Orientation with reference to the time-compensated sun-azimuth compass has been established for the homing pigeon Columba livia. Previous qualitative studies claim that pigeons are sensitive to the orientation of a polarizer and it has been suggested that these animals are able to use sky-light polarization as an indirect reference to the sun’s position when the latter is shielded from view. We report experiments which were undertaken to quantify the sensitivity of the homing pigeon to the orientation of linearly polarized light. The results of our initial experiments suggested that the animals responded to secondary cues. Further experiments were carried out to avoid such artefacts. Under circumstances where secondary cues were rigorously avoided, we were, however, not able to demonstrate any directional response that was caused by the E-vector orientation of the illumination. These results throw doubt on the suggested polarization-sensitivity of birds in general.

To choose a compass direction, a bird preferentially utilises the time-compensated sun-azimuth compass (Schmidt-Koenig, 1979; Neuss and Wallraff, 1988; Wiltschko and Balda, 1989). Release experiments with clock-shifted pigeons clearly show that even when the information provided by the sun’s position is in conflict with other possible sources of information, such as the direction and inclination of the geomagnetic field, the former primarily determines the direction of departure. Moreover, a deterioration of visual cues, such as that caused by mist during a flight, reduces the success of homing (Schietecat, 1988a, b), which emphasises the importance of visual information during flight.

The pattern of partially polarized light in the (blue) sky is linked to the position of the sun. Information about the position of the sun is present in the direction of the electric vector (E-vector) of this linearly polarized light, albeit with an uncertainty of 180°. This angle of polarization depends on the relative positions of the sun, the observer and the point observed. The ability to perceive the orientation of the E-vector could thus be regarded as an extension of the sun compass. It would, for instance, allow an individual to estimate the sun’s position, given a few clear blue patches in an otherwise overcast sky. Since von Frisch’s discovery (von Frisch, 1949) of the perception of polarized light by the honeybee, behavioural as well as electrophysiological responses to linearly polarized light have been demonstrated for many invertebrates and for some species of lower vertebrates (for reviews, see Waterman, 1981; Wehner, 1989).

Evidence that birds are sensitive to the orientation of the electric vector stems from two kinds of experiments. Polarotactic experiments with manipulated sky light were carried out using dusk migrators in Emlen cages covered with a polarizer (for a review, see Able, 1989). The direction of the Zugunruhe was parallel to the E-vector. The polarizer changes all aspects of the incident sky light, so these experiments do not exclude the possibility that other factors that covary with the E-vector manipulation determine the results. For instance, when the polarizer is rotated, the perceived intensity distribution across the ‘sky’ undergoes large changes. Helbig and Wiltschko (1989) did not exclude the possibility that a type of phototaxis may have been responsible for their results.

Further information is obtained from experiments in which the animal’s ability to discriminate between different orientations of the E-vector is tested by employing a forced-choice procedure in the laboratory. Two such studies (Kreithen and Keeton, 1974; Delius et al. 1976) led to the generally accepted conclusion that homing pigeons are able to perceive the E-vector direction of linearly polarized light. From the latter study, it is apparent that the illumination must reach a photopic level (higher than approximately 20 cd m−2). Moreover, ultraviolet components are apparently not necessary because they are not produced in large amounts by tungsten lamps (even less when intensity is reduced by lowering the supply voltage) and they are not readily transmitted by polarization filters (except specialized types). Emphasis is laid on the stimulation of the ventral part of the retina (the dorsal visual field). Delius and Emmerton (1979) claim that this is why Montgomery and Heinemann (1952) did not find any polarization-sensitivity in the pigeon.

In contrast to the situation in invertebrates, in birds there are no known structures which could permit polarization-sensitivity. However, Young and Martin (1984) describe a model in which tight coupling between both members of the double cone could provide the basis for such a mechanism. An examination of the spectral properties of the polarization-sensitive mechanism might provide a test for this model. More specifically, we could expect the sensitivity to be determined by the absorption of light by the P567 pigment, which is screened by the yellow oil droplet of the principal cone. According to Bowmaker’s (1977) calculations, this combination would peak at approximately 567 nm. Previous studies have not conclusively shown how the polarization-sensitivity of the pigeon depends on wavelength, so we set out to measure this function (experiment 1).

