Journal of Comparative Physiology A

, Volume 190, Issue 1, pp 61–68

Behavioral analysis of polarization vision in tethered flying locusts


    • Fachbereich Biologie, TierphysiologieUniversität Marburg
  • U. Homberg
    • Fachbereich Biologie, TierphysiologieUniversität Marburg
Original Paper

DOI: 10.1007/s00359-003-0473-4

Cite this article as:
Mappes, M. & Homberg, U. J Comp Physiol A (2004) 190: 61. doi:10.1007/s00359-003-0473-4


For spatial navigation many insects rely on compass information derived from the polarization pattern of the sky. We demonstrate that tethered flying desert locusts (Schistocerca gregaria) show e-vector-dependent yaw-torque responses to polarized light presented from above. A slowly rotating polarizer (5.3° s−1) induced periodic changes in yaw torque corresponding to the 180° periodicity of the stimulus. Control experiments with a rotating diffuser, a weak intensity pattern, and a stationary polarizer showed that the response is not induced by intensity gradients in the stimulus. Polarotaxis was abolished after painting the dorsal rim areas of the compound eyes black, but remained unchanged after painting the eyes except the dorsal rim areas. During rotation of the polarizer, two e-vectors (preferred and avoided e-vector) induced no turning responses: they were broadly distributed from 0 to 180° but, for a given animal, were perpendicular to each other. The data demonstrate polarization vision in the desert locust, as shown previously for bees, flies, crickets, and ants. Polarized light is perceived through the dorsal rim area of the compound eye, suggesting that polarization vision plays a role in compass navigation of the locust.


Compass navigationCompound eyeLocust flightPolarization visionSchistocerca gregaria


Scattering of sunlight in the earth’s upper atmosphere leads to partial polarization of light from the blue sky (Waterman 1981). As a result, the blue sky shows a polarization pattern consisting of e-vectors oriented in concentric circles around the sun. The degree of polarization varies from 0 (direct sunlight) to a maximum (under optimal conditions) of about 0.75 or 75% along the great circle of the sky at an angular distance of 90° from the sun (Coulson 1988). Since the e-vector pattern of the sky is linked to the solar position, directional information in the sky is independent of the visibility of the sun, and can also be obtained from a small patch of blue sky. Many insects are able to detect the polarization pattern of the blue sky and use this capability for spatial orientation (Wehner 1984; Wehner et al. 1996). The use of skylight polarization for spatial navigation was shown in field studies on honeybees and ants (von Frisch 1949; Wehner 1982; Rossel and Wehner 1986) and, using a spontaneous orientation response, in laboratory studies in crickets and flies (Brunner and Labhart 1987; von Philipsborn and Labhart 1990).

Burghause (1979) reported ommatidia specialized to detect skylight polarization in a small dorsal rim area (DRA) of the cricket’s compound eye, and similar DRAs have meanwhile been demonstrated in many other insect species (reviewed by Labhart and Meyer 1999). Specializations of DRAs favoring a role in polarized light detection are similar in all species studied and include (1) orthogonal arrangement and high alignment of microvilli, (2) reduced length and enlarged cross-sectional area of rhabdoms, (3) high polarization sensitivity of DRA photoreceptors, and finally (4) degraded optics by light scattering structures in the cornea or missing screening pigment (Labhart and Meyer 1999, 2002).

The compound eye of the desert locust, Schistocerca gregaria, has a particularly prominent DRA (Homberg and Paech 2002). It largely consists of blue-sensitive photoreceptors with high polarization sensitivity (Eggers and Gewecke 1993). Tracing studies have mapped central processing stages which are supplied by input from the DRA and showed that these include dorsal areas in the lamina and medulla, the anterior lobe of the lobula, the anterior optic tract and tubercle, the lateral accessory lobe, and the central complex (Homberg et al. 2003). Single-cell recordings demonstrated polarization sensitivity in interneurons at the level of the medulla (Homberg and Würden 1997), the anterior optic tubercle, the lateral accessory lobe, and the central complex (Vitzthum et al. 2002). In order to gain further insight into the mechanisms of polarized-light-dependent spatial orientation in the locust, we have developed a behavioral assay for polarization vision in this insect. This study demonstrates polarotaxis in stationary flying locusts and shows that photoreceptors in the DRA of the eye are essential for this behavior. Parts of this study have been published in abstract form (Mappes and Homberg 2003).

