- Email: [email protected]
Binocular depth vision in arthropods
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
E ome insects demonstrate their capacity for accurate depth vision by leaping on to objects or by catching prey. While it has been known for several decades that locusts can exploit the monocular depth cue of motion parallax to judge the distance of an object 1, only recently has it become clear that insects can also use binocular cues. Earlier evidence for binocular depth vision was circumstantial, based principally on the observations that in many insects the visual fields of the two eyes show substantial overlap (e.g. Fig. 1) and that monocularly blinded predatory insects have difficulty in catching prey. Indeed, sceptics doubted that insects possessed the elaborate neural connections needed for binocular vision. Experiments by Rossel 2 on the praying mantis have now settled this question. He has shown that mantids must use binocular cues to extract depth and suggests a simple way in which they might do so without straining a scientist's credulity or an insect's nervous system. The mantid's method of catching prey makes it an ideal animal for studies of depth vision, as Maldonado 3 realized in the 1960s. It sits in wait to ambush a passing insect. When one comes into view, the mantid turns its head to place the image of the insect on its fovea. Should the insect then wander within range, the mantid shoots out its foreleg to grasp the insect. The bulk of the mantid is immobile during this sequence, and it continues to catch prey if it is anchored firmly to a holder and base-out prisms are placed in front of its eyes (Fig. 2). Since the animal's eyes are fixed in its head, these prisms increase the horizontal binocular disparity of the target image on the two retinae. Mantids viewing prey through base-out prisms strike accurately at its apparent binocular position2; the prey is seen to be nearer than it
really is. The precision with which apparent distance is controlled by the power of the prisms suggests that at close range the mantid relies entirely upon binocular disparity to gauge depth. The range over which binocular cues can operate depends upon the distance between the eyes and the accuracy with which the position of a target image can be measured. The interocular distance in the mantid, Tenodera, is about 4 ram. Work by Maldonado et al. 4 indicates that binocular disparities might be used to discriminate an a~ian predator seen at 45 cm from one seen at 60 cm. These figures suggest that mantids can assess disparity to within about 0.13 deg, which in turn indicates a somewhat finer resolution than that suggested by the 0.6 deg difference between the lines of sight of adjacent foveal ommatidia 2. What neural mechanism mediates binocular vision in insects? One possibility is that there are more binocular connections than have hitherto been supposed. In flies, signs have been found of a retinotopic projection between the lobulae of the two optic lobes 5, although the extent and function of such connections in mantids remain to be worked out. Since the mantid only strikes when the two xget images fall within the 30 deg circular foveas, binocular connections need only exist between foveal regions of the lobula. However, Rossel 2 has proposed an attractive mechanism that does not require any local binocular interactions. His suggestion is that each optic lobe measures the vertical and horizontal positions of the target image with respect to the centre of the fovea of the ipsilateral eye. For simplicity, suppose that there is a convenient Cartesian representation with four
TINS January- Vol.lO No. 1 [103]
~)1987, EJsevierSciencePublishersB.V.Amsterdam 0378-5912/87/$0200
S
Binocular depth vision in arthropods output neurons from each optic lobe, one encoding the angular horizontal distance to the left of a reference point in the fovea, one the distance to its right, one the vertical distance above it and the last the vertical distance below. An average of the horizontal signals coming from the two optic lobes would then provide a measure of the azimuth of the target with respect to the head, while the difference between the horizontal signals would carry information about target distance. Direct evidence for a scheme of this type is so far only available for the assessment of target direction. If vision is restricted to one eye, the mantid makes a saccade-like
T. 5. Colle~
Schoolof Biological Sciences, The Universityof Sussex, Falmer,Brighton, SussexBN19QG, UK.
Fig. 1. Tenodera australasiae. The binocularfield of this mantid extends2400 verticallyand 350 ho&,ontally. (Kindly provided by S. Rossel.)