The usefulness of polarization perception depends on the resolving power of the detection system. The experiments mentioned above are of a qualitative nature. So far, no-one has measured discrimination thresholds (just noticeable difference; j.n.d.) for two E-vector orientations. Experiment 2 was designed to measure the j.n.d. for sequentially presented stimuli.

Although the experiments were initially designed to test the spectral sensitivity of the polarization-sensitive mechanism of the pigeon (experiment 1) as well as its E-vector discrimination threshold (experiment 2), their purpose, as will be explained below, gradually changed.

Part of this work was presented at the third international symposium of the Northern Eye Institute, Manchester, UK (Coemans and Vos, 1989) and was also part of a short communication (Coemans et al. 1990).

Experiment 1

The pigeon’s sensitivity to the E-vector direction of polarized light was tested in a symmetrical Y-maze. In the centre of the maze, the axis of a stimulus of linearly polarized light matched the axis of a linear polarizer on the ceiling of one of the corridors. Pigeons had to choose this matching corridor.

Apparatus

The apparatus consisted of a wooden Y-maze (Fig. 1A) which was 60 cm in height. Each corridor was covered by a weakly illuminated sheet of a linear dichroic polarizer (HN22, Polaroid Corporation) oriented parallel to the corridor’s longitudinal axis. The orientation of the polaroid filters was checked with a polarizer of known orientation. To obtain diffuse light, these polarizers were covered with a sheet of semi-opaque Perspex. Luminance measured under the filter ranged between 30 and 50 cd m−2. At the point where the three corridors met, the ceiling consisted of a circular tray (20 cm inner diameter) containing a polarizer covered with a sheet of semi-opaque Perspex. This combination could be rotated to any orientation in either direction by a stepper motor. Its absolute orientation was encoded by means of three photodetectors, each of which corresponded with one of the corridors. The change to a new orientation took place in total darkness. A random trajectory was used in order to exclude auditory cues.

Fig. 1.

(A) Y-maze for testing the perception of polarization. The bird was trained to walk from one corridor (I, II or III) into another on the basis of the orientation of the central polaroid filter (CP). Each corridor had its own polarizers (P) in order to facilitate training (matching to sample). Ph, photodetectors; F, pneumatic feeders; CL, illumination of the corridor; L, central illumination. (B) Schematic course of the experiment. See text for further details.

Fig. 1.

(A) Y-maze for testing the perception of polarization. The bird was trained to walk from one corridor (I, II or III) into another on the basis of the orientation of the central polaroid filter (CP). Each corridor had its own polarizers (P) in order to facilitate training (matching to sample). Ph, photodetectors; F, pneumatic feeders; CL, illumination of the corridor; L, central illumination. (B) Schematic course of the experiment. See text for further details.

The central polaroid was illuminated from above with a 100 W, 3400 K halogen lamp (Osram) shielded by a heat-reflecting filter (Kodak IR301). Intensity (maximum luminance of the polarization source was 240 cd m−2) and colour could be varied by inserting nickel alloy neutral-density filters (Oriel Corporation) and interference filters (Balzers) into the light path. Unpolarized light passing non-normally through a flat surface will be polarized to some extent. As a result of this, the surface of a subsequent polarizer will not be illuminated homogeneously. A weak pattern coupled to the orientation of the polarizer, and maximal at the edges of the filter (about 4% intensity modulation in our case), will therefore be formed. Changing the distance of the illuminator from the surface alters the maximum angle of incidence and thus the visibility of this pattern. To check whether the animal responded to this unwanted cue, a test was designed in which the intensity of another light source was varied by changing its distance from the filter surface or its power (100 W, 120 W or 200 W, tungsten, Philips). Preliminary studies of the effect of colour were carried out with this source as well. We used the following broadband colour filters: cinemoid no. 24 (green; passband with peak at 510 nm; approximately 90 nm half-width); no. 45 (blue, very broad transmission characteristic); and no. 46 (orange; 50% cut-off point at 550 nm).

