Although a small number of Aprosthema larvae were collected from additional sites nearby, there were four main sites (Fig. 1) in Dracénie Provence Verdon (Var, France) from which this study stems. Some are rather diffuse but each is believed to support an essentially self-contained population, more or less isolated (or at least discontinuous) from those in the other areas.

Site 1. Taradeau (200 masl; N 43.2931, E 6.2555). A rather small area comprising a linear fire-break of slightly disturbed limestone grassland and scrub (half of the approximately 25 m wide strip being mown along its 350 m length in alternate winters), within a moderately open mixed dry forest.

Site 2. Callas, Colle Blanche (400 masl; N 43.3519, E 6.3328). Along a forest road (900 m × 10 m) through Pinus and Quercus with intermittent open spaces. A nearby similar road on the same slope of Colle Blanche at Clavier is included. Also, La Garidelle, a nearby open mixed forest with fields and forest roads; although there were many host plants, only one larva was found in 2016 and none in 2017 and no further effort was made.

Site 3. Bargemon, Les Estangs (420 masl; N 43.3721, E 6.3435). A large (5 hectares) and very diverse wild garden with ponds and hedgerows, in 300 × 50 m of which Lathyrus grows. The forest road leading to it (Le Plan) with many roadside Lathyrus plants is included (a strip of approximately 50 × 5 m).

Site 4. Bargemon, Favas (623 masl; N 43.3714, E 6.3149). An abandoned clearing, measuring about 200 m × 50 m, along a road, with numerous shrubs surrounded by Quercus and Pinus.

Other sites involved in the study included: (i) Le Muy. A wooded area along a road between vineyards where many Lathyrus plants grow. A number of larvae were collected but were unable to form their cocoons, possibly owing to the use of insecticides, so parasitism was not ascertained. The site was visited only once, in 2018, and is not considered further. (ii) Apricale, Italy (302 masl; N 43.5342, E 7.3906), which is a wild garden in an olive grove with many Lathyrus plants, surrounded by a mixed forest of Quercus, Castanea and Pinus. Although this place is almost 100 km from the locations in Var, the similar findings from small collections made there are also presented.

Although L. latifolius is certainly the dominant Lathyrus in the region, on which the vast majority of the Aprosthema collected had undoubtedly fed, the scarcer but closely similar L. sylvestris has also been recorded locally and we cannot discount the possibility that larvae may sometimes have been collected from that. The two plants are not easy to tell apart, and no effort was made to do so.

All field-work (2014–2020) was conducted by PK and BKvLS. In the early years it was focussed on ascertaining the identity of the cocoon from which Thibetoides aprosthemae was reared in 2014 (Shaw et al. 2018), and larvae of Aprosthema tardum were first found in 2016. The first parasitoids were not reared from hosts known at the time of collection to be A. tardum until 2017. From 2018 to 2020, the four main locations were visited regularly every week (1 to 3 times for between one and four hours), from the middle of March to the first week of August. Adults of A. tardum emerge from around the middle of March and larvae appear in the field from mid-April continuously until about mid-July. After then larvae were no longer found in the field, although at a much-reduced level some cultured larvae continued developing successfully until mid-August, then producing only over-wintering cocoons.

No formal sampling protocols were employed; rather, collections were opportunistic and approaches adjusted when new observations and questions arose. For much of the period covered by this paper the host was not very numerous and many tens of hours were spent searching in each year. Binoculars (Pentax 6.5 × 21 Papilio II) were often used to search plants for characteristic signs of feeding in order to protect the sites from trampling. Although cocoons were always collected when found, not all encountered larvae of Aprosthema were collected immediately: the locations of some of the smaller ones were marked with a view to collecting the larvae at a later stage, with varying levels of success. Some of the host larvae we collected were reared in a straightforward way, but others were used for experiments with parasitoids. Most of the adult A. tardum reared were returned to their site of origin.

The total numbers of host larvae and cocoons found at the four principal sites, and at Apricale, are presented in Table 1. Effort was broadly similar for Site 1 for all years and at other principal sites in the three years 2018–2020, so those totals at least roughly reflect the considerable fluctuations in the host populations over the period.

Host larvae in their fourth and fifth instars (i.e. within 6 days of forming a cocoon) were reared individually in 9 cm diameter × 14 cm tall glass containers closed with a screwed lid, with freshly cut sections of Lathyrus latifolius (which stays fresh under these conditions for about seven days). Details of their progress were recorded daily. Four days after becoming cocooned, they were individually removed to a 1 × 9.5 cm glass tube closed with cotton wool.

Second and third instar larvae were treated similarly except that the Lathyrus was stood in a buried pot of water with a cylinder inverted over it. Cocoons were kept under conditions of natural light, and as far as possible kept outdoors and shaded.

While some observations were made (and even filmed) in the field, much of the behavioural information on the parasitoid species, and most of the filming, was obtained indoors in front of a large closed window to facilitate recovery of the parasitoid adult. Host larvae were always offered in their feeding positions on foodplant. The adult female parasitoids involved were either reared or collected in the field; in either case they were fed on dilute honey and also given flowers.

Except for a small number of specimens of L. erythrocephalus deliberately selected for dissection (see below) all adult parasitoids were killed by placement in a domestic freezer (ca -20 °C) for about an hour, then transfer to 96% ethanol for shipment to Edinburgh.

Dissections (in water) of both adult female parasitoids and Aprosthema larvae to recover parasitoid eggs or 1^st instar larvae were performed on material preserved in 70% ethanol in France then sent to Edinburgh. Minor distortions (and changes in appearance of egg content, in particular) inevitably will have taken place.

In order to provide a basis for future phylogenetic analysis of the reared parasitoids, some of which are very rarely encountered, we sequenced the D2 and partial D3 segments of 28S rRNA (28S) and the “barcoding portion” of cytochrome oxidase subunit 1 (CO1). DNA extraction and sequencing followed standard protocols, as detailed in Johannson and Klopfstein (2020). The sequences are available in Genbank under the accession numbers recorded under each of the determined parasitoid species here. The relevant specimens are deposited in the National Museums of Scotland (NMS) and labelled accordingly.

Methods of preparation are those of Wahl (1989). The measurements of the labial sclerite and mandible use the landmarks in Finlayson (1975: figs 2 and 4). DBW’s notation for larval preparations follows the museum acronym: it consists of his initials, the day, month, year, and a letter designating the individual preparation. The terminology for the cephalic sclerites of the mature larva is that of Wahl (1990) and Sime and Wahl (1998). A new terminology for spiracles is included in the Discussion.

It should be noted that almost all drawings of ichneumonid cephalic sclerites will involve elements of reconstruction, due to vagaries of the mounting process which can result in tears, skewing of sclerite positions, and structural distortions. The method employed by DBW, here and elsewhere, is to use a drawing tube to make accurate outlines, and then flip and trace structures so as to produce a bilaterally symmetrical result, which aims to be a faithful rendition of the original. For this study, setae have been placed to accurately reproduce their position and number.

In a series of papers culminating in Short (1978), J.R.T. Short described and drew the final instar cephalic sclerites of a wide range of Ichneumonidae. This was pioneering work, but unfortunately Short’s drawings are often at odds with the original slide mounts in terms of proportions and details. His drawing methodology is unknown, although he may have used an ocular grid. Reinterpretations of some of his slides are presented.