Materials and methods


Animals were kept in a crowded laboratory colony under LD 12:12 h at 28°C and 50% relative humidity (RH). Adult male and female desert locusts (S. gregaria) at least 10 days after imaginal molt were used. A small metal pin was attached ventrally to the pterothoracic sternum by rosin wax. The pin was used to attach the locusts to a friction free thrust/yaw-torque meter (see below). In all experiments, the heads of the animals were fixed to the thorax by rosin wax.

Experimental setup

Locusts were flown in a laminar air current within the flight chamber (34 cm×25.5 cm×26 cm) of a horizontal wind tunnel (Gewecke 1975), so that the body axis was nearly horizontal (Fig. 1). The sides and bottom of the chamber were constructed from sheets of matt-black plastic material. For optimal flight performance, wind speed was adjusted at 3.0 m s−1. The yaw-torque meter was provided by Dr. R. Preiss and has been described in detail by Preiss and Gewecke (1991). It consists of a pair of transducers and allows to measure thrust and yaw torque simultaneously but was used here only for measurement of instantaneous yaw torque by electronic subtraction of the signals from the two transducers. The subtracted signal was filtered (low-pass, 1 Hz) and fed into an interface (CED 1401 plus; Cambridge Electronic Design) and a laboratory computer (Pentium III processor) equipped with Spike2 software for graphic depiction of turning tendency and for later analysis. Stimulation, testing procedure and data analysis largely follow the procedure developed by Brunner and Labhart (1987) and von Philipsborn and Labhart (1990) for analysis of polarization vision in crickets and flies.
Fig. 1

Experimental setup for testing polarotaxis of tethered flying locusts (animal not drawn to scale). A mounting rod connected to the yaw-torque meter (for more details see Gewecke 1975; Preiss and Gewecke 1991); B flight chamber; C rotating polarization filter, diffusor, or combined polarizer-diffusor; D diffusor; E fluorescent lamp; F cover to ensure optimal lighting; G wind tunnel producing laminar air current

Light was produced by a fluorescent lamp (Osram Dulux EL 16 W; Fig. 1). It passed through a horizontal diffuser and evenly illuminated a circular window in the ceiling of the flight chamber (Fig. 1). The window was fitted with different rotating filters and provided a wide field stimulus of 87.3° visual angle centered around the zenith. The following rotating stimuli were used: (1) polarized light was produced by inserting a linear polarization filter (Polaroid HN 38S, irradiance: 25.8 μW cm−2); (2) unpolarized light was produced by inserting a diffuser (two sheets of translucent drawing paper, irradiance: 30.2 μW cm−2); (3) to control for the possible effects of intensity gradients produced by the polarizer, filter (2) was overlaid with the polarization filter (irradiance: 9.8 μW cm−2). All light intensities were measured with a radiometer (Photodyne 18XTA, silicium detector) at the position of the locust head. Filter rotation was monitored through a photodiode covered by a polarization filter (invisible to the locust). Its output was fed into the second channel of the data recording equipment (see above).

Testing procedure

Experiments were carried out at 30–32°C and 40–50% RH. In order to obtain regular flight behavior and an even flight position (Preiss and Gewecke 1991), animals were flown in complete darkness for a few minutes. After locusts became accustomed to the situation, light was turned on and the polarizer/diffuser started rotating. A single experiment consisted of three full revolutions of the polarizer (speed: 5.3° s−1) either clockwise or counter-clockwise. Therefore, each e-vector orientation of the pattern occurred six times.

