The chromatograms obtained from both prepupa and adult of Wainia sexsignata were clean. Translations to amino acids revealed that none of the sequences contained stop codons. Blast searches on Genbank and identification requests on BOLD indicated that neither sequence appeared to be close (>90% similarity) to any previously published sequence. The closest matches (approximately 88% similarity) were to diverse osmiine bee taxa (no sequence of Wainia (Caposmia) is available for this mitochondrial fragment). Both sequences were nearly identical to each other and differed by a single nucleotide on position 214, corresponding to an uncorrected genetic distance of 0.18%. This single point mutation is in a third codon position and is silent. Consequently, even if both sequences are not 100% identical it can confidently be concluded that the analyzed prepupa belongs to W. sexsignata.

The two nests of W. sexsignata were already closed when they were discovered. Each nest was built in an empty snail shell of Sphincterochila sp. (Sphincterochilidae) with a maximal diameter of 19 mm (Figs 2, 3). The two nests were not hidden among vegetation or under stones but lay openly on the stony and sparsely vegetated ground in a distance of few meters from each other. No traces of leaf pulp were visible on the outside of the shells, indicating that the female bees did not glue patches of leaf pulp onto the shell surface as many other snail shell nesting osmiine bees do (see Discussion). The two nests were not sealed against their rear end with a basal wall. Both had a very similar architecture except for minor differences in the construction of the nest plug and consisted of three brood cells each delimited towards the nest entrance by a cell partition (Fig. 3) and the nest plug that occupied the space between the outermost cell partition and the shell opening (Fig. 6).

Figures 1–3. 

Wainia sexsignata. 1 Nesting habitat 2.75 km southwest of Ifrane Atlas-Saghir in southern Morocco (photo V. Mauss) 2 Nest in a shell of Sphincterochila spec. closed with a plug built from chewed leaves 3 Opened nest with three linearly arranged brood cells each containing a prepupa.

The brood cells were arranged in a linear series and had a length of 15–17 mm (Fig. 3). Each cell contained a dead prepupa spun in a dark brown cocoon of 11–13 mm length except for one cell that harboured a meloid beetle pupa (see below). The space between the front end of the cocoon and the cell partition was densely packed with faecal pellets over a distance of 3–5 mm. The cell partitions were very thin and consisted of two to four pieces of yellow petals, which either slightly overlapped or formed two petal layers tightly contacting each other. Two non- or only slightly overlapping petal pieces were sufficient to close the shell whorl. The petal pieces were arranged parallel to each other and almost perpendicular to the shell columella (Fig. 4). Small portions of leaf pulp were used to glue the petal pieces on either end to the lateral sides of the inner shell surface (Figs 4, 9). The petal pieces contacting the ceiling or the floor of the shell whorl were additionally fixed with leaf pulp to the shell surface along their lateral margins. However, the petal pieces were never agglutinated medially with leaf pulp among each other. Petal pieces used to construct the cell partitions were invariably yellow ligules of Asteroideae ray florets, which were usually bitten off close to their base (Fig. 9). In contrast, leaf pulp used to glue the petal pieces to the inner shell surface was collected on green leaves as could be judged from its greenish to brownish colour and the presence of star-shaped hairs in the pulp matrix.

Figures 4–9. 

Wainia sexsignata. 4 Brood cell partition built from yellow petal pieces of Asteroideae, which are partly covered with dark brown remnants of the larval cocoon; note the portion of leaf pulp at the right upper side of the partition used to glue the petal pieces to the inner shell surface 5 Partition of the nest plug 6 Nest plug consisting of a series of four partitions constructed from yellow Asteroideae petal pieces immediately followed by the nest closure at the shell opening, which consists of leaf pulp; the partition at the extreme right is the outermost brood cell partition 7 Yellow petal pieces adhering to the rear side of the nest closure 8 Anterior part of the nest plug with an additional wall of leaf pulp built in front of the second outermost plug partition 9Asteroideae petal pieces used to construct cell and plug partitions; petals in the upper row with traces of leaf pulp, which glued them to the inner shell surface.

