Orientation of rainbow trout (Oncorhynchus mykiss) to multiple patches ...

Color profile: Disabled Composite Default screen

407

Orientation of rainbow trout (Oncorhynchus mykiss) to multiple patches of linearly polarized light Shelee L. Degner and Craig W. Hawryshyn

Abstract: Orientation responses of juvenile rainbow trout (Oncorhynchus mykiss) to two linearly polarized light patches were examined under controlled laboratory conditions. Fish were trained to swim the length of the training tank under a polarized light field created by two linearly polarized stimuli that were oriented either parallel or perpendicular to the length of the tank. Trained fish were released in a circular tank and their angular responses were recorded. For each testing paradigm, the E-vector (electric vector) orientation of one of the two linearly polarized light patches was varied by 15° between 0° and 90°. Each fish was therefore tested in seven different paradigms in which the two E-vector orientations differed by 0°, 15°, 30°, 45°, 60°, 75°, and 90°. Rainbow trout oriented in a bimodal distribution when the two E-vector orientations differed by 0°, 15°, 30°, 45°, and 90°. These results suggest that rainbow trout perceived the two stimuli as being the same when the two E-vector orientations differed by 45° or less. Conversely, rainbow trout did not significantly orient when the two E-vector orientations differed by 60° and 75°. Rainbow trout may be able to discriminate two E-vector orientations that differ between 60° and 75°, and therefore they do not significantly orient, since they perceive two distinct E-vectors to orient to instead of one. When rainbow trout were exposed to a depolarized light field, they did not exhibit significant orientation subsequent to the E-vector cue. Résumé : L’orientation de Truites arc-en-ciel (Oncorhynchus mykiss) en réponse à deux taches de lumière polarisée rectiligne a été examinée dans des conditions contrôlées de laboratoire. Les poissons ont été entraînés à nager le long de l’aquarium sous un champ de lumière polarisée créé par deux stimulus polarisés rectilignes, orientés parallèlement ou perpendiculairement à la longueur de l’aquarium. Les poissons entraînés ont été relâchés dans un aquarium circulaire et l’angle de leurs réactions a été enregistré. Pour chacun des paradigmes à la base des tests, l’orientation du vecteur E (vecteur électrique) de l’une des taches de lumière polarisée a été modifiée de 15E entre 0E et 90E. Chaque poisson a donc été testé d’après sept paradigmes différents, dans lesquels l’orientation d’un vecteur E est décalée de 0E, 15E, 30E, 45E, 60E, 75E et 90E par rapport à celle de l’autre vecteur E. Les orientations des truites se répartissent selon une distribution bimodale lorsque les deux vecteurs E diffèrent de 0E, 15E, 30E, 45E et 90E. Ces résultats indiquent que les truites perçoivent les deux stimulus comme s’il s’agissait d’un seul lorsque les orientations des vecteurs E diffèrent de moins de 45E. Inversement, les truites ne réagissent pas de façon particulière lorsque les vecteurs E sont orientés avec une différence de 60E à 75E. Les truites sont peut-être capables de distinguer deux orientations avec une différence de 60E à 75E et elles n’adoptent pas d’orientation particulière parce qu’elles perçoivent les deux vecteurs E plutôt qu’un seul pour s’orienter. Lorsque les Truites arc-en-ciel sont par la suite exposées à un champ de lumière dépolarisée, elles ne changent pas le orientation pas de nouveau et gardent celle qu’elles ont adoptée lors de l’exposition aux vecteurs E. [Traduit par la Rédaction]

Introduction

415

polarized stimuli that are matched for intensity but differ in Degner and Hawryshyn

Even though animals such as fish may demonstrate polarization sensitivity, they may not be capable of polarization vision (Kirschfeld 1973). When an animal behaviourally orients to an E-vector (electric vector), it is demonstrating polarization sensitivity. Polarization vision is more complex in that it involves the ability to discriminate between two linearly Received June 13, 2000. Accepted December 1, 2000. Published on the NRC Research Press Web site on February 26, 2001. S.L. Degner and C.W. Hawryshyn.1 Department of Biology, University of Victoria, P.O. Box 3020, Station CSC, Victoria, BC V8W 3N5, Canada. 1

Corresponding author (e-mail: [email protected]).