In control experiments (1) the polarization filter was replaced by the diffuser (see above), (2) the polarization filter was placed directly behind the diffuser (intensity gradient), and (3) animals were flown under a stationary e-vector (parallel to their longitudinal body axis). Each animal was first flown under the polarizer and subsequently under one of the control conditions. In order to identify the eye region involved in the response, locusts were flown under the polarization pattern with their DRAs painted black (Marabu Decorlack matt, water-based). In a second group, the compound eyes were painted black except for the DRAs.

Data analysis

Recordings from each flight were transferred from Spike2 to Microcal Origin 8.0 and Clampfit 8.0 (Axon Instruments, Foster City, Calif., USA) for further analysis. For each experiment, the yaw-torque responses of the six periods were averaged within 5° bins; an inherent turning tendency, defined as yaw-torque average of all bins, was subtracted from the data. For each histogram, a periodicity score P as a measure of the animal’s response to the rotating e-vector was determined empirically (for details see Appendix) following a procedure developed by von Philipsborn and Labhart (1990). It describes the degree of periodicity in the histograms by detecting periodic behavior both in cases with strong and weak modulation of yaw torque. Periodicity scores obtained under different stimulus conditions were statistically compared using the non-parametric U-test after Wilcoxon, Mann and Whitney (Clampfit 8.0).

At two or more positions of the histograms the locusts showed no turning tendency. The preferred e-vector orientation was defined as being followed by a turning response in the direction of e-vector rotation. Turning in the opposite direction indicates the avoided e-vector orientation. For statistical evaluation of the distributions of the preferred and avoided e-vectors, the χ2-test (SPSS for Windows 11.0; SPSS, Chicago, Ill., USA) was applied.



In this study we tested yaw-torque responses of desert locusts to polarized light in stationary flight. The animal’s readiness to fly was highly variable and did not depend on sex. Also the motivation to respond consistently to the rotating polarizer varied and was sometimes low, even in good flyers. Only flights without any interruption were, therefore, evaluated and included in the data. There was no significant difference between flights under the polarizer rotating clockwise or counter-clockwise (non-parametric U-test after Wilcoxon, Mann and Whitney).

Rotating polarizer

For analysis of yaw torque under the rotating polarizer, 142 flights of 92 animals were evaluated. Rotation of the polarization filter induced periodically changing yaw-torque responses in the locusts which corresponded to the 180° periodicity of the rotating polarizer (Fig. 2A). During 180° rotation a single period of yaw torque in the direction of the rotating polarizer was usually followed by a period of yaw torque in the opposite direction or vice versa (Fig. 2A, B). The transitions between right and left turning were moments when no yaw torque occurred. Periodicity scores of these flights ranged from 0.08 to 259.2 with a mean score of \( \overline{{\text{P}}} = 48.9 \) (Fig. 4A).
Fig. 2

Flight traces and yaw-torque histograms under rotating polarizer (A, B), rotating diffuser (C, D), rotating intensity gradient (E, F), and stationary e-vector (G, H). Positive and negative yaw-torque values correspond to right- and left-hand turning, respectively. Inherent turning tendency was subtracted in the flight traces and in the yaw-torque histograms. A Test situation with rotating polarizer. The animal shows a 180° periodicity. B Average yaw-torque responses from the six 180° rotations in A, divided in 5° bins. The periodicity score calculated from the histogram (see Appendix) is P=100. C Rotating diffuser. D Histogram from flight in C. The periodicity score is 0.9. E Intensity gradient; the animal shows a shift in yaw torque during the experiment, but no periodic changes. F Histogram from flight in E; P=4.1. G Stationary e-vector. H Histogram from flight in G. The periodicity score of this flight is 5.6. The units on the abscissae in G and H are virtual numbers, corresponding in time to the angular filter positions in AF

