes , and Nebulosa contain species with relatively simple wing patterns, but ones that do not adhere to our classical image of warning coloration.

With a cladogram of genera now in hand, what generalizations can be made regarding the evolution of wing pattern in the Dioptinae ? The attempt to outline observable trends below is presented against a backdrop of incredible diversity. When the details of this complexity some day become understood, a fascinating story will undoubtedly emerge. For now, I can only offer a crude summary of information.

APOSEMATISM: In Lepidoptera , aposematic coloration in adults is usually thought to be accompanied by chemical protection, conferred by sequestration of toxins from the larval host plant (e.g., Brower, 1984) or by synthesis as adults ( Brown, 1984, 1985). Nishida (2002) pointed out that, while moth and butterfly caterpillars come in contact with a huge array of plant secondary chemicals, sequestration involves a relatively small subset of those. Of the notodontids that have been tested, all are highly palatable to birds ( Jones, 1932; MacLean et al., 1989). However, not a single dioptine species has been presented to potential predators in an experimental context, nor have the tissues of dioptine larvae or adults been examined for sequestered toxins. It is tempting to infer that Passiflora -feeding Josiini are protected. Some heliconiines sequester cyanogenic glycosides from their Passiflora hosts, while still others can synthesize their own cyanogens ( Nahrstedt and Davis, 1983; Raubenheimer, 1989; Engler-Chaouat and Gilbert, 2007). Howev- er, assumptions about chemical protection and palatability must be carefully tested on a case-by-case basis ( Nishida, 2002). For example, it had long been assumed that the Australian whistling moth, Hecatesia exultans ( Noctuidae : Agaristinae), being aposematic and diurnal, is protected. However, when chemical analyses were done ( Talianchich et al., 2003), it was discovered that their larval and adult tissues do not contain isoquinoline alkaloids found in the host plant, Cassytha ( Lauraceae ). Furthermore, feeding trials demonstrated that H. exultans is palatable to its potential predators. At this time, we can do little more than speculate whether chemicals sequestered from their hosts protect Dioptinae .

Aposematism in Lepidoptera is furthermore assumed to be associated with diurnal flight, as typified by the Papilionoidea. Dioptines are commonly referred to as diurnal (e.g., Köhler, 1930). Recently it has been shown that, although they are generally less nocturnal than other Notodontidae , dioptines display a spectrum of flight-activity patterns, ranging from completely diurnal to completely nocturnal. In a sample of Venezuelan species, Fullard et al. (2000) found that josiines are mostly diurnal, whereas Scotura is almost entirely nocturnal. Polypoetes circumfumata flew at nearly all hours of the day and night. Thus, simply characterizing the Dioptinae as a diurnal group masks underlying complexity. Associated with this tendency toward diurnal activity, the metathoracic ears of dioptines, when compared to the ears of more typical nocturnal notodontids, have lost their sensitivity to the echolocation frequencies of bats ( Fullard et al., 1997; Fullard et al., 2000). These authors noted that dioptine ears have retained function, but at frequencies more probably related to interspecific communication. The Dioptinae perhaps fit a model analogous to that found in some Arctiidae (e.g., Sanderford and Conner, 1995; Sanderford et al., 1998; Sanderford, 2009), where sounds are produced and received during courtship, but are probably not involved in defense.

It is similarly difficult to neatly fit particular dioptine species into simple wing-pattern types, such as aposematic vs. cryptic. Using decidedly simplified and subjective criteria, aposematic wing patterns would include those with bold markings and bright colors such as black, yellow, white, and blue. This would certainly apply to three clades in the Dioptinae —the Josiini , Erbessa , and Phaeochlaena . Less clear-cut would be the wings of genera such as Xenorma (pl. 8), Brachyglene (pl. 15), and Pseudoricia (pl. 23) or those of the Divisa Group in Hadesina (pl. 14). Seemingly isolated cases of potential aposematism, such as S. franclemonti in the otherwise drab-colored genus Scoturopsis (pl. 22), are even more problematic. All told, 40–50% of dioptine species exhibit wing patterns that could safely be characterized as aposematic. Even in mapping the most obvious candidates onto the dioptine cladogram, the trait shows multiple origins. At the very least, judging from the phylogenetic positions of Erbessa , Phaeochlaena , and the Josiini (fig. 7), aposematic wing patterns evolved no fewer than three times within the subfamily.

By applying a broader definition of warning coloration to include the wings of Cleptophasia , Xenorma , Brachyglene , and others, mapping aposematism onto the dioptine cladogram becomes much more complex. One might propose that the ancestral dioptine made a dramatic transition from a nocturnal life with camouflage wings to a diurnal and aposematic life. Presumably, this would have been accompanied by a dramatic shift in host-plant use ( Brower, 1984; Rothschild, 1984). For example, it might have involved a transition from feeding on nontoxic trees, as occurs in most Notodontidae , to feeding on toxic plants such as Passiflora ( Miller, 1992a; Powell et al., 1999). Under this scenario, warningly colored wings would be plesiomorphic for the Dioptinae . Cryptic coloration, the derived condition, would then have evolved multiple times. Within the Dioptini , camouflage, or at least drab, wings occur in the Phryganeata Group of Oricia (pl. 2), as well as in the genera Scotura (pls. 1, 2), Phryganidia (pl. 8) and Chrysoglossa (pl. 15). The wings of Xenomigia are certainly cryptic (pls. 22, 23), as are those of most species in the clade that includes Dolophrosyne , Scoturopsis , and Xenormicola (pls. 21, 22). Using any possible regime of character definition and optimization, a cryptic wing pattern would have evolved no fewer than five times within the Dioptinae .

HYALINE WINGS: It is difficult to offer a simple explanation for the evolution of hyaline wings in the Dioptinae . The most famous clear-winged dioptines are Dioptis . Bates (1862) formulated the theory of mimicry by drawing on his accumulated personal observations of Amazonian Lepidoptera . To illustrate his ideas, he showcased the remarkable mimetic resemblance among clear-winged butterflies in the Pieridae ( Dismorphia ), Riodinidae ( Stalachtis ), and Ithomiinae ( Ithomia ). Dioptis figured prominently in his examples. Bates noted that certain Dioptis species exhibit precise resemblance to the particular ithomiine butterflies with which they co-occur, ascribing this to ‘‘the process of the origination of a mimetic species through variation and natural selection’’ (1862: 564). As part of that seminal paper, he described three new Dioptis species, naming each after the ithomiine with which it ‘‘flies in company’’. These three taxa belong together in the Cyma Group of Dioptis , an almost exclusively Amazonian clade containing 28 species.