Can. J. Zool. 79: 407–415 (2001)

J:\cjz\cjz79\cjz-03\Z00-221.vp Wednesday, March 07, 2001 8:44:50 AM

polarization angle and (or) degree of polarization (Bernard and Wehner 1977). For instance, Shashar and Cronin (1996) trained Octopus sp. to select a square that was made from contrasting polaroid filters. Octopuses were then simultaneously presented with two square polaroid filters to test whether they could discriminate between them. One of the squares contained contrasting polaroid filters, the other did not. The octopuses continued to select the training square that contained contrasting polaroid filters. Migrating salmon may use celestial polarized light patches as navigational cues. Celestial hemisphere simulation has been used to study orientation of rainbow trout (Oncorhynchus mykiss) to a linearly polarized light field. Rainbow trout were trained to orient either perpendicular or parallel to an imposed E-vector orientation. When the imposed E-vector orientation was rotated 90°, the fish continued to orient to

DOI: 10.1139/cjz-79-3-407

© 2001 NRC Canada


408

the E-vector in accordance with the trained E-vector orientation (Hawryshyn and Bolger 1990; Hawryshyn et al. 1990; Parkyn 1998). Because there can be multiple polarized light patches that differ in polarization angle, intensity, and (or) degree of polarization in the celestial hemisphere at any given time, fish with polarization vision may be able to detect differences between certain E-vector orientations and then orient accordingly. Recent research has demonstrated single-neuron E-vector coding in rainbow trout. Coughlin and Hawryshyn (1995) measured the polarization sensitivity curves of colour-opponent biphasic units. They determined that the ultraviolet (UV)sensitive-cone on-response component of biphasic units was sensitive to vertically polarized stimuli. Conversely, the red-sensitive-cone off-response component was sensitive to horizontally polarized light. The polarity of response to a UV-polarized stimulus changed from an on-response at an E-vector of 0°–30° to an off-response at 60°–120° and then back to an on-response at 150°–180°. Coughlin and Hawryshyn (1995) concluded that rainbow trout have a twochannel polarization visual system and the resultant polarization sensitivity curve, termed a W function, is due to the interplay between the two differentially sensitive polarization channels. The description of a two-channel polarization system by Coughlin and Hawryshyn (1995) has allowed us to make some predictions about the ability of rainbow trout to discriminate E-vector orientation in the stimulus field. In attempting to mimic more realistic conditions, we felt it was important to conduct an experiment where trout would be presented with more than one overhead linearly polarized stimulus. The experimental paradigm used to test for Evector discrimination performance was one where rainbow trout were trained to orient under two patches of polarized light with the same E-vector orientation. Once the trout were trained to orient correctly under the training conditions, the E-vector orientation in one patch was systematically varied in 15° angular deviations from the other patch. Based on our knowledge of how visual neurons process E-vector information, we made the following predictions: (i) As E-vector is varied, trout would continue to orient along the trained axis when the two imposed E-vector orientations differed by 30° or less because only the vertical detector would be activated. Consequently, only one E-vector orientation would be perceived. (ii) When the two imposed E-vector orientations differed between 30° and up to 75°, the trout would not significantly orient along the trained axis, since there may be an interaction between vertical and horizontal detectors over this range of E-vector orientation differences. Consequently, both the vertical and horizontal detectors would be stimulated and rainbow trout might be capable of perceiving two E-vector orientations. In such situations trout would choose one or the other E-vector, hence their orientation would resemble a random distribution. (iii) When the two imposed E-vector orientations differed by 90°, both the vertical and horizontal detectors would be stimulated and one would observe orientation along two orthogonal axes (vertical and horizontal). (iv) When a diffuser was used, thus depolarizing the linearly polarized light field, rainbow trout would be randomly distributed. Our analysis was directed at examining these specific predictions.