Control experiments

Controls were executed between test flights under the rotating polarizer in order to demonstrate that e-vector responses were present at that time. Yaw-torque responses during rotation of the diffuser were analyzed in 55 animals. During these flights, yaw torque was irregular and, in many flights, altered frequently between left and right turns (Fig. 2C, D). Similar results were obtained when placing the polarization filter behind the diffuser, tested in 20 experiments (Fig. 2E, F). In these controls, periodicity scores ranged from 0.03 to 129.6 (diffuser only, \( \overline{{\text{P}}} = 14.2 \)) and from 0.06 to 86.4 (polarizer behind diffuser, \( \overline{{\text{P}}} = 11.1 \)) (Fig. 4B, E). Statistical comparison of the periodicity histograms shows significantly increased P scores in the test situation compared to the two controls (P<0.001, diffuser; P<0.001, diffuser plus polarizer). The histograms of the two control situations are not significantly different from each other (P<0.5).

In addition to the two controls described above, 32 experiments were performed under a stationary e-vector, parallel to the animals’ longitudinal axis. Since this stimulation provides no periodic information, P scores were low and ranged from 0.02 to 28.8 with \( \overline{{\text{P}}} = 7.6 \) (Figs. 2G, H and 4D). Periodicity scores of flights under stationary e-vector are significantly different from the test flights (P<0.001). No significant difference occurs between the P scores of flights under the stationary e-vector and the two other controls (in all cases, P<0.1).

Determination of eye region involved in polarotaxis

For many insect species, a specialized DRA of the compound eye was shown to be responsible for the detection of polarized light (see Introduction). The presence of a DRA in S. gregaria likewise suggests that this region plays an important role in polarization vision. To determine the eye region involved in the polarization response animals with different parts of their eyes occluded were tested under the rotating polarizer. Experiments with DRAs painted black or with compound eyes except the DRAs painted black were executed between regular test flights to show that e-vector responses existed in the untreated animal. In 31 flights the compound eyes except the DRAs were overlaid with black paint. These animals showed no difference in flight behavior from untreated animals under the rotating polarizer (P<0.1) (Fig. 3A, B). Periodicity scores ranged from 2.1–324, with \( \overline{{\text{P}}} = 35.3 \) (Fig. 4C). In 32 flights the DRAs were occluded and the rest of the eyes was left open. In these experiments, the animals’ turning tendency changed irregularly and was sometimes minimal (Fig. 3C, D), and periodicity scores ranged from 0.007–81, \( \overline{{\text{P}}} = 15.7 \) (Fig. 4F). Statistical comparison shows that P scores of experiments with free DRAs are significantly larger than those of experiments with DRAs painted black (P<0.05). No significant difference occurred between experiments under the polarizer in untreated animals and in animals with compound eyes except DRAs painted black (P<0.1). Likewise, P scores of control experiments with diffuser and with diffuser plus polarizer are not significantly different from the experiments with painted DRAs (P<0.1, diffuser; P<0.1, diffuser plus polarizer).
Fig. 3

Flight traces and yaw-torque histograms under rotating polarizer with compound eyes painted black and dorsal rim areas (DRAs) free (A, B), and with DRAs painted black (C, D). A, B Painting the compound eyes except the DRAs results in yaw-torque responses corresponding to the 180° periodicity of the rotating polarizer; P=108. C, D With DRAs painted black, the flight trace is irregular and P is low (0.2)

Fig. 4A–F

Comparison of P scores of all test flights and controls. Number of flights and means of P scores are given for each histogram. A Rotating e-vector; B rotating diffusor; C rotating e-vector, compound eyes painted black, but DRAs open; D stationary e-vector; E rotating intensity gradient; F rotating e-vector, DRAs painted black, rest of the eyes open. Flights under rotating polarizer (A) have significantly higher P scores than the controls in B (P<0.001), D (P<0.001), and E (P<0.001). Likewise, flights with DRAs open (C) have significantly higher P scores than flights with DRAs painted black (F, P<0.05). The distributions in A and C (rotating polarizer, DRAs open) are not significantly different from each other (P<0.1), and no significant difference exists between the four control situations (B versus D, P<0.5; B versus E, P<0.5; B versus F, P<0.1; E versus D, P<0.1; E versus F, P<0.1; F versus D, P<0.1)