The nest plug, which occupied the space between the outermost cell partition and the shell opening, had a length of 9 mm and 11 mm, respectively. It consisted of four and six partitions, respectively (Fig. 6), which were evenly distributed over a distance of 6 mm and 8 mm, respectively, and the 2–3 mm thick nest closure at the shell opening (Figs 6, 8). The plug partitions were constructed in exactly the same way and from exactly the same material as the cell partitions (Fig. 5) with the exception that they were slightly thicker consisting of five to six yellow petal pieces arranged in three layers immediately behind each other. The nest closure consisted of rather roughly masticated greenish and reddish leaf pulp without the addition of sand grains (Figs 2, 6, 8). While the former was collected on green leaves of unknown origin, the latter was probably harvested on the reddish perigons of a Rumex (Polygonaceae) species growing at the study site. The nest closure was built immediately adjacent to the outermost plug partition with the petal pieces of the latter tightly adhering to its rear side (Figs 7, 8). In the nest containing six plug partitions, an additional and slightly less than 1 mm thick wall of greenish leaf pulp was built in front of the second outermost partition (Fig. 8); again, the yellow petals of the plug partition adhered to the rear side of this wall.

The innermost brood cell of one nest harboured a pupa of the blister beetle Sitarobrachys thoracica (Kraatz) (Meloidae) encapsulated in the cast skins of its last larval instars. This species parasitizes brood cells of andrenid and megachilid bees (Bologna and Pinto 2002). It was found to develop in soil nests of Andrena spec. and Hoplitis (Anthocopa) longispina (Pérez), in mud nests of Megachile (Chalicodoma) sicula (Rossi) and in snail shell nests of Hoplitis (Hoplitis) pallicornis (Friese) (Stahlberg 1909, Andres 1910, Martínez de la Escalera 1914, Cros 1924, Kaszab 1957, Faucheaux 2011). Wainia sexsignata is a new host of this blister beetle species.

The microscopical analysis of digested pollen grains in the larval faeces of Wainia sexsignata revealed that pollen of the Asteroideae (Asteraceae) was the near exclusive constituent of the former pollen provisions (Tab. 1). 96.6% of all pollen grains belonged to this subfamily, while 1.7% was assigned to the Scrophulariaceae and the remaining 1.7% could not be identified. The Asteroideae pollen in the larval faeces was exclusively of the Aster type. This suggests that Asteriscus graveolens, which richly flowered at the site where the Wainia nests were discovered (Fig. 10), was the main pollen host, which is in line with the observation of a female of W. sexsignata near Sidi Ifni visiting A. graveolens (see Methods). Cladanthus arabicus - the only other representative of the Asteroideae flowering at the study site - can be excluded as a pollen source since it possesses pollen grains of the Anthemis type.

Wainia eremoplana is mesolectic on Asteraceae and Brassicaceae (Tab. 2). Among the Asteraceae, the subfamily Asteroideae is by far the most important pollen source. Species of the Carduoideae are also frequently exploited (Fig. 11), whereas representatives of the Cichorioideae appear to be only very rarely used. Based on field data and label records, Anthemis spec., Chrysanthemum coronarium, Pulicaria incisa and Senecio spec. serve as pollen hosts among the Asteroideae and Centaurea hyalolepis and thistles among the Carduoideae. Pollen uptake from the Asteraceae flower heads involves rapid up and down movements of the metasoma against the pollen-bearing florets. On the Asteroideae, pollen is taken up directly into the metasomal scopa without the aid of the hind legs, whereas on Carduoideae the hind legs are involved in directing the florets under the seesawing metasoma. Pollen of Brassicaceae was recorded in 27 out of the 50 pollen loads analysed (Tab. 2). However, none of the loads was composed of Brassicaceae pollen alone, only one load contained more than 50% Brassicaceae pollen and the proportion of Brassicaceae pollen was overall rather low with an average of 19% per load. Nectar is occasionally also taken from other plant taxa than Asteraceae and Brassicaceae. Based on label records, several specimens of W. eremoplana were collected on Commicarpus hazeva in Israel, suggesting that this Nyctaginaceae species has served as a nectar source.

