Cell and Tissue Research

, Volume 316, Issue 3, pp 377–390

Retinal and optical adaptations for nocturnal vision in the halictid bee Megalopta genalis


    • Department of Cell and Organism BiologyLund University
  • Willi A. Ribi
    • University of Human Sciences of the Principality of Liechtenstein
  • Eric J. Warrant
    • Department of Cell and Organism BiologyLund University
Regular Article

DOI: 10.1007/s00441-004-0883-9

Cite this article as:
Greiner, B., Ribi, W.A. & Warrant, E.J. Cell Tissue Res (2004) 316: 377. doi:10.1007/s00441-004-0883-9


The apposition compound eye of a nocturnal bee, the halictid Megalopta genalis, is described for the first time. Compared to the compound eye of the worker honeybee Apis mellifera and the diurnal halictid bee Lasioglossum leucozonium, the eye of M. genalis shows specific retinal and optical adaptations for vision in dim light. The major anatomical adaptations within the eye of the nocturnal bee are (1) nearly twofold larger ommatidial facets and (2) a 4–5 times wider rhabdom diameter than found in the diurnal bees studied. Optically, the apposition eye of M. genalis is 27 times more sensitive to light than the eyes of the diurnal bees. This increased optical sensitivity represents a clear optical adaptation to low light intensities. Although this unique nocturnal apposition eye has a greatly improved ability to catch light, a 27-fold increase in sensitivity alone cannot account for nocturnal vision at light intensities that are 8 log units dimmer than during daytime. New evidence suggests that additional neuronal spatial summation within the first optic ganglion, the lamina, is involved.


Visual systemNocturnal visionApposition compound eyeRetina structureDim lightMegalopta genalis (Insecta)


A true nocturnal lifestyle has emerged only eight times within the bees (Engel 2000). One of these fascinating transitions from diurnal to nocturnal behaviour can be found in the facultatively social halictid bee Megalopta genalis (Fig. 1). Although our current knowledge about nocturnal bees is limited, M. genalis has been well studied, and we know a good deal about its nest structure (Sakagami 1964; Janzen 1968) and biology (Wolda and Roubik 1986; Roulston 1997; Hopkins et al. 2000; Arneson and Wcislo 2003; Smith et al. 2003). Behavioural studies have shown that M. genalis females are only active 1 h before dawn and 20 min after dusk when light intensities under the dense forest canopy are equivalent to starlight intensities above the canopy (Almut Kelber, unpublished data). Females navigate through the tangled rainforest at night and are even able to orient and learn landmarks visually, implying that they can see well in dim light (Warrant et al., in preparation).
Fig. 1

Female nocturnal halictid bee Megalopta genalis drawn after mounted material (by L. Ribi)

Arthropods possess two major types of compound eyes: apposition eyes and superposition eyes. Apposition eyes (Fig. 2a) feature small apertures consisting of a single, tiny facet. The visual units, called ommatidia, are optically isolated from each other by a thick sheath of pigments absorbing light reaching the eye off-axis. Due to the small aperture, the apposition eye design works best in high light intensities, usually restricting the animal to a diurnal lifestyle. Low light intensities result in a poor photon catch and unreliable visual signals. Superposition eyes (Fig. 2b), in contrast, feature wide “superposition apertures” that comprise a large number of facets (typically several hundred). By employing specialised optics (Land 1981), the lenses bend incoming parallel light rays across a wide pigment-free “clear zone” (Fig. 2b), where they are superimposed onto a single rhabdom. Each rhabdom therefore receives light through a much wider aperture, via the optical action of many ommatidia, and this greatly improves photon catch and thus sensitivity (Nilsson 1989). Due to this enhanced sensitivity, superposition eyes are typically found in nocturnal insects. Interestingly, the benefits of improved sensitivity are not restricted to nocturnal insects: many diurnal insects also possess superposition eyes (Warrant 2001).
Fig. 2

Schematic drawings of a an apposition and b a superposition compound eye. a In the apposition compound eye each rhabdom (Rh) receives only light from its respective corneal facet (large arrow), representing the aperture of the eye and having a facet diameter D. Off-axis light (small arrow) is absorbed by pigment cells surrounding the crystalline cone (CC). b In superposition compound eyes (here of the refracting type, see Nilsson 1989) light travels through wide apertures (A) composed of numerous corneal facets (large arrows). This significantly increases the amount of light entering the eye and passing through the clear zone (CZ) to be focused onto a single rhabdom (Rh) (C cornea). (Adapted from Nilsson 1989)

Considering Megalopta’s remarkable nocturnal visual abilities, superposition eyes would be of great advantage to this bee. However, the more than 16,000 described bee species worldwide (Michener 2000) have formed a clearly established monophyletic lineage (Jander and Jander 2002) with the apposition eye as its dominant eye design. Even though Megalopta’s nocturnal visual abilities using apposition eyes are an extreme example, the African race of the honeybee Apis mellifera and the Asian apid A. dorsata can continue to forage on nights with a half-full moon present in the sky (Fletcher 1978; Dyer 1985). Theoretically, however, light levels at early dusk should render their insensitive apposition eyes blind (Warrant et al. 1996). Due to this clear contradiction, Megalopta’s eyes must have evolved distinct nocturnal adaptations. Several possibilities exist that may allow Megalopta to increase light capture at night. These include the development of (1) more sensitive optics, (2) larger and more sensitive rhabdoms, and/or (3) an eye design that changes from apposition during the day, to superposition at night, via morphological and optical circadian changes as found in the eyes of some tiger beetles (Brännström 1999). We explored these possible nocturnal adaptations by means of histology and optical measurements. Our study represents the first detailed description of the apposition compound eye of a nocturnal bee. The results are analysed and discussed in comparison with the well-studied worker honeybee Apis mellifera and females of the more closely related diurnal halictid bee Lasioglossum leucozonium.