The associated videos were filmed using a Canon XL2 with a 20× zoom, XL 5.4–108 mm lens, supplemented when appropriate with the addition of a Canon 72 mm close-up 500D lens. For macro a Canon EF 100 mm 1:2.8 with an EF Canon XL adaptor was used. The footage is recorded on mini DV tapes of 60 minutes. Photos of living insects were taken in the field with a Lumix HD Panasonic DMC-TZ10 or abstracted from the filmed sequences. Photos of eggs and early instar larvae obtained by dissection were taken as single shots down one arm of a Wild M5A binocular microscope with ×20 eyepieces using a Canon PowerShot S110. Photos of mounted adults are stacked. DBW’s images of cocoons and whole larvae were taken with an EntoVision micro-imaging system, consisting of a Leica M16 zoom lens attached to a JVC KY-75U 3-CCD digital video camera that feeds image data to a desktop computer. The program Archimed 5.3.1 was used to merge an image series (representing typically 15–30 focal planes) into a single in-focus image. Lighting was provided by an EntoVision dome light. Photographs of larval slides were taken by a Nikon D810 body attached to a Nikon Labophot compound microscope with a trinocular head; photograph series were assembled into a single image using Helicon Focus 7.6.4 Lite.

Aprosthema tardum was initially determined by Andrew D. Liston. Material is deposited in NMS and Senckenberg Deutsches Entomologisches Institut (SDEI). All parasitoids were determined by MRS and are mostly deposited in NMS, with the following exceptions: One female of Terozoa quadridens has been donated to each of Canadian National Collection of Insects (CNC), Zoologische Staatssammlung München (ZSM) (especially to enable both sexes of T. quadridens and T. anatolica to be compared within the ZSM collection), and Natural History Museum London (NHMUK). One female of Ischyrocnemis goesi has been donated to NHMUK. Slides of larval preparations were examined from National Museum of Natural History, Smithsonian Institution (NMNH), Utah State University Collection (EMUS) and Australian National Insect Collection, CSIRO (ANIC).

We use the term “plurivoltine” to indicate that a species has more than one generation per annum, i.e. that it does not have a fully obligatory diapause but rather can develop successive generations in a season. Often it may in practice be largely bivoltine, but the term plurivoltine allows for the possibility that under favourable circumstances there may be more than two annual generations to a significant extent.

Liston et al. (2018) described and illustrated the biology of Aprosthema tardum; minor additions are given in the following summary. There are five larval instars and, in line with other investigated Argidae (Kontuniemi 1965), there is no post-feeding instar for cocoon construction. The winter is passed as a cocooned prepupa. At relatively low altitude and in years with an early spring, up to four partly overlapping generations per annum can occur (Fig. 2), with first generation adults (i.e. emerging from overwintering cocoons) first appearing in early March. No adults of A. tardum were seen visiting flowers. From observations on released adults, which initially fly (in ascending circles in the case of males) to the canopy, it is presumed that courtship and mating take place there: this assumption was reinforced by the observation of a dead male in a spider’s web at 3 m height. Females were regularly observed among the foodplant, but on only two occasions did we see a male, in each case briefly approaching an ovipositing female but without success. We were unable to induce mating in captivity. If constrained, captive virgin females commence oviposition after about 24 hours, and continue to do so without deviation if released. The egg is placed between the upper and lower epidermis of leaves of Lathyrus latifolius (Fig. 3); generally two eggs are deposited per leaf, but sometimes more. Depending on weather, the early spring eggs develop in around 14 days but, as the season progresses, this can be in as little as five to six days. One or two days before the egg hatches, the area in the leaf edge becomes conspicuous owing to the development of the relatively large first instar larval head (Fig. 4).

The first instar larva (Fig. 5), which immediately starts feeding at the leaf margin, is about 3 mm in length. By its final instar the larva feeds across the width of the leaf, leaving a characteristic stepped profile. The larvae move little: sometimes the entire larval development takes place from a single leaf. After about 10–18 days and four moults the fully fed final instar larva (Fig. 6) measures about 11–14 mm (males) to 14–17 mm (females).

An important feature is the clear difference in cocoon structure and positioning of overwintering cocoons (“winter cocoons”) from those from which adult sawflies will emerge in the year of formation (“summer cocoons”). Summer cocoons (Fig. 7) are constructed among more or less aerial vegetation, though often somewhat concealed from view. They have a coarse open weave and are rather springy, and pale yellowish in colour. The winter cocoon (Fig. 8, showing a cocoon constrained to be spun exposed) is made in or just below the litter layer. While still having a more or less open weave it is denser and generally smaller (about 10–13 mm as opposed to 12–15 mm), almost hard owing to its much thicker (almost wiry) strands, and usually darker (buff or light brownish). Both morphs have a thin more densely spun inner envelope, that of the summer cocoon being still somewhat open, but more parchment-like in the winter version. In practice it is immediately obvious from its construction whether or not a cocoon is destined to overwinter. As expressed in Fig. 2, an increasing proportion of the cocoons made in each generation comprises winter cocoons, thereby entering diapause in summer. Emergence from summer cocoons takes 7–12 days. A total of 33–36 days development from oviposition to adult for the non-diapausing fraction of the second generation (i.e. first complete cycle of the season) is typical, but summer development in the following generation may take less than 30 days.

Differences in the head morphology of adults of winter and summer generations are rather pronounced, particularly in females, and similar to the dimorphism described by Vikberg (2004) in Aprosthema melanurum (Klug) (Liston et al. 2018). As noted by these authors this is a reflection of the relatively tougher winter cocoon, necessitating stronger musculature for the mandibles, which are larger in the winter generation although there is no major difference in their structure between generations. The colour of the adult (Fig. 9) is fairly constant.

Some way into this research we discovered that a second species of Aprosthema, probably undescribed (Andrew Liston, pers. comm.), also feeds on Lathyrus in the area, but it is evidently present in very much smaller numbers. We observed (but could not collect) one female in the act of oviposition at Site 2 in 2018, and reared another female (Fig. 10) from a summer cocoon collected at Site 4 in 2019, in contrast to the roughly 100 adults of A. tardum reared from larvae and cocoons collected in the field. No differences were noted among the Aprosthema larvae collected, but it seems likely that any differences between the two species would be slight and easily overlooked: however, it should be noted that we have not seen a larva known to belong to the second species (unfortunately the tagged plant on which oviposition was observed had been destroyed by wild boar by the time of our next visit). For the purpose of this paper we are treating all the parasitoids reared as using A. tardum, and additionally presuming that the second species is also parasitized equally; i.e. quantitative statements do not distinguish between the two. Although any rearing from wild-collected larvae or cocoons unfortunately cannot be fixed to either of the possible hosts with certainty on morphological grounds, some experimental rearings of parasitoids using male A. tardum larvae resulting from known and certainly determined virgin females have provided unequivocal host associations of L. erythrocephalus, I. goesi and T. quadridens with A. tardum. Adult females of the two Aprosthema species differ conspicuously in colour (Figs 9, 10) and also in a range of morphological features, but the (as yet undecided) identity of the second species is under investigation by A. D. Liston and is not further considered here.

No parasitoid taxa ovipositing into the cocoon (as opposed to the larva) were found (36 summer cocoons collected overall) and, rather surprisingly, no evidence of parasitism by Tachinidae (Diptera) was seen. Four species of Ichneumonidae that could be identified to species were found, all of which attack the larval stage, are strictly solitary, and emerge as adults from the host cocoon. A fifth species (henceforth Ctenopelmatinae sp. X) was detected only from a single cocooned prepupa resulting from a host larva collected at Site 2.

The sites in which the parasitoids are known to have been active are recorded in Table 2. Given that the only years of intensive sampling were 2018–2020, it is clear that the most abundant parasitoid, Lathrolestes erythrocephalus, was universally and consistently present (also occurring at Apricale). The next most regularly recorded (although comparatively rare) parasitoid, Terozoa quadridens, occurred at all four of the main sites and was found in all of them in most years. The other two regular parasitoids were found only in smaller numbers, but Thibetoides aprosthemae was present at three sites (and additionally at Apricale) and Ischyrocnemis goesi at two. Discounting the unidentified Ctenopelmatinae sp. X, only Site 3 was found to support all four of the other parasitoid species, though (sometimes different) combinations of three species were present at the other main sites. The taxonomy, biology and phenology of each parasitoid species is outlined in separate sections below.