Can. J. Zool. Vol. 79, 2001

Materials and methods Animals Wild stock rainbow trout (Badger Lake, British Columbia), 4– 5 g body mass, were obtained from the Fraser Valley Fish Hatchery (British Columbia Ministry of Environment), Abbottsford, B.C., Canada. Fish were held individually in 20-L continuous-flow-through aquaria under a 12 h light : 12 h dark photoperiod with a mean water temperature of 15 ± 1°C for a minimum of 4 weeks prior to training. Fish were fed Vextra Pelago® (2.0 mm) pellets (Ewos, Vancouver, B.C.) every second day during that month. While in training, fish were fed only during training sessions. During testing, fish were fed once a week (at least 2 days prior to being tested) to maintain health and motivation.

Apparatus To generate two patches of linearly polarized light, two identical but separate optical systems were assembled parallel to each other, perpendicular to the ground, and spaced 20.0 cm apart. The angular displacement between the two stimuli was 15.7° and the angular subtense of the patch for a test fish located beneath one stimulus was 4.3°. Each stimulus channel was illuminated by a 250-W lamp (quartz tungsten–halogen). A parabolic mirror was placed behind the lamp and a condensing lens (Edmund Scientific) was placed in front of the lamp to form an image of the lamp’s filament on a projection lens (Edmund Scientific). The beam of light was subsequently projected onto the following filters in a filter tray: heat glass (Corion), BG-18 filter (Schott), and 0.3 neutral-density filter (Corion). The BG-18 filter (Schott) was chosen for its close approximation to the spectral characteristics of Lake Cowichan, British Columbia, Canada, which is a salmonid nursery lake (Novales Flamarique et al. 1992). The beam of light was then focused on a UV-grade HNP′B linearly polarizing filter (Polaroid Corp.). To avoid the interaction of inherent polarized light from the optical system and the polarizing filter, a sheet of tracing paper was placed before the polarizer as a diffuser to ensure uniform illumination of the polarizer; in the control experiment only, an additional sheet of tracing paper was placed after the polarizers to effectively depolarize the light fields. To minimize spurious cues, all experiments were conducted in a self-contained dark room (the walls and ceiling were painted flat black). To further reduce any spurious visual cues, a black felt curtain was placed around the tank and the optical system. Two 7.5 cm diameter holes were cut out of a piece of black Coroplast® sheeting that was horizontally centred in front of the polarizers to allow only light from the polarizers to pass through to either the training or the testing tank.

Characterization of the light field The spectral distribution of each stimulus condition was measured with an Li-1800 underwater spectroradiometer (Biggs 1984). Both stimulus channels exhibited peak output at approximately 540 nm and appeared blue–green to the experimenter (Fig. 1). The spectral distribution of the control paradigm is similar to that of the training paradigm, but is 0.5 log unit darker. A Photodyne radiometer was used to measure light intensity across the testing tank in the testing and control paradigms, following the protocol of Hawryshyn et al. (1990) (Fig. 2, left-hand side). The detector was fitted with a cylinder (9.4° acceptance solid angle) to measure radiance at several points within the test tank. Figure 2 (right-hand side) illustrates the measured percent polarization of the linearly polarized and depolarized light fields (for calculations see Hawryshyn and Bolger 1990).

Training In addition to the optical system, a 60-W background light was suspended from the ceiling outside of the curtain to allow the © 2001 NRC Canada



Degner and Hawryshyn

409

Fig. 1. Spectral irradiance of stimuli for training and testing paradigms. Spectral irradiance (log photons·m–2·s–1) was measured directly under each stimulus channel using a Li-1800 underwater spectroradiometer. The spectral irradiance of the stimuli for the control paradigm is shown beneath the trained paradigm.