Materials and methods

Females of Megalopta genalis (Meade-Waldo 1916) were collected in a 2-week period (25 March to 5 April 2003) on Barro Colorado Island (9°N, 79°W), Republic of Panama. Their distribution ranges from Mexico to southern Brazil, and they were common at our study site in Panama. Females live solitarily or in small groups (of up to ten bees), constructing nests and brood cells with their strong mandibles in small wooden branches. These can be primarily found above the ground and within the thicket of the rainforest. As these bees are only active at night during two activity periods shortly after dusk and before dawn, we used UV light tubes mounted over white linen sheets to attract females during these peak activity times. In this tropical semideciduous moist forest the dry season generally lasts from mid-December until mid-May (Leigh 1999). M. genalis is most abundant during the dry-to-wet season transition and during the first half, and at the end of, the wet season (Wolda and Roubik 1986).

As comparison species we chose the more closely related diurnal halictid bee Lasioglossum leucozonium (Schrank 1781) and the European worker honeybee Apis mellifera carnica (L. 1758) (Table 1). L. leucozonium was collected near the ecological field station of Uppsala University, situated on the Swedish island of Öland, and in southern Sweden near Lund. This halictid bee is only active during the day (J. Tengö, personal communication) and we caught them on flowers during their foraging flights. Depending on the year, the highest abundance of L. leucozonium lasts from the beginning of May until mid-August (Amiet et al. 2001). A. mellifera workers were caught on flowers around the Zoology building in Lund.
Table 1

Taxonomy of the bee species used in this study












Apoidea ⇒






M. genalis








L. leucozonium








A. mellifera


Bees were immobilized by cooling them down at 4°C. Their head capsules were subsequently opened by cutting the mouthparts and the frontal head cuticule away. Optimal fixative penetration into the retina was achieved by cutting the most ventral rim of the eye away. In order to study the effect of dark adaptation vs. light adaptation in the retina in M. genalis, one group of bees was dissected under dim light conditions for dark adaptation and another group in bright light at 6:30 a.m., at the end of their activity period. The diurnal comparison species A. mellifera and L. leucozonium were dissected and fixed in bright light during daytime. We used a fixative consisting of a mixture of 2.5% glutaraldehyde and 2% paraformaldehyde in phosphate buffer (pH 7.2–7.5). The heads were fixed for 2–3 h at 4°C before being osmicated (2% OsO4 in distilled H2O) for 1 h. The heads were subsequently dehydrated in an ethanol series, transferred to propylene oxide and embedded in epoxy resin (Fluka).

Frontal, longitudinal and tangential serial sections, 1 μm thick, were cut on a Reichert Ultracut microtome using glass knives. Four tangential and six longitudinal section series were cut from the whole eye of light adapted M. genalis females, as well as two tangential and two longitudinal section series from the whole eye of dark adapted M. genalis. In A. mellifera and L. leucozonium two tangential and longitudinal sections were sectioned from the whole eye of both species. Moreover, we applied a re-embedding method (Ribi 1976, 1978) that enabled us to follow single ommatidia throughout their whole length in different regions of the eye of M. genalis. In order to do this, serial frontal sections, 25–50 μm thick, were cut on a Reichert sledge microtome, and embedded with epoxy between two acetate sheets. Sections containing complete and parallel ommatidia were selected and small areas containing six to ten ommatidia were cut out and re-embedded for tangential and frontal section series. Two tangential and four longitudinal section series of the whole length of the retina were cut from the dorsal and medial regions of the eye. The 1-μm-thick section series were placed on microscope slides and flattened on a 60°C hot-plate. These sections were then stained with toluidine blue and viewed under a Zeiss photo-microscope. Colour pictures were taken with Fuji AS100 film and the negatives scanned in digitally for further analysis.

The corneal micro-surface and facets were studied in the three bee species using a JEOL JSM-5600 scanning electron microscope. Furthermore, corneas of two M. genalis females and two L. leucozonium females were painted with a thin layer of nail polish to produce cornea replicas. After letting them dry for a few hours the cornea replicas could be carefully and completely removed by cutting along the corneal border of the eye with a microscalpel (Ribi 1978). Several incisions were necessary to flatten the retina on the microscope slide. The replicas were coated with coverslips, sealed with melted wax, and viewed under a Zeiss Axiophot microscope. Each replica was photographed with an Olympus DP 50 digital camera and analysed on a Compaq EvoW4000 PC using Adobe Photoshop 7.0. The largest possible circle was then fitted into the hexagonal facets across different eye regions to calculate the respective facet diameters (Praagh et al. 1980). Measurements of facet diameter were also made on M. genalis females mounted in a Leitz goniometer and viewed with a macroscope (for full details see Land and Eckert 1985; Rutowski and Warrant 2002).


