A new molecular phylogeny offers hope for a stable family level classification of the Noctuoidea (Lepidoptera)

Introduction

The classification of the diversity of life on Earth is one of the major ongoing undertakings of human society (Wilson 2000), but is still far from completion, both in terms of the inventory of species and of the classification of those species in a hierarchical system that has a phylogenetic basis. Although there has been some discussion of abandoning the 250-year-old Linnaean system for such classification, focusing more on genetic diversity, most researchers still prefer to use this hierarchical system. It provides a framework for access to a massive information resource on the biology, ecology and economic importance of all species, perhaps epitomized today by the Encyclopaedia of Life initiative (Wilson 2003). The use of molecular data, in particular DNA sequences, is becoming increasingly important for testing and improving classifications, especially for highly diverse groups of organisms such as insects.

Lepidoptera are one of the four major Orders of insects, and Noctuoidea are the largest superfamily within the Order with between 42 000 (Heppner 1991) and 70 000 (Kitching & Rawlins 1998) described species. The next most numerous superfamily is Geometroidea with about 21 500 species, followed by Pyraloidea, Papilionoidea and Gelechioidea each with between 15 000 and 16 000 species (Heppner 1991). In contrast, the total number of terrestrial vertebrates is approximately 21 500 species (Maddison 2007).

The monophyly of Noctuoidea, based on the presence of a single apomorphic character, the metathoracic tympanal organ (Miller 1991) and its associated abdominal structures, seems well established (Kitching & Rawlins 1998; Mitchell et al. 2006). However, the limits and content of the constituent families, and the evolutionary relationships amongst and within these, are very poorly understood (Mitchell et al. 2006).

Noctuoid species are placed in approximately 4200 genera (Kitching & Rawlins 1998) but there are numerous undescribed species, particularly from tropical regions. For example, the noctuoid species total for Borneo now stands at about 2800, an increase of over 50% on a total estimated from a previous survey (1980) of the collections of The Natural History Museum in London (J. D. Holloway, unpublished data). This increase is the result of a quarter of a century of taxonomic effort on a major input of fresh survey material from Borneo, leading to a series of monographs on the Macrolepidoptera, those on the noctuoids mostly cited in this paper. Taking as a minimum the calculations of the total Bornean Lepidoptera fauna by Robinson & Tuck (1993, 1996), this figure for noctuoids represents about one quarter of that total. The proportion of Noctuoidea in the global total is likely to be similar.

The larvae of many noctuoid genera include armyworms, cutworms, bollworms and stem borers, that collectively have a massive economic impact annually (Kitching 1984). The adults of other genera damage fruit crops by piercing the skins to suck juices (Bänziger 1982). Noctuoids constitute one quarter of the approximately 6000 Lepidoptera species noted to be of economic importance by Zhang (1994). Though many of these can be assigned to what Mitchell et al. (2006) termed the ‘pest clade’, many more are distributed across the whole superfamily in over 500 genera (Zhang 1994). Therefore, resolution of stable, extrapolative higher level classificatory structure for the superfamily may prove to be an important prerequisite for studies of pest bionomics across the group.

Numerous classifications of the family groups of Noctuoidea have been proposed. The fundamental distinction between the different systems is based on the use of unsatisfactory (occasionally plesiomorphic) characters in phylogenetic reconstruction. Various authors have recognized between five and thirteen families, and strikingly, no two publications have agreed on the same divisions of the superfamily into families (Kitching & Rawlins 1998; Lafontaine & Fibiger 2006). Miller (1991) recognized seven families: Oenosandridae, Doidae, Notodontidae, Lymantriidae, Arctiidae, Aganaidae and Noctuidae. Scoble (1992) included six families, placing Aganainae as a subfamily within Noctuidae. Kitching & Rawlins (1998) later recognized three fundamental lineages of Noctuoidea: Oenosandridae, Doidae + Notodontidae, and the quadrifid families (those where vein MA2 arises very close to, or is stalked with, MP1 in the forewing, i.e., Arctiidae, Lymantriidae, Noctuidae, Nolidae and Pantheidae). Most recently, three landmark publications (Fibiger & Lafontaine 2005; Lafontaine & Fibiger 2006; Mitchell et al. 2006) presented detailed phylogenies and revised the classification of Noctuoidea three times, each classification having its own limitations and strengths (Roe et al. 2010). Fibiger & Lafontaine (2005) proposed a new classification with ten families: Oenosandridae, Doidae, Notodontidae, Strepsimanidae, Nolidae, Lymantriidae, Arctiidae, Erebidae, Micronoctuidae and Noctuidae. Lafontaine & Fibiger (2006) proposed a further revision to the classification of the families of Noctuoidea, in which Nolidae, Strepsimanidae, Arctiidae, Lymantriidae and Erebidae sensuFibiger & Lafontaine (2005) were downgraded to subfamily status within an expanded family concept of Noctuidae based on the quadrifid venation of the forewing and the presence of a tympanal sclerite in the tympanal membrane. In their view, the superfamily should consist of five families: Oenosandridae, Doidae, Notodontidae, Micronoctuidae and Noctuidae.