Lathrolestes erythrocephalus (Ctenopelmatinae, Perilissini)

Figs 11–28; FilmingVarWild videos 2, 3, 10–13

Taxonomy. Lathrolestes is a medium sized genus with about 25 European species, remarkable among Ctenopelmatinae for including parasitoids of leaf-mining Lepidoptera and (outside Europe) Coleoptera (Barron 1994) in addition to the otherwise apparently universal parasitism of Symphyta by the subfamily (Broad et al. 2018). There is no controversy concerning the identity of L. erythrocephalus, which is a distinctive species (Figs 11, 12). In contrast, its generic placement is less certain. Much of the literature, including most databases, has placed it in Perilissus Holmgren, 1857, but Aubert (2000) transferred it to Lathrolestes Förster, 1869 and the most recent taxonomic revision (Reshchikov 2015) of Lathrolestes included erythrocephalus. On the grounds that this is the latest generic placement in a formal taxonomic work, we follow that here, despite the more numerous previous references to it as a species of Perilissus (in which genus it might arguably have best been left pending a more thorough and wider review (Gavin Broad, pers. comm.)). The phylogeny of Ctenopelmatinae is beyond the scope of this paper but, because Perilissus as currently construed is almost certainly polyphyletic, arguing for a simple return of erythrocephalus to that genus would not constitute a real advance. It is the type species of the genus Polyoncus Förster, 1869, currently regarded as a synonym but which might one day be resurrected. Lathrolestes erythrocephalus (as Perilissus) has previously been recorded from France, and Reshchikov (2015) cited a rearing record from an undetermined Aprosthema species in Kazakhstan.

Genbank accession numbers of L. erythrocephalus from our study (SK_19_50): CO1 OK393909; 28S OK393942.

Biology. This is by far the dominant species in the parasitoid complex of A. tardum, consistently present at all well-sampled sites including Apricale (Table 2) and having several, probably overlapping, generations through the season. Over 40 have been reared from wild-collected hosts (plus many more in culture). The earliest date of emergence from a winter cocoon was 7 March. Adults were regularly encountered in the field, searching the Lathyrus plants from late March to the end of July, especially in somewhat overcast weather. Its distinctive dark eggs were seen in many host larvae in the field, from their second instar onwards, often more than one in a host (Fig. 13). Although the hatched eggshells remain visible, the growing host becomes progressively more opaque and this external sign of parasitism is less evident in the late instar hosts. In experiments L. erythrocephalus freely oviposited into 2^nd to 5^th instar larvae, with successful outcome from each instar. As also observed in the field, often more than one egg was laid in a host during essentially the same visit to a well-grown larva, which seemed deliberate, and in cases of separated visits or successive visits by different females we saw no avoidance of either superparasitism or multiparasitism. After brief antennal exploration that seems to be of the recent feeding damage as much as of the host itself, oviposition takes from 7 to 28 seconds; the host, especially when large, sometimes reacts by raising its abdomen and thrashing, but smaller hosts are more often unreactive. The host is not temporarily paralysed and, during oviposition, the female parasitoid is in contact with it only by its ovipositor. On one occasion a female L. erythrocephalus was seen to bite and eat most of an early instar Aprosthema larva, this destructive event being the only host-feeding seen.

Dissections showed that freshly emerged females (N = 2) have no mature eggs in the common oviduct, but after 4 days with access to dilute honey there are usually about 20, with more to come (N = 4). The initially white egg, ca 0.75 × 0.18 mm (Fig. 14), darkens within two or three hours of being laid (Fig. 15), eventually to become practically black. It bears a remarkable hooked terminal protrusion, probably in fact a set of protrusions (Figs 15–18), unequivocally deduced to be at its head end from four lines of evidence: (i) the position of these hooks at the broader end of the egg (Figs 14, 15); (ii) the consistent appearance of hatched eggshells recovered by dissection (Figs 17, 18); (iii) dissection of a well-developed but dead ruptured egg found in a host, probably killed by a competing first instar larva, in which the head capsule was discerned at the end with the hooks, and the caudal appendage was seen to be curled under (or possibly over) the posterior segments at the other pole; and (iv) the orientation of the egg as it leaves the ovary such that the narrow (unadorned) caudal end passes into the oviduct first in accordance with Hallez’s Law (Hallez 1886), confirmed by dissection of a 4 day-old female with mature eggs present in the common oviduct. Once in the host’s haemocoel the toothed protrusions seem to act as grappling hooks to weakly fasten the egg to a tissue such as the gut or muscle following free (random) placement. The egg hatches on the third day and the caudate first instar larva (Fig. 19) leaves the chorion completely, and measures ca 1.2 mm. It has large sharp mandibles (Fig. 20) and weak (easily overlooked) ventrolateral tooth-like projections particularly on the more posterior abdominal segments (Fig. 21). It can often be seen through the host’s integument as it moves freely within the host’s body and it has sometimes been observed apparently re-visiting its own eggshell, perhaps as part of a routine inspection to eliminate competitors using its sharp mandibles. The larva stays in its first instar until the host is cocooned. After then completing its feeding rapidly, it makes its orange-brown weakly banded cocoon (Fig. 22) inside that of the host.

Generally the adult parasitoid emerges during the summer from summer cocoons of the host, after about 26 days, and often it will overwinter in the host’s winter cocoon, but this phenological relationship is evidently not under endogenous control as the parasitoid has often emerged in the same summer from an intended winter host cocoon, or sometimes during the winter when hosts would be unavailable. Several times it has remained in its cocoon within a winter host cocoon through the following summer to emerge later (this happened in several cases, emergence eventually being prompted following an airline flight to Edinburgh and a period of darkness then subsequent indoor warmth in the early part of the next winter): presumably, left to themselves, in both the foregoing cohorts these adults would usually have emerged in spring, having skipped a year, although in two cases we have witnessed emergence in late September and early October, at a time when no hosts would be present, from undisturbed cocoons made in June and July of the previous year. It is difficult to ascribe this entirely to captive conditions. Sometimes the host successfully encapsulated a L. erythrocephalus egg (Fig. 23) and developed to the adult stage, which could be construed as evidence that the very frequently observed self-superparasitism may be adaptive in significantly improving the likelihood of success (see Discussion). We observed a single copulation in captivity, with the stance (Fig. 24) permitting antennal contact typical of many Ichneumonidae (cf. Shaw et al. 2021). Morphological features of mandibles, briefly discussed later, and ovipositor are given in Figs 25 and 26 respectively.

In several cases of observed multiparasitism, the faster-developing L. erythrocephalus generally triumphed over both Thibetoides aprosthemae and Terozoa quadridens; though sometimes the host was unable to moult successfully when parasitised by both T. aprosthemae and L. erythrocephalus, with the result that host and parasitoids all perished. Adults of L. erythrocephalus successfully developed on several occasions following oviposition by I. goesi into A. tardum larvae in which L. erythrocephalus had already been present, but the reverse situation was not investigated.

Final instar cephalic sclerites (Figs 27, 28). Specimens examined: DBW 14.vi.2020d, 27.vi.2020a, 27.vi.2020b, 27.vi.2020c, 27.vi.2020d, 13.vii.2020b, 13.vii.2020c, 13.vii.2020d (all NMS).