In order to measure the focal length f of M. genalis and the diurnal bees, a small slice of the frontomedial cornea was cut and placed in distilled water. All pigments were removed with a fine brush and the corneal piece was then placed, with its external surface directed outwards, on a droplet of water located on a coverslip. To prevent evaporation we constructed a chamber by mounting an O-ring on a microscope slide. We then covered the upper side of the ring with a thin layer of Vaseline. The coverslip was placed upside-down (with the corneal piece directed downwards) onto the O-ring. The corneal piece was viewed with the ×40 objective of a Nikon Labophot-2 microscope. A linear, square grating, with a spatial object wavelength λo of 6.7 mm, was placed on the light source in the microscope foot (Nilsson et al. 1988). The square grating was viewed through the corneal piece and the position of best focus of the cornea’s image determined. The spatial image wavelength λi of the image (in mm) was used with the object distance s (the distance between the corneal piece and the object) to calculate the focal length f of the corneal lens:
$$f = s{\left( {\lambda _{i} \lambda _{o} } \right)}$$

The nodal point N was situated at a distance of one focal length f in front of the image and its position within the corneal lens was determined by measuring the distance from the image to the back of the lens and then subtracting this distance from the focal length.


Anatomical aspects of the retina in nocturnal bees

The hymenopteran visual system consists of an optical apparatus and a peripheral retina contained within the compound eye and three optic ganglia—the lamina, medulla and lobula in the optic lobe of the brain (Fig. 3). There are two chiasmas between the optic ganglia—the outer chiasma between the lamina and medulla and the inner chiasma between the medulla and the lobula. The third optic ganglion, the lobula, is not divided into two separate ganglia (the lobula and the lobula plate), as found in dipterans.
Fig. 3

The visual system of the worker honeybee featuring a peripheral retina (R) and three optic ganglia—the lamina (L), medulla (M) and lobula (Lo). The lamina and medulla are connected by the outer chiasma (OCh), whereas the inner chiasma (ICh) is situated between the medulla and the lobula. The retina consists of a high number of repeating visual units, the ommatidia (O). Each ommatidium contains a dioptric apparatus—the corneal facet (C) and crystalline cone (CC)—and a light-sensitive fused rhabdom (Rh), made up of rhabdomeres from nine retinula cells. The retinula cells end in six short visual fibres (svf) and three long visual fibres (lvf). These reach through the basement membrane (BM) and the cell body layer (CBL) in order to synapse with first-order interneurons (L-f) in the lamina. (Adapted from Ribi 1987). Scale bar 200 μm

Females of the nocturnal halictid bee M. genalis feature large eyes relative to their body size compared to diurnal bees (Jander and Jander 2002). The compound eyes of two M. genalis females consisted of 4,883 and 4,876 facets, respectively, with facet diameter ranging from 28 μm in the dorsal rim area of the eye to its maximum value of 36 μm in the frontomedial region (Figs. 4, 8c). The outer surfaces of the facets are covered with corneal protuberances (Miller 1981; Miller et al. 1966) (Fig. 8a).
Fig. 4

Facet diameters in female M. genalis shown on a spherical plot derived from optical measurements. The three-dimensional sphere contains the visual field of the bee’s left eye, and shows lines of latitude and longitude at 10° intervals (thin lines), and the boundary of the eye’s visual field (bold line). Facet diameters within the eye are shown as isolines (D dorsal, V ventral, A anterior, L lateral)

The ommatidium, the optical unit of compound eyes, features a dioptric apparatus composed of the corneal facet and the crystalline cone, and nine retinula cells forming the fused rhabdom (Fig. 5). In M. genalis females, the lamellated cornea exhibits an enormous thickness of 102 μm, the distal most 9 μm building the surface layer (Fig. 5j). The outer corneal curvature of the eye is flattened (radius of curvature r1≈∞), whereas the corneal underside facing the crystalline cone is highly convex (r2=−22 μm) (Fig. 9c). Furthermore, the concave apex of the crystalline cone is in tight connection with the convex cornea, and consequently features the same radius of curvature of −22 μm as the corneal underside of the cornea (Figs. 5j, 9c).
Fig. 5a–k

Fine structure of the retina in M. genalis females. Semi-schematic drawing of an ommatidium (centre) with photographs of transverse (a–i) and longitudinal (j, k) sections at representative levels. The dioptric apparatus of the ommatidium consists of the prominent corneal facet (C) (j) and the crystalline cone (CC) (a, b). Note the flattened outer, and highly convex, inner, corneal curvature as well as its tight connection to the crystalline cone (j). The crystalline cone is surrounded by two primary pigment cells (PPC) (b, j) and 19 secondary pigment cells (SPC), their cell bodies located distally (a). The SPCs optically insulate the ommatidia from each other throughout the whole length of the ommatidia. Nine retinula cells (RC) build up the light-sensitive fused rhabdom (Rh) and have their cell bodies situated in three distinct layers (arrows in e–g). The ninth retinula cell (RC 9) only appears in the proximal third of the ommatidium (k) and has its cell body in layer three (g). Four crystalline cone extensions (CCE) reach from the crystalline cone to the proximal retina. Their crystalline cone extension pigments (CCEP) surround the most distal tip of the rhabdom (inside the dashed ellipse in h) before the retinula cells pass the basement membrane (BM) in the form of receptor axon bundles (RCA) (k). A ring of retinula cell pigments tightly surrounds the rhabdom in the light-adapted state (arrow in d and j). Transverse and longitudinal 1-µm sections, toluidine-blue stained. Scale bars for a–i (shown in i) and for j, k (shown in k) 25 μm. Schematic pigments are not to scale

