J. exp. BioL (1978), 76, a37-*4i 2 3 7 With 3 figures
Printed in Great Britain
SHORT COMMUNICATION
PEERING - A LOCUST BEHAVIOUR PATTERN FOR
OBTAINING MOTION PARALLAX INFORMATION
BY T. S. COLLETT
School of Biological Sciences, University of Sussex, Brighton BNi gQG, Sussex, U.K.
(Received 12 April 1978)
Motion parallax is probably the most important method available to insects for judging the distance of objects, yet with one exception very little is known about the ways in which insects exploit this cue. Almost twenty years ago Wallace (1959) proposed that the side-to-side peering movements of juvenile locusts are performed specifically to obtain parallax information. Leg movements cause the whole body to pivot about the abdomen, shifting the head laterally about 0-5 cm and displacing the retinal image of an object in front of the insect. Provided that the locust knows how far (or fast) it moves its head and that it can measure image displacement (or velocity) it can compute the distance of the object (Fig. 1 a). The following suggests locusts do indeed do something of this kind. When an object on to which a locust is about to jump is moved horizontally in a frontal plane, in synchrony with peering, locusts are fooled into misjudging the object's distance. For instance, if the object is moved in the opposite direction to that in which the locust peers, so enhancing image movement, the locust jumps short (Wallace, 1959).
Two features of peering are analysed here which suggest that the behaviour pattern is carefully designed for extracting parallax information.
(1) If the head were to rotate during peering there would be added an unwanted component to movement of the retinal image. This would be independent of the distance of objects in the environment, and would distort the simple relation-between image displacement and distance shown in Fig. 1 (a). Films of head movement taken during peering show that the angular orientation of the head remains constant despite rotation of the body.
(2) The precision with which distance can be estimated during peering is limited by the amplitude of head movement and by the insect's horizontal acuity. The accuracy also decreases as the square of the distance between locust and target (e.g. Horridge, 1977). It is shown here that the locust increases its amplitude of peering when object distance is raised and so to some extent is able to compensate for the inherent decline in accuracy with distance.
238 T . S. COLLBTT
1 cm
H r- H 1 1 1 1 1 H 1 h- H 1 1
10 Time (s)
15 20
Fig. i. (a) The geometry of motion parallax in locusts. Head moves laterally through distance
S cm, displacing the retinal image of the target D cm from the locust by aa°. (6) Lateral
head movements (Sh^a) result from leg movements which rotate body (<f'tna>tvm)- Head orientation ((4>«d) is nonetheless constant, (c) Plot against time of the orientation and lateral position of the head and orientation of the body during a series of peers (the beginning of the sequence was not caught on film). The movement of the head with respect to the body is given by (^i«d — ^prooemm). Anti-clockwise angles are taken as positive.
Short Communication
239
Lateral vision occluded Lateral drum
0,
(c)
+ 0-2
'head + 0 |
(cm)
2 3 Time (s)
Frontal drum
t
o pooood o
„ o o c o o W Y —
Time (s)
Fig. a. (a) T w o peers with lateral vision occluded showing the imprecision of head counter rotation. In the first peer the body rotates through 6° and the head counter rotates 2° too far. Locusts peer less readily when vision is occluded and the amplitude of peer (St**^ is usually small, (ft) Peering with moving drums in lateral visual field. T h e two drums rotate in the same sense reversing direction about every 2 8. The image velocity as seen by a
ttationary locust ( ^d nJ is about $".r1 (locusts do not tolerate higher velocities). When the
head moves laterally it also rotates in the direction of drum movement. Average closed-loop gain (^h»d/^<iram) i> then about o-6, though given the small movements and times involved this figure is not very accurate, (c) Peering with moving drum in the frontal visual field. #dmm >8 about 15°.s"1. Head orientation is unaffected by drum rotation both at this drum velocity and at s0.*"1.
of tape. The signal corresponding to sideways head position (<Shead) was fed into
240 T . S. COLLETT Target at:
5 cm
50 peers
10cm
30 cm
i i i i
0-5 Amplitude of peer
1 cm
Fig. 3. The effect of target distance on peering amplitude. Mean peering amplitude increases slightly but significantly (P < o-oi) with distance (0-21, 0-19 and 0-35 cm at target distances of 5, 10 and 30 cm respectively). Vertical calibration is 50 peers.
mechanisms contribute to the compensatory neck movement that maintains head orientation constant.
(1) The internal motor programme which drives leg movements also causes the head to counter-rotate. When the visual input influencing head rotation is removed by occluding lateral retina, the head compensates for changes in body orientation, but does so less precisely than before. Sometimes the head turns by just the right amount, but sometimes it rotates too far (Fig. 2 a), and sometimes not far enough. Visual feedback is thus needed to correct the motor programme.
Short Commxtmcation 241
regions of retina across which there is no image slip are those facing along the direction of movement, i.e. lateral retina. Thus image slip over lateral retina can provide the locust with an unambiguous index of head rotation, whereas frontal retina sums the effects of translational and rotational movement.
Fig. 3 (a) presents amplitude distributions of all the peers produced by eight locusts at three different object distances. There is a slight but clear increase in the amplitude of peering as the target distance is increased which is seen in individual as well as in pooled data. This growth of peering amplitude with target distance enhances image movement and so delays the target distance at which image displace-ment becomes undetectably small. This effect is more marked in sequences of peers ending in a jump. Typically, when a locust is placed on the bar used in these experiments it walks a short distance, stops, sometimes gives a series of peers, after which it may or may not jump at its target. The number of peers in a sequence is very variable, ranging from one to eleven with a modal value of about three. Despite great variability there is a significant tendency (P < o-oi) for the amplitude of peering to increase over the first few peers of a series. On the third peer of a series the mean amplitude is less than 0*3 cm (3"4° on retina) when the target is 5 cm away and more than o-6 cm ( I * I ° on target) for a target distance of 30 cm. It is as though the locust measures image displacement (or velocity) during a peer and, if it is too little, increases the amplitude (or velocity) of the subsequent peer. This finding raises the intriguing possibility that locusts could extract distance information, not by measuring image motion resulting from a given head movement, but simply by monitoring the amplitude of the largest peer they produce.
Peering is thus under the control of two distinct visual reflexes which employ separate regions of retina. One limits image rotation and is governed by visual input to lateral retina, whereas the second increases image translation and is controlled by frontal retina.
I thank Brian Cartwright for designing and building the device to measure head position and Mike Land and Peter Slater for a critical reading of the manuscript. Financial support came from the S.R.C.
REFERENCES
HORRIDOB, G. A. (1977). Insects which turn and look. Endeavour, N.S. i, 7-17.