Figures 10–12. 

10 Asteriscus graveolens (Asteroideae, Asteraceae), an important pollen host of Wainia sexsignata (photo V. Mauss) 11 Female of Wainia eremoplana collecting pollen on Centaurea spec. (photo G. Pisanty) 12 Habitat of Wainia eremoplana in the central Negev Desert of Israel at Nahal Hatira (the large crater), 10 km east of Yeroham.

The use of large petal pieces to construct both cell and plug partitions and the presence of a series of partitions evenly distributed in the space between the outermost cell partition and the nest closure as observed in Wainia sexsignata are unique characteristics among the Palaearctic osmiine bees. Both characteristics, however, were also recorded in two snail shell nesting Afrotropical W. (Caposmia) species (Gess and Gess 1999, 2008, 2014, Kuhlmann et al. 2011). In W. atrorufa (Friese) and W. gessorum Kuhlmann, the thin partitions between the brood cells also consist of large pieces of petals arranged at right angles to the shell columella (Gess and Gess 2014), and the space in front of the outermost brood cell is filled with a succession of partitions constructed from petal pieces (Gess and Gess 1999, but see Gess and Gess 2008, 2014 for a slightly deviating description). In contrast to W. sexsignata, however, the nest closure at the shell opening does not consist solely of leaf pulp but rather is a combination of “cemented sand and plant matter” (Gess and Gess 1999). The petals used by the two Afrotropical species for nest construction possibly originated from Grielum (Neuradaceae) (S. Gess personal communication). However, as Grielum species have yellow flowers like several Asteraceae species growing in the same habitat, it cannot be excluded that the bees actually harvested yellow Asteraceae flowers for nest construction as in the Palaearctic W. sexsignata (see discussion on possible pollen hosts of W. atrorufa below). In summary, apart from slight differences in the material used for the outermost nest closure and possibly differing petal sources, the nesting biologies of the two Afrotropical W. (Caposmia) species and the Palaearctic W. sexsignata are identical, suggesting close relatedness.

In contrast, the snail shell nests of the Afrotropical Wainia (Caposmia) elizabethae (Friese) substantially differ from those of the three W. (Caposmia) species addressed above (Gess and Gess 2008, 2014, Kuhlmann et al. 2011). In W. elizabethae, the brood cell partitions and the single-walled nest closure consist of sand grains, which are cemented together with an unidentified resinous substance (which might possibly be saliva; S. Gess personal communication), and the space between the outermost brood cell and the nest closure is loosely filled with sand. The differing nest building material and deviating nest architecture of W. elizabethae indicate a considerable variability in the nesting biology of W. (Caposmia) species, which is unusual for osmiine bee subgenera, where nest architecture and nest building material are often highly conserved (Müller 2018a).

Beside the four species discussed above, the South African Wainia (Caposmia) braunsi (Peters) was reared from snail shells (Peters 1984). A further Afrotropical species, W. (Caposmia) algoensis (Brauns), is probably also a snail shell nester (Kuhlmann et al. 2011, Gess and Gess 2014). Therefore, the selection of empty snail shells as a nesting site appears to be a subgeneric trait of Wainia (Caposmia).