//
Fig. 2. Diagram of mantid viewing a fly through base-out prisms. (Adapted from Ref. 10.) 1
Fig. 3. Someintriguing eye-stalksin arthropods. (A) Fema/eCyrtodiopsis white (eyeseparation 6.5 mm). (B) Macropthalmus setosus(eye-
separation, 7.2 mm). (A) Kindlyprovidedby I. de la Motte, (B) by J.Zeil,and adapted from Ref.9.
turn of its head to fixate the target with the viewing fovea. However, when both eyes see the target, the saccadic turn is a weighted average of the saccades driven separately by the left and right eyes 2. One apparent limitation of this model is that there would be confusion should several prey be viewed simultaneously. There is no guarantee that the two eyes would decide to fixate the same one. Intriguingly though, Cloarec6 has recently found that the predatory water stick-insect, Ranatra, has no problems with two prey arriving at once. It deftly catches one with each foreleg. Insects may have methods for sorting out which images belong together. Two prey would not be equally attractive; for instance, one may move more and so 'capture the attention' of both optic lobes. Also, since the vertical disparities between the two eyes will normally be zero, two images with the same elevation will very likely represent the same prey in object-space. Two interesting studies have taken a comparative approach to problems of depth vision. The first deals with a puzzling curiosity of the entomological world, the stalkeyed fly (family Diopsidae), some species of which seem to have gone to exuberant lengths to increase their interocular separation (Fig. 3A). There have been many speculations that the stalks are an adaptation for enhanced binocular vision. Burkhardt and de la Motte 7'~ have tackled this question by examining the behaviour and dimensions of the eye-stalks of nine diopsid species found in Malayan rain forest. In some species, males and females have modest eye-stalks of equal length and Burkhardt and de la Motte conclude that initial selection for wide-set eyes was probably 2
driven by the problems of relying on vision to find mates in tropical rainforest, where conspecifics are rarely encountered and species diversity is high. In dimorphic species the male stalks are hugely hypertrophied. Here it seems sexual selection has taken over and acted primarily to make the stalks conspicuous. Among a group of different sized males of one species, the length of the eye-stalk increases so steeply in relation to body length that it acts not only as an advertisement but also a sensitive index of an individual male's size. Features of the social behaviour of both males and females depend on accurate assessment of a male's size. Dimorphic species of the genus Megalobopsaggregate at dusk on hanging plant-fronds in sleeping groups of about a dozen with one large male, ten or so females and perhaps one or two male runts. The male is usually the first to arrive and the females choose the frond with the largest male on it. Males arriving later are either chased off, or take over the harem, or are so small that they can hide safely among the females. Competition between males is settled by size, with fights seen only among similar-sized males. Otherwise males remain peaceful and size up their competitors from afar. The significance of eye-stalk size and separation has also been explored in crabs by Zeil, Nalbach and Nalbach 9. They have contrasted species living high on beaches in the complexly structured world of mangrove roots and rocks with species living on extensive mud-fiats: Crabs on mud-flats have eyes on long, vertically held stalks, which are set relatively close together (Fig. 3B). The eyes are elongated vertically with strongly enhanced vertical acuity at
the equator of the eye - the region looking at the horizon - but, significantly, with no corresponding increase in horizontal acuity. One advantage of long eyestalks (and legs) for animals confined to a fiat plane is in computing distance. Distance of an object on the plane is given by H/tan 0, where H is the height of the eye above the plane and 0 is the angular position of the image of the object relative to the horizon. The bigger H and the better an animal's vertical resolution, the greater will be the range and accuracy of its depth vision. And it is just these parameters that the flat-landers have been at pains to optimize. The vertical-horizontal anisotropy makes it difficult to argue that these are simply adaptations for expanding the crab's horizon. Species inhabiting a more complex threedimensional world and that cannot rely on such simple geometrical tricks, have short stalks set further apart. Their eyes are spherical, with less anisotropy - a system perhaps more suited for binocular vision.