Cephalic structures generally moderately to strongly sclerotized. Epistoma lightly sclerotized; epistomal band present; dorsal margins of both poorly defined and merging with general light sclerotization of frons. Labral sclerite absent; clypeolabral area with lightly sclerotized central region, lenticulular in shape, with moderately sclerotized dorsal margin, bearing setae and two clypeolabral plates next to its ventral margin. Stipital sclerite present and strongly sclerotized, more or less horizontal; cardo absent. Pleurostoma lightly to moderately sclerotized; posterior struts of inferior mandibular processes not connected by band; accessory pleurostomal area weakly present; lateral margins of pleurostoma and pleurostomal area weak and fading into general cephalic area. Hypostoma long and strongly sclerotized, lateral end not divided in two at posterior tentorial pit and without extensions; accessory hypostomal area present on dorsal margin and lightly sclerotized. Hypostomal spur present, about 2.5–2.9 × as long as its basal width. Labial sclerite weakly ovoid, lightly to strongly sclerotized. Salivary orifice U-shaped. Prelabial sclerite present as weakly sclerotized transverse band, connected to interior ventral margin of labial sclerite by weakly sclerotized projection of labial sclerite. Labial sclerite with 6 setae. Prelabial area with 4 setae. Maxillary and labial palpi each bearing two sensilla. Mandible strongly sclerotized; blade about 0.6 × as long as full mandibular length, without fine denticles. Antenna without papillus. Parietal band present as lightly to moderately sclerotized vertical oblong with irregular margins. Spiracle with closing apparatus absent; intercalary trachea absent, subatrium about as long as atrium. Skin covered with small, bubble-like protuberances; spines absent; setae present, short and scattered.

As noted below under Ctenopelmatine species X, Short’s (1978) key to tribes of ctenopelmatine larvae is woefully inadequate. But, as discussed in the Ischyrocnemis goesi section below, there are a number of ctenopelmatine genera that have a clypeolabral structure that superficially resembles a labral sclerite: Lathrolestes, Opheltes, and Perilissus in the Perilissini, Euryproctus in the Euryproctini, and Protarchus in the Mesoleiini. Short (1978) illustrated Lathrolestes luteolator (Gravenhorst) as having this structure, and it is interesting that L. erythrocephalus has it as well. Unfortunately, Short’s characterization of Lathrolestes is at odds with what is presented here for L. erythrocephalus: he states the labial sclerite is ventrally incomplete and the epistomal band absent (‘Median dorsal part of the epistoma not lightly sclerotized’). Perhaps his interpretation was due to a poor preparation: examination of Short’s slides reveals many of the heads to be essentially dismembered or twisted about, resulting in partly imaginative illustration.

Perhaps the most surprising finding was the lack of a spiracular closing apparatus. The other two ctenopelmatines in this study lack it as well. Short (1978) illustrates 57 ctenopelmatine species, with the spiracle shown in 55 of them; of these 55, only six were depicted as lacking the closing apparatus. DBW has been able to examine 16 of Short’s original Ctenopelmatinae slides that were deposited in NMNH: Euryproctini (Hypsantyx lituratoria (Linnaeus), Phobetes striatus (Davis), Synodites olympiae (Ashmead), Synomelix sp.); Mesoleiini (Lamachus albopictus Cushman, L. contortionis Davis, L. eques (Hartig), L. lophryii (Ashmead), L. ruficoxalis Cushman, L. tsugae Cushman, L. virginianus (Rohwer), Mesoleius tarsalis (Cresson), M. tenthredinis Morley); Perilissini (Lophyroplectrus nipponensis Cushman, Opheltes glaucopterus flavipennis (Provancher)); Pionini (Rhorus clapini (Provancher)), for all of which Short portrayed a closing apparatus; but re-examination showed that only two (Synomelix sp. and Rhorus clapini) actually had one. Finlayson (1960a, 1960b, 1963) illustrated the spiracles of eight species of Mesoleius and Lamachus and all also lack a closing apparatus. It would appear that this condition is quite widespread among ctenopelmatines. A number of other subfamilies of higher Ophioniformes appear to lack the closing apparatus as well, although in a different way: Ophioninae, Campopleginae, and Cremastinae (DBW, pers. obs.). Elucidation of exactly what is going on in these endoparasitoids will require further study (see also Discussion).

Ischyrocnemis goesi (Ctenopelmatinae)

Figs 29–41; FilmingVarWild videos 4, 12

Taxonomy. (See also section on Terozoa, below). Ischyrocnemis has been variously treated as a member of Ctenopelmatinae (Perkins 1962; Townes and Townes 1959) or Metopiinae (Townes 1971; albeit doubtfully), and both Perkins (1962) and Aubert (2000) suggested it might best be placed in the ctenopelmatine tribe Pionini by envisioning a relationship with Rhorus Förster, 1869. Nevertheless, its placement, even to subfamily, has remained problematical and has not been resolved by molecular and morphological phylogenetics (Quicke et al. 2009), although the biological evidence presented here comes down firmly in favour of Ctenopelmatinae. Whether the uncertainty has been compounded by the confusion between Ischyrocnemis (Ctenopelmatinae) and Terozoa (Tryphoninae) is unclear but possible. The two were regarded as synonyms since Townes (1971) [despite Perkins 1962] until Kasparyan (2019a) resolved the matter. At species level our determination of I. goesi is based on the key to the five Eurasian species given by Kasparyan (2019a), of which only I. goesi is known in Western Europe. It has not previously been recorded in France.

Genbank accession numbers of I. goesi from our study (SK_19_49): CO1 OK393908; 28S OK393941.

Biology. Ischyrocnemis goesi (Figs 29, 30) was found at just two sites (Table 2) and only in small numbers. It was only ever reared from winter cocoons (N = 7), from host larval collection dates 8–13 June with host cocoon formation 14–20 June, and emergence of the adult parasitoid 3–18 April the year after. It was reared in culture only twice (both from winter cocoons, following oviposition into late instar hosts on about 26 June with emergence of one adult male on 11 May the next year, and a female that died failing to eclose at about the same time) from several ovipositions into unparasitized hosts in captivity. Most of these hosts, readily oviposited into from 2^nd to 5^th instars, unfortunately died as prepupae in host winter cocoons (never in summer cocoons), or sometimes as larvae before cocooning. The capture of two adult females in the field, on 12 May and 25 June, on the face of it might suggest that I. goesi is plurivoltine, but we have not been able to show that, neither from field rearings nor experimentally, and it seems more plausible that it is univoltine and that either there is a prolonged emergence period from winter cocoons (cf. 3–18 April and also 11 May, above) and/or the adult females are unusually long-lived. It is possible that oviposition by I. goesi provokes the host into forming a winter cocoon, but the evidence is only circumstantial. Oviposition takes place after only brief antennal contact with the host, which is usually grasped by at least the front two pairs of legs during a process lasting from 7–22 seconds (Fig. 31). The ovipositing female orientates along the long axis of the host, but both head-to-head and head-to-tail orientations were seen and the egg is randomly placed in the haemocoel. There is no paralysis, but the host generally fails to react until later.

The small white egg is about 0.4 × 0.11 mm long, strongly curved, and without any appendage (Fig. 32). In small hosts it is visible through the cuticle (Fig. 33), and remains white until the caudate first instar larva hatches within 3 days. On dissection from the host the first instar larvae seen were curled (Fig. 34) but if arranged straight would be about 1.1 mm long (including tail) with a large head (about 0.2 mm long) soon after hatching, in overall appearance typical of many endoparasitoid ichneumonids. We could not detect any sign of a surrounding trophamnion, although from the relative sizes of freshly laid egg and first instar larva the possibility that the egg is alecithal should not be ruled out. The larva probably remains in its first instar until the host has formed its cocoon, though we did not ascertain that. The first instar larvae depicted in Fig. 34 were dissected from a single host and the dark material adhering to the left hand larva might be a result of its having been wounded by the other. The frail parchment-like cocoon is light brown with its strong central band frequently orangish (Figs 35, 36). The ovipositor is shown in Figs 37 (dissection) and 38 (valves separated in dead specimen). The apex of the clypeus bears a pronounced tooth, and the teeth of the broad mandible have internal flanges (Fig. 39).

Although self-superparasitism is easily induced in captivity (cf. Figs 32, 34), it seems not to be deliberate and requires successive visits (though these can be almost immediate). On one occasion a male I. goesi was reared from a host carrying three Terozoa quadridens eggs when collected.