Pollen analysis of larval faeces of Wainia sexsignata and female scopal contents of W. eremoplana revealed a strong affiliation of both species to the Asteraceae, particularly to the Asteroideae and to a lesser extent also to the Carduoideae. However, neither species is strictly oligolectic on Asteraceae. W. eremoplana regularly collects pollen also on Brassicaceae and pollen of Scrophulariaceae was recorded in the larval faeces of W. sexsignata. The finding that the larval faeces from four out of the six brood cells were exclusively composed of Asteroideae pollen indicates that the larvae of W. sexsignata are capable of developing on a pure Asteraceae pollen diet, thus being physiologically adapted to cope with the unfavourable properties of Asteraceae pollen, which has repeatedly been shown to hamper larval development of bee species not specialized on this plant family (Müller and Kuhlmann 2008, Praz et al. 2008b, Sedivy et al. 2011, Vanderplanck et al. 2018).

The pollen host preferences of the southern African Wainia (Caposmia) species have not been studied. Except for the observation of a single female of W. atrorufa visiting flowers of Hermannia (Malvaceae), no flower records exist (Kuhlmann et al. 2011, Gess and Gess 2014). However, the larval provisions in the brood cells of five nests of W. atrorufa from two different localities were deep yellow and consisted of heavily sculptured and spiny pollen grains of 25 to 40 μm diameter (Gess and Gess 1999). These pollen characteristics clearly point to the Asteroideae, suggesting that W. atrorufa and possibly also other Afrotropical W. (Caposmia) species might exhibit a preference for Asteraceae pollen as the two Palaearctic species do.

Empty snail shells offer nesting opportunities for bees that build their brood cells within preexisting cavities. They appear to be well suited for nest construction for several reasons: i) the narrow and linear, albeit curved form of the shell cavity allows for the arrangement of the cells in a linear series and easy closure of the nest at the shell aperture; ii) the progeny are well protected by the considerable hardness of the shell and the usually white shell colour, which reflects the sunlight and thus might reduce the risk of overheating; iii) xeric habitats often harbour large numbers of empty snail shells providing numerous nesting opportunities; iv) the number of colonizable snail shells does not diminish over time due to the permanent renewal of the supply of dead shells; and v) the presence of empty snail shells enables colonization of habitats, e.g. (semi)arid areas, which naturally lack preexisting linear cavities such as insect burrows in dead wood or hollow plant stems used by innumerable bee species as nesting sites. On the other hand, nesting in exposed and movable snail shells may also entail disadvantages compared to nesting in the ground, in stems, in dead wood or in rock crevices: i) the larvae may have a higher risk of experiencing lethal temperatures by direct sunlight, particularly when they are still young and do not yet have reached the diapausing stage (see Müller et al. 2017); ii) nests, larval provisions or progeny may be more prone to parasitization or to predation, e.g. by small mammals or birds; and iii) flood events e.g. in dry riverbeds, which are important nesting habitats of bees in (semi)deserts, may destroy a larger proportion of the nests by sweeping them away.

Records of snail shell nesting species exist for all three main bee taxa that occupy preexisting cavities as nesting sites, i.e. Megachilini, Anthidiini and Osmiini (Megachilidae). However, the proportion of snail shell nesters substantially differs among these three taxa. Snail shell nesting in the Megachilini appears to be almost absent. The only record of a snail shell nesting megachiline bee refers to the Palaearctic Megachile (Chalicodoma) lefebvrei (Lepeletier), which usually constructs its brood cells in holes of stones and rocks (Ferton 1908), but facultatively also colonizes snail shells (Ferton 1920). Among the Anthidiini, snail shell nesting has been observed only in four Palaearctic species of the genus Rhodanthidium (Pasteels 1977 and references therein, Erbar and Leins 2017) and in two Afrotropical species of the genus Afranthidium (Gess and Gess 1999, 2008, 2014), most of which seem to be obligate snail shell nesters. The great majority of bees colonizing snail shells belongs to the Osmiini. Snail shell nesting has been recorded in 52 osmiine bee species belonging to 5 genera and 14 subgenera (Tab. 3). Among these species, representatives of the genus Osmia predominate with 35 species that exclusively or facultatively nest in snail shells. Snail shell nesting is much more common in Palaearctic osmiine bees (4 genera/11 subgenera/43 species) than in the Osmiini from the Afrotropics (2/2/5) and the Nearctic (2/2/4). The low proportion of snail shell nesters in the Afrotropics may be explained by the distinctly lower diversity of osmiine bees south of the Sahara compared to both the Palaearctic and the Nearctic fauna (Ascher and Pickering 2018). However, the almost complete absence of snail shell nesters in the Nearctic, which encompasses about 30% of the worldwide osmiine bee diversity in terms of described species (Ascher and Pickering 2018), remains enigmatic and can neither be explained by an overall scarceness of snails in North America nor by a poorer knowledge of Nearctic osmiine bee biology.