When the entire Dioptinae is taken into account, my phylogenetic analyses suggest that transparent wings have evolved independently no fewer than five times in the subfamily (fig. 355). Remarkably, with the exception of Dioptis (45 species), each case involves a small, highly derived, clear-winged clade arising from an ancestor with opaque wings. For example, two species of Erbessa possess transparent wings— pales and capena (pl. 7). These almost certainly evolved from an opaque-winged ancestor. Similarly, in Phanoptis , the Cyanomelas Group shows hyaline wings (pl. 9), whereas in the Fatidica Group the wings are densely scaled (pl. 8). Hadesina (pl. 14) provides yet another example. The Limbaria Group exhibits transparent wings, but in the Divisa Group the wings are dark brown and densely scaled. Two examples of hyaline wings seem to have evolved through a transitional form with translucent wings. In Monocreaga (pl. 21) the wings are transparent, while in their sister group, Momonipta , the wings are translucent (pl. 21). Four Isostyla species exhibit hyaline wings (pls. 24, 25), whereas in the remaining taxa—and in their sister taxon, Tithraustes — the wings are translucent. Approximately 60 species of Dioptinae possess hyaline wings, only 13% of the subfamily. These are currently placed in six genera. Thus, although their ranks are relatively small, clear-winged taxa are conspicuous in showing multiple evolutionary origins. When the entire Lepidoptera is surveyed, wing transparency shows homoplasy at higher taxonomic levels as well, occurring in nearly all butterfly families ( Punnet, 1915), as well as in at least nine families of moths ( Kristensen and Simonsen, 2003). According to my survey, hyaline wings show identical scale morphology across all dioptine groups (figs. 65, 191, 217). Transparency is apparently achieved by the same means in each of these clades.

MIMICRY: It is difficult to assess the extent to which mimicry has driven the evolution of wing pattern in the Dioptinae . The application of cladistic methodology has opened up new avenues for understanding mimicry, and the approach is now being used in a variety of insect groups (e.g., Zrzavý and Nedvêd, 1999). In Lepidoptera , phylogeny-based studies have revealed intricate Müllerian mimicry schemes ( Müller, 1879), whereby species evolve resemblance to other members of their own clade, or to a related clade, through convergence ( Brower, 1996; Simmons and Weller, 2002; Willmott, 2003; Yen et al., 2005b; Simmons, 2009). Based on the cladogram produced here (fig. 283), the plesiomorphic pattern in Josiini consists of a blackish-brown forewing with an orange-yellow transverse band (e.g., pl. 26). Rare longitudinal FW stripes appear in two separate clades—once in the Patula Group of Lyces (pls. 30, 31), and a second time in Josia (pls. 32, 33). This is the first example of Müllerian mimicry discovered in the Dioptinae ( Miller, 1996) , but species-level cladograms for additional groups will undoubtedly reveal many more cases.

It has been suggested that mimetic appearance is regulated by ‘‘supergenes’’ ( Clarke and Sheppard, 1960; Joron and Mallet, 1998; Mallet and Joron, 1999), according to which, accurate resemblance is under the control of a relatively few tightly linked elements. Simmons and Weller (2002) invoked supergenes to explain the remarkable wasp mimicry found among Sphecosoma and relatives ( Arctiidae : Ctenuchinae ). In that system, different moths have evolved precise resemblance to particular vespid models; imperfect mimics do not occur. Such theories perhaps explain wing patterns that seem to appear de novo in the Dioptinae . Cases can be found throughout the group where a striking, novel mimetic wing pattern seems to have arisen abruptly within a clade. An example is the Josia mimic, Erbessa mimica (pl. 4), which has evolved orange longitudinal wing stripes within a tight monophyletic group, Erbessa , where nothing even remotely similar occurs. The red-banded wings of Phanoptis miltorrhabda (pl. 8) diverge dramatically from those of other Phanoptis species (pls. 8, 9). In southeastern Brazil, the wing pattern of Proutiella vittula (pl. 26) mimics that of Lyces constricta (pl. 29), a distant relative in the Josiini (fig. 7). Its pattern is novel for Proutiella .

A mimicry ring is defined as a group of sympatric species with a common aposematic pattern ( Punnet, 1915; Brown, 1988; Joron and Mallet, 1998; Mallet and Joron, 1999). Mimicry rings are extremely prevalent among the butterfly fauna of South American rainforests ( Brower and Brower, 1964; Papageorgis, 1975). The Dioptinae participate in many of these, but are also engaged in complexes involving day-flying moths as well as butterflies. The wings of the Amazonian species Phaeochlaena solilucis are boldly marked with yellow and black (pl. 9). This pattern recurs in Geometridae ( Xanthyris , Smicropus ), Arctiidae ( Ephestris ), agaristine Noctuidae ( Seirocastnia ), and Riodinidae ( Chamaelimnas , Pachythone ). Erbessa citrina (pl. 7) and Phaeochlaena hazara (pl. 9), on the other hand, belong in the famous tiger-stripe mimicry ring ( Beccaloni, 1997; DeVries, 1997). This ring spans a huge range of taxonomic groups, and includes species of pericopine Arctiidae ( Chetone ), Pieridae ( Dismorphia ), Riodinidae ( Stalachtis ), and Papilionidae ( Pterourus ), as well as members of at least four subfamilies of Nymphalidae —the Danainae ( Lycorea ), Heliconiinae ( Heliconius , Eueides ), Charaxinae ( Consul ), and Ithomiinae (e.g., Mechanitis ). The orange and yellow wing patterns of E. citrina and P. hazara diverge dramatically from those of their congeners, yet the resemblance to other tiger-stripe mimics is precise.

Multiple mimicry rings can occur ( Beccaloni, 1997; Mallet and Gilbert, 1995; DeVries et al., 1999). In butterflies, as many as 28 mimicry rings can be found together at a single tropical locality ( Turner, 1984; Brown, 1988). The dioptine fauna of Tambopata Reserve, in Amazonian Peru, participates in no fewer than four mimicry rings. Dioptis species, members of the hyaline-winged group made famous by Bates (1862), are abundant. The Phaeochlaena species discussed above— P. solilucis and P. hazara — members of two additional mimicry rings, occur there as well. Finally, there exists a poorly documented mimicry ring involving small- to medium-sized, day-flying moths in which the forewing is black with a yellow transverse band, and the central area of the hind wing is white. This complex includes three Dioptinae —two from the Josiini ( Proutiella tegyra , Ephialtias abrupta ) and one in the Dioptini ( Erbessa tegyroides )—as well as an arctiid ( Ordishia klagesi ; pl. 29) and a remarkable diurnal pyralid in the subfamily Chrysauginae ( Nachaba sp. ; pl. 26). Resemblance among these species is startling, and all of them can be collected along the same forest trail.

This discussion notwithstanding, exercises to posit generalizations about the evolution of wing pattern in the Dioptinae tend to obscure potentially remarkable discoveries that lie hidden within the details. An important corollary of mimicry theory is that novel warning patterns can multiply within already warningly colored clades ( Brower, 1996; Mallet and Joron, 1999). Erbessa , with its fantastic diversity of aposematic coloration, may be a prime case of this phenomenon. Analysis of wing-pattern transformation in Erbessa , using a phylogenetic hypothesis for all 60 species, would undoubtedly yield more about the evolutionary process than could be matched by any attempts to generalize using bold strokes at the genus level.