Within the superfamily, the most controversial family group taxon is Noctuidae. Many of the traditional subfamilies are now recognized as unnatural (Kitching 1984; Beck 1991, 1992; Lafontaine & Poole 1991; Speidel et al. 1996; Kitching & Rawlins 1998; Fibiger & Lafontaine 2005; Mitchell et al. 2006). Indeed, the composition and monophyly of many subfamilies is still open to question, and in particular, the taxonomic composition of the quadrifine noctuoids (those with a strong vein MA2 in the hindwing) has remained notoriously difficult to establish. The situation has been reviewed recently by several authors (Speidel & Naumann 1995; Fibiger 2003; Kühne & Speidel 2004; Holloway 2005, 2008), who have suggested that the monophyly of the group was highly doubtful. At a noctuid workshop in Denmark in 2002 (Holloway 2005), it was decided that, prior to any further attempts to redefine family groups across the superfamily, it was necessary to gain a clearer understanding of the higher taxonomic diversity involved by attempting to identify on morphological grounds more potentially monophyletic groupings of genera within the immense diversity of the trifid section of the superfamily, particularly amongst the much less well worked quadrifine richness in the tropics. Studies of this kind would provide a basis for a sampling strategy for future phylogenetic studies across the group as a whole, exemplars being selected from significant groupings of genera and morphologically well-supported concepts of higher taxa (and subgroups thereof) such as the traditional Arctiidae and Lymantriidae. This approach was adopted by Holloway (2005) when exploring a broad cross-section of the Oriental tropical quadrifine fauna from Borneo, relating it as far as possible to type taxa of available family group names globally.

Several molecular studies have examined higher level relationships within the Noctuidae sensu lato. Weller et al. (1994), using partial sequences of nuclear 28S rRNA (300 bp) and mitochondrial ND1 (320 bp) from 26 noctuoid species, including 10 noctuids, noted that, despite low levels of support, parsimony analyses consistently grouped quadrifine noctuids with Arctiidae, and often Lymantriidae, rather than with trifine noctuids (those with vein MA2 in the hindwing usually vestigial or absent so that the cubital vein appears to branch into three veins), suggesting paraphyly of Noctuidae. However, vein reduction in this region also occurs in some Arctiidae and Nolidae as we discuss later. Subsequent studies based on sequences of two nuclear genes, Elongation Factor-1α (EF-1α) and Dopa Decarboxylase (DDC) (Friedlander et al. 1994; Mitchell et al. 1997, 2000, 2006; Fang et al. 2000) provided further evidence for the paraphyly of Noctuidae. Mitchell et al. (2006) found a strongly supported clade of quadrifine noctuid moths that included the families Lymantriidae and Arctiidae. They termed this the L.A.Q. clade (Lymantriidae, Arctiidae and Quadrifine Noctuidae).

Two recent molecular studies on ditrysian Lepidoptera sampled members of Noctuoidea and found that the enigmatic family Doidae did not group with the other noctuoids, but appeared to be related to Drepanoidea (Regier et al. 2009; Mutanen et al. 2010). Otherwise both studies found Noctuoidea to be monophyletic, with Oenosandridae being sister to the rest and Notodontidae the next lineage branching off.

However, all these studies had very poor sampling of the higher taxa putatively belonging to the L.A.Q. clade, and critically they did not sample type genera of many higher taxa. Given that the monophyly of many named groups remains in question, it is crucial to sample the type genera of each family, subfamily and tribe to assess the taxonomic limits of a given category.