The inside of the Y-maze was initially covered with matt black paint. At a later stage, different materials were used (sandpaper and blotting paper) for reasons that are discussed below. In some experiments, masking lights, consisting of small tungsten lamps (3 W), were mounted above the entrance of each corridor. The filaments of these lamps were shielded from view to prevent glare.

To prevent a bird simply walking from one corridor to another, a 4 cm high cardboard platform was placed in the centre. The bird therefore had to jump onto and down from this obstacle. At the end of each corridor, an illuminated air-pressure-driven foodhopper delivered mixed grains. At the entrance of each corridor, photodetectors monitored the bird’s presence. All functions of the apparatus were controlled by a microprocessor.

Subjects

Initially, four mature homing pigeons, all of which had free-ranging experience, were used in the experiment. They were housed indoors with daylight illumination. The birds were deprived of food (at 80% of their free-feeding weight) and kept on water and grit. Three of the pigeons were successfully trained manually (more than 90% correct choices). Two of them performed well in the fully automated procedure.

Procedure

A session started when a subject was placed into one of the corridors. It had to choose between the two other corridors on the basis of the orientation of the central filter. Choosing the corridor lined up with the polarizer resulted in a 6 s access to food at the end of the corridor (Fig. 1B). An incorrect choice resulted in 22 s of total darkness, during which all activity was suspended (a time-out). The chosen corridor was taken as the starting point for the next trial. One of the other two corridors was randomly assigned as the next goal. The next trial started when the filter had rotated to its new position, which took about 5 s. Results are presented as proportions of correct choices. Each session lasted 1 h and consisted of approximately 100 trials.

Experiment 2

The sensitivity of pigeons to the E-vector direction of linearly polarized light was tested in a single-key Skinner box with an overhead source of polarized light. Pigeons were rewarded with food when they pecked the key when the stimulus had a chosen orientation [the number of pecks when the positive stimulus (S+) was presented is designated as RS+]. Pecking the key when the stimulus’ orientation was orthogonal to this led to a time-out [the number of pecks when the negative stimulus (S) was presented is designated as RS-].

Apparatus

The apparatus was made of white plastic-covered chipboard (40 cm×30 cm×70 cm; Fig. 2A). Its inner walls were sandpapered in order to obtain a matt white appearance. At a later stage (during experiment 2b and all subsequent experiments, as explained below) these surfaces were covered with large sheets of white blotting paper. A pecking key was located 15 cm above the floor of the box. In order to minimize reflections, it was either illuminated from behind or covered with rough plaster. Underneath it, sunk into both the wall and the floor, was a pigeon feeder driven by air pressure.

Fig. 2.

(A) Apparatus for testing polarization perception in a geometrically simple environment. It employed a single pecking key and feeder apparatus. The number of pecks per second (x) given during stimulus presentation determined the duration of access to food (time=2x+1) after S+ (to a maximum of 8 s), or the length of a time-out (time=8x+2) after S-(to a maximum of 20 s). P, polarizer; K, pecking key; R, room lights; L, light source; F, feeder; A, additional cue (coloured spots); the arrows indicate that the position of the help cue could change from session to session (see text). (B) Schematic course of experiments a and b described in the text. (C) Schematic course of experiments c and d.

Fig. 2.

(A) Apparatus for testing polarization perception in a geometrically simple environment. It employed a single pecking key and feeder apparatus. The number of pecks per second (x) given during stimulus presentation determined the duration of access to food (time=2x+1) after S+ (to a maximum of 8 s), or the length of a time-out (time=8x+2) after S-(to a maximum of 20 s). P, polarizer; K, pecking key; R, room lights; L, light source; F, feeder; A, additional cue (coloured spots); the arrows indicate that the position of the help cue could change from session to session (see text). (B) Schematic course of experiments a and b described in the text. (C) Schematic course of experiments c and d.