HOST PLANTS

The Dioptinae are intriguing because their larvae tend to feed on plants atypical for moths ( Kitching and Rawlins, 1999). Instead, dioptine hosts mirror those of many butterfly groups ( Miller, 1992a). Early generalizations regarding host-plant associations in the Dioptinae were based on few points of reference. When I began studying the group, published host records were available for only nine species— Passiflora -feeding Josiini accounted for six of those. Today, the picture is much more complicated and far more interesting. Tables 4 and 6 summarize our current knowledge of life histories in the Dioptinae . They list verified host records for 87 species in 26 genera, but the Dioptini and Josiini are not represented in equal proportions. Life histories for 31 josiine species are now known, roughly 30% of the fauna, whereas in the Dioptini , immatures have been discovered for 56 species, only 16%. As new host records for the Dioptinae have accumulated, a correspondence with butterflies still seems applicable, but the boundaries for comparison have expanded dramatically.

DIOPTINAE AND HELICONIINAE : By far the most compelling similarities between dioptines and butterflies involve the subfamily Heliconiinae ( Nymphalidae ). According to recently accepted, more broadly defined classifications ( Ackery et al., 1999; Penz and Peggie, 2003; Wahlberg et al., 2003), the Heliconiinae includes approximately 400 species in 40 genera, a subfamily nearly equal in size to the Dioptinae . An obvious difference is that, while dioptines are restricted to the New World tropics, heliconiines are found worldwide. Nevertheless, the two groups show startling correspondence in their host-plant affiliations.

Caterpillars of the Heliconiinae are famous for specializing on Passifloraceae (e.g., Ehrlich and Raven, 1964; Benson et al., 1976). Worldwide, over 110 heliconiine species, scattered across 14 genera, feed on plants in this family ( Ackery, 1988; Robinson et al., 2007). The bulk of these records can be attributed to a single New World genus, Heliconius , containing 42 species ( Beltrán et al., 2007), all of which are restricted to Passifloraceae ( Michener, 1942; Smiley, 1982; DeVries, 1987). In the Dioptinae , Passifloraceae-feeding has been recorded for 28 species in six genera, all confined to the tribe Josiini (table 6). A conservative estimate suggests that between 75 and 85 species of Josiini will ultimately be found in association with this plant family. The Josiini and Heliconiinae are the only two groups of Lepidoptera showing sizeable radiations on Passifloraceae .

Violaceae appear in the host lists of both the Dioptinae and Heliconiinae . Many heliconiines, such as the fritillaries (subtribe Argynniti ), feed as larvae on Violaceae ( Ackery, 1988) , with a particularly strong preference for two genera, Hybanthus and Rinorea . These plants, hardly reminiscent of garden violets, grow as understory shrubs or small trees (pl. 47C) in lowland forests throughout Central and South America. The first reports that some dioptine caterpillars feed on Hybanthus and Rinorea ( Wolda and Foster, 1978; Harrison, 1987) gave rise to speculation that this association might be as widespread in the Dioptinae as it is in Heliconiinae ( Miller, 1992a) . Four dioptine genera— Scotura , Oricia , Phanoptis , and Pseudoricia —have now been recorded on Violaceae (table 4). Together these account for 36 species, approximately 8% of the subfamily. When Violaceae-feeding is mapped onto the dioptine cladogram, the association appears to have evolved at least three separate times (fig. 356). While the evidence at hand does not indicate Violaceae as a dominant host-plant relationship for the Dioptinae , this affiliation, otherwise rare in Lepidoptera , is conspicuous nevertheless.

The Achariaceae , another plant family playing a significant role in the host list of the Heliconiinae , contains genera formerly in the Flacourtiaceae , a group that has largely been dissolved according to newly revised concepts of plant phylogeny ( Chase et al., 2002; Sosa et al., 2003; Soltis et al., 2005). Members of the pantropical heliconiine tribe Acraeini are particularly strong on Achariaceae . Their caterpillars feed on a diversity of genera, such as Caloncoba , Hydnocarpus , Kiggelaria , Rawsonia , and Xylotheca ( Ackery, 1988) . When I initiated my research, published host records for the Josiini referred to a single plant genus— Passiflora . The breadth of the Josiini diet is now known to extend beyond Passifloraceae into other plant families. In a fascinating recent discovery, the larvae of Ephialtias dorsispilota (pl. 39G) were found in Panama feeding on Lindackeria laurina (table 6). Lindackeria , a small genus with worldwide distribution ( Gentry, 1993), is currently placed in the Achariaceae ( Chase et al., 2002) .

The final host-plant link between the Dioptinae and Heliconiinae involves the Turneraceae , a family closely related to the Violaceae and Passifloraceae ( Cronquist, 1981; Soltis et al., 2005). The larvae of two neotropical heliconiines— Euptoieta hegesia and Eueides procula —have been recorded on Turneraceae ( Janzen, 1983; Ackery, 1988). While Euptoieta and Eueides are not particularly close relatives within the Heliconiinae ( Brower, 2000; Penz and Peggie, 2003), the other included species in both genera specialize on Violaceae and Passifloraceae ( DeVries, 1987) . A similar pattern occurs in the Josiini ( Miller, 1996) , where species in Josia and Ephialtias have been recorded on Turnera (table 6), a genus of aromatic shrubs (pl. 41D) endemic to disturbed habitats in Central and South America. Interestingly, a third case of colonization occurs within Heliconiinae ; eight species in the African genus Acraea (tribe Acraeini ) have been reported on Turneraceae ( Robinson et al., 2007) .

In a broad sense, Acraea mirrors host associations across the entire Dioptine. There are approximately 240 species of Acraea ( Pierre, 1987; Larsen, 1991; Williams, 2006), and their larvae have been recorded from nearly 20 different plant families ( Ackery, 1988). However, host-plants reminiscent of the Dioptinae figure prominently. These include: Rinorea and Hybanthus ( Violaceae ), Tricliceras and Turnera ( Turneraceae ), various genera of Passifloraceae ( Adenia , Basananthe , Passiflora ), as well as Kiggelaria and Rawsonia (both Achariacae). For Acraea , the Urticaceae and Malvaceae are also important hosts. Within the Dioptini (table 4), Xenorma specializes on Cecropia ( Urticaceae ) while at least four Polypoetes species have been discovered on Malvaceae . Host utilization in Acraea seems to act as a map for the Dioptinae .

Returning to the overall picture, it is remarkable that both the Dioptinae and Heliconiinae contain species specializing on Passifloraceae , Violaceae , Achariaceae , and Turneraceae . According to current theory, these plant families together constitute a subclade within the large order Malpighiales ( Chase, 2004; Soltis et al., 2005). Their interrelationships are only now being explored (e.g., Hearn, 2006). These host-plant parallels could not possibly have evolved by chance. Instead, during their evolutionary histories, the two Lepidoptera groups must have independently converged on similar sets of chemical cues, employed for finding their hosts. Furthermore, the caterpillars may have evolved similar metabolic abilities, enabling them to detoxify, and perhaps sequester, particular plant compounds. Over 45 years ago, Ehrlich and Raven (1964), describing patterns of host utilization in the Heliconiinae , predicted the existence of chemical similarities uniting these four plant families, even though none were known at the time. Their prediction is now being born out.