The snail shell nesting osmiine bees can be divided into two groups (Tab. 3). The representatives of the first group nest in various preexisting cavities and occasionally also colonize snail shells. Some species of these facultative snail shell nesters, such as Osmia (Osmia) bicornis, O. (Osmia) cornuta, O. (Osmia) tricornis or O. (Helicosmia) notata, only exceptionally utilize snail shells. If they do, they usually colonize large shells and often place their brood cells side by side rather than in a linear series (Tab. 4, Fig. 14), conforming to their ability to use cavities of widely different shapes and sizes for nesting. Other species of this first group, such as Osmia (Pyrosmia) ferruginea or O. (Pyrosmia) viridana (Fig. 20) frequently nest in snail shells but are not entirely dependent on them and occasionally also colonize other preexisting cavites. The representatives of the second group exclusively nest in empty snail shells and appear to have lost the flexibility to colonize alternative nesting sites. For these species, no nesting sites other than snail shells have ever been reported and they often belong to clades entirely composed of snail shell nesters. Though the assignment to one of these two groups is tentative for several species due to the low number of nests reported so far (Tab. 3), at least 56 osmiine bee species belonging to 4 genera and 14 subgenera are assumed to be obligate snail shell nesters, which corresponds to about 5% of the described osmiine bee species diversity (Ascher and Pickering 2018, Müller 2018a). Included in these figures are all species of Hoplitis (Tkalcua), Osmia (Allosmia), O. (Erythrosmia), the O. spinulosa species group of O. (Hoplosmia), O. (Neosmia) and Wainia (Caposmia), irrespective of whether they have been recorded to nest in snail shells or not. In these taxa, snail shell nesting is most probably a subgeneric or species group trait as all their representatives, for which information on the nesting biology is available, utilize snail shells as exclusive nesting sites. The percentage of obligate snail shell nesters is certainly higher than currently known. Several speciose taxa, such as Hoplitis (Hoplitis), Osmia (Helicosmia), O. (Pyrosmia) or Protosmia (Protosmia) are expected to contain additional species that exclusively nest in snail shells. Furthermore, species of other taxa, for which the nesting sites are still unknown, e.g. Osmia (Nasutosmia) or Wainia (Wainiella), might possibly turn out to be also snail shell nesters.

Based on available molecular phylogenies (Praz et al. 2008a, Rightmyer et al. 2013, Sedivy et al. 2013a), obligate snail shell nesting has likely evolved independently at least twelve times in the evolutionary history of the osmiine bees (Fig. 13): i) at least once in Protosmia (Protosmia), ii) once in Hoplitis (Anthocopa), iii) once in Hoplitis (Hoplitis), iv) once in the ancestor of Hoplitis (Tkalcua), v) once in the ancestor of Wainia (Caposmia), vi) once in the ancestor of the clade composed of Osmia (Allosmia) and Osmia (Neosmia) (but see below), vii) once in the ancestor of the Osmia (Hoplosmia) spinulosa species group (but see below), viii) once in the ancestor of Osmia (Erythrosmia), ix) at least once in Osmia (Pyrosmia), x) at least once in Osmia (Diceratosmia) and probably at least twice in Osmia (Helicosmia), i.e. in xi) O. aurulenta and xii) O. orientalis, which substantially differ in both adult morphology and nest architecture suggesting only distant relatedness.

Figure 13. 