entire training tank to be equally illuminated (Fig. 3a). The motordriven gate, which separated the response compartment from the neutral compartment of the Plexiglas training tank, partially blocked the light from the stimuli from reaching the neutral compartment. This background illumination ensured that the test rainbow trout were not dark-adapted, as polarization vision disappears under dark adaptation (Parkyn and Hawryshyn 1993). The background light bulb was hidden from the field of view of the fish so that it could not be used as a directional cue. This background light was not used during testing, since the entire test tank was well illuminated by the optical system. The fish were confined to the neutral compartment for at least 30 min prior to training sessions, and were trained following the procedure outlined in Hawryshyn et al. (1990). The behaviour of each fish was observed from behind an occluding curtain. Twelve fish were trained to swim parallel to the length of the training tank using operant conditioning. The E-vector orientation was set on both polarizers to be either parallel or perpendicular to the length of the training tank. Five rainbow trout were individually trained to swim parallel to the E-vector orientation, while an additional seven rainbow trout were trained to swim perpendicular to the E-vector orientation. Training sessions for individual fish were spaced at least 1 day apart. To allow the fish to light-adapt to ambient light levels, each fish was held for at least 30 min prior to each training session in a 20-L continuous-flow-through aquarium located near the experimental area. In session 1, the fish was placed in the neutral compartment of the training tank with the motordriven gate left open, allowing it to swim throughout the tank. If the fish swam freely throughout the tank within an hour, it was returned to its individual tank. If the fish either made fast-starts throughout the tank or did not swim freely throughout the tank, it was discarded and not used again. In session 2, the motor-driven gate was closed when the fish was placed in the neutral compartment. A food pellet was placed in the response compartment by the gate while the fish was adapting. After 20–30 min the gate was opened. If the fish did not move towards the gate within 10 min, it was guided with a glass rod. When the fish consumed the food pellet in the response compartment, a second pellet was given for positive reinforcement. Pellets were always given as positive reinforce-

ment, and no negative reinforcement was used. The fish was then guided back towards the neutral compartment and the gate was closed. This pattern was repeated six more times, effectively giving each fish seven trials with a total training time of 45 min. In session 3, the fish was placed in the neutral compartment with the gate closed. After 20 min the gate was opened and the fish was given one food pellet once it successfully swam through the gate without being guided. Upon eating the first pellet it received a second for further reinforcement. This pattern was repeated for a total of seven trials also. In session 4 and subsequent training sessions, the fish was placed in the neutral compartment with the gate closed for approximately 20 min. The gate was then opened and the fish was given a pellet only if it swam in the training direction for at least half of the length of the response compartment. Once the fish was swimming freely in the trained direction at least halfway through the response compartment in all seven trials, a partial reinforcement schedule was started. In the first partial reinforcement session the number of trials was increased to 12. The fish was fed one pellet for the first correct response and every second thereafter. Once the fish was responding correctly (i.e., swimming freely at least halfway into the response compartment) at least 80% of the time, it was fed on the first correct response and every third correct response thereafter. Once this level of performance was attained, the number of trials was decreased to 10 and the fish was fed on the first and every fourth correct response. This allowed the fish to be fed three times during a session, receiving a pellet on the last or second last trial. This kept the test fish reinforced until the next training session. Once fish responded correctly at least 80% of the time in the final partial reinforcement paradigm, testing began. Training sessions were interspersed between testing sessions. The number of training sessions required to train each fish was 17.2 ± 4.5 (mean ± SD; n = 12 fish).

Testing The testing tank (diameter 76.3 cm) consisted of a round tank centred directly beneath both stimulus channels (Fig. 3b). The inside of the tank was painted flat neutral gray, and a scoring line (width 1.0 mm, radius 32.0 cm) was drawn with a black felt marker © 2001 NRC Canada



410


Fig. 2. Radiance of the polarized light field. Radiance (log W·cm–2·s–1·sr–1, where sr is steradian), shown in the left-hand column, was measured along eight arms in the test tank at the water surface using a Photodyne radiometer. Percent polarization of the polarized light field is shown in the right-hand column. (a) Both E-vector orientations set at 0°. (b) Both E-vector orientations set at 90°. (c) Diffuse depolarized light field.