Final instar larval cephalic sclerites (Figs 40, 41). Specimens examined: DBW 17.vi.2020b, 17.vi.2020a (both NMS).

Cephalic structures generally moderately to strongly sclerotized. Lateral portions of epistoma lightly sclerotized, epistomal suture unsclerotized. Labral sclerite absent; clypeolabral area with lightly scerotized central region, lenticular in shape, with moderately sclerotized dorsal margin, bearing setae and two clypeolabral plates near its ventral margin. Stipital sclerite present and strongly sclerotized, more or less horizontal; cardo absent. Pleurostoma lightly sclerotized; posterior struts of inferior mandibular processes not connected by band; accessory pleurostomal area absent. Hypostoma long and strongly sclerotized, lateral end not divided in two at posterior tentorial pit and without extensions; accessory hypostomal area absent. Hypostomal spur present, about 1.5 × as long as its basal width. Labial sclerite nearly circular, moderately to strongly sclerotized. Salivary orifice U-shaped. Prelabial sclerite present as lightly sclerotized transverse band, connected to interior ventral margin of labial sclerite by weakly sclerotized projection of labial sclerite. Labial sclerite with six setae. Prelabial area with four setae. Maxillary and labial palpi each bearing two sensilla. Mandible strongly sclerotized; blade about 0.6 × as long as full mandibular length, without fine denticles. Antenna without papillus. Parietal band not visible. Spiracle with closing apparatus absent; intercalary trachea absent, subatrium about 2.0 × as long as atrium. Skin covered with small, bubble-like protuberances; spines absent; setae present, short and scattered.

Using the key to subfamilies in Short (1978), almost all the above characters would place I. goesi in Ctenopelmatinae. The one ambiguous feature is the sclerotized area on the clypeolabral surface. Short (1978) interprets this as the labral sclerite, recording it in the following ctenopelmatines: Euryproctini – Euryproctus luteicornis (Gravenhorst); Mesoleiini – Protarchus testorius (Thunberg); Perilissini – Lathrolestes luteolator, Opheltes glaucopterus (Linneaus), Perilissus filicornis (Gravenhorst) and P. rufoniger (Gravenhorst). He acknowledged this anomalous character for ctenopelmatines in a footnote but did not work it into his key. We were not able to examine Short’s slides of these species but we did examine Lathrolestes erythrocephalus as part of this study. It too has the sclerotized clypeolabral feature (Fig. 28). However, the labral sclerites, as found in ectoparasitoid ichneumonids (Pimplinae, Xoridinae, Cryptinae, Phygadeuontinae, Tryphoninae, etc.), are quite different from the sclerotized areas in I. goesi and L. erythrocephalus. Labral sclerites have well-defined borders and end with downturned ‘knobs’ that overlap the mandibles; the labral interior is not noticeably sclerotized. The structures in I. goesi and L. erythrocephalus have lateral ends that merge into clypeolabral plates, and have lightly to moderately sclerotized interiors. The structure is situated high on the clypeolabral area and does not overlap with the mandibles. Short (1978 figs 311–312) indicates something of this nature for his Perilissus species. The Opheltes example (Short 1978: fig. 316) is only a poorly defined arc. In summary, we do not believe that labral sclerites are present in the Ctenopelmatinae. Sharanowski et al. (2021) present a phylogenomic tree for the Ophioniformes. In the subfamilies beyond the Tryphoninae (which would best be called the ‘higher Ophioniformes’) the labral sclerite is absent in all known larvae, and we do not know of any reversals of this loss, notwithstanding Short (1978). The forthcoming study by Bernardo Santos and his collaborators (in prep.), using genomic ultraconserved elements (UCE), has sequenced some 500 genera of Ichneumonidae, though unfortunately not Ischyrocnemis. It would appear from their results that the sclerotized clypeolabral structure has arisen at least twice in ctenopelmatines: once in Perilissini and once in the rather distantly related Euryproctus (Protarchus was not sequenced).

Unsurprisingly, in view of its biology, the larva of I. goesi lacks the derived suite of characters found in Metopiinae, which all pupate within Lepidoptera pupae. The only subfamily into which Ischyrocnemis can reasonably be placed on larval grounds is the Ctenopelmatinae, which is entirely compatible with its developmental biology. However, its biological and larval characters are too generalized to fit easily into any of the recognized tribes.

Ctenopelmatinae sp. X

Figs 42, 43

This species was found only once, as a cocoon with a dead prepupa in a winter host cocoon resulting from an Aprosthema larva collected at Site 2. The unbanded light brown cocoon (Fig. 42) has a felted appearance quite different from cocoons of the other parasitoids encountered.

Final instar larval cephalic sclerites (Fig. 43). Specimen examined: DBW 21.vi.2020a (NMS).

Cephalic structures generally moderately to strongly sclerotized. Epistoma lightly to moderately sclerotized; epistomal band present; dorsal margins of both poorly defined and merging with general light sclerotization of frons. Labral sclerite absent; clypeolabral area with weakly sclerotized central region, poorly defined crescentic band present; clypeolabral plates absent. Stipital sclerite present and strongly sclerotized, more or less horizontal; cardo absent. Pleurostoma lightly sclerotized; posterior struts of inferior mandibular processes not connected by band; accessory pleurostomal area absent. Hypostoma long and strongly sclerotized, lateral end not divided in two at posterior tentorial pit and without extensions; accessory hypostomal area partially present and lightly sclerotized. Hypostomal spur present, about 3.8 × as long as its basal width. Labial sclerite ovoid, ventral portion 0.6 × as long as length of labial sclerite, lightly to strongly sclerotized. Salivary orifice U-shaped. Prelabial sclerite present as weakly sclerotized transverse band, connected to interior ventral margin of labial sclerite by weakly sclerotized projection of labial sclerite. Labial sclerite with six setae. Prelabial area with four setae. Maxillary and labial palpi each bearing two sensilla. Mandible strongly sclerotized; blade about 0.6 × as long as full mandibular length, without fine denticles. Antenna without papillus. Parietal band present as lightly to moderately sclerotized vertical oblong with irregular margins. Spiracle with closing apparatus absent; intercalary trachea absent, subatrium about 1.5 × as long as atrium. Skin covered with small, bubble-like protuberances; spines absent; setae present, short and scattered.

This specimen can be unequivocally placed to Ichneumonidae using the keys and illustrations in Short (1952), Čapek (1970, 1973), and Finlayson (1987). As with Ischyrocnemis goesi, this specimen runs to Ctenopelmatinae in Short (1978) but cannot be placed further. Short used tribes based on adult characters and then created keys based upon character combinations that appear to be useless. While some ctenopelmatine genera or groups of genera seem distinct, no match is evident for Ctenopelmatinae sp. X.

Terozoa quadridens (Tryphoninae, Tryphonini)

Figs 44–47, 50–68; FilmingVarWild videos 5–7, 13

Taxonomy. Terozoa is a Eurasian genus of only three or four species that are taxonomically ill-understood. Nothing is published on the biology of any species. The taxonomic history of Terozoa is complex. The genus was first described (Förster 1869) without an included species, but one was designated by Perkins (1962) by the description of the single specimen standing as Terozoa in Förster’s collection, a male, as Terozoa quadridens Perkins, 1962. That remained a little-known and poorly understood species, and Townes (1971) erroneously treated Terozoa as a junior synonym of the genus Ischyrocnemis Holmgren, 1858 (type species Iscyrocnemis goesi Holmgren, 1858), that Townes (1971) placed, but only tentatively, in the subfamily Metopiinae. At that time both nominal genera each contained a single species, and nothing was known of their biology. Townes’s (1971) mistaken synonymy led to Scaramozzono (1995: 56) incorrectly placing Ischyrocnemis in Tryphoninae following his detection of an externally-borne egg on the ovipositor of a specimen he recognised as “Ischyrocnemis” quadridens; ironically confounding the credit due to him of first recognising Terozoa as a tryphonine. Perkins (1962) had in fact previously recognised them as distinct genera, but placed both in Ctenopelmatinae.