Phylogenetic distribution of snail shell nesting in osmiine bees. Red = taxa that probably exclusively consist of obligate snail shell nesters; blue = taxa that contain both obligate and facultative snail shell nesters (in Hoplitis (Anthocopa), only one (exclusive) snail shell nester is known); green = taxa that probably contain only facultative snail shell nesters; black = taxa that do not contain any snail shell nesters to the present knowledge (nesting biology of Hoplitis (Megahoplitis), Hoplitis (Platosmia), Wainia (Wainia), Wainia (Wainiella) and Osmia (Nasutosmia) still unknown). sp. gr. = species group. Phylogeny based on Praz et al. (2008a), Rightmyer et al. (2013) and Sedivy et al. (2013a).

Among the snail shell nesting osmiine bees, Osmia (Allosmia) and O. (Neosmia) are the only subgenera containing species that seal their nests with a mixture of leaf pulp and fragments of mollusc shells and that transport and bury their shells after nest closure (Tab. 4, Figs 23, 24). In addition, representatives of both subgenera glue patches of leaf pulp onto the shell surface, which is otherwise observed only in one O. (Helicosmia) species (Fig. 16). These peculiar characteristics common to species of both subgenera suggest close relatedness. In fact, the two subgenera were found to be sister taxa in molecular phylogenetic studies (Praz et al. 2008, Rightmyer et al. 2013). Thus, snail shell nesting in O. (Allosmia) and O. (Neosmia) has probably evolved only once in the ancestor of the two subgenera. Given the unsettled phylogenetic relationships among the species groups of Osmia (Hoplosmia) and the unknown nesting biology of O. (Nasutosmia) nasuta (see Fig. 13), it cannot be excluded that snail shell nesting is the ancestral condition in the Nasutosmia-Allosmia-Neosmia-Hoplosmia clade. Under this scenario, the minimal number of independent origins of snail shell nesting in osmiine bees would reduce to eleven.

Two species of Osmia (Allosmia), probably all species of O. (Neosmia) and one species of O. (Helicosmia) glue patches of leaf pulp onto the outer surface of the nest shell before and often also during cell construction and provisioning (Tab. 4, Figs 16, 23). This behaviour has likely evolved twice, probably once in the ancestor of O. (Allosmia) and O. (Neosmia) and once in O. (Helicosmia) aurulenta. Its function is still enigmatic. It has alternatively been interpreted as an evolutionary relict inherited from an ancestor that did not colonize preexisting cavities but constructed free standing brood cells (Bellmann 1981), as a camouflage strategy to reduce the optical conspicuousness of the white shells (Grozdanic 1969), or as a means to facilitate the movement of the shell (Malyshev 1937), which, however, is unlikely as several snail shell transporting O. (Allosmia) species do not show this behaviour (Tab. 4). Interestingly, instead of gluing patches of leaf pulp on the outer surface of the shell, some O. (Allosmia) species, such as O. melanura and O. rutila, attach portions of leaf pulp to the inner shell surface few millimeters behind the shell opening immediately after selection of a suitable shell for nesting (Müller 1992, Haeseler 1997). Experiments showed that this leaf pulp markings near the nest entrance allow the females to recognize their own nests (Haeseler 1997). Thus, covering the outer shell surface with patches of leaf pulp as observed in other O. (Allosmia) species, in O. (Neosmia) and in O. (Helicosmia) aurulenta might possibly serve the same purpose, i.e. to individually mark the shell and to signal other females that the shell is already occupied.

Species of Osmia (Allosmia) invariably construct only one brood cell per snail shell (Tab. 4, Fig. 18). All other obligate snail shell nesters build two or more cells per shell if the available space allows. Most of these species arrange their cells in a linear series (Fig. 15), resulting in the selection of predominantly small to medium-sized shells with narrow whorls. Exceptions among the obligate snail shell nesters are H. (Hoplitis) fertoni and O. (Helicosmia) aurulenta, which usually colonize medium-sized to large shells and build their brood cells side by side rather than in a linear series, at least in the wide outer shell whorls (Fig. 17).