Previous molecular studies have used only a small number of molecular markers, usually one to three gene regions (Wahlberg & Wheat 2008). Here, we present a phylogenetic hypothesis for higher taxa of Noctuoidea using new molecular data from eight gene regions.

Materials and methods

We sampled 152 representatives of many major lineages of the Noctuoidea complex. These comprise four outgroup taxa and 148 Noctuoidea species representing four families (Oenosandridae, Notodontidae, Noctuidae and Micronoctuidae), 50 subfamilies and 51 tribes, as recognized by Lafontaine & Fibiger (2006), as well as 16 taxa of uncertain position (Table 1). Based on the results of Regier et al. (2009) and Mutanen et al. (2010), as well as our own preliminary analyses, we did not include the family Doidae. We were unable to sample some scarce taxa with restricted distributions and/or low species richness (e.g., subfamilies Cocytiinae, Eucocytiinae and Strepsimaninae and the type genera of a few tribes/subtribes). To test the monophyly of the Noctuoidea, we included four species from three other superfamilies, namely Drepanoidea, Bombycoidea and Geometroidea. We rooted the cladograms with Thyatira batis (Drepanidae).

We extracted DNA from one or two legs, dried or freshly preserved in 96% ethanol, using the DNeasy tissue extraction kit (QIAGEN, Hilden, Germany). For each specimen, we sequenced the cytochrome oxidase subunit I gene (COI) from the mitochondrial genome, and the EF-1α, Ribosomal protein S5 (RpS5), Carbamoylphosphate synthase domain protein (CAD), Cytosolic malate dehydrogenase (MDH), Glyceraldehyde-3-phosphate dehydrogenase (GAPDH), Isocitrate dehydrogenase (IDH) and wingless genes, from the nuclear genome. All genes are protein coding and have been found to be highly informative for phylogenetic analyses at the level of families and superfamilies (Wahlberg & Wheat 2008; Wahlberg et al. 2009; Mutanen et al. 2010). PCR and sequencing protocols follow Wahlberg & Wheat (2008). Resulting chromatograms were checked and DNA sequences aligned by eye using the program BioEdit (Hall 1999). Alignment was trivial and the few insertion/deletion events that were detected, were of entire codons (in CAD, IDH and RpS5) and could be easily aligned.

The gene regions were analysed separately and combined in various partitions using parsimony and maximum likelihood (ML) methods. The data were combined in three ways: all gene regions together, all nuclear genes together (i.e., the mitochondrial gene COI excluded) and all gene regions together with third codon positions excluded.

Parsimony analyses were undertaken by performing New Technology heuristic searches in the program TNT (Goloboff et al. 2003). All characters were treated as unordered and equally weighted. Clade robustness was estimated by Bremer support (Bremer 1988, 1994) using a script (Peña et al. 2006) in TNT. Model-based phylogenetic analyses were implemented using ML and a GTR+G+I model was chosen as the most appropriate model of sequence evolution for each gene partition using FindModel (http://www.hiv.lanl.gov/content/sequence/findmodel/findmodel.html). However, we assigned all partitions with the GTR+G model, as the parameters I (proportion of Invariant positions) and G (Gamma distribution) are strongly correlated and deeply intertwined such that it is impossible to tease them apart (Ren et al. 2005; Kelchner & Thomas 2007), making it likely that complications arise in estimating values for these parameters. The gamma function is enough for correcting for the rate variations amongst sites, including sites which do not change at all in the dataset. ML analyses were conducted using the web-server RAxML (Stamatakis et al. 2008). ML bootstrap analysis with 1000 pseudoreplicates (Felsenstein 1985) was conducted with RAxML.

Results

Our analyses are based on sequence data from seven nuclear gene regions (1240 bp of EF-1α, 400 bp of wingless, 617 bp of RpS5, 850 bp of CAD, 410 bp of MDH, 691 bp of GAPDH and 710 bp of IDH) and one mitochondrial gene region (1477 bp of COI), for a total of 6407 aligned nucleotide sites (Table 2). We were not able to amplify some genes for some taxa (Tables 1 and 2).