Soon after Townes’s generic classification, an unusual genus of Tryphoninae, Parablastus Constantineanu, 1973, monobasic with type species Parablastus bituberculatus Constantineanu, 1973, was described and placed in Tryphonini, again with nothing known of its biology. Bennett (2015) included Parablastus in his generic treatment of Tryphoninae but, as he had not seen a specimen, and because the original description was not particularly rich in detail, the genus does not run smoothly in Bennett’s (2015) key, largely because at least some species have an apparent glymma (Fig. 44, but see below).

Kasparyan (2019a), in a paper treating both Terozoa and Ischyrocnemis at species level, resurrected Terozoa from the synonymy of Ischyrocnemis, recognised Terozoa as a genus of Tryphonini, and placed Parablastus as a junior synonym of it. Because three nominal species had been described in Parablastus, new combinations were established for Terozoa bituberculata (which he synonymised with Terozoa quadridens, but see below), T. iberica (Kasparyan, 1999) and T. anatolica (Gürbüz & Kolarov, 2005). No species that would now be classified as Terozoa has previously been recorded from France.

The identification of Terozoa species is bedevilled by firstly quite profound and unusual sexual dimorphism (except for T. iberica, males have predominantly or entirely black faces while in females the face is largely yellow), and secondly the fact that descriptions of new species have not been supported by examination of types of species already described, and some characters have been misinterpreted or expressed in vague terms. Thus keys, which have not been based on examination of specimens believed to represent all species, have perpetuated earlier mistakes. Through the kindness of Stefan Schmidt (ZSM) we have been able to borrow the holotypes of both T. quadridens (a male, in fairly good but not perfect condition) and the nominal Parablastus anatolicus, a female which was accompanied by a clearly conspecific male paratype (both in good condition). The pair of T. anatolica has been crucial for understanding aspects of sexual dimorphism in the genus. Unfortunately our attempt to borrow (or to obtain photographs of) the female holotype of the nominal Parablastus bituberculatus from the Constantineanu collection was unsuccessful, but photographs of the female holotype and male paratype of T. iberica have been examined.

The determination of our species was problematical because for most of the duration of the project we had only females (N = 6). In 2021, however, we succeeded in rearing a male specimen, which agrees well with the type of T. quadridens, the main difference being its entirely black first metasomal tergite (postpetiole becoming brownish in holotype of T. quadridens, as indeed is the case for all other species except for T. ibericus in which T1 is more extensively yellowish posteriorly (Kasparyan 2019a)). We had already concluded that our females are not T. bituberculata, as they all have a completely black propodeum, and claws strongly pectinate basally, unlike the original description (Constantineanu 1973) of the female of T. bitubercuata which indicates orange markings on the propodeum and claws with only a comb of setae. Contantineanu (1973) makes no mention of a glymma, but whether or not this is because it is lacking in T. bituberculata is unclear. Through examination of the female holotype we also ruled out T. anatolica, which is a shorter/broader insect than our material, the female holotype having a noticeably shorter mesosoma and broader metasomal tergites (T2 0.9 and T3 0.7 × as long as wide, as against respectively 1.2–1.3 and 0.8–1.0 in our females), weaker pectination of claws and interstitial 1cu-a in the fore wing (postfurcal in the type of T. quadridens and in all the specimens resulting from our field-work, though in one female nearly interstitial). Also the male of T. anatolica has yellow marks at the upper corners of the clypeus and in the lower part of the inner orbits and malar space (face completely black in male T. quadridens; only the clypeus with a yellow band, which is also present but nearly broken centrally in the male paratype of T. anatolica). The Transcaucasian T. iberica, which is much more richly yellowish marked, as illustrated by Kasparyan (2019a), can also be excluded. Accepting our species as T. quadridens refutes Kasparyan’s (2019a) synonymization of T. bituberculata, which was based largely on his assumption that there would be only one species in the relatively well-studied strictly European fauna (Dmitri Kasparyan, pers. comm.), and we here resurrect T. bituberculata (Constanineanu, 1973) (stat. rev.).

The difference in body colour between the two sexes of T. quadridens from our study are as follows. Male: head black except apical band on clypeus and small mark adjoining vertical orbit yellow (mandible subapically and palpi in part yellowish); mesosoma entirely black. Female: Head black except the following yellow: face almost entirely (except for a small area ventral of antennal socket between it and yellow inner orbit) and clypeus, vertical mark in orbit, malar area and genae to above the lower level of eye, mandible except blackish teeth, and palpi (partly); mesosoma black but with propleuron (in one specimen diffusing into pronotal collar), subalar prominence, posterior band on scutellum, band on metanotum (sometimes nearly divided into two spots) yellow.

Genbank accession numbers for T. quadridens from our study (SK_19_48): CO1 OK393910; 28S OK393940.

While it is our view that more material is needed to inform the taxonomy of European and Anatolian species of Terozoa, to make the most of our opportunity to examine some relevant specimens we give a brief provisional key to species below, and then comment on some characters that have previously been used to separate species that appear to have little value, or have been misstated:

Note: The characters for T. bituberculata are from the original description, and the male is unknown.

The major surprise was the egg, which bears a kind of grappling hook at the head end. When first seen this appeared to be a single structure, but it later became clear, from a dissected-out hatched egg, that it is (or at least can be) at least three separate hooks, each carried on its own strand (Fig. 18). The egg, which darkens and toughens after being laid, becomes fastened to an internal structure of the host by means of this and it then stays put (even after hatching) during the host’s larval life. Being almost black in a nearly translucent host, it is easily seen from the exterior (particularly when the host is in an early instar). Both the black visibility of the egg and its fixed position after hatching might be explained by the following facts and (substantial!) speculation: we observed that the females of L. erythrocephalus seem to search for hosts visually, and also appear regularly to engage in a level of superparasitism (including self-superparasitism in single visits to larger hosts) that we postulate might be adaptive (cf. van Alphen and Visser 1990) in exhausting the host’s defensive response. We have seen that single eggs are sometimes encapsulated, and there is reasonable evidence that Ctenopelmatinae are always synovigenic and, in general, time-limited rather than egg-limited (Cummins et al. 2011). Heitland and Pschorn-Walcher (1992) noted a level of superparasitism much higher than would be expected by chance in several species of another ctenopelmatine tribe, Euryproctini, but in studied species of that tribe the first instar larva is surrounded by a trophamnion, which is not recorded for Perilissini. In fact these authors mention in contrast Perilissini species in which eggs seemed to be randomly distributed – unlike our belief for the perilissine L. erythrocephalus. As this species seems to recognise its host by sight, it seems conceivable that the visible presence of many eggs in a host, as opposed to just one or two, might deter the female from laying more eggs in such a host as, if several were already present, more eggs would almost invariably be wasted. This would explain the conspicuously dark egg colour in adaptive terms. The fixed position of the egg chorion might be interpreted as beneficial to the first instar parasitoid larva, which we observed to be highly active and mobile within the host – presumably in search of its potential competitors which it must kill in order to survive – and we frequently saw it coming up against its own eggshell. Possibly having a fixed array, rather than a changing scatter, of already dealt-with items within the host’s haemocoel aids its searching efficiency through some kind of spatial recognition, enabling it to concentrate on new items. It is otherwise very difficult to conceive of any adaptive benefit in the grappling hooks.