The Passifloraceae have long been recognized as a rich source of secondary compounds, including harmane alkaloids and flavonoids ( MacDougal, 1994). Perhaps because of such chemicals, these plants appear to be strongly defended against herbivorous insects. Other than the Josiini and Heliconiinae , a single clade, the leaf beetle subtribe Disonychina ( Chrysomelidae : Galerucinae : Alticini), specializes on Passiflora ( Duckett, 1999) . Recent research has begun establishing precise phytochemical links between the Passifloraceae , Violaceae , Turneraceae , and Achariaceae . While cyanogenesis is widespread in the plant world, a conspicuous novelty for members of this clade is the ability to produce cyanogenic glycosides with a cyclopentene moiety ( Spencer, 1988). For example, cyclopentanoid cyanohydrin glucosides are produced by both Passiflora ( Jaroszewski et al., 2002) and Lindackeria ( Jaroszewski et al., 2004) . Turneraceae manufacture precisely these same compounds ( Saupe, 1981; Spencer, 1988; Shappert and Shore, 1995).

Butterflies in the Limenitidinae ( Nymphalidae ) provide further indication that plant chemistry has guided the evolution of these host-plant associations. In this taxonomically complex group, two host families predominate, the Euphorbiaceae and Sapindaceae ( Ackery, 1988) . Nothing seems to presage the host affiliations of heliconiines or dioptines. However, tucked within the morass of limenitidine genera are two African taxa— Cymothoe and Harma . Most species in these genera feed on Rinorea , while others specialize on a range of plants in the Achariaceae , including Lindackeria ( Amiet and Achoundong, 1996; Larson, 1991). In yet another fascinating case of convergence, Cymothoe and Harma seem to have broken through a biochemical barrier allowing them to colonize two plant families rarely utilized by other Lepidoptera , but important in the host-plant histories of the Dioptinae and Heliconiinae .

It is now generally acknowledged that parallel cladogenesis with host lineages is rare in Lepidoptera ( Janz and Nylin, 1998; Powell et al., 1999). An alternative explanation for observed host use patterns suggests that certain phytophagous insect groups have overcome phytochemical barriers, allowing them to track particular classes of plant secondary chemicals over evolutionary time (e.g., see Miller, 1987c). Major shifts in specialization can in turn lead to cases of adaptive radiation, driven by ecological opportunities in the novel host-plant zone ( Mitter and Brooks, 1983; Mitter et al., 1988; Farrell et al., 1992; Lees and Smith, 1992; Miller and Wenzel, 1995; Willmott and Freitas, 2006). Phylogenetic studies further demonstrate that, far from being an evolutionary dead end, specialization and recolonization are extremely dynamic processes ( Janz et al., 2001). Host shifts mediated by chemistry will often involve related plants, because secondary chemical arrays, at least to some degree, reflect botanical relatedness. Ultimately, however, host transformations in phytophagous insects seem to more accurately reflect chemical similarities among plants, rather than their cladogenesis. The most likely explanation for the observed convergence in host-utilization patterns between dioptines and heliconiines seems to lie in this tracking hypothesis. In accordance with the theme so eloquently put forth by Ehrlich and Raven (1964), cyanogenic cyclopentanoid gylosides are likely candidates as female oviposition cues or larval defenses, shared by the Dioptinae and Heliconiinae .

OTHER HOST-PLANT TRENDS: Surprisingly, monocots have become the new frontrunners in the race for preeminent dioptine host plants. The first monocot record came from the newly described Costa Rican species Tithraustes snyderi , reared at Las Alturas Field Station in 1992 by Cal Snyder (AMNH). The caterpillars were feeding on an understory palm in the genus Chamaedorea . Since that time, additional Tithraustes species have been discovered on palms, as have members of Isostyla and Stenoplastis (table 4). These three genera constitute a clade near the top of the dioptinae phylogeny (Clade 17; fig. 7). At least some palm-feeders are oligophagous, showing a fairly broad diet breadth within the Arecaceae (table 4). Altogether, six understory palm genera are now included in the host lists of Tithraustes and Isostyla .

Recent discoveries of monocot-feeding in Dioptinae are attributed to Dyer et al. (2009), working in eastern Ecuador. Their team has discovered species in three genera (table 4) specializing on cloud forest bamboo—caterpillars of Dolophrosyne , Scoturopsis and Xenomigia (pl. 38N) have been recorded on Chusquea (pl. 44C, E, F). According to my studies, these three genera form a cloud-forest clade (Clade 14; fig. 7), along with Xenormicola . If all the members of Clade 14 are eventually revealed to be Chusquea - feeders, then this represents a major radiation of monocot specialists in the Dioptinae .

Finally, two separate teams in Costa Rica have discovered larvae of Dioptis longipennis on the palm Geonoma cuneata ( Dyer and Gentry, 2002; Janzen and Hallwachs, 2008). This dioptine, newly added here to Dioptis after transfer from its previous position in Tithraustes , represents a basal element of the genus (fig. 3). In January 2009, I was part of an expedition that discovered two Cyma Group species, these being more derived members of Dioptis , on Geonoma in Amazonian Ecuador (see table 4). It thus seems highly probable that all 45 Dioptis species utilize understory palms as their larval hosts.

When the palm- and bamboo-feeding dioptine genera are considered together, they currently total 87 described species, roughly a quarter of the Dioptini . However, this number is misleading. Xenomigia and Dioptis are two of the most poorly known genera in the subfamily. Both contain vast numbers of undescribed species (table 7). For example, only 11 Xenomigia have been described to date (appendix 2), but when the genus is thoroughly revised it will ultimately become one of the largest in the Dioptinae , probably totaling well over 50 species. Monocots will undoubtedly be revealed as one of the most important host-plant associations in the Dioptinae , of equal significance to the Passifloraceae .

Continuing our pursuit of comparisons between the Dioptinae and butterflies, a major butterfly lineage, the Satyrinae (2400 spp.), is dominated by monocot-feeders ( Ackery, 1988). Across the Satyrinae a wide range of monocots are utilized, with palms and bamboos well represented. Three closely related, basal tribes of the satyrine clade (see Peña et al., 2006)—the Old World Amathusiini (92 spp.), the New World Brassolini (92 spp.), and the New World Morphini (42 spp.)—are associated with monocots. Interestingly, Morpho is found mostly on Fabaceae , but a few Morpho species specialize on Chusquea ( Penz and DeVries, 2002) , the same cloud-forest bamboo fed upon by Dolophrosyne , Scoturopsis , and Xenomigia . Higher up in the satyrine phylogeny, an extremely diverse radiation, the Andean tribe Pronophilini , consists almost entirely of Chusquea specialists ( Pyrcz and Wojtusiak, 2002; Viloria, 2003). In summary, at least roughly speaking, monocot-feeders form a single butterfly assemblage ( Ehrlich and Raven, 1964). Similarly, preliminary indications are that monocot-feeding arose in Clade 11 of the Dioptinae , the large terminal lineage (fig. 7). The picture is complicated by apparent loss of the monocot association in Pseudoricia (table 4).