The optimal cladograms found by the two methods (parsimony and ML) for the combined, complete datasets are very similar, but show novel relationships not previously suggested (1, 2). The monophyly of Noctuoidea is strongly supported (BP ≥96; BS ≥18), within which we find six strongly supported clades that we feel deserve family status. These are Oenosandridae, Notodontidae, Erebidae, Nolidae, Euteliidae stat. rev. and Noctuidae (1, 2). The Notodontidae are found to be the sister group of all other Noctuoidea, with the Australian family Oenosandridae branching off next. However, this pattern of relationships relative to the rest of Noctuoidea is not well supported. Both Oenosandridae and Notodontidae have a trifid forewing venation similar to that of Geometridae, a character state that appears to be plesiomorphic relative to the quadrifid forewing venation found in the other noctuoid families. Relationships amongst the remaining four families are not clear, although they form a monophyletic group with very strong support (Fig. 1). Euteliidae are sister to Noctuidae in ML analyses (Fig. 1), and sister to the other three families together in parsimony analyses. Similarly, Nolidae are sister to Erebidae in ML analyses, but form a trichotomy with Erebidae and Noctuidae in parsimony analyses.

The six strongly supported clades are also found when only nuclear gene regions are analysed (i.e., the mitochondrial gene is excluded) and when third codon positions are excluded (Supporting information, Appendices S1 and S2). However, the relationships amongst the clades we are designating as families (Oenosandridae, Notodontidae, Euteliidae, Erebidae, Nolidae and Noctuidae) are not stable. The two analyses now place Oenosandridae as sister to the rest of Noctuoidea, with Notodontidae branching off next, and this arrangement is quite well supported (BP ≥87 for the node Notodontidae + the rest) when the third codon positions are excluded (Supporting information, Appendix S2). The four remaining families always form a well-supported clade, but their relationships once again vary. Nuclear gene regions place Nolidae as sister to the rest and Erebidae as sister to Euteliidae + Noctuidae, whereas when third codon positions are removed, Euteliidae are sister to the rest and Erebidae are sister to Nolidae + Noctuidae. Of the single gene analyses only CAD recovers all six family clades as monophyletic (Supporting information, Appendix S3), although most members of the clades recovered in the combined analyses do tend to remain together in a clade with the other genes, and the non-monophyly of the families is not strongly supported. As the results of these analyses do not show strongly supported conflict with the combined, complete analysis, we henceforth describe in detail only the latter results.

Nolidae come out as a well-supported monophyletic clade (BP ≥97; BS ≥10) contra Beck (2009). Nolinae (represented by type genus) are placed as sister to the rest of the family. Eligminae are confirmed as belonging to this family, though they lack most of the diagnostic characteristics listed by Holloway (1998, 2003) and are associated with Selepa, a genus unassigned to any subfamily by Holloway. The subfamily Chloephorinae forms a major clade that also includes two further groups given subfamily status by Holloway: Bleninae, placed as sister to Sarrothripini; and Eariadinae, placed with representatives of Chloephorini and Ariolicini. The subfamily Westermanniinae remains separate; the subfamilies Afridinae, Collomeninae and Risobinae were not sampled.

Euteliidae are strongly supported as a monophyletic group (BP ≥100; BS ≥9). The clade consists of two subfamilies, Stictopterinae and Euteliinae.

The Noctuidae clade, as delimited in Fig. 1, has strong support (BP ≥99; BS ≥10), the most striking feature of which is the inclusion of a clade comprising two quadrifine tribes, Arcteini (represented by type genus) and Dyopsini (represented by type genus), which were previously assumed to be related to Erebinae sensuLafontaine & Fibiger (2006). Otherwise, this clade is made up of trifine noctuids. The ‘pest clade’ of Mitchell et al. (2006) (Heliothinae to Xyleninae) has strong BP (100) but weak BS (1) support.