Selected references 1 Wallace, G. K. (1959) J. Exp. Biol. 36, 512-525 2 Rossel, S. (1986) J. Exp. Biol. 120, 265-281 3 Maldonado, H., Levin, L. and BarrosPita, J.C. (1967) Z. Vgl. Physiol. 56, 237-257 4 Maldonado, H., Benko, M. and Isern, M. (1970) Z. Vgl. Physiol. 68, 72-83 5 Strausfeld, N. J. (1979) Verb. Dtsch. Zool. Ges. 167-179 6 Cloarec, A. (1986) J. Exp. Biol. 120, 59-77 7 Burkhardt, D. and de la Motte, I. (1983) J. Comp. PhysioL 151, 407-421 8 Burkhardt, D. and de la Motte, I. (1985) Naturwissenschaften 72, 204-206 9 Zeil, J., Nalbach, G. and Nalbach, H. O. J. Comp. Physiol. (in press) 10 Rossel, S. (1983) Nature 302, 821-822
TINS-January 1987 [10]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
E ome insects demonstrate their capacity for accurate depth vision by leaping on to objects or by catching prey. While it has been known for several decades that locusts can exploit the monocular depth cue of motion parallax to judge the distance of an object 1, only recently has it become clear that insects can also use binocular cues. Earlier evidence for binocular depth vision was circumstantial, based principally on the observations that in many insects the visual fields of the two eyes show substantial overlap (e.g. Fig. 1) and that monocularly blinded predatory insects have difficulty in catching prey. Indeed, sceptics doubted that insects possessed the elaborate neural connections needed for binocular vision. Experiments by Rossel 2 on the praying mantis have now settled this question. He has shown that mantids must use binocular cues to extract depth and suggests a simple way in which they might do so without straining a scientist's credulity or an insect's nervous system. The mantid's method of catching prey makes it an ideal animal for studies of depth vision, as Maldonado 3 realized in the 1960s. It sits in wait to ambush a passing insect. When one comes into view, the mantid turns its head to place the image of the insect on its fovea. Should the insect then wander within range, the mantid shoots out its foreleg to grasp the insect. The bulk of the mantid is immobile during this sequence, and it continues to catch prey if it is anchored firmly to a holder and base-out prisms are placed in front of its eyes (Fig. 2). Since the animal's eyes are fixed in its head, these prisms increase the horizontal binocular disparity of the target image on the two retinae. Mantids viewing prey through base-out prisms strike accurately at its apparent binocular position2; the prey is seen to be nearer than it
really is. The precision with which apparent distance is controlled by the power of the prisms suggests that at close range the mantid relies entirely upon binocular disparity to gauge depth. The range over which binocular cues can operate depends upon the distance between the eyes and the accuracy with which the position of a target image can be measured. The interocular distance in the mantid, Tenodera, is about 4 ram. Work by Maldonado et al. 4 indicates that binocular disparities might be used to discriminate an a~ian predator seen at 45 cm from one seen at 60 cm. These figures suggest that mantids can assess disparity to within about 0.13 deg, which in turn indicates a somewhat finer resolution than that suggested by the 0.6 deg difference between the lines of sight of adjacent foveal ommatidia 2. What neural mechanism mediates binocular vision in insects? One possibility is that there are more binocular connections than have hitherto been supposed. In flies, signs have been found of a retinotopic projection between the lobulae of the two optic lobes 5, although the extent and function of such connections in mantids remain to be worked out. Since the mantid only strikes when the two xget images fall within the 30 deg circular foveas, binocular connections need only exist between foveal regions of the lobula. However, Rossel 2 has proposed an attractive mechanism that does not require any local binocular interactions. His suggestion is that each optic lobe measures the vertical and horizontal positions of the target image with respect to the centre of the fovea of the ipsilateral eye. For simplicity, suppose that there is a convenient Cartesian representation with four
TINS January- Vol.lO No. 1 [103]
~)1987, EJsevierSciencePublishersB.V.Amsterdam 0378-5912/87/$0200
S
Binocular depth vision in arthropods output neurons from each optic lobe, one encoding the angular horizontal distance to the left of a reference point in the fovea, one the distance to its right, one the vertical distance above it and the last the vertical distance below. An average of the horizontal signals coming from the two optic lobes would then provide a measure of the azimuth of the target with respect to the head, while the difference between the horizontal signals would carry information about target distance. Direct evidence for a scheme of this type is so far only available for the assessment of target direction. If vision is restricted to one eye, the mantid makes a saccade-like
T. 5. Colle~
Schoolof Biological Sciences, The Universityof Sussex, Falmer,Brighton, SussexBN19QG, UK.
Fig. 1. Tenodera australasiae. The binocularfield of this mantid extends2400 verticallyand 350 ho&,ontally. (Kindly provided by S. Rossel.)