Preferred and avoided e-vectors

Under the rotating polarizer, the animals’ turning tendency changed periodically every 180°. However, at two positions of the histogram the stimulus induced no turning tendency. After the appearance of these e-vector positions, the locusts either tried to follow the rotating polarizer (preferred e-vector) or to turn against the rotating direction (avoided e-vector). Preferred and avoided e-vectors, determined in 82 animals, were broadly distributed from 0 to 180° (Fig. 5A). Statistical evaluation (χ2-test) shows no significant difference of both histograms from random distributions (P<0.5). For all animals, however, the difference between preferred and avoided e-vector is approximately 90° (Fig. 5B).
Fig. 5

A Distribution of preferred and avoided e-vector orientations of 82 animals under the rotating polarizer. Both histograms are not significantly different from a uniform distribution (χ2-test, P>0.5). B Preferred versus avoided e-vector orientations from 82 animals


In the present study we developed a behavioral assay to analyze polarization vision in the desert locust S. gregaria. We examined the flight behavior of tethered flying locusts and showed that a slowly rotating polarizer above the animals induced periodic changes in turning responses with and against the direction of rotation. Locusts flying under a rotating diffuser or under a rotating intensity gradient showed yaw-torque responses which were indistinguishable from those to a stationary e-vector. Thus, intensity gradients inherently connected to polarized light stimulation do not contribute to the observed behavior. In an earlier study, Eggers and Weber (1993) demonstrated polarotaxis in locust larvae (S. gregaria and Locusta migratoria) walking on a Kramer treadmill. Polarization vision in the locust is, therefore, present in larvae and in adults and is not restricted to a specific behavioral context such as flight.

For a better comparison with previous studies, our experimental design and procedure for data evaluation closely followed experiments performed in crickets, Gryllus campestris, and flies, Musca domestica (Brunner and Labhart 1987; von Philipsborn and Labhart 1990). In contrast to polarotaxis in flight, as demonstrated here, Labhart and coworkers used spontaneous changes in walking direction correlated with changing e-vector directions. In the present study (Fig. 4) and that by von Philipsborn and Labhart (1990) on walking flies, there is considerable overlap between the periodicity histograms for polarized light and those for control experiments. Possible reasons for this may lie in the choice and balance of the different parameters that contribute to evaluation of the periodicity scores. Flying locusts showed occasional brief but strong turning movements during tethered flight, apparently irrespective of sensory stimulation. These led to considerable increases in the periodicity score, particularly when they occurred against an overall low yaw-torque background. Addition of a low-pass filter in the averaging procedure, as done by Brunner and Labhart (1987) and von Philipsborn and Labhart (1990), had no consistent effect on the statistical evaluation of the data and was therefore not included in the present study.

A difference between this study and the work on crickets and flies is the lack of general preferred and avoided e-vector orientations in locust polarotaxis. Crickets and flies showed a slight preference for e-vector orientations perpendicular to their longitudinal axis (Brunner and Labhart 1987; von Philipsborn and Labhart 1990), but the reason for this behavior is unclear. While a general preferred e-vector orientation might be expected for genetically programmed compass directions, as found in seasonal migrants such as the monarch butterfly (Mouritsen and Frost 2002), laboratory-reared insects used here or central place foragers which change their navigational directions depending on the availability of a food source would be expected to lack generally preferred directions.

