The recurrent veinlet is known in the following extant taxa of Neuroptera: all species of Psychopsidae and Ithonidae (including Polystoechotidae and Rapismatidae), the vast majority of Hemerobiidae; rarely in Berothidae (including Rhachiberothinae) and Mantispidae (only Symphrasinae).

The recurrent veinlet has not been detected in any Permian Neuroptera (Permithonidae, Archeosmylidae) and the majority of Triassic taxa represented by these two families, and ‘Mesoberothidae’; however the basal portion of the forewing in any reported specimen of the latter is poorly preserved (e.g., Riek 1955; Lambkin 1988; Vilesov and Novokshonov 1994; Vilesov 1995; Novokshonov 1996; Jell 2004; VM, pers. obs.). In Triassic Neuroptera, the recurrent veinlet is only found in those taxa that have the costal space strongly dilated basally, i.e., the psychopsoids (Psychopsidae, Osmylopsychopidae, and broad-winged taxa whose family affinities are not yet clear). These are: (1) several undescribed Osmylopsychopidae from the Ladinian/Carnian Madygen Formation (Kyrgyzstan) (see Shcherbakov 2008: fig. 6); (2) Triassopsychops superbus Tillyard, 1922: fig. 89; (3) Archepsychops triassicus Tillyard, 1919: fig. 27; (4) possibly in Osmylopsychops spillerae Tillyard, 1923: fig. 93 (however see Lambkin 1992: fig. 2), and (5) Petropsychops superbus Riek, 1956: pl. 1, fig. 2 (see also Grimaldi and Engel 2005: fig. 9.15) (all from the Carnian Ipswich Series of Australia). Of these taxa, the recurrent veinlet is most clearly visible in one of the Osmylopsychopidae wings from the Madygen Formation, where it is strongly curved towards the wing base and has eight branches (VM, pers. obs.). In the Australian taxa, the recurrent veinlet is poorly preserved.

Of the Early Jurassic Neuroptera, the recurrent veinlet is detected in all psychopsoids whose forewing base is well preserved (e.g., Tillyard 1933: fig. 2; Ansorge 1996; fig. 52) and in Prohemerobiidae. The latter family, however, remains undefined; we treat it as containing only the type genus Prohemerobius Handlirsch, 1906 (probably paraphyletic), recorded from the Early Jurassic of Germany and England (Handlirsch 1906–1908, 1939; Bode 1953; Whalley 1988; Ponomarenko 1995) and the Late Jurassic of Mongolia (Khramov 2011). The majority of species in this genus have the basal portion of the forewing costal space missing or poorly preserved. Still, some species (including the type species Prohemerobius dilaroides Handlirsch, 1906) possess the recurrent veinlet, configured similarly to that of the Middle Jurassic Nymphidae, i.e., with few (or one) simple branches (J. Ansorge, VM, pers. obs.). The wings of Prohemerobius are not of the psychopsoid type; they are small, relatively narrow with hemerobiid-like venation, and the costal space is narrow compared with those of psychopsoids. All known Early to Middle Jurassic Parakseneuridae possess a very well developed recurrent veinlet, along with up to 15 dichotomously forked branches (Yang et al. 2012: figs 1, 25, 30). All Jurassic to Cretaceous Osmylopsychopidae and Psychopsidae have the recurrent veinlet (e.g., Peng et al. 2010: fig. 3A; VM, pers. obs.). Some species of the Middle Jurassic to Early Cretaceous family Kalligrammatidae possess the well-developed recurred veinlet (e.g., Yang et al. 2009: fig. 3, Yang et al. 2011: figs 2, 3A, Ponomarenko 2002: fig. 254); however, this condition is obviously absent in other species, even from the Middle Jurassic (VM, QY, pers. obs.).