As in other hymenopterans, the crystalline cone is of the eucone type, secreted by four Semper cells, and surrounded by two primary pigment cells (Fig. 5b). The entire crystalline cone measures 48 μm in base-to-apex length and its distal base is 25 μm in diameter (Fig. 5a, j). The apex of the crystalline cone has a width of 8 μm. The crystalline cone sends four extensions, between each second retinula cell, along the rhabdom down to the basement membrane. These crystalline cone extensions expand at their endings and are packed with pigments that surround the proximal tip of the rhabdom (Fig. 5h). Each ommatidium is surrounded by 19 secondary pigment cells along its entire length, from the cornea to the basement membrane (Fig. 5a, j). The nine retinula cells in each ommatidium contribute their rhabdomeres to a light-absorbing structure, the fused rhabdom. The distal end of the rhabdom measures 8 μm across (Fig. 5c–f) and decreases in width to 5–6 μm in the proximal third of the ommatidium (Fig. 5g). Three distinct retinula cell body layers are visible. The first layer, situated 90 μm below the crystalline cone apex, features the cell bodies of the retinula cells 1, 3, 5 and 7 (Fig. 6a). The second layer, with the cell bodies of the retinula cells 2, 4, 6 and 8, can be found 150 μm below the cone apex (Fig. 6b) and the third layer, with the cell bodies of the ninth retinula cell, 240 μm below the cone apex (Fig. 6c: for classification see Ribi 1987). The ninth cell is, apart from in the dorsal rim area of the eye, only present in the proximal third of the ommatidium, where it replaces the rhabdomere of retinula cell 4 (Fig. 6c). In M. genalis females the entire ommatidium in the medial part of the eye is 500 μm long, with a rhabdom length of 350 μm, and measures 30 μm in diameter.
Fig. 6a–d

Schematic drawings of transverse sections through the ommatidium at the three retinula cell body (RCB) layers and d at the level of the pseudocartridge formed by retinula cell axons (RCA). a The first layer features the RCBs of the retinula cells 1, 3, 5 and 7; b the second layer the RCBs of the retinula cells 2, 4, 6 and 8; and c the third layer the RCB of retinula cell 9, which replaces the rhabdomere of retinula cell 4. The small circle represents the axon of retinula cell 4. d Within the pseudocartridge one can observe the large RCAs 3 and 7, the intermediate RCAs 2, 4, 6 and 8—forming together the six short visual fibres, as well as the small RCAs 1, 5 and 9—representing the long visual fibres (Rh rhabdom). Scale bars 15 μm (a–c, shown in c), 5 μm (d)

In the dark-adapted retina pigment migration could not be observed in the primary and secondary pigment cells. Following dark adaptation, the rhabdoms retained the shape and position they possess in bright light and no clear zone was created between the crystalline cones and the rhabdoms (as developed in the dark-adapted eyes of some tiger beetles: Brännström 1999). However, the numerous retinula cell pigments which tightly ensheath the rhabdom in the light-adapted state (Fig. 7a) disperse into the cytoplasm when dark adapted (Fig. 7b). Furthermore, intracellular cisternae, also called palisades, surround the rhabdom in the light-adapted state but reduce significantly in size during dark adaptation (Fig. 7a, asterisk).
Fig. 7a, b

The effect of a light and b dark adaptation of the retina in M. genalis females. No pigment migration in the primary pigment cells (PPC) or secondary pigment cells (SPC) is visible in this apposition eye. The schematic longitudinal drawings and transverse histological sections through the crystalline cone (CC) and the distal rhabdom (Rh) show the effect of the longitudinal pupil. a In the light-adapted state retinula cell pigments tightly ensheath the rhabdom (small arrow) and cisternae surrounding the rhabdom are visible (asterisk). b During dark adaptation the retinula cell pigments migrate away from the rhabdom and disperse into the cytoplasm (large arrow). The cisternae reduce significantly in size. The majority of the thick cornea (C) is omitted. Schematic pigments are not to scale. Scale bar 25 μm (a, b, shown in a)

The nine retinula cells of each ommatidium give rise to retinula cell axons: six short visual fibres (svf) and three long visual fibres (lvf), ending in a straight projection to the lamina or the medulla, respectively. These retinula cell axons penetrate the basement membrane in a distinct axon bundle (Fig. 5h, i), forming pseudocartridges (Fig. 6d). On the way to the lamina, the first optic ganglion, these pseudocartridges pass through the fenestrated layer, a region below the basement membrane containing trachea and the cell bodies of lamina cells. After reaching the lamina, a network of tangential fibres keeps their hexagonal arrangement according to their assembly in the retina. Retinula cells originating in the overlying ommatidium target the same cartridge in the lamina.

Distal to the basement membrane a thick layer of pigments is present (Fig. 5i, k), which can be found within the secondary pigment cells as well as in the crystalline cone extensions that envelop the most proximal tip of the rhabdom (Fig. 5h). Consequently, no tapetum is visible in the retina of M. genalis females.