Neurobiology of polarization vision in the locust

The present study demonstrates that polarotactic yaw-torque responses of tethered flying locusts are mediated exclusively through photoreceptors of the DRA of the compound eye. Similar results were obtained in behavioral studies on larval locusts (Eggers and Weber 1993), honeybees, ants, crickets, and flies (reviewed by Labhart and Meyer 1999, 2002). The data strongly suggest that the DRA of these insects serves a role in sky compass orientation. As in many other insect species (Labhart and Meyer 1999) the DRA of the locust shows a number of morphological specializations favoring a role in polarization vision (Homberg and Paech 2002), and single cell recordings demonstrated high polarization sensitivity of DRA photoreceptors with maximum spectral sensitivity in the blue range (Eggers and Gewecke 1993). Several central processing stages involved in polarization vision have recently been identified in the locust. Tracing studies showed that visual pathways originating in the DRA of the compound eye include specially modified dorsal rim areas in the lamina and medulla, a ventral layer of the anterior lobe of the lobula, the lower unit of the anterior optic tubercle, certain subunits of the lateral accessory lobe, and the lower division of the central body (Homberg and Paech 2002; Vitzthum et al. 2002; Homberg et al. 2003). Polarization-sensitive interneurons have been characterized especially in the optic tubercle, lateral accessory lobe, and central complex (Vitzthum et al. 2002; Pfeiffer and Homberg 2003), but have also been found at the level of the medulla (Homberg and Würden 1997). All neurons studied so far show polarization opponency, i.e., they are maximally excited at a particular e-vector orientation (Φmax) and are maximally inhibited at an e-vector orientation perpendicular to Φmax. The central complex appears to serve an important role in motor control and Vitzthum et al. (2002) specifically proposed a function of the central complex as an internal compass for spatial orientation.

Functional significance

Although locusts perform long-range migratory flights, both in swarms and as solitary animals (reviewed by Uvarov 1977; Baker 1978; Farrow 1990), the potential involvement of a polarization compass in spatial orientation of these insects has not been studied. Even more surprising, the necessity for sky compass orientation in migrating gregarious locusts has been denied by several investigators (Draper 1980; Baker et al. 1984). On the other hand, most studies agree that wind speed and wind direction plays the dominant role in controlling flight direction (Kennedy 1951; Baker 1978; Riley and Reynolds 1986). At low wind speeds locusts often fly upwind, but may turn into the wind as wind speed increases (Kennedy 1951). Interestingly, however, many studies both on solitary animals and on migrating swarms note that the mean course angle in relation to wind direction can be significantly different from zero suggesting active orientation mechanisms (Kennedy 1951; Riley and Reynolds 1986; Baker et al. 1984; reviewed by Farrow 1990). Moreover, flying swarms as well as migrating walking larvae (“marching hopper bands”) usually maintain a particular migratory direction for several hours irrespective of temporary wind shifts (Kennedy 1945, 1951; Baker et al. 1984). Although sky compass orientation has been suspected to play a role in maintaining these course directions (Baker 1978), only two field studies include experiments to address this question (Kennedy 1945, 1951). In observations on marching hopper bands, Kennedy (1945) used a blanket and a mirror to change the sun’s position artificially by 180° (from left to right). As a result of this, all marching locusts stopped and marched off in the opposite direction. Kennedy even noted that the angle of orientation to the sun shifted gradually during the course of the day, as expected for time-compensated sun compass orientation. While these experiments on walking locusts were quite striking, similar experiments on flying individuals were more difficult. In one fortuitous attempt, however, five animals coming out of the shadow of a tree instantly changed flight directions in the predicted way, when experiencing the altered sun position (Kennedy 1951). Although these experiments are anecdotal, they clearly suggest the presence of sun compass orientation in migratory locusts, but studies over the last 50 years have not investigated this aspect of spatial orientation any further.

In view of our behavioral evidence in tethered flight as well as the neurobiological evidence for polarization vision in the desert locust, field studies concluding a predominant role of wind directions on flight orientation in migratory locusts may need serious reconsideration. We propose that the locust, like many other insect species, has a well developed sense for spatial navigation based on a sun/polarization compass. Field studies will be essential to reveal when and how this compass is used during spatial orientation and migration.


We are particularly grateful to Drs. Michael Gewecke und Reinhard Preiss for providing the wind tunnel and yaw-torque meter. Sincere thanks are given to Dr. Jan Dolzer for helpful and essential suggestions on data evaluation. This research was supported by DFG grants HO 950/13.

Copyright information

© Springer-Verlag 2004