to facilitate recording of the final response of the fish. Prior to testing, each fish was left in the adaptation tank for 30 min. The fish was then retained under a release box constructed of clear Plexiglas in the centre of the testing tank for 20 min. A trial began when the release box was lifted. The behaviour of the fish was observed

and recorded from behind an occluding curtain. For each trial, the location and time when the test fish crossed the scoring line was recorded. A trial ended when the eye of the test fish passed the scoring line. The test fish was then netted and returned to the release box, where it was held for 1–3 min before the next trial

© 2001 NRC Canada




411

Fig. 3. (a) Side view of the optical system and training tank. Stimulus channel 1 (ST.CH. 1) is the same as stimulus channel 2 (ST.CH. 2) and is directly behind it. (b) Front view of the optical system and testing tank. B, plywood box; B.C., black Coroplast®; B.G., BG-18 color filter; B.L., background light; C, condensing lens; D, diffuser; E1, orientation of E-vector in stimulus channel 1; E2, orientation of E-vector in stimulus channel 2; F.C., felt curtain; H.G., heat absorbing glass; N.C., neutral compartment; N.D., neutral density filter; P, polarizer; P.L., projection lens; P.M., parabolic mirror; R.C., response compartment; S, source.

began. This procedure was repeated for a total of 10 trials for each fish. Fish were not fed during testing and were retrained every 7– 10 days. Each fish was individually tested, with at least 24 h between testing sessions. To determine whether fish were fully trained to the polarized light field, they were first tested with the training paradigm conditions (both polarizers were set at the same E-vector orientation). If a fish significantly oriented in the trained direction (Hotelling’s one-sample test, P < 0.05), it was then tested in the remaining paradigms. For each of the remaining tests, the E-vector orientation of one of the two linearly polarized light patches was varied by 15° between 0° and 90°. Each fish was therefore tested in seven linearly polarized light paradigms in which the two E-vector orientations differed by 0°, 15°, 30°, 45°, 60°, 75°, and 90°. The polarized light patch in which the E-vector orientation was varied was randomly switched from stimulus channel 1 to stimulus channel 2 using a random number table. Once all 12 fish had been tested in the seven paradigms, each fish was tested under a diffuse depolarized light field following the same testing protocol.

Analysis For all eight testing paradigms, the mean axis of orientation and vector length (r), or first-order mean angular responses (MARs),

were determined for each fish (Batschelet 1981; Zar 1984). Vector length is directly proportional to the strength of orientation of a test fish within a testing paradigm. Although there is variation within the responses of an individual fish, a strong orientation can still be observed. If an individual was strongly oriented, its MAR would show a strong vector length along the mean axis of orientation. If an individual randomly chose one E-vector orientation over the other, the vector length would be shorter and the mean angle would shift from the trained axis (i.e., 0°–180°) towards the average angle. Likewise, if an individual oriented to the average angle of the two stimuli, it would show a strong vector length along the average angle. The MARs of each fish were doubled to test for significant bimodal distributions. Since E-vector orientation is not based on a single direction such as a geographical bearing, the angular data were doubled to convert biomodal distributions (Zar 1984). Hotelling’s one-sample test was used to determine whether the population had a significant mean direction within each of the testing paradigms (i.e., the population centre deviated significantly from the origin (P < 0.05)). Following Batschelet (1981) and Zar (1984), Hotelling’s two-sample test was then used to determine whether the population centres (previously determined using Hotelling’s one-sample test, P < 0.05) of the two groups of fish (parallel-trained versus perpendicular-trained) deviated significantly from each other. © 2001 NRC Canada



412 To determine if population variance increased as the two polarized light patches became increasingly different, the angular deviation of each testing paradigm was calculated following the method outlined in Zar (1984). The Mann–Whitney test was used as a rank test to determine whether there was a significant difference in dispersion between the training and testing paradigms. The doubled angles for all 12 fish tested under the training paradigm (i.e., the two E-vector orientations were the same) were ranked against the doubled angles for all 12 fish in the remaining six testing paradigms (i.e., the two E-vector orientations differed by 15°, 30°, 45°, 60°, 75°, or 90°).