Biological studies on Ctenopelmatinae are rather few, as partly summarised and referenced by Broad et al. (2018). Among Perilissini, there is however considerable variation between species assigned to Lathrolestes. The univoltine Lathrolestes ensator (Brauns) is reported to have initially white but eventually black comma-shaped eggs (Zijp and Blommers 1993; Vincent et al. 2019 and references therein), though in this case eggs were essentially randomly distributed (i.e. there was neither positive superparasitism nor avoidance of it) and the egg did not hatch until the host was a cocooned prepupa in the ground. As in L. erythrocephalus, some L. ensator developed without diapause, progressing to the adult stage gradually through the winter to emerge the following spring, while others entered prepupal diapause in autumn and (under some circumstances) spent an extra year or more before becoming adult and emerging. Lathrolestes ensator was found in hosts as early as the first instar with an apparent preference for oviposition into the second instar, but later instar hosts were not attacked, being inaccessible. Many eggs were dissected from hosts in Zijp and Blommer’s (1993) study, and no mention of any ornamentation was made. The plurivoltine Lathrolestes nigricollis (Thomson) was studied in complementary ways by Eichhorn and Pschorn-Walcher (1973), who figured the first instar larva and the head capsule of the final instar, and by Quednau and Guevrement (1975, as Priopoda) who figured the female’s reproductive organs. The host is a leaf-miner and, after discovering the mine apparently at random, the parasitoid jabbed repeatedly into it, finally locating the host through its movements. Non-destructive host feeding occurred opportunistically but was not obligatory. The cucumber-shaped egg is white and no fastening structure was reported; the first instar larva hatched in from 3 to 7 days, but the authors do not say whether or not it then delays its development until the host becomes prepupal (but it would be surprising if this is not the case). Oviposition was into all instars, but especially into moderately well-grown hosts. Eggs suffered a high rate of encapsulation; although this was alleviated by superparasitism, eggs were distributed at random rather than there being a functional response. Only a proportion of first-generation offspring emerged in the same summer, the rest entering diapause as prepupae to emerge the following year, earlier than those from the second generation. Although with similar phenology, Lathrolestes luteolator (Gravenhorst), investigated by Carl (1976), seems to have a rather different biology and first instar larva. The female antennates the exposed host for up to a minute before ovipositing, preferring third to final host instars, less commonly attacking the second. The egg is white and bean-like, and no mention is made of any fastening device. The neonate first instar larva is said to remain partially embedded in the egg chorion, freeing itself as it grows, and it has ventral outgrowths on abdominal segments and spines on thoracic segments as well as an unusual head structure (illustrated by Schönrogge (1991), although without mention of any continued association of the first instar larva with its egg chorion), which have not been noted in other Lathrolestes species by the above authors. It is also said to have a distinctively unusual head with small mandibles. As in most ctenopelmatines it delays its further (rapid) development until the host is a prepupa.

As far as we are aware, fastening devices on Ctenopelmatinae eggs have not been reported, apart from the eggs of species of Euryproctus Holmgren (not a Perlissini) illustrated by Schönrogge (1991) and reported also by Cummins et al. (2011) having a small stalk at the caudal pole, akin to many Tryphoninae. In the extensive survey of ovarian eggs of Ichneumonidae undertaken by Iwata (1958, 1960) none that belong to Ctenopelmatinae had any ornamentation. We dissected eggs from a female of the very large Perilissini species Opheltes glaucopterus for careful examination, and the 1.25 × 0.25 mm egg certainly does not bear any trace of a projecting structure. It was confirmed that the slightly wider end is the capital pole (i.e. the narrower end was directed so as to issue first from the oviduct in accordance with Hallez’s (1886) Law). It is important not to conceive the appendage in L. erythrocephalus eggs as homologous to the stalked eggs of Tryphoninae and Lycorininae (and/or Euryproctini): quite apart from anything else, they are at different poles of the egg. They also appear to bear no relation to the more lateral modifications seen in the eggs of certain Anomaloninae (Broad et al. 2018 and references therein). If the structure of the egg of Lathrolestes erythrocephalus proves to be an autapomorphy, it should be borne in mind in any reclassification of Perilissini that this is the type species of the generic name Polyoncus Förster, 1869.

The repeated emergence of adults at times of year when no hosts would be available is noteworthy. Although captive conditions might have been contributing in some cases, it seems that asynchrony might arise naturally after L. erythrocephalus had entered prolonged diapause. It is possible, though we have no evidence for it, that in these cases the adult parasitoids could overwinter successfully, in view of their being synovigenic.

The really significant discoveries regarding I. goesi are firstly its host, and secondly the fact that it is an endoparasitoid that kills its sawfly host in the prepupal stage typical of Ctenopelmatinae rather than Metopiinae (which are larva-pupal parasitoids of Lepidoptera, cf. Broad et al. 2018). The larval cephalic sclerites amply confirm its placement to Ctenopelmatinae, though not to any of the currently recognised tribes. As there appears to be no trophamnion surrounding the first instar larva, and as the fine ovipositor lacks a dorsal notch (Fig. 37), it is almost certainly not placeable to Euryproctini, and from our biological data, including its small eggs, we concur with Perkins (1962) and Aubert (2000) in suggesting that it would most comfortably fit in Pionini of the ctenopelmatine tribes currently recognised; although both the ovipositor having the upper valves apically swollen and the short cerci would be anomalies. However, the internal classification of Ctenopelmatinae, and even whether the subfamily as currently understood is monophyletic, remains in doubt. The unusually strong tooth on the clypeus of adult I. goesi might aid the positioning of the powerful mandibles against the host cocoon as the adult cuts its way out, perhaps supporting our belief that I. goesi is a univoltine species always needing to deal with the tough winter cocoon of the host.

No biological information on Terozoa (= Parablastus) has hitherto been available. Unusual features of T. quadridens include regular oviposition onto the host’s head capsule. Although this is known in a few other species of Tryphonini (Zinnert 1969), frequent choice of the eye is even more unusual. In such cases the behaviour of the female, once the eye is chosen as the target, is absolutely resolute. Oviposition onto the head restricts T. quadridens to parasitising final instar hosts as the egg would be unable to transfer through the head capsule as it is cast, in contrast to some Tryphoninae with eggs deeply anchored into the epidermis of a body segment thus allowing the egg to tear through as the host moults (see Thibetoides, below, and also Shaw 2001 for Netelia). This explains the observed behaviour of the adult parasitoid monitoring penultimate instar hosts, waiting for them to become suitable for oviposition – a behaviour reminiscent of that of the campoplegine ichneumonid Hyposoter horticola (Gravenhorst) which oviposits into first instar larvae of the nymphalid butterfly Melitaea cinxia (Linnaeus) just before they hatch, having monitored the developing egg batches of the host for the previous several days (Nouhuys and Kaartinen 2008). Egg placement on the host’s head presumably dictates that the first instar T. quadridens larva must leave the eggshell and move to softer tissue before it can commence feeding. It is interesting that in none of our six female specimens (all killed by deep-freezing prior to immersion in ethanol) is there an egg exposed on the ovipositor (in three of our 6 dead specimens the egg can be discerned in the hypopygium and the stalk is just visible), unlike the several illustrations of females of Terozoa species in the literature (Kasparyan and Tolkanitz 1999; Kasparyan 2019a), which we presume must be of females that had experienced a less peaceful, chemically convulsion-inducing death. Holding the egg back in life may be necessary in view of its anchor being so tiny and stud-like: we never saw eggs held ready on the ovipositor in the living females seen searching or ovipositing, but rather it appeared to take some effort to force the egg through the hypopygium during the oviposition process, suggesting that the egg may not have fully emerged from the oviduct beforehand. Against this, however, is the strong concertina movements of the apex of the metasoma seen in the videos immediately after oviposition, that might signify re-arming with another egg. The quadridentate mandibular structure of T. quadridens must be a specialization aiding emergence by cutting through the particularly tough strands of the (inevitably) winter cocoon of the host.

Preferential parasitism of the host in an early instar (it seems that attempted oviposition on later instars is generally unsuccessful) is very unusual in Tryphonini (Bennett 2015), and the apparently rather precise and consistent egg placement, low down behind the third (or sometimes second) thoracic leg, is also notable.