Anatomical aspects of the retina in diurnal bees

Anatomical structures within the compound eye were also studied in the more closely related diurnal halictid bee L. leucozonium and the well-studied worker honeybee A. mellifera in order to investigate the effect of taxonomy on eye design. The compound eyes of two L. leucozonium females contained 3,381 and 3,461 facets, respectively, with a maximal facet diameter of 20 μm in the frontomedial region of the eye. In the worker honeybee a maximal facet size of 20 μm was also measured (Fig. 8d). Values for facet numbers (4,752 facets: Table 2) originate from Praagh et al. (1980). Comparing 15 bee species of varying size and taxonomy, Jander and Jander (2002) also discovered a tight relationship between body size and the number of ommatidia for both nocturnal and diurnal bees. Bees of equal size have approximately the same number of ommatidia. As A. mellifera and M. genalis have a similar body size and very similar facet numbers (see Table 2), any difference in ocular morphology in these species cannot be an effect of body size. L. leucozonium, in contrast, is a much smaller bee and the body size difference is well supported by nearly 1,500 fewer facets. As in M. genalis, corneal protuberances are also found on the outer surface of honeybee facets, but these are restricted to the facet borders and show a distinct separation line towards a smooth central facet surface (Fig. 8b). In the diurnal halictid, weakly developed and coalesced protuberances were noted.
Fig. 8a–h

Comparisons of eye structure in M. genalis females and the worker honeybee A. mellifera. a The corneal outer surface features a distinct corneal nipple-array in M. genalis. b In the worker honeybee this corneal nipple-structure is restricted to the facet borders and shows a distinct separation line (arrow) towards the smooth facet surface. c, d The corneal facets, here shown in the frontomedial eye region, are approximately twice as wide in M. genalis (c) as in the worker honeybee (d). The black squares mark the regions shown in a, b. The thick flat cornea of M. genalis (e) compared to the thin, convex cornea of the worker honeybee (f). The diameter of the rhabdom in M. genalis (g) is considerably larger than in the honeybee (h). a–d Scanning electron micrographs. e, f Longitudinal and g, h tangential 1-μm sections, toluidine-blue stained. Scale bars 1 μm (a, b), 25 μm (c–h)

Table 2

Anatomical and optical parameters in the eye of the nocturnal halictid M. genalis, the diurnal halictid L. leucozonium and the worker honeybee A. mellifera. All values are taken from the frontomedial region of the eye in both species




M. genalis

L. leucozonium

A. mellifera

Number of facets per eye






Maximal corneal facet diameter






Corneal thickness






Crystalline cone length






Distal rhabdom diameter






Rhabdom length






Ommatidial length






Focal length






Absorption coefficient of the rhabdom












Optical sensitivity


μm2 sr




Theoretical acceptance angle






Thorax length






aAveraged from two bees; see “Results”

bFrom Praagh et al. (1980)

cFrom Bruno et al. (1977)

Regarding the dioptric apparatus, the diurnal halictid bee has a 41-μm-thick cornea with a convex outer corneal curvature (r1=26 μm) and a convex inner corneal curvature (r2=−24 μm). The slightly concave base of the crystalline cone (r3=69 μm) is in tight connection with the cornea (Fig. 9b). The cornea of the worker honeybee has a thickness of 26 μm, a convex outer corneal curvature (r1=28 μm) and a convex inner corneal curvature (r2=−34 μm). The base of the crystalline cone is strongly convex (r3=69 μm) and touches the cornea only slightly (Fig. 9a). The crystalline cone has a base-to-apex length of 35 μm and 55 μm, and an apex width of 1.6 μm and 2 μm, in L. leucozonium and A. mellifera, respectively (Figs. 9a, b, 8f). The rhabdom has a width of 1.6 μm in L. leucozonium and 2 μm in A. mellifera (Fig. 8h), and a length of 220 μm and 320 μm, respectively (Table 2). The entire ommatidium is 300 μm in length and 14 μm in diameter in the diurnal halictid compared to 400 μm and 17 μm in the worker honeybee (Table 2). No other structural difference could be found in the retina of the three bee species studied.
Fig. 9

Schematic drawing of the dioptric apparatus and rhabdom in A. mellifera (a), L. leucozomium (b) and M. genalis (c). The respective radii of curvature (r) and lengths are given for the cornea (grey), the crystalline cone (white), and the diameter of the rhabdom (black). Note the wide rhabdom in M. genalis, and the prominent cornea with its flattened outer curvature

Optical aspects of the nocturnal and diurnal bee retina

Clear images of grating patterns were formed behind the corneal lens (Fig. 10). In M. genalis females, the posterior nodal point N is located 56 μm beneath the corneal surface. With a focal length f of 97±5 μm (±SD, n=6), measured from a mean facet diameter of 35±2 μm, the focal plane is located at the distal tip of the rhabdom. The same measurements were performed in the honeybee (f=66±5 μm; ±SD, n=6) and L. leucozonium (f=57±5 μm; ±SD, n=5). Knowing the focal length f and the facet diameter D (Table 2), we can now calculate the F-number F of the ommatidium:
$$F = f/D$$
Fig. 10

A linear, square grating (λo=6.7 mm) viewed at its plane of best focus behind corneal pieces of a M. genalis and b the worker honeybee A. mellifera. Scale bar 10 μm

The F-number predicts the light gathering capacity of an eye, with lower F-values indicating a more sensitive eye. However, in some eyes with F<2.0, such as diurnal arthropods with heavy screening pigmentation between rhabdoms (Warrant and McIntyre 1991; Stavenga 2003b), sensitivity is not improved by having a lower F-number (Stavenga 2003b). The apposition eye of M. genalis females has an F-number of 2.7 compared to 3.3 in the worker honeybee and 2.9 in the diurnal halictid (Table 2). A better measure of sensitivity is the optical sensitivity S of the eye, which defines the amount of light energy that is absorbed by a photoreceptor when it views an extended source of light (Kirschfeld 1974; Land 1981). Equation 3 is the Kirschfeld-Land sensitivity equation, with a term [kl/(2.3+kl)] that assumes that an extended source of white light is being viewed (Warrant and Nilsson 1998):
$$S = {\left( {\pi /4} \right)}^{2} D^{2} {\left( {d/f} \right)}^{2} {\left[ {kl/{\left( {{\text{2}}{\text{.3 + }}kl} \right)}} \right]}$$