Results Testing fish trained with two linearly polarized light patches The bearings of individual fish within each testing paradigm were doubled to test for significant bimodal distributions. The doubled-angle distributions (Hotelling’s one-sample test, P < 0.05) of the two groups of fish did not differ significantly from each other (Hotelling’s two-sample test, P > 0.05), demonstrating that presentation of the stimuli had little or no effect. Therefore, the data for all 12 fish were combined. The first-order MARs of the 12 fish in the training paradigm (i.e., both E-vectors were set at the same orientation) are shown in Fig. 4a. Both E-vector orientations were set at 0° to test the parallel-trained test fish, whereas they were both set at 90° to test the perpendicular-trained test fish. Since both groups of test fish were trained to swim towards 0° (i.e., the E-vector orientation was adjusted), all data are plotted with 0°, or the trained direction, at the top of each circle plot. The bimodal behavioural data are shown at the left-hand side. Each solid line represents the MAR of one fish. This was determined from an individual fish’s bearings within one testing paradigm. Given that the vector lengths for most individuals are long and are along the trained axis, most of the test fish were strongly oriented along the trained axis. Broken lines across the diameter of the bimodal behavioural circle plots indicate the axis of the second-order MAR of the population, which was determined from the first-order MARs of the individuals. In Fig. 4, the doubledangle distribution of the 12 test fish is shown to the right of the bimodal behavioural circle plot for each testing paradigm. The population is significantly oriented when the 95% confidence ellipse created by Hotelling’s one-sample test does not include the center of the circle plot (Hotelling’s onesample test, P < 0.05). All 12 test fish significantly oriented in a bimodal distribution along the predicted axis when the two E-vector orientations were the same (Hotelling’s onesample test, P < 0.05) (see Table 1 for statistical analysis of the second-order MAR of the population). In the test paradigm where the two E-vector orientations differed by 15° (Fig. 4a), one E-vector orientation was set at 0° and the other was set at 15° for the parallel-trained test fish. To test the perpendicular-trained test fish, one E-vector orientation was set at 90° and the other at 75°. This format was followed for the remaining test paradigms. Significant orientation was observed in the test paradigms where the two E-vector orientations differed by 0°, 15°, 30°, 45° (Fig. 4a), and 90° (Fig. 4b). As shown by the vector lengths, most individual fish were strongly oriented along the trained axis when the two E-vector orientations differed by 0°, 15°, 30°,


45°, and 90°. Alternatively, fewer individuals were strongly oriented along the trained axis when the two E-vector orientations differed by 60° and 75°. This is shown in the bimodal behavioural circle plots, where more of the individual vector lengths are shorter and are not along the trained axis. Consequently, fish did not exhibit significant orientation when the two E-vector orientations differed by 60° and 75° (Fig. 4b). Low angular dispersion was observed when rainbow trout showed a bimodal distribution (i.e., the two E-vector orientations differed by 0°, 15°, 30°, 45°, and 90°) (Fig. 5). Although there was not a significant difference in angular dispersion between the seven testing paradigms (Mann– Whitney test, P > 0.05), angular dispersion was lowest when the two E-vector orientations differed by 0° and 15°. Angular dispersion was greatest when the population did not significantly orient (i.e., the two E-vector orientations differed by 60° and 75°). Testing fish under a diffuse depolarized light field Twelve juvenile rainbow trout that were trained and tested under a linearly polarized light field presented by two stimuli were tested under a diffuse depolarized light field. All of the vector lengths of the individual fish were weak, and most were not oriented along the trained axis. The population was not significantly oriented (Hotelling’s one-sample test, P > 0.05) (Fig. 4b). See Table 1 for the results of the statistical analysis.