Our results provide strong support (BP ≥99; BS ≥9) for the monophyly of Erebidae. Strikingly, many of the basal divergences in the family show very short branches with no or low support. Some traditionally recognized families, subfamilies and tribes show clear evolutionary relationships with strong BS and BP support. For example, there is clear support for a clade with Pangraptinae (represented by its type genus) as sister to another well-supported clade comprising Aganainae (represented by its type genus) + Herminiinae (represented by its type genus and three other core herminiines) and Arctiinae (represented by 12 genera including the type genus) (Fig. 2). The systematic position of what traditionally has been recognized as the family Lymantriidae is clearly within Erebidae (Fig. 2), consistent with the findings of previous molecular studies. Oxycilla, a genus recently excluded from Rivulinae by Fibiger & Lafontaine (2005), is grouped with Rivula (type genus of Rivulinae) with strong support values (BP ≥96; BS ≥20). Both ML and parsimony analyses place the recently discovered family Micronoctuidae within subfamily Hypenodinae (BP ≥89; BS ≥14). A strongly supported clade (BP ≥98; BS ≥18), which we term the boletobiine clade, places together a large number of taxa with very diverse feeding habits, including detritivory, fungivory, lichenivory, frugivory, pod-boring, predation on other insects and carrion feeding as well as defoliation; association of many of these taxa was suggested by Holloway (2005, 2009).

Our analyses fail to recover some previously recognized subfamilies within Erebidae as monophyletic groups (Fig. 2). Both parsimony and ML analyses suggest that some recent concepts of subfamilies Calpinae, Catocalinae, Erebinae and Phytometrinae are polyphyletic. For example, Calpinae sensuLafontaine & Fibiger (2006) consisted of four tribes, Anomini, Scoliopterygini, Calpini and Phyllodini. Our results placed two of these tribes, Anomini and Scoliopterygini, into a strongly supported monophyletic group (BP ≥100; BS ≥23) as in Holloway (2005). This is well separated from the clade that comprises the five genera of Calpini and two Phyllodini (BP ≥83). Within this clade, four Calpini (Calyptra thalictri, the type genus; Gonodonta; Plusiodonta; Oraesia) constitute a monophyletic group with some support (BP ≥93; BS ≥3), whereas the fifth calpine, Eudocima, was clustered with the two genera of Phyllodini (Phyllodes and Miniodes), albeit with low branch support (BP ≥67; BS ≥1).

Discussion

Until recently, the higher classification of the Noctuoidea has been based on morphological characters with a predominantly phenetic approach until the review by Kitching (1984). Higher taxa have been defined on alternative states of particular adult characters such as ones of wing venation mentioned earlier, but also: the orientation of the thoracic tympanum (ventral vs. posterior); the position of the counter-tympanal hood relative to the spiracle of A1 (anterior vs. posterior); spining of the mid-tibia (present vs. absent); genitalic structures including everted vesicae and expanded corpus bursae (Fibiger & Lafontaine 2005). However, these characters were often found to be in conflict, or both states were found to occur in taxa that otherwise would be considered congeneric, leading to debate about the relative weight each should be accorded. Genital characters, including structures of the everted vesica and within the corpus bursae, have also augmented the body of morphological information available, providing autapomorphies for all or major parts of many higher taxa within the superfamily, though not significantly so at the family level, except possibly for Euteliidae and Nolidae. For Erebidae, Fibiger (2003) and Holloway (2005) came to somewhat different conclusions about classification of the traditional quadrifine noctuids, even though these were based in part on the same collection of over 2000 genitalia slides from the Oriental fauna and from the type species of genera on which global family group names were based (Holloway 2005). There has also been disagreement about the relative value of larval vs. adult characteristics (e.g. Beck 2009), although several larval characteristics have been found to be of significance for higher classification (Kitching & Rawlins 1998). Besides early stages and adult characters, there is now another important new set of characters provided by molecular sequences. The increasing availability of molecular information has brought new insights into the relationships of noctuoid taxa (Weller et al. 1994; Mitchell et al. 1997, 2006), such as the paraphyly of the old concept of Noctuidae.

Our results, with more molecular data than have been used previously for this group of moths, point to six well-supported major lineages that we have defined as families. Two of the major lineages are well-recognized taxa that have often been considered families within Noctuoidea, i.e., Oenosandridae and Notodontidae. Oenosandridae are a small family, only known from Australia, comprising eight species in four genera (Nielsen et al. 1996), which mainly feed on Myrtaceae (Miller 1991). Miller (1991) excluded Oenosandra from Notodontidae based on the non-homology of scale tuft structures on the female tergite 7 (the hairs are used to cover the egg masses), placing it in a separate family. Notodontidae contain approximately 4200 species and occur worldwide. The other four lineages have been split into as many as 10 families, with arctiines, lymantriines and nolines frequently being considered to be sufficiently distinct from the rest to warrant full family status.