Several Palaearctic and Afrotropical species of the subgenera Osmia (Allosmia), O. (Erythrosmia), O. (Neosmia), O. (Pyrosmia) and Wainia (Caposmia) construct a barrier between the outermost cell partition and the nest closure, which most probably contributes to impede the invasion of the nests by predators and parasites (Fig. 19). Depending on the species, this barrier consists of small and often densely packed particles (small pebbles, earth crumbs, plant particles etc.), sand or a succession of partitions (Tab. 4). The construction of such barriers, however, is not restricted to snail shell nesting species but has also been recorded from many osmiine bees nesting in insect burrows in dead wood or in plant stems, particularly of the subgenera Hoplitis (Alcidamea) and Osmia (Pyrosmia) (Müller 2018a and references therein).

Figures 14–19. 

14 Opened nest of Osmia (Helicosmia) notata with brood cells arranged side by side 15 Opened nest of Osmia (Hoplosmia) pinguis with brood cells arranged in a linear series 16 Nest of Osmia (Helicosmia) aurulenta densely covered with patches of leaf pulp 17 Opened nest of Hoplitis (Hoplitis) fertoni with brood cells arranged side by side (photo G. Le Goff) 18 Opened nest of Osmia (Allosmia) rufohirta with single brood cell (photo P. Westrich) 19 Opened nest of Osmia (Neosmia) bicolor with barrier composed of small pebbles and earth crumbs (photo A. Krebs).

One of the most fascinating aspects of the biology of snail shell nesting osmiine bees is the way the females treat their nests after having sealed them (Tab. 4, Figs 22–25). While numerous species leave the closed shell without changing its position or orientation, some Palaearctic species belonging to three subgenera exhibit specialized behaviours: i) O. (Hoplosmia) spinulosa turns the shell in situ into a particular position with the shell aperture being directed tightly towards the ground (Müller 1994; Fig. 22); ii) O. (Allosmia) rufohirta rolls the shell often over a considerable distance to hide it under vegetation or stones (Bellmann 1981; Fig. 23); iii) several species of O. (Allosmia) and O. (Neosmia) transport the shell to a suitable place where it is buried few centimeters deep into the sandy ground (Ferton 1890, Müller 1992, Haeseler 1997, Vereecken and Le Goff 2012; Fig. 24) and iv) two species of O. (Neosmia) either turn the shell in situ until the aperture is directed towards the ground or bury it shallowly into the ground, before it is completely covered with a dense pile of coniferous needles, grass blades or the like (Grozdanic 1971, Bellmann 1981; Fig. 25). Interestingly, O. (Allosmia) sybarita usually buries the shell into sandy ground (Vereecken and Le Goff 2012), but hides it under low vegetation if the ground is too hard to dig (G. Le Goff personal communication). A similar intraspecific flexibility is observed in O. (Allosmia) rufohirta and O. (Neosmia) bicolor, which occasionally bury the shell at least partly in the ground (Bonelli 1972, Bellmann 1981). These specialized behaviours of turning, hiding, burying or covering of the sealed shells are certainly linked to the protection of the progeny. However, whether they have specifically evolved as a protection against heavy rain, overheating, optically oriented predators or brood parasites remains unclear and needs experimental testing.

Figures 20–25. 

20Osmia (Pyrosmia) viridana at nest entrance (photo N. Vereecken) 21Protosmia (Protosmia) paradoxa on snail shell (photo N. Vereecken) 22Osmia (Hoplosmia) spinulosa turning its nest in situ 23Osmia (Allosmia) rufohirta rolling its nest towards a protected place 24Osmia (Allosmia) sybarita rolling its nest into a self excavated hole in sandy soil (photo N. Vereecken) 25Osmia (Neosmia) bicolor covering its nest with plant stalks (photo A. Krebs).