The ceiling of the box contained the source of polarized light and two room lights. The polarized stimulus consisted of light (from a 50 W reflector tungsten–iodide lamp, Philips) passing through a white semi-opaque Perspex diffusor and a linear dichroic polarizer (HN22, Polaroid Corporation). The ceiling had a luminance of approximately 200 cd m−2 (in the photopic luminance range). At a later stage, the ceiling was moved up to a height of 1 m above the box floor, leaving a gap of 30 cm between the cover and the rim of the box. This removed the reflections on the wall directly below the filter and made it possible to use a video camera in order to monitor the bird’s behaviour. The polarizer could be rotated to any position in either direction with the aid of a stepping motor. Its position was controlled by means of four fixed photodetectors. The filter was rotated to a new position in complete darkness, following a random trajectory in order to exclude auditory cues.

As the experiments proceeded, additional cues were paired with the polarization cue, to facilitate training. Four types of experiment were conducted. (a) No extra cue; (b) flicker; implemented by switching S on and off (10 Hz under software control); (c) colour; delivered as circular spots (4 cm in diameter), initially projected next to the key and later next to the polaroid filter (Fig. 2A), produced by 50 W a.c.-driven tungsten–iodide lamps in combination with an OG570 or BG38 (Schott) filter and neutral-density filters (cinemoid-and inconel-coated glass substratum); (d) orientation; formed by rectangular pieces of black cardboard 20 mm wide and of different lengths between 20 and 200 mm. These were placed on top of the semi-opaque Perspex, parallel to the E-vector axis.

All functions of the Skinner box, as well as the course of the experiment, were computer-controlled. Initially, sessions lasted 1 h. At the end of the experiments, they lasted up to 5 h (approximately 50 trials per hour).

Subjects

Four inexperienced homing pigeons, all having several years of free ranging experience, were used in the experiment. They were kept indoors with daylight illumination (north-facing window). The birds were deprived of food (at 80% of their free-feeding weight) and kept on water and grit.

Procedure

A schematic course of a trial in experiments a and b is depicted in Fig. 2B. Initially, the room lights were on. After the first key peck, they were dimmed and the polarization source was switched on. During the next 14–20 s (randomly determined), the number of key pecks was counted and the mean number of pecks per second calculated. This pecking activity determined the duration of food access (1–8 s) after S+ as well as the length of a time-out (total darkness during 2–20 s) after S (see the legend of Fig. 2B). In order to maintain a strong coupling between pecking activity and food access, food was delivered immediately after an additional obligatory peck when the counting period had ended. After a reward or time-out, all lights were extinguished and the polarizer was rotated to a new position. S+ and S were presented in random order (50% each). The number of consecutively presented S+ stimuli was limited to four.

In experiments c and d, the procedure was slightly different (Fig. 2C). In these cases, we measured the reaction time (RT) as well. This is the period between the onset of the stimulus and the first key peck. The reaction time was used as an indication of the direction of the bird’s attention: frontal vision (the neighbourhood of the pecking key) and the upper field of vision (the polarizer). This is explained below. Pecking activity was counted in the 15 s following the first peck, and was used to determine food access or time-out duration. An example of pecking activity using an additional cue of coloured spots near the pecking key is given in Fig. 3A. Because there are two positions of the filter wheel for each E-vector orientation, these positions should be indistinguishable from each other. To make sure that spots, or irregularities in the polaroid filter, did not interfere with the experiment, we regularly compared the responses to these stimulus pairs. Fig. 3B shows an example.

Fig. 3.

(A) Typical recording of pecking responses, as pecks per 15 s, during an experiment. The additional cue consisted of coloured spots next to the pecking key. Filled symbols, responses to the positive stimulus (RS+); open symbols, responses to the negative stimulus (RS). (B) E-vector orientations repeat themselves after half a turn of the filter wheel (180°). This means that, for a specific direction of the plane of polarization, there are two possible orientations of the filter wheel. To diminish the possibility that irregularities in the polarizer or diffusor provided cues for the pigeons, both possibilities were used, each with an equal probability. The figure shows means and standard deviations of the suppression ratios depicted in A, when these are categorized for each orientation of the filter wheel.