Wing venation has been thought to be informative of the phylogenetic relationships of noctuoids, with the quadrifine hindwing defining a group comprising our Nolidae, Erebidae and Euteliidae and the trifine hindwing defining Noctuidae. However, our results suggest that trifine moths have evolved from quadrifines multiple times, e.g., amongst some Erebidae, such as Arctiinae: Syntomini (Griveaud 1964; Holloway 1988), and Nolidae (Holloway 1998, 2003) show hindwing vein reduction in addition to the Noctuidae sensu stricto. Furthermore, the tribes Arcteini and Dyopsini, previously classified as quadrifines, are part of a well-supported clade otherwise consisting of trifines (some of the more basal ‘trifine’ subfamilies, e.g., Plusiinae, Pantheinae, Bagisarinae also have a quadrifine hindwing). The placement of Arcte is corroborated by morphological characters (see Holloway 2005, 2009), whereas Dyops has not yet been investigated morphologically in this context.

We follow Fibiger & Lafontaine (2005) in employing the family name Erebidae, which was previously used as a subfamily name by Forbes (1954). The family group names Arctiidae, Herminiidae and Erebidae were described in the same publication (Leach 1815) and thus have equal priority, and all three have priority over Catocalidae. However, both Arctiinae and Herminiinae are well-known taxa generally considered to be subfamily level (Holloway 2008) or family level (e.g. Kitching 1984) taxa, and thus we consider that it would generate less confusion to adopt the name Erebidae for the larger group of noctuoids that includes both Arctiinae and Herminiinae.

Within Erebidae, relationships of only a few lineages are well supported. The monophyly of Aganainae + Herminiinae + Arctiinae clade is strongly supported by our analyses. The association of Herminiinae (Renia) with arctiines was found in an earlier morphological study by Jacobson & Weller (2002). The clade also has a morphological synapomorphy in the prespiracular position of the counter-tympanal hood. Adults of many aganaines and arctiines are visually striking and aposematic, and aganaines and herminiines share long labial palps and a bare lower frons. Kitching (1984) was the first to exclude Aganainae and Herminiinae from Noctuidae based on the prespiracular hood, which was then thought to be plesiomorphic. In addition, the clade may be characterized by two further synapomorphies: modified foretibia in the males of most genera and a swollen metepimeron ventral to pocket IV (Kitching 1984). Holloway (2008) also discussed this grouping, commenting on biological differences between the groups, Herminiinae being generally cryptic, feeding on vegetable detritus, and Aganainae being aposematic, feeding on the same suite of toxic cardenolide-synthesizing plant families (Apocynaceae, Asclepiadaceae and Moraceae) as the danaine Nymphalidae and other moth genera such as Glyphodes (Crambidae) and Agathia (Geometridae). We have also found that Pangraptinae + Masca, a member of the Episparis group of genera of Holloway (2005), are sister to the Aganainae + Herminiinae + Arctiinae clade with good support. Pangraptinae were previously thought to be associated with Eublemminae (Fibiger & Lafontaine 2005), but see Holloway (2005, 2008, 2009) and below.

The boletobiine clade (Fig. 2) has several potential morphological synapomorphies. Many of the included taxa have the central part of the valve heavily sclerotized with several processes arranged transversely, with the cucullus reduced to a membranous flap. In addition, the female genitalia often have a ring of claw-like spines in the corpus bursae, the base of which is frequently narrow and coiled to some degree. The larvae have the first two pairs of prolegs reduced and are often warty or pubescent. They feed on a diversity of resources as mentioned above. Much of this clade is made up of a revised and enlarged concept of Aventiinae, together with Eublemminae (Holloway 2009), as well as many Araeopteroninae and Phytometrinae. We identify the clade as boletobiine from the oldest family group name amongst the included taxa (Boletobiinae). Phylogenetic relationships within the clade are in much need of study, and promise to reveal considerable insight into morphological and life history evolution in quadrifines.