Another host-plant comparison between butterflies and Dioptinae was initiated by an early reference to pipevine-feeding in Phaeochlaena gyon (table 4). Ehrlich and Raven (1964) heralded the association between pipevines ( Aristolochia ; Aristolochiaceae ) and swallowtail butterflies in the tribe Troidini as one of the strongest pieces of evidence in favor of coevolution. At a time when our understanding of host-plant associations in the Dioptinae was in its infancy, I speculated that an Aristolochia association might be important for the group ( Miller, 1992a). However, two lines of evidence suggest otherwise: First, there are no modern host-plant records for P. gyon ; published references suggesting Aristolochia as the host of this moth (e.g., Prout, 1918; Hering, 1925; Biezanko, 1962a) seem to have simply perpetuated the original citation by Mabilde (1896). Mabilde described the host plant of P. gyon as a lactiferous vine, using two Brazilian colloquial names— timbó and baba de touros. According to Carla Penz (personal commun.), the first may refer to either Ateleia ( Fabaceae ) or Serjania ( Sapindaceae ). The second, which translates as ‘‘bull’s spit’’, is thought to refer to Araujia sericifera (G. Lamas, personal commun.) in the Asclepiadaceae . The pipevine reference is thus in doubt. Since butterfly workers studying the Troidini frequently search Aristolochia foliage for larvae, one might suspect that caterpillars of P. gyon would have been discovered in recent years. Secondly, Solanum ( Solanaceae ) has been established as a new host for Phaeochlaena lampra , the sister species of P. gyon (table 4). This in itself is fascinating, since Solanaceae is the most important host for the butterfly subfamily Ithomiinae ( Ackery, 1988; Willmott and Freitas, 2006). No additional Phaeochlaena hosts are known, but their discovery is crucial, since this genus constitutes one of the most prominent diurnal clades in the Dioptinae (see Wing Pattern, above).

Some of the newly discovered host-plant associations do little to bolster the Dioptinaebutterfly paradigm. These instead will provide challenges for future understanding of host-plant evolution. For example, it now appears that many species of Erbessa are associated with Melastomataceae —eight species have so far been recorded on that plant family (table 4), and more will undoubtedly be found. Two other Erbessa species feed on Myrtaceae . There is no published chemical or taxonomic link between these two plant families, so this transition is difficult to explain. It is equally difficult to comprehend the switch from Violaceae , the apparent ancestral host plant of Erbessa (fig. 356), to either Melostomataceae or Myrtaceae . It will be interesting to see what patterns unfold as hosts for the remaining Erbessa species come to light.

Another seemingly anomalous association involves Cecropia ( Urticaceae ), the host of Xenorma . Xenorma ’s position on the dioptine cladogram (fig. 356) suggests that the transition to Cecropia took place in an ancestor that fed on Violaceae . Similarly, Bauhinia ( Fabaceae ) makes little botanical sense within the context of known host associations for the larger clade to which Brachyglene belongs; that plant does not appear in the host lists of Chrysoglossa or Nebulosa (table 4). However, an understanding of host use in Brachyglene is hampered by a dearth of information. In addition to having few records for Chrysoglossa and Nebulosa , we know nothing about the biology of its nearest relatives— Cacolyces and Hadesina (fig. 356). Perhaps as we learn more about the life histories of these groups a cohesive picture will begin to emerge.

The known host plants of Chrysoglossa , Quercus and Alfaroa , are interesting for two reasons: First, most Dioptinae , even those endemic to northern portions of the group’s range, feed on plants more typical of the tropical flora. The association between Phryganidia and oaks has been well documented ( Herbert, 1920; Beutelspacher, 1986; Puttick, 1986; Casher, 1996). Chrysoglossa is the only other dioptine found on them. Chrysoglossa norburyi is the lone dioptine recorded from Juglandaceae , another plant group characteristic of the temperate realm. Second, these plant associations provide a parallel with the hosts of other notodontid subfamilies. Many species of Notodontidae feed on oaks. The large North American genus Datana ( Phalerinae ) includes a species restricted to Juglandaceae , D. integerrima , as well as one on Quercus , D. contracta ( Forbes, 1948; Miller, 1992a; Wagner, 2005). As additional hosts are discovered for Chrysoglossa , it will be interesting to see whether this pattern holds true.

There are several large dioptine genera for which the accumulated host records so far show no obvious pattern. For example, the host list of Polypoetes tends toward Malvaceae , with a collection of unrelated families thrown in. However, the genus is so large and complex that speculation concerning patterns of host-plant utilization is premature; three of five species groups are not represented at all on the Polypoetes host list (table 4). One trivial comparison with butterflies can be seen in the use by two Polypoetes species of Celtis , or Hackberry, which is fed upon by all members of the Libytheinae , a small nymphalid subfamily with worldwide distribution ( Michener, 1943).

There is likewise no apparent pattern in the host list of Nebulosa . Of the 30 described species, food plants are known for only three (table 4), but at least two of those are reminiscent of hosts in unrelated dioptine groups: Tibouchina belongs in the Melastomataceae along with Miconia , the host of Erbessa ; Cestrum is in the Solanaceae with Solanum , the host of Phaeochlaena lampra . Interestingly, Cestrum and Solanum are important food plants of ithomiine butterflies ( Ackery, 1988). One hypothesis is that Nebulosa has colonized a series of plants for which it was biochemically preadapted. Obviously, more data are needed.

THE JOSIINI AND PASSIFLORACEAE : The relationship between Heliconiinae and Passifloraceae has been touted as one of the most important test cases for coevolution ( Ehrlich and Raven, 1964; Mitter and Brooks, 1983; Brower, 1997). The plants themselves present a complicated scenario. The Passifloraceae comprises roughly 630 species in 18 genera ( Vanderplank, 2000). However, the great bulk of this diversity, approximately 450 species, is contained in a single genus, Passiflora ( Ulmer and MacDougal, 2004) . Only 20 Passiflora species occur in the Old World ( Krosnick and Freudenstein, 2005; Hearn, 2006), while the remainder is endemic to the Neotropics ( Holm-Nielsen et al., 1988; Gentry, 1993). Because of the group’s taxonomic complexity, Passiflora classification has been controversial. The earliest and most influential treatise proposed 22 subgenera ( Killip, 1938), but more recently that number has been greatly reduced. In the current classification, based on a cladistic analysis of molecular characters ( Yockteng and Nadot, 2004), only eight Passiflora subgenera are recognized. The most primitive of these is Astrophea ( Escobar, 1994), which contains shrubs and treelike forms (pls. 40A, C; 41B, C) with relatively inconspicuous, white flowers (pl. 40D). The more derived Passiflora subgenera, such as Granadilla , include vines exhibiting some of the most spectacular flowers in the plant world (e.g., pl. 41E).