The optical sensitivity of a compound eye is greater if it views an extended scene with facets of greater area [(π/4)D2 μm2], with photoreceptors that view larger solid angles of visual space [(π/4)(d/f)2 sr, from the definition of solid angle] and with longer (l μm) and more absorptive (k μm−1) photoreceptors. Recent work (Stavenga 2003b) has shown that Eq. 3 has limitations in eyes operating according to the principles of waveguide optics, and in some eyes with F-numbers lower than 2. Therefore, some caution needs to be applied when interpreting sensitivity values obtained in the wave-guiding eyes of the diurnal bees. The eye of M. genalis, with 8-μm-wide rhabdoms, can clearly be approximated by geometric optics, and with F=2.7, Eq. 3 can be applied with safety.

With known values of facet diameter D, rhabdom diameter d, rhabdom length l and the absorption coefficient of the rhabdom k (Table 2), the optical sensitivity of M. genalis is found to be 2.7 μm2 sr, 27-fold higher than for both diurnal bee species (which had optical sensitivities of around 0.1 μm2 sr). Moreover, we can theoretically predict the eye’s spatial resolving power, which describes the ability of an eye to resolve details in space. Generally, a wider receptive field (or acceptance angle) can capture more light at the cost of lower spatial resolution. Using values given in Table 2, one can theoretically approximate the dark-adapted acceptance angle Δρ of a photoreceptor using the following expression (Stavenga 2003a, 2003b):
$$\Delta p \cong {\left( {d/f} \right)}$$

After conversion from radians to degrees, M. genalis females have a theoretical acceptance angle Δρ of 4.7°, which is nearly 3 times wider than found in worker honeybees (Δρ=1.7°) and the diurnal halictid bee (Δρ=1.6°).


The retina of the nocturnal bee Megalopta genalis has the same general anatomical organisation as the typical diurnal bee apposition eye of the worker honeybee Apis mellifera and the diurnal halictid bee Lasioglossum leucozonium. However, there are a number of specific differences that have most likely evolved for Megalopta’s purely nocturnal activity. The most distinct nocturnal adaptations are (1) the 1.8 times larger ommatidial facets and (2) the 4–5 times wider rhabdom (Fig. 8). A nocturnal transition to a superposition eye design, as found in some tiger beetles, was not found. Furthermore, having determined the focal lengths, we can, with limitations, discuss Megalopta’s nocturnal optical adaptations in terms of sensitivity and spatial resolution.

The advantages of being nocturnal

A nocturnal lifestyle has two major advantages: lower competition for scarce food resources and fewer predators (Roubik 1989; Warrant et al. 1996). With many flowers opening at night, the purely nocturnal bee M. genalis has the great advantage of being able to exploit abundant food sources when diurnal bees are inactive. Pseudobombax septanatum (Bombaceae), a large canopy tree common on Barro Colorado Island in Panama, has been reported as being a possible pollen host of M. genalis (Roulston 1997). These plants produce large white flowers that open at sunset and are highly attractive to diurnal bees in the mornings (Roulston 1997). A similar beneficial effect of lower competition has also been reported for matinal, as well as crepuscular, bees (Kerfoot 1967; Linsley and Cazier 1970; Shelly et al. 1993). As usual predators of large bees, one can consider birds, toads, dragonflies, spiders, robber flies, lizards and mantids. Many of these predators are active during the day and most likely contribute little to the mortality of bees active at night (Wolda and Roubik 1986). Furthermore, nest parasites are known to mainly use visual cues and only enter the nests when the host is absent (Wolda and Roubik 1986).

An apposition eye with nocturnal adaptations

According to Kerfoot (1967) and Engel (2002), the particular features that distinguish bees with nocturnal activity are enlarged eyes and ocelli, a large body size, and pale body pigmentations. Jander and Jander (2002) have shown that the surface area of the eyes is 1.8 times larger in nocturnal bees (including Megalopta) than in diurnal bees. These results fit well with our 1.8-fold increase in facet diameter found in the nocturnal bee compared to the diurnal bees (Fig. 8c, d). A larger eye with larger facet diameters will, under the same light conditions, accept more light than a smaller eye with smaller facets, leading to greater sensitivity (Warrant and McIntyre 1993). We will return to these larger facets later.

In addition, we could observe that the ratio of corneal thickness to crystalline cone length is 1.8 and 4.5 times larger in the nocturnal bee compared to the diurnal halictid and the worker honeybee, respectively. As the mechanical strain on the cornea should be similar in both halictid bees (M. genalis burrows out a nest in a wooden stick and L. leucozonium burrows in the ground), this pronounced difference in thickness is hard to understand. Furthermore, the outer corneal surface is distinctly flat in the nocturnal bee compared to the clearly convex cornea of diurnal bees (Fig. 9). This may possibly indicate an optical function for the thick flat cornea. As nocturnal mosquitoes possess nearly hemispherical lenses (Land et al. 1997, 1999), the thick and flat cornea of Megalopta may not be an adaptation to dim light. Nevertheless, the focussing power of the nocturnal bee cornea could be due to the presence of an inhomogeneous refractive index, creating a lens cylinder that might gather light more effectively than the homogeneous, thin and curved cornea of mosquitoes. We can speculate about the possible presence of a corneal lens cylinder by applying the thick lens formula (Eq. 5: Land et al. 1999):
$$1/f = {\left( {n_{1} - n_{0} } \right)}/r_{1} + {\left( {n_{2} - n_{1} } \right)}/r_{2} - s{\left( {n_{1} - n_{0} } \right)}{\left( {n_{2} - n_{1} } \right)}/n_{1} r_{1} r_{2} $$
where n are refractive indices, r are the radii of curvature of corneal surfaces, and s is the axial separation of the corneal curvature centres. If all the refractive power comes from a homogeneous lens, then the focal length f calculated with the thick lens formula would need to be in accordance with our experimental focal length measurements for Megalopta (f=97±5 μm, see “Results”, “Optical aspects of the nocturnal and diurnal bee retina”).