Prior to this study, the closing apparatus had been regarded by the one of us (DBW) concerned with larval morphology as a pair of bow-like longitudinal structures found below the atrium and before the onset of the coarsely annulated spiracular trachea (Fig. 79A, B). The sections of trachea containing the closing apparatus typically had finer annulations than the following spiracular trachea. Sometimes a section of coarsely annulated trachea (similar to the spiracular trachea) was seen between the atrium and closing apparatus (Fig. 79A). Short (1978: 124) summarized what was known about the distribution and possible functions of these morphologies: “The closing apparatus of the spiracle adjoins the atrium in most endoparasites, whereas in ectoparasites it may adjoin the atrium or be situated some distance from the atrium. There are no obvious explanations for this. It is generally assumed that the spiracles are open in endoparasites in the final larval instar which is carnivorous, feeding on the tissues of the host and liberating air from its tracheae. If the spiracle is open there is possibly some biological advantage in having the closing apparatus situated close to the atrium. If one function of the closing apparatus is to prevent the entry of body fluids of the host into the tracheal system, then it is best situated to perform this function when it adjoins the atrium. If, however, such a precaution were necessary, one would more readily expect to find hydrofuge structures at the opening of the tracheal system, as in many aquatic insects.” This is the only published speculation on the function of the closing apparatus. We and others had never closely considered how the closing apparatus actually functioned, vaguely supposing it was some sort of valve; Andrew Bennett (pers. comm.) said that he “just assumed that the thin longitudinal structure was some sort of muscle or tendon that attached to the distal end of the trachea and somehow constricted the spiracle to close it.”

We were therefore surprised to find that the three ctenopelmatine species in this study (Lathrolestes erythrocephalus, Ischyrocnemis goesi, Ctenopelmatinae sp. X) lack a closing apparatus, and that this state is apparently widespread in ctenopelmatines (see comments under L. erythrocephalus). This led us to consider the closing apparatus in larval ichneumonids more closely. As far as we can determine, the term was first used by Beirne (1941). He provided a set of detailed drawings (his fig. 3) and said that the closing apparatus was either thick-walled or thin-walled. From his drawings, the thick-walled type gave the appearance of the longitudinal structures whilst the thin-walled type was just a simple cylinder (his m, n, p, q). This interpretation was accepted by Thelma Finlayson in her numerous papers treating ichneumonid larvae, and apparently by Short (1959, 1978) by his use of ‘thin walled’ to describe ctenopelmatine spiracles (although he then depicted internal longitudinal structures).

There are problems, however, with accepting the appearance of the closing apparatus to be the result of tracheal wall thickness. Looking at larger specimens such as Xorides rileyi (Ashmead) (Fig. 79A), the longitudinal structures can often be traced up into the section of trachea that separates Beirne’s closing apparatus from the atrium. Yet more troubling are the cases where there are clear indications of internal spines in the closing apparatus. Banchini exhibit this, and although some genera (Banchus Fabricius, Ceratogastra Ashmead, Philogalleria Cameron) lack longitudinal internal structures, others such as Exetastes Gravenhorst do have them (Wahl 1988). NMNH slides of Exetastes show this very clearly: E. illinoiensis (Walsh), E. ? bifenestratus Cushman (Short slides 188 and 190) and E. sp. (DBW 31.x.1985k). The section of trachea containing the longitudinal structures are flask-shaped and the walls in cross-section are thin, not thickened (Wahl 1988: figs 6–7). Some of the longitudinal structures are distinctly bent or otherwise off the median axis. It would seem that a better model for the closing apparatus is a tube that runs from the spiracular trachea to the atrial base, thus accommodating internal spines in the finely-annulated section of trachea that surrounds it. Another striking specimen is Metopius dentatus (Fabricius) (NMNH, Short slide 158). The subatrial region is spinose and what is clearly a tube runs from the atrium to the spiracular trachea. Several of the spiracles have the tube constricted and off-center, and one has the tube apparently expanded and partially contiguous with the tracheal wall.

More evidence for the closing apparatus being a tube is found in the Ichneumoninae. Ichneumonine larvae are highly derived (Bennett et al. 2019; Santos et al. 2021: supplementary material S8), yet since DBW was not interested in spiracular morphology, he had ignored (Sime and Wahl 1998; Bennett et al. 2019) one of their most interesting features: the heavily spinose region below the atrium, discussed in detail by Gillespie and Finlayson (1983) as the closing apparatus. Gillespie and Finlayson illustrated the spiracles of 42 ichneumonine species but did not show internal longitudinal structures (evidently considering them to have closing apparatuses of the ‘thin-walled’ type). Conversely, Short (1978) illustrated the spiracles of 111 species and showed longitudinal structure within all of them. DBW was able to examine 21 of Short’s slides and nine of Gillespie and Finlayson’s. The slide of Alomya semiflava Stephens (Hinz and Short 1983), the sister group to the rest of the Ichneumoninae (Bennett et al. 2019; Santos et al. 2021), was also to hand. Alomya semiflava has: (i) the area below the atrium with internal spines, and (ii) unmistakable longitudinal structures in that region (Fig. 79D). Of the 21 Short species, 14 show internal longitudinal structures. Gillespie and Finlayson slides show the internal structures in 4 species. Additional specimens [NMNH, EMUS] showing similar internal structure were also examined: Araeoscelis spp., Gnamptopelta obsidianator (Brullé), Holcojoppa mactator (Tosquinet), Mokajoppa respinozai Ward and Gauld, Pedinopelte sp., and Psilomastax pyramidalis Tischbein. In all cases, the longitudinal structure resolves as a tube within a spinose subatrial region (similar to Fig. 79C). Gnamptopelta obsidianator and M. respinozai have the tube itself distinctly spinose. Most ichneumonine specimens have the tube presenting as lacking in the majority of spiracles. This appears to be due to two factors: (i) the tube is apparently rather weakly attached to the atrium, and (ii) the tube continues down into the spiracular trachea, where it presumably merges at some point. Between these two aspects, the tube is often detached and retracted down the spiracular trachea – either from the moult from larva to pupa or the stretching and manipulation of the exuviae during slide mounting. As to why Short and Gillespie and Finlayson were at such odds in what they portrayed, Robertson Davies said it best: “… the eye sees only what the mind is prepared to comprehend”.

It is hard to believe that the longitudinal structure in most ichneumonid larval spiracles is the result of thickened tracheal walls, whilst the Ichneumoninae, Banchini, and Metopius have intraspiracular tubes. Instead it seems simpler to judge that Beirne misinterpreted what he saw. In short subatrial sections, the tube is broad where it meets the spiracular trachea, narrows in the middle, and then expands again at the atrial base – giving the impression of thickened walls that are internally convex. We propose a new spiracular terminology based upon the work of Michener (1953), who introduced morphological terms used throughout the aculeate community. The new usage is illustrated in Fig. 79. The finely annulated section of trachea below the atrium is the ‘subatrial area’. It connects with the more coarsely annulated ‘spiracular trachea’. If a coarsely annulated tracheal section occurs between the atrium and subatrial area, this is known as the ‘intercalary trachea’. The ‘closing apparatus’ is now defined as a tubular structure running within the subatrial area and connecting the atrium and spiracular trachea.

How the closing apparatus functions to regulate moisture loss (for ectoparasitoids) or fluid ingress (endoparasitoids) is an open question. A survey of Short’s NMNH slides reveals that the higher Ophioniformes (Cremastinae, Anomaloninae, Ophioninae, Campopleginae; Bennett et al. 2019) all lack a closing apparatus (Short generally drew them with a closing apparatus present, at least in a rudimentary fashion). This clade of larval endoparasitoids (larval-pupal in the case of anomalonines) comprises a large and successful radiation, despite this lack of a closing apparatus.