Benson et al. (1976) compared a heliconiine phylogeny with one for the Passifloraceae and argued that primitive butterflies show a historical predilection for colonizing Astrophea . Subsequent reevaluation of the case ( Mitter and Brooks, 1983) suggested that the Astrophea association evolved at least four times within the Heliconiinae , in derived positions on the butterfly phylogeny. My own examination of the most recent heliconiine cladogram ( Beltrán et al., 2007), along with study of published host-plant records (e.g., Benson et al., 1976; DeVries, 1987; Ackery, 1988), confirms multiple origins for Astrophea -feeding. No fewer than four separate colonization events took place within Heliconius , and at least two more occurred within its sister genus, Eueides . One could conclude that the Heliconiinae show evolutionary flexibility for host transfer between Passiflora subgenera. There is no evidence for mutual descent between these butterflies and plants.

Compared to the Heliconiinae , where hosts have been recorded for nearly every butterfly species, the picture in the Josiini is incomplete. To date, life histories have been discovered for 31 josiine species (30% of the tribe) in seven genera (table 6). According to the cladogram of josiine relationships (figs. 7, 283), Turnera (pl. 41D) has been colonized twice independently—once in Josia and once in Ephialtias . These represent derived transitions, apparently from a Passiflora -feeding ancestor. Within Ephialtias , a host shift to Achariaceae also occurred. It will be interesting to obtain additional host-plant records for Ephialtias to better understand the sequence of events leading to the evolution of these novel associations. The host list for the Josiini also demonstrates use of Passiflora species across a broad taxonomic spectrum, representing six of the subgenera defined by Killip (1938). Unlike the situation in Heliconiinae , an association with Astrophea appears to have arisen only once. Polyptychia and Getta , the Astrophea -feeding genera, belong in a clade along with Phavaraea , whose hosts are as yet unknown, near the base of the tribal phylogeny (figs. 7, 283). Interestingly, according to this phylogenetic hypothesis, primitive Josiini are indeed found on primitive Passiflora . However, before making bold generalizations, it will be important to fill in several missing data entries for the Josiini . In particular, it is crucial that we discover host plants for Proutiella , the basalmost element of the tribe. Other glaring holes include the hosts of Phintia and Notascea , pivotal genera in the josiine phylogeny for which life history information remains unknown.

FUTURE RESEARCH

This publication provides the taxonomic context to support many areas of future inquiry. Several of the subcategories touched on above could be expanded into research projects of their own. For example, developing a comprehensive hypothesis of host-plant evolution in the Dioptinae , along with the concordant field studies needed to complete the host list, could occupy decades of intensive research. In the near term, it is important to identify methods for resolving a more restricted problem—understanding the evolution of dioptine genera. As is acknowledged throughout this paper, while I place considerable confidence in the robustness of the generic classification proposed (appendix 2), the phylogenetic hypothesis of genus interrelationships is provisional. It relies on a single character set, adult morphology. Below, I outline the potential utility of incorporating characters from immature morphology and DNA characters into future analyses.

MORPHOLOGY OF IMMATURE STAGES: Characters from larvae and pupae can be immensely important for unraveling phylogenetic relationships in holometabolous insect groups (e.g., DeVries et al., 1985). Although in most studies, immature traits are fewer in number than those from adults, in combined analyses the immature data invariably increase overall tree resolution and tree support ( Meier and Lim, 2009). These facts have been born out by research on families throughout the Lepidoptera (e.g., Epstein, 1996; Yen et al., 2005a), including the Notodontidae ( Miller, 1991) , and have been documented for other insect orders as well (e.g., Alexander, 1990). Immature stages in the Notodontidae have long been recognized as a particularly rich source of taxonomic and phylogenetic information ( Packard, 1890, 1895; Nagano, 1916; Godfrey and Appleby, 1987). For reasons not yet explained, notodontid immatures exhibit more structural diversity than any other family in the Noctuoidea ( Wagner, 2005).

An analysis of relationships among dioptine genera that incorporates immature and adult characters thus has a high probability of yielding a strongly supported and wellresolved cladogram. A basic problem at present is that immatures are known for only 26 of the 43 genera. Obtaining life histories and preserved immatures for species representing the remaining genera may not seem an insurmountable problem. However, many of these taxa are extremely rare or live in remote locations. One could nevertheless hope that fieldtrips to strategically chosen sites will provide crucial material. Below, I review several morphological characters from larvae and pupae that show promise for phylogenetic analysis. These examples are simply meant to highlight the utility of immature stages. Comprehensive study will undoubtedly yield a large data set.

It has already been established that larval morphology provides important synapomorphies at the subfamily and tribal levels in Dioptinae . For example, the presence of cuticular microprojections covering the body (figs. 358C–F, 359B, 359C) provides an instantly recognizable feature for the subfamily ( Forbes, 1939a; Stehr, 1987; Godfrey and Appleby, 1987). This unusual surface structure, observed by early Lepidopterists ( Fracker, 1915) and termed ‘‘shagreened’’ ( Peterson, 1962; Nichols, 1989), provides an important synapomorphy for the Dioptinae ( Miller, 1991) . The larval head surface in Notodontidae shows a variety of microsculpture types. Typically it is covered with tiny pebblelike projections, giving it a bumpy appearance ( Miller, 1991), as occurs in the Dioptini (fig. 357B, C, E). Josiine caterpillars instead exhibit a derived condition—presence of a smooth, almost glassy head surface (fig. 358A, B). This trait seems to unite the entire tribe ( Miller, 1996). The Josiini is characterized by a second synapomorphy from larvae, presence of only four instars ( Spitz, 1931; d’Almeida, 1932 b; Markin et al., 1989; Miller and Otero, 1994; Miller, 1996), rather than the five or six found in all other Notodontidae ( Packard, 1890, 1895; Godfrey and Appleby, 1987), including the Dioptini ( Herbert, 1920; Wolda and Foster, 1978).

Certain features of the larvae may prove useful for establishing relationships among genera. Modified A10 prolegs characterize the Notodontidae ( Peterson, 1962; Godfrey and Appleby, 1987; Wagner, 2005). Usually these are reduced and not used for walking, but are held aloft. In some notodontids, such as Cerura ( Notodontinae ), the A10 prolegs form greatly elongated, whiplike tails at the end of the abdomen. These specialized structures, termed stemapods ( Gerasimov, 1952) or lashes ( Holloway et al., 1987), bear eversible, warningly colored glands at their apices ( Wagner, 2005), presumably used as a defense against parasitoids ( White et al., 1983; Chow and Tsai, 1989). Most dioptine caterpillars possess small A10 prolegs (e.g., fig. 359A, B), typical of Notodontidae (e.g., Polypoetes villia , pl. 38B; Josia megaera , pl. 39L). However, stemapods appear in at least three genera. Their most dramatic development occurs in Erbessa (pls. 36D, F, I; 37A, B, D), where the A10 prolegs are greater than half as long as the body. The stemapod surface is shagreened (fig. 358C, D, F), and there is an invaginated gland at its apex (fig. 358E). Caterpillars of Phanoptis (pl. 37G) and Phaeochlaena ( Bastelberger, 1908) exhibit shorter stemapods. It will be interesting to document the distribution of stemapods across the Dioptini . They never occur in Josiini . This modification has evolved by convergence in four different notodontid subfamilies ( Miller, 1991). Furthermore, the stemapods of Dioptinae show fundamental structural differences from those of other Notodontidae ( Miller, 1991) . Our understanding of these fascinating structures would benefit from thorough comparative study.