As mentioned in the “Results”, the outer corneal curvature of the eye is flat and thus has an infinite radius of curvature (r1≈∞), whereas the inner corneal surface is convex, with a curvature of r2=−22 μm. Assuming realistic refractive indices for the cornea (n1=1.45) and the crystalline cone (n2=1.34) (Stavenga 2003b), defining n0=1 as the refractive index of the outer medium (air), noting that the axial separation of the corneal curvature centres (s) is 48 μm and using the radii of curvature of the corneal lens given above, we can calculate a focal length of 200 μm. This theoretical value is twice as large as the experimentally measured focal length, indicating that an inhomogeneous lens might be present in order to focus light onto the tip of the rhabdom. To clarify this, exact refractive index measurements of the corneal lens cylinder are needed, and these will be the subject of further investigations.

The corneal outer surface of M. genalis features distinctly noticeable corneal protuberances or nipple-array (Fig. 8a). In comparison, worker honeybee facets are smooth (Fig. 8b), and protuberances are only formed at the facet borders; a clear separation line towards the smooth facet surface can be distinguished (Fig. 8b). Regarding the function of this corneal nipple-array, it has been shown that for moths in particular these are likely to function as an anti-reflection coating when they reach a height of 50 nm or greater (Miller 1981). Corneal nipples not only decrease reflections from the cornea, effectively camouflaging the animal, they also increase the intensity of a focused image by as much as 4%. Corneal protuberances of less than 50 nm height show a tendency towards coalescence, as seen on the corneas of flies (Miller 1981). A similar coalescence could be observed on the facets of L. leucozonium. In this case, the protuberances are not thought to have an anti-reflection effect. Although the exact height of the corneal nipples of M. genalis still needs to be determined, we present the first indications that the corneal protuberances in M. genalis could be involved in increasing light flux into the eye, an important factor for a photon starved animal.

The organisation of the nine retinula cells in M. genalis is very similar to the retinal structure in the diurnal bees. No anatomical differences based on taxonomy were found. As L. leucozonium is of much smaller body size, rhabdom size and volume are only compared between the equally sized A. mellifera and M. genalis. In M. genalis the distal tip of the fused rhabdom, measured in the frontomedial region of the eye, is a remarkable 4 times wider than in the worker honeybee (Fig. 8e–h). This wider and also longer rhabdom (for values see Table 2) leads to a 17.5-fold greater photoreceptive volume in M. genalis, representing a significant improvement in sensitivity. Apart from the effect of an increased rhabdom width, a wide pupil during dark adaptation also leads to a greater sensitivity.

During dark adaptation, pigment migration in the nocturnal bee was only seen inside the retinula cells (Fig. 7b). No retraction of the rhabdoms and secondary pigments—necessary for the formation of a clear zone—was visible. This disproves our hypothesis that Megalopta might possess a functional superposition eye at night. The retinula cell pigment migration (Fig. 7b) represents a so-called “longitudinal pupil” and this type of pupil is also found in day-active insects such as flies, butterflies (Stavenga 1979) and the honeybee (Kolb and Autrum 1972, 1974; Stavenga and Kuiper 1977; Autrum 1981). During light adaptation, the retinula cell pigments tightly surround the rhabdom, absorbing much of the propagating light and leading to higher spatial acuity (Fig. 7a). During dark adaptation the pigments migrate away from the rhabdom, resulting in higher sensitivity, but at the cost of spatial resolution (Fig. 7b) (Land and Osorio 1990). Furthermore, in the light-adapted state one can observe large cisternae, also known as palisades, surrounding the rhabdom. In locusts, where they are prominent in the dark-adapted retina, these palisades are thought to improve the light-guiding properties of the rhabdoms (Horridge and Barnard 1965). This is probably not the case in M. genalis, as the palisades almost disappear during dark adaptation.

An adaptation to dim light that is particularly common in deep sea and nocturnal animals is a mirror-like structure called the tapetum. Tapeta are mostly developed at the proximal base of the retina and reflect the light back into the rhabdoms, giving the retina a second chance to absorb light (Land and Nilsson 2002). In many insects, tapeta are constructed of modified tracheal structures ensheathing the proximal end of the rhabdom and are capable of producing a characteristic eye glow (Ribi 1980). Due to its nocturnal lifestyle, one might expect M. genalis to have a tapetum at the proximal base of the retina. Indeed, by viewing the eye from the direction of illumination we were able to observe a bright eye glow that could easily arise from a tapetum. However, the pigmented enlargements of the crystalline cone extensions optically isolate the most proximal end of the rhabdom (Fig. 5h) and secondary pigment cells construct a thick layer of pigments above the basement membrane absorbing all stray light (Fig. 5i, k). These features can also be observed in the diurnal bees studied, which have neither a tapetum nor an eye glow. From this we conclude that no tapetum exists in the retina of M. genalis females. However, the cause of the bright eye glow is still unclear and is currently being investigated.