The separation in the molecular results of the calpine clade from the scoliopterygine clade and the inclusion of Phyllodini in the former provides an object lesson on how the sharing of peculiar morphological adaptations may mislead in classification, and how shared features of a more subtle nature may be overlooked in an unchallenged traditional classification. A robust, highly developed fruit-piercing (and in some cases skin piercing and blood sucking) proboscis is found widely in Calpini and in some more robust scoliopterygines such as Anomis, with many similar features in the structure. But the molecular results point to this being a homoplasy.

Phyllodini, with a spined adult mid-tibia, were traditionally catocalines, and Calpini, with an unspined mid-tibia, were traditional ophiderines, indeed including the type genus thereof. The grouping of Phyllodini with Eudocima in the molecular analysis brings into greater prominence several features shared by these groups (Holloway 2005): leaf mimicry in the forewing and flash colouration in the hindwing of the adults; Menispermaceae as a favoured larval host family; a similar method of pupation in a leaf shelter.

The Euteliidae lineage is undoubtedly a well-defined group of noctuoids and is raised here to family level, thus supporting the findings of Mitchell et al. (2006). Kitching (1987) demonstrated that Euteliinae and Stictopterinae are related and form a monophyletic group, based on a large number of synapomorphies including: reduced female frenulum, modified basiconic sensilla on the proboscis, presence of a small oval plate in the ductus ejaculatorius, anal papillae modified so that their inner surfaces are directed posteriorly and the counter-tympanal hood has a unique double structure (Richards 1932; Holloway 1985; Kitching 1987; Kitching & Rawlins 1998). The host plant range of the two subfamilies embraces mostly lactiferous plants, with Euteliinae favouring Anacardiaceae and Stictopterinae having distinct lineages on Calophyllaceae/Clusiaceae, Dipterocarpaceae and Euphorbiaceae (Holloway 1985).

The relationships of the six clades remain somewhat ambiguous, although it is clear that Euteliidae, Nolidae, Erebidae and Noctuidae together form a monophyletic group. The position of Oenosandridae varies in our analyses, with the nuclear only and third codon position excluded analyses placing it as sister to the rest of Noctuoidea. This position has also been found in the two recent studies on Ditrysia (Regier et al. 2009; Mutanen et al. 2010). Our full, combined dataset places Notodontidae in this position, with Oenosandridae diverging next, but this is not well supported. It appears that all poorly supported, unstable relationships are characterized by short basal branches, especially within Erebidae (Fig. 2). Such patterns of short branches and low support values have been thought to indicate rapid radiations (Tajima 1983; Wiens et al. 2008; Kodandaramaiah et al. 2010).

Elucidating the evolutionary history of the massive Erebidae clade (potentially including 40 000 species) will require more intensive sampling. Further studies are also needed to identify the reasons for the short basal branches, i.e., whether there is a historical explanation behind them (e.g., rapid radiation), or whether it is simply an artefact of insufficient data. Noctuoidea represent a unique opportunity to investigate the reasons underlying massive diversification in phytophagous insects, and a first step in such investigations is the identification of monophyletic groups and their interrelationships.

The sampling strategy that we have adopted, described in the introduction, is being used to establish priorities for ongoing studies, to test further the robustness of the four major clades, both in relation to each other and internally. Work in progress on an increased sample of taxa, particularly in Erebidae, shows that the clades remain distinct and robust, and that ‘known unknowns’, when included, fall within these clades and tend to reinforce the well supported parts of their internal structure, rather than perturbing it (R. Zahiri et al. unpublished data).

In summary, we have shown that there are six strongly supported lineages in Noctuoidea that can be assigned family status. Four additional groups that we were unable to sample, Dilobinae, Cocytiinae, Eucocytiinae and Strepsimaninae, may prove to be further independent lineages. There are also a few genera of a similar nature that appear to be noctuoid but have yet to be assigned to a family, such as Kenguichardia, Plagerepne and Clytophylla (Holloway et al. 2001; Holloway 2003). Within the newly circumscribed families, Erebidae require much more attention; many traditional subfamilies and tribes (e.g. Calpinae and Chloephorinae of previous authors) were found to be polyphyletic, even with our limited sampling. The conclusion that the Euteliidae are not closely related to Erebidae was surprising and also needs further investigation.