The primary setae of Lepidoptera caterpillars provide extremely useful taxonomic characters (e.g., Fracker, 1915; Hinton, 1946; Kitching, 1984b). In the Noctuoidea, chaetotaxy is an important source of traits for diagnosing families, and for resolving relationships among families ( Kitching and Rawlins, 1999; Fibiger and Lafontaine, 2005). Within the Notodontidae , one of the key synapomorphies supporting a relationship between the Dioptinae and their sister group, the Nystaleinae , is the location of seta L2 on abdominal segment 8 (see Introduction: Phylogenetic Position of the Dioptinae ). The primary setal arrangements of Dioptinae have not been comprehensively studied, but such effort will undoubtedly yield valuable character information. For example, some Josiini possess two setae rather than one in the L position on segment A9 ( Miller and Otero, 1994; Miller, 1996). Characters from chaetotaxy are apparently more conservative than those from adults, so they thus offer strong potential for stabilizing an hypothesis of relationships among dioptine genera.

As far as is known, all Lepidoptera caterpillars possess three enlarged setae at the base of each thoracic tarsus, located on the mesal surface (fig. 359D). These are thought to function as adhesive devices ( Hasenfuss and Kristensen, 2003). Taxonomic value, derived from variation in the shape and number of these tarsal setae, has been demonstrated in groups spanning the Lepidoptera ( MacKay, 1972) , including the Zygaenidae ( Yen et al., 2005a) , Noctuidae ( Beck, 1960; Godfrey, 1972), and Notodontidae ( Miller, 1991) . A survey across the Dioptinae would be useful; subtle shape differences are shown by various species of the Josiini ( Miller and Otero, 1994) . Other morphological features in larvae worthy of attention include the spinneret ( Miller, 1991; Miller and Otero, 1994), antenna ( Dethier, 1941), and internal mouthpart structure ( Kitching, 1984b).

A study of representative Josiini ( Miller, 1996) suggested that body coloration in larvae could provide pattern characters of equal importance to those in adult wings. Larval patterns seem to be relatively conservative, and in fact might outperform those on the wings, which are notoriously labile and difficult to interpret. Final instar Polypoetes larvae are diagnosable by a pair of conspicuous, longitudinal yellow stripes—one running subdorsally and the other laterally (pl. 37H, I; 38A–C, F). Considering the huge spectrum of morphological variation shown by Polypoetes adults, simple larval coloration characters such as these might offer useful diagnostic tools. Within genera, it seems likely that larval body patterns could reflect species group boundaries based on adults. Pinkish purple dorsal patches appear on segments A3, A5, and A 8 in caterpillars of Nebulosa elicioi (pl. 38K), a member of the Nervosa Group. On the other hand, Nebulosa yanayacu , from the Fulvipalpis Group, exhibits pinkish dorsal patches on all abdominal segments, as well as on T2 and T3 (pl. 38L). Within species groups, head-pigmentation patterns are effective for separating closely related species ( Miller and Otero, 1994; Miller, 1996). Overall, a comparative analysis of larval coloration holds great promise for future phylogenetic work.

Study of pupal morphology will also yield fruitful results. This life stage has long been known to provide characters effective for resolving Lepidoptera classification ( Mosher, 1916). The pupae of Notodontidae are particularly variable, and offer a wealth of taxonomically informative traits ( Mosher, 1917; Nakamura, 2007). In a study of notodontid subfamilies, a set of only 24 pupal characters was remarkable for accurately retrieving the same phylogenetic hypothesis as one based on a much larger data set from larvae and adults ( Miller, 1992b). Except for Phryganidia californica , which has been treated in detail ( Packard, 1895; Mosh- er, 1916; Forbes, 1939a; Miller, 1992b), the pupae of Dioptinae have not been studied. Nevertheless, judging from what little has been documented so far, this life stage will be an important component of future analyses.

The pupae of most Notodontidae are formed in loose cocoons, usually hidden inside a cell, either in the ground or on the ground surface ( Forbes, 1939a; Forbes, 1948). This is generally true of the Josiini , which pupate within a shelter composed of leaves or debris tied together by larval silk ( Miller and Otero, 1994). The shiny pupae of josiines are brown to reddish brown (pl. 39I), with black markings. In contrast, chrysalides of the Dioptini are contrastingly patterned, without cocoons (pls. 36H; 37C, J–M; 38D, J; 39C, F). These are attached to the substrate in the exposed manner of many Nymphalidae ( Kitching and Rawlins, 1999) . This mode of pupation in dioptines, derived within Notodontidae , seems to be correlated with a morphological attribute unique in the Lepidoptera —the presence of hook-shaped setae on the dorsum of abdominal segments 7–9 (figs. 360E, F; 361A, C, D). These setae, serial homologues of those on the cremaster (fig. 360F), provide additional anchors when the chrysalis attaches itself to its silk mat (e.g., pl. 37J–M; 38J). Packard (1895) first documented abdominal anchoring setae in pupae of Phryganidia californica (see also, Mosher, 1916; Miller, 1987a). However, these occur in additional dioptine genera, including Dioptis , Polypoetes , and Nebulosa . A survey of their taxonomic distribution across the Dioptini will be valuable.

Erbessa and Scotura belong in a clade, along with three additional genera— Cleptophasia , Oricia , and Eremonidia —identified by analysis of adult morphology (figs. 7, 30). Pupae of these latter genera are unknown, but those of Erbessa and Scotura exhibit two unique characteristics, potentially adding evidence for monophyly. First, although primary setae, analogous to those of larvae, are found universally in Lepidoptera pupae ( Common, 1990), they are usually microscopic and difficult to locate. In the pupae of some Dioptinae , on the other hand, the primary setae are large and conspicuous (fig. 360B). This condition occurs in Erbessa and Scotura ( Miller, 1992b) . Interestingly, these genera do not posses abdominal hook-shaped setae. Secondly, the pupal heads of Erbessa and Scotura bear anterior processes found nowhere else in the Notodontidae ( Miller, 1992b) . In Scotura , the head process varies in its shape and size, but is usually short and globose (fig. 360A). The homologous process of Erbessa is remarkable in being elongate and bifid (pl. 37C); in E. lindigii it forms a pair of extremely long, thin horns (pl. 36H). The function of this structure, if there is one, is open to speculation.