Optical adaptations to dim light

Insects living in dim habitats generally possess compound eyes of lower F-number, indicating greater sensitivity to extended sources of light (Warrant and McIntyre 1991). The F-numbers of nocturnal insects with superposition eyes usually range between 0.5 and 1.2 (Warrant and McIntyre 1993). Despite its nocturnal activity, the high F-number of M. genalis indicates that its eye is more typical of diurnal apposition eyes, which generally have F-numbers higher than 2.1 (Warrant and McIntyre 1993). There are, however, apposition eyes with exceptionally low F-number, such as those of the deep sea isopod Cirolana borealis with an F-number of 1.0 (Nilsson and Nilsson 1981), implying that dim habitats can lead to specialised apposition eyes with high sensitivity. Megalopta’s high F-number suggests that its eye is not only adapted to high sensitivity, but also maintains a construction that supports reasonable spatial resolution, no doubt necessary for its difficult orientation tasks at night. An F-number of 2 or greater also ensures that in eyes with wide rhabdoms all of the light that reaches the rhabdom remains trapped within it (Stavenga 2003b). With an F-number of 2, light strikes the rhabdom with an angle of incidence of 10°, the maximum angle that ensures total internal reflection within the rhabdom (Warrant and McIntyre 1993; Stavenga 2003b). In eyes with low F-numbers (i.e. with a large aperture relative to the focal length), a large amount of light can escape and spread to neighbouring rhabdoms, leading to a degradation of image quality and a loss in spatial resolving power. The nocturnal mosquito Anopheles gambiae, with thin near-hemispherical lenses and an F-number of 0.78, avoids this problem by possessing unique conical rhabdoms that are able to trap light from much wider angles (Land et al. 1997, 1999). Consequently, with an F-number of 2.7, M. genalis can capture all of the incident light within its wide rhabdoms.

Despite the fact that a higher F-number indicates the importance of spatial resolution for M. genalis, its apposition eyes must of course be sensitive enough to see at night. This is reflected in the optical sensitivity S of the eye, a quantity that takes account of other factors when considering sensitivity (Eq. 3). In M. genalis, S is considerably higher than in the diurnal bees. In the nocturnal bee, the photoreceptors absorb 27 times more light when viewing an extended light source, a finding confirmed in another study that relies on electrophysiological measurements (Warrant et al., in preparation). With an optical sensitivity of 2.69 μm2 sr, the nocturnal apposition eye of M. genalis is nevertheless considerably less sensitive than that of the nocturnal superposition eye of the elephant hawkmoth Deilephila elphenor, which has an optical sensitivity of 69 μm2 sr (Warrant, unpublished data). Theoretical approximations of the acceptance angle Δρ resulted in a value that is nearly 3 times larger in M. genalis (Δρ=4.7°) than in the diurnal bees (for values see Table 2). This greater acceptance angle in M. genalis indicates higher sensitivity at the cost of poorer resolution. This approximate value is relatively close to the value measured experimentally in the frontal part of Megalopta’s eye (Δρ=5.6°; Warrant et al., in preparation).

Considering the fact that M. genalis is constrained by its unsuitable apposition eyes, it performs astonishing orientation and homing tasks at night. One might imagine that they have exploited all possible retinal and optical adaptations to find the ideal trade-off between maximised sensitivity and spatial resolution. In fact, the adaptations we have described for this nocturnal bee are extreme and unique for an insect apposition eye, as one might expect for an animal that must cope with up to a hundred million times less light than a diurnal bee. Although a 27 times higher sensitivity to light in M. genalis is a remarkable improvement, this alone cannot account for their ability to see at light levels 8 log units dimmer. Therefore, we must assume that higher neuronal mechanisms, increasing sensitivity via summation of photons in space and time, are involved (Warrant 1999). Temporal summation, which takes place in the retina and in the brain, involves the summation of photons over a longer time period, similar to a camera with a long exposure time. However, its cost in making vision slower could be detrimental for fast moving animals like bees. For spatial summation, the eyes could sum signals from large groups of ommatidia, in order to improve their photon catch (Warrant et al. 1996; Warrant 1999). However, this only occurs at the cost of courser spatial resolution. Our current idea is that spatial summation takes place very early in visual processing, and therefore most probably resides in the lamina, the first optic ganglion in the optic lobe of the brain. Recent findings of wide laterally branching first-order interneurons in Megalopta’s lamina (Greiner et al., in preparation) indicate that spatial summation may play a major role in the impressive visual abilities of this nocturnal bee.


We would like to thank Almut Kelber, Doekele Stavenga, Mark Holdstock and the two anonymous referees for critically reading the manuscript, Ladina Ribi for the taxonomic drawing of Megalopta, Rita Wallén and Carin Rasmussen for histological and technical support, Rikard Frederiksen for help with the corneal replicas, Mikael Sörenson and Jan Tengö for taxonomical help, as well as Victor Gonzales and Sara Juhl for help with fieldwork. We thank William T. Wcislo and the staff of the Smithsonian Tropical Research Institute, Panama City, for their help and the Autoridad Nacional del Ambiente of the Republic of Panama for permission to export bees. The histological work was partly done at the Center for Visual Sciences, Research School of Biological Science, Australian National University, Canberra.

Copyright information

© Springer-Verlag 2004