Discussion Previous studies have shown that fish such as juvenile rainbow trout (Hawryshyn and Bolger 1990; Hawryshyn et al. 1990; Parkyn 1998), juvenile sockeye salmon (Groot 1965), halfbeaks (Forward et al. 1972; Forward and Waterman 1973), and goldfish (Kleerekoper et al. 1973) are capable of orienting to linearly polarized light fields. The present study has extended the work on juvenile rainbow trout by simultaneously presenting the test fish with two linearly polarized light patches. Although there have been studies of polarized light discrimination in fish, the present study is unique in that the stimuli were presented from above and not from the side, the latter being more appropriate for testing species that may use polarization vision for contrast enhancement (Davitz and McKaye 1978). The celestial hemisphere exhibits a predictable pattern of E-vectors where polarization angle, intensity, and (or) degree of polarization differ in a point-for-point representation over the sky vault. Rainbow trout may be able to detect E-vector differences in different positions in the sky, and orient accordingly. The data for all 12 fish were combined, since the Hotelling’s one-sample test results did not differ. This demonstrated that the orientation of the imposed E-vector during training (i.e., either parallel or perpendicular to the fish) had little overall effect on the orientation response of rainbow trout. All 12 fish significantly oriented when the two stimuli had E-vector orientations, which differed by 0°, 15°, 30°, 45°, and 90°. However, the test fish were not significantly oriented when the two E-vector orientations differed by 60° and 75°. This observation is consistent with the observation that rainbow trout have two-channel polarization vision, recently proposed by Coughlin and Hawryshyn (1995). Coughlin and Hawryshyn © 2001 NRC Canada




413

Fig. 4. Mean angular responses (MARs) of rainbow trout (Oncorhynchus mykiss) to various testing paradigms. The bimodal behavioural data are shown at the left-hand side, whereas the doubled-angle distribution is shown at the right. Each solid line represents the first-order MAR of one fish. Solid lines without circles represent fish trained to orient parallel to the E-vector orientation, whereas solid lines with solid circles represent fish trained to orient perpendicular to the E-vector orientation. The second-order MAR of the population is shown by a broken line across the diameter of the bimodal distribution circle plot. Vector length (r) is proportional to the strength of orientation with the diameter of the bimodal distribution circle plot, and the radius of the doubled-angle distribution circle plot, corresponding to r = 1. Data are plotted with respect to the direction that both groups of fish were trained to swim towards (the top of each circle plot, or 0°). (a) The two E-vector orientations differed by 0°, 15°, 30°, and 45°. (b) The two E-vector orientations differed by 60°, 75°, and 90°; fish were tested under a diffuse depolarized light field.

(1995) measured the polarization sensitivity curves of colour-opponent biphasic units in rainbow trout. By making single-unit recordings, they determined that the UV-sensitive cone on-response component of biphasic units was sensitive to vertically polarized light, whereas the red-sensitive cone off-response component was sensitive to horizontally polarized light. Sensitivity was lower when the E-vector orienta-

tion was between 0° (i.e., vertical) and 90° (horizontal) and between 90° (horizontal) and 180° (i.e., vertical) than when the E-vector orientation was set at 0°, 90°, or 180°. This polarization sensitivity, termed a W function, is due to the interaction of two E-vector channels in the polarization visual system of juvenile rainbow trout. Because the resolution of this sensitivity curve has not yet been clearly defined, it is © 2001 NRC Canada



414

Can. J. Zool. Vol. 79, 2001 Table 1. Statistical analysis of the second-order mean angular response (MAR) data concerning the bimodal distribution of rainbow trout (Oncorhynchus mykiss), using Hotelling’s one-sample test from circular statistics. Difference in E-vectors (deg.)

n

Mean axis

Mean vector length

Mean angular deviation

P (Hotelling’s test)

0 15 30 45 60 75 90 Depolarized

12 12 12 12 12 12 12 12

0–180 2–182 1–181 17–197 na na 7–187 na

0.34 0.29 0.26 0.26 na na 0.29 na

32.9 34.1 34.9 34.9 na na 34.1 na