Enlarged primary setae and head processes are not found in other Dioptinae , and thus seem to be synapomorphic for Erbessa + Scotura . I predict that they will be found in pupae of Cleptophasia , Eremonidia , and Oricia as well. However, an anomaly presents itself. The only record of pupal morphology for the genus Phaeochlaena is represented by a single exuvium of P. gyon , in the BMNH collection ( Brazil, São Paulo, Ypiranga, Aug 1928 , leg. R. Spitz). This specimen exhibits long, slightly club-shaped primary setae and a narrow, beak-shaped head process. Thus, although evidence from adult morphology places Phaeochlaena in a clade with Polypoetes and relatives (figs. 3, 7), pupal characters indicate otherwise; the genus may more accurately fall within the Erbessa / Scotura clade. Only when larvae and pupae representing all dioptine genera become available, and the characters from these are added to an adult matrix, can we begin to place confidence in an hypothesis of generic relationship. Such study may ultimately lead to a more refined tribal classification as well.

CHARACTERS FROM DNA: In recent years, molecular analyses have dominated the field of insect systematics ( Meier and Lim, 2009). Studies across insect orders have been extremely productive ( Caterino et al., 2001). Perhaps no group has received more attention from DNA systematists than the butterflies. A staggering amount of effort has been applied at a wide range of taxonomic levels. Nearly all the families have been examined, including the Papilionidae ( Caterino et al., 2001) , Pieridae ( Braby et al., 2006) , Riodinidae ( Campbell et al., 2000; Campbell and Pierce, 2003), and Hesperiidae ( Warren et al., 2009) . Family-level work on the Nymphalidae (e.g., Wahlberg and Nylin, 2003) has been supplemented by phylogenetic analyses of its various subclades, such as parts of the Heliconiinae ( Brower and Egan, 1997; Beltrán et al., 2007) and Satyrinae ( Peña et al., 2006) .

Unfortunately, comparatively few studies have addressed Lepidoptera groups other than butterflies. This is particularly frustrating considering the massive existing diversity. Most estimates put the number of butterflies, including Hedyloidea, Hesperioidea, and Papilionoidea, at 17,500 species ( Robbins, 1982; deJong et al., 1996; Ackery et al., 1999), with only a small percentage remaining to be described ( Lamas et al., 2004). Determining the total number of Lepidoptera species is problematic; roughly 150,000 have been described ( Kristensen and Skalski, 1999), but the number of extant species may reach as high as 500,000 (N. Kristensen, personal commun.). Thus, an inordinate amount of phylogenetic work has focused on butterflies, a group constituting somewhere between 4%–12% of the Lepidoptera . Thankfully, recent progress is being made on moths. A combined molecular/morphological analysis ( Wiegmann et al., 2002) has outlined considerable phylogenetic detail, with strong support, for relationships among the basalmost groups within the order. Molecular analyses have also addressed higher-level phylogenetic relationships in the Lasiocampidae ( Regier et al., 2000) and Sphingidae ( Regier et al., 2001) . The Noctuoidea have been the subject of important DNA studies (e.g., Weller et al., 1994), with the Noctuidae receiving particularly strong attention ( Mitchell et al., 2005). This brief review is not meant to downplay elegant morphology-based phylogenies, recently undertaken in a range of lepidopteran groups, such as the Zygaenidae ( Yen et al., 2005a) , Noctuidae ( Speidel et al., 1996b) , and Arctiidae (Jacobson and Weller, 2002) . Even butterflies have yielded to important morphological research within the past decade (e.g., de Jong et al., 1996; Penz, 1999; Penz and DeVries, 1999; Penz and DeVries, 2002; Penz and Peggie, 2003; Freitas and Brown, 2004; Willmott and Freitas, 2006).

Nevertheless, the systematic community has witnessed sweeping methodological changes in recent years, to the point where some researchers have questioned the relevance of morphology as a phylogenetic tool. These critics would relegate morphology to a limited role, either in species recognition ( Givnish and Sytsma, 1997), or to be examined a posteriori within the context of an existing molecular phylogeny ( Scotland et al., 2003). According to both schools of thought, molecular data should serve as the sole guide to phylogenetic relationships. As someone who has spent his entire scientific career awestruck by the power of comparative morphological analyses and all they can teach us about insect evolution, I find such suggestions specious. Baker and Gatesy (2002) and Jenner (2004) are eloquent advocates of a universal approach. These authors point out that all classes of systematic characters possess their own strengths and weaknesses; no data should be dismissed. Furthermore, they argue that concerns about combining molecular and morphological data are unfounded.

Much of the ‘‘morphology vs. molecules’’ controversy revolves around comparing the relative performance of partitioned data sets. All acknowledge that partitioning can be instructive (e.g., Nylin et al., 2001). For example, it can reveal the phylogenetic level or levels at which different character sets are most informative. The contentious issue is whether to choose cladistic hypotheses built on a subset of data or ones constructed from all data combined. A large body of empirical evidence has demonstrated the strength of combined analysis. Concordantly, these studies reemphasize the utility of morphological characters. Morphology and DNA utilized together can have synergistic effects ( Wahlberg et al., 2005), creating a methodology for deciphering even the most difficult phylogenetic problems. The potential of this approach has been born out by research across the animal kingdom, and the Lepidoptera provide poignant examples. One of the early test cases involves the Dioptinae . A study of the Josiini analyzed characters from larvae, pupae, adults and DNA ( Miller, Brower, and DeSalle, 1997). Bucheli and Wenzel (2005) examined higher-level relationships in the Gelechioidea, utilizing morphology as well as characters from mitochondrial DNA. Finally, combined analyses have investigated butterfly phylogeny at both the tribal ( Wahlberg and Nylin, 2003) and superfamily levels ( Wahlberg et al., 2005). All these studies converge on the same conclusion: Combined analysis provides the most robust phylogeny. Differing character sources dilute the bias of homoplasy inherent within individual data partitions ( Wahlberg et al., 2005).

Willi Hennig (1966) advocated the then controversial ‘‘holomorphological approach’’ in insect phylogenetics, whereby characters from all life stages are combined to produce the best-supported phylogeny. There is little doubt that, were he alive today, Hennig would further argue for combining characters from DNA with holomorphology. Future research on dioptine evolution should utilize characters from adults, larvae, pupae and DNA, as well as traits provided by any other novel forms of data, should they become available. Careful, scientifically sound analyses are the key. Our common goal of achieving stable phylogenetic hypotheses to understand broad evolutionary issues will emerge through study of characters from all available sources.

Figures

Fig. 304. Known distributions of newly described South American Josiini: Josia neblina, Notascea brevispinula, Notascea loxa, Phintia broweri, Polyptychia hermieri, and Scea torrida.

Fig. 353. Genitalia of Scea torrida, sp. nov. (³ JSM-501, ♀ JSM-502). A, ³ genitalia; B, ³ Tg8; C, ³ St8; D, ♀ genitalia; E, aedeagus.