//
Fig. 2. Diagram of mantid viewing a fly through base-out prisms. (Adapted from Ref. 10.) 1
Fig. 3. Someintriguing eye-stalksin arthropods. (A) Fema/eCyrtodiopsis white (eyeseparation 6.5 mm). (B) Macropthalmus setosus(eye-
separation, 7.2 mm). (A) Kindlyprovidedby I. de la Motte, (B) by J.Zeil,and adapted from Ref.9.
turn of its head to fixate the target with the viewing fovea. However, when both eyes see the target, the saccadic turn is a weighted average of the saccades driven separately by the left and right eyes 2. One apparent limitation of this model is that there would be confusion should several prey be viewed simultaneously. There is no guarantee that the two eyes would decide to fixate the same one. Intriguingly though, Cloarec6 has recently found that the predatory water stick-insect, Ranatra, has no problems with two prey arriving at once. It deftly catches one with each foreleg. Insects may have methods for sorting out which images belong together. Two prey would not be equally attractive; for instance, one may move more and so 'capture the attention' of both optic lobes. Also, since the vertical disparities between the two eyes will normally be zero, two images with the same elevation will very likely represent the same prey in object-space. Two interesting studies have taken a comparative approach to problems of depth vision. The first deals with a puzzling curiosity of the entomological world, the stalkeyed fly (family Diopsidae), some species of which seem to have gone to exuberant lengths to increase their interocular separation (Fig. 3A). There have been many speculations that the stalks are an adaptation for enhanced binocular vision. Burkhardt and de la Motte 7'~ have tackled this question by examining the behaviour and dimensions of the eye-stalks of nine diopsid species found in Malayan rain forest. In some species, males and females have modest eye-stalks of equal length and Burkhardt and de la Motte conclude that initial selection for wide-set eyes was probably 2
driven by the problems of relying on vision to find mates in tropical rainforest, where conspecifics are rarely encountered and species diversity is high. In dimorphic species the male stalks are hugely hypertrophied. Here it seems sexual selection has taken over and acted primarily to make the stalks conspicuous. Among a group of different sized males of one species, the length of the eye-stalk increases so steeply in relation to body length that it acts not only as an advertisement but also a sensitive index of an individual male's size. Features of the social behaviour of both males and females depend on accurate assessment of a male's size. Dimorphic species of the genus Megalobopsaggregate at dusk on hanging plant-fronds in sleeping groups of about a dozen with one large male, ten or so females and perhaps one or two male runts. The male is usually the first to arrive and the females choose the frond with the largest male on it. Males arriving later are either chased off, or take over the harem, or are so small that they can hide safely among the females. Competition between males is settled by size, with fights seen only among similar-sized males. Otherwise males remain peaceful and size up their competitors from afar. The significance of eye-stalk size and separation has also been explored in crabs by Zeil, Nalbach and Nalbach 9. They have contrasted species living high on beaches in the complexly structured world of mangrove roots and rocks with species living on extensive mud-fiats: Crabs on mud-flats have eyes on long, vertically held stalks, which are set relatively close together (Fig. 3B). The eyes are elongated vertically with strongly enhanced vertical acuity at
the equator of the eye - the region looking at the horizon - but, significantly, with no corresponding increase in horizontal acuity. One advantage of long eyestalks (and legs) for animals confined to a fiat plane is in computing distance. Distance of an object on the plane is given by H/tan 0, where H is the height of the eye above the plane and 0 is the angular position of the image of the object relative to the horizon. The bigger H and the better an animal's vertical resolution, the greater will be the range and accuracy of its depth vision. And it is just these parameters that the flat-landers have been at pains to optimize. The vertical-horizontal anisotropy makes it difficult to argue that these are simply adaptations for expanding the crab's horizon. Species inhabiting a more complex threedimensional world and that cannot rely on such simple geometrical tricks, have short stalks set further apart. Their eyes are spherical, with less anisotropy - a system perhaps more suited for binocular vision.
Selected references 1 Wallace, G. K. (1959) J. Exp. Biol. 36, 512-525 2 Rossel, S. (1986) J. Exp. Biol. 120, 265-281 3 Maldonado, H., Levin, L. and BarrosPita, J.C. (1967) Z. Vgl. Physiol. 56, 237-257 4 Maldonado, H., Benko, M. and Isern, M. (1970) Z. Vgl. Physiol. 68, 72-83 5 Strausfeld, N. J. (1979) Verb. Dtsch. Zool. Ges. 167-179 6 Cloarec, A. (1986) J. Exp. Biol. 120, 59-77 7 Burkhardt, D. and de la Motte, I. (1983) J. Comp. PhysioL 151, 407-421 8 Burkhardt, D. and de la Motte, I. (1985) Naturwissenschaften 72, 204-206 9 Zeil, J., Nalbach, G. and Nalbach, H. O. J. Comp. Physiol. (in press) 10 Rossel, S. (1983) Nature 302, 821-822
TINS-January 1987 [10]