The costal space in all numerous undescribed Ithonidae from the Middle Jurassic locality of Daohugou is basally narrowed, and therefore, the recurrent veinlet (where well developed) has short simple branches. Sometimes, the humeral veinlet is not branched and crossvein-like (VM, pers. obs.). Younger Ithonidae (Early Cretaceous to Recent) have the costal space usually broader basally and have a well developed recurrent veinlet (e.g., Archibald and Makarkin 2006: figs 5B, 8D, 9A; Makarkin and Archibald 2009: fig. 3; Makarkin et al. 2012: figs 3D, E). All species of the Middle Jurassic to Early Cretaceous subfamily Mesomantispinae possess a rather well developed recurred veinlet (Makarkin 1996: fig. 1; Jepson et al. 2013: figs 1C, 2D, 4B). A similar recurred veinlet is present in the extant subfamily Symphrasinae, the single fossil species of which is known from the middle Eocene of Messel, Germany (Wedmann and Makarkin 2007). Other Mantispidae (both fossil and extant) do not possess the recurred veinlet. All fossil Hemerobiidae, including the oldest known specimens from the Late Jurassic and Early Cretaceous have well-developed recurred veinlet (e.g., Henriksen 1922: fig. 4; Panfilov 1980: fig. 91; Makarkin 1991: fig. 2a; Ponomarenko 1992b: fig. 4). Of the Berothidae, the well-developed recurrent veinlet is detected in the Late Jurassic to Early Cretaceous Mesithoninae (Panfilov 1980: figs 86, 87, 90; Ren and Guo 1996: fig. 6; Makarkin 1999: figs 2a, b, 4), and few other genera (e.g., Plesiorobius Klimaszewski & Kevan, 1986: fig. 2). In other berothids, the recurrent veinlet is poorly developed, only slightly recurrent and bearing few branches, or does not develop at all (e.g., Engel and Grimaldi 2008: figs 19, 21, 24, 25).

In the Mesozoic family Brongniartiellidae, the well-developed recurred veinlet with long branches is detected in two species from the Early Cretaceous of the Baissa locality, Transbaikalia (Makarkin 2010: fig. 3). The taxonomic affinities of several Mesozoic genera which bear a well developed recurrent veinlet are uncertain, e.g., Sibithone Ponomarenko, 1984: fig. 5a (Early / Middle Jurassic of Novospasskoe, Siberia); Osmylogramma Ponomarenko, 1992b: fig. 5 (Early Cretaceous of Tsagan-Tsab, Mongolia); Meilingius Ren et al., 2002: fig. 2 and Jurapolystoechotes Ren et al., 2002 (both from Daohugou; pers. obs.).

The presence of the recurrent veinlet was long considered a plesiomorphic condition in Neuroptera (e.g., MacLeod 1970; Oswald 1993; Aspöck and Nemeschkal 1998) because the most ancient, ‘primitive’ extant families possess this condition (e.g., Ithonidae, Psychopsidae), and but it is absent in the youngest, more derived families (e.g., Myrmeleontidae, Ascalaphidae). More recently published phylogenetic analyses of the group place these families in more derived positions and thus the presence of the recurrent veinlet is now considered an apomorphic state evolving independently at least three times (Winterton et al. 2010; Winterton and Makarkin 2010; Yang et al. 2012). The origin of the recurrent veinlet throughout Neuroptera most likely occurred during in the late Permian to earliest Triassic (Figure 7). This condition was secondarily lost at least five times in those families of the clade ‘R’ that lack this condition; and was never possessed in families basal to this clade.

Figure 7.

The occurrence of the recurrent veinlet through Neuroptera families. The phylogeny of the order is based on that of Yang et al. (2012), but modified according to the most probable relationships of families as discussed by Yang et al. (2012, 2013) and Makarkin et al. (2013).The families that we propose originally lacked the recurrent veinlet are represented in orange; R, families possessing the recurrent veinlet in blue, and those that we propose lost it secondarily in brown. I the ithonoid clade; C Chrysopoidea (the chrysopoid clade); M Myrmeleontoidea (the myrmeleontoid clade); P Psychopsoidea (the psychopsoid clade).

The two species described above from the Middle Jurassic locality of Daohugou are the oldest known record of Nymphidae, indeed, of Myrmeleontoidea. Myrmeleontoidea are believed to belong to the suborder Myrmeleontiformia, which also includes Psychopsidae (Aspöck et al. 2001) or all families of Psychopsoidea (Engel and Grimaldi 2008). According to our phylogeny (Yang et al. 2012: fig. 32), however, the clade comprising Myrmeleontoidea and Chrysopoidea is sister to the ithonoid clade, and that combined clade is sister to the psychopsoid clade. Known psychopsoid fossils are much older than those of Myrmeleontoidea, and the recurrent veinlet first appears in their Triassic representatives (see above). Therefore, the presence of the recurrent veinlet in the Middle Jurassic Nymphidae may be interpreted as a remnant of ancient condition occurring in the psychopsoid + ithonoid + chrysopoid + myrmeleontoid clade before separation of the clade Myrmeleontoidea + Chrysopoidea. This character condition was lost in other families of Myrmeleontoidea and Chrysopoidea, and in the younger Nymphidae.

The recurrent veinlet is known mainly in species whose costal space is basally dilated. The strong dilation of the basal portion of the costal space is characteristic of Neuroptera. It never occurs in other Holometabola including other Neuropterida orders, although in some taxa, the costal space is markedly dilated at some distance from wing base (e.g., Raphidioptera: Aspöck 1986; the mecopteran family Dinopanorpidae: Archibald 2005: figs 3A–D). The costal space of earlier Neuroptera (basal to the clade ‘R’) is also narrowed basally. The appearance and development of the recurrent veinlets and the strong dilation of the basal costal space are probably functionally and structurally tied. However, in some Mesozoic Neuroptera which possess the recurrent veinlet the costal space is narrowed basally: the two species of the Middle Jurassic Nymphidae described here, and at least some Early Jurassic Prohemerobiidae and the Middle Jurassic Ithonidae. It seems reasonable that these taxa probably retained the recurrent veinlet after the costal space was secondarily narrowed in those lineages.

The ScA in all Neuroptera (when present) has a similar configuration: a short vein structure terminating on ScP before the recurrent veinlet. The structure of ScA of Daonymphes bisulca gen. et sp. n. is very similar to that found in extant Nymphidae, especially Nymphes Leach, 1814 (VM, pers. obs.) and closely related genera (e.g., Austronymphes sp.: see Riek 1967: fig. 3D). A similar ScA occurs in most extant Hemerobiidae (e.g., Makarkin 1993: fig. 13). Other extant Neuroptera have no distinct ScA, but a convex sclerotized bulge presented in the humeral area of most other families is considered as the modified ScA (Kukalová-Peck and Lawrence 2004). Of the fossil taxa, the ScA is present in the Mesozoic Kalligrammatidae (e.g., Yang et al. 2011: fig. 2) and Parakseneuridae (Yang et al. 2012: figs 24, 25, 30; QY, VM, pers. obs.). According to the hypothesis of Kukalová-Peck (1983), the ScA of the pterygote wing venation ground plan is a relatively short vein running to the costal margin with a strong subcostal brace connecting ScA and ScP midway (Kukalová-Peck 1983, fig. 15), or by other interpretation the ScA is divided midway into the anterior branch (ScA1+2) running to the costal margin and the posterior branch (ScA3+4) terminating on ScP (Kukalová-Peck 1991: fig. 6.3C). Such a primitive ScA is believed to present in the gigantic Carboniferous Bojophlebiidae (Riek and Kukalová-Peck 1984: fig. 20; Kukalová-Peck 1991: fig. 6.14A), although it appears to be poorly preserved (Prokop et al. 2010). In general, the configuration of ScA in some Paleozoic Palaeoptera is similar to that of Neuroptera in that ScA is terminating on ScP, i.e., in Dictyoneuridae (Palaeodictyoptera; see Kukalová 1970: figs 55, 59, 60, 73) and some families of Ephemeroptera (e.g., Carpenter 1979: figs 1, 8, 11); however, the ScA is branched in the latter order.

The ground-plan Neoptera wing venation is hypothesized to have ScA consisting of two separate veins with no common stem, ScA1+2 and ScA3+4; the former runs to the costal margin, the latter is terminated on ScP (Haas and Kukalová-Peck 2001: fig. 1; Kukalová-Peck 2009: fig. 14). Therefore, the ScA of Neuroptera may be interpreted as the homologue of this hypothesized ScA3+4, i.e., as the ground-plan neopteran ScA lost its anterior branch ScA1+2. ScA in other Neuropterida orders (Megaloptera, Raphidioptera) is configured similarly to most other winged insects, i.e., a vein terminating on the costal margin (Kukalová-Peck 1983: fig. 17A; 1991: fig. 6.16; Liu et al. 2013). Such a configuration of ScA is most developed in Orthoptera (e.g., Béthoux and Nel 2001: fig. 1). This state of ScA may be interpreted as the homologue of the hypothesized ScA1+2, i.e., as the ground-plan Neoptera ScA lost its posterior branch ScA3+4.

The configuration of ScA characteristic of Neuroptera (i.e., short and terminating on ScP) does not occur in other orders of Neoptera, and therefore may be considered as an autapomorphy. On the other hand, the similarities of ScA in Neuroptera and some Paleozoic Palaeoptera (see above) may indicate their homology, and therefore a symplesiomorphy. The latter, however, appears to be less probable than the former.