Fig. 3.

(A) Typical recording of pecking responses, as pecks per 15 s, during an experiment. The additional cue consisted of coloured spots next to the pecking key. Filled symbols, responses to the positive stimulus (RS+); open symbols, responses to the negative stimulus (RS). (B) E-vector orientations repeat themselves after half a turn of the filter wheel (180°). This means that, for a specific direction of the plane of polarization, there are two possible orientations of the filter wheel. To diminish the possibility that irregularities in the polarizer or diffusor provided cues for the pigeons, both possibilities were used, each with an equal probability. The figure shows means and standard deviations of the suppression ratios depicted in A, when these are categorized for each orientation of the filter wheel.

During all experiments, we computed the moving average and standard deviation of the pecking rates over the last five trials when the positive stimulus was presented. The mean minus 1 S.D. was taken as a criterion level, which was used during the experiment to evaluate the bird’s response. When the pecking rate elicited by the next negative stimulus was below this criterion, the response was considered correct. In experiments c and d this meant that the next trial was an S+ (a postponed reward); otherwise an S- was given as a correction trial. The results of these trials were discarded.

To analyze the data, we used the suppression ratio (SR; Smith, 1970), as well as the difference in reaction time (DRT). In both cases, the response to S (RS) was compared with the response to the immediately preceding S+ (RS+). We calculated the suppression ratio as follows:
The equation is used to ensure that pooled results have a distribution which is symmetrical around zero, when RS+ is approximately equal to RS-; otherwise, random fluctuations would bias the ratio to negative values. The reaction time difference was defined as:
We computed 95% confidence intervals (t-test) for the reaction time data as well as the suppression ratio data. The validity of P-values computed by means of the t-test when testing the significance of the difference between RS+ and RS was regularly checked by a permutation test (a random sample of N=1000 or N=10 000 was drawn from the population of all possible permutations). Both tests were in good agreement, especially when the differences between RS+ and RS were small.

Experiment 1

After installation of the cardboard platform, the birds learned (>80% correct choices in five subsequent sessions) to choose the corridor indicated by the orientation of the polarizer in the automated procedure. The luminance of the polarization source was 100 cd m-2. We next measured the performance as a function of the intensity of the central light source when this was varied by changing the lamp height (Fig. 4A). Discrimination performance was found to be a function of luminance, which resembled the earlier finding of Delius et al. (1976). This suggested to us that the above-mentioned intensity patterns on the filter did not have any undesirable effect. However, visual inspection of the experimental chamber, while the central filter was rotating at a constant moderate speed (0.5 Hz), revealed distinct brightness cues reflected from the inner walls of the Y-maze.

Fig. 4.

The combined psychophysical functions of the two birds from the fully automated procedure. Discrimination performance is plotted as a function of intensity (I). Error bars indicate 95% confidence limits. (A) Walls covered with matt black paint. Circles, intensity was changed by varying the height of the light source; squares, results of Delius et al. (1976) for comparison. Both results are very similar. (B) Walls covered with yellowish-white coarse sandpaper, containing grains with specular reflection properties. The intensity of the stimulus source was changed by means of neutral-density filters. The whole curve is shifted somewhat to the left compared with the curve in A. Asterisks indicate differences between corridors, as explained in the text.

Fig. 4.

The combined psychophysical functions of the two birds from the fully automated procedure. Discrimination performance is plotted as a function of intensity (I). Error bars indicate 95% confidence limits. (A) Walls covered with matt black paint. Circles, intensity was changed by varying the height of the light source; squares, results of Delius et al. (1976) for comparison. Both results are very similar. (B) Walls covered with yellowish-white coarse sandpaper, containing grains with specular reflection properties. The intensity of the stimulus source was changed by means of neutral-density filters. The whole curve is shifted somewhat to the left compared with the curve in A. Asterisks indicate differences between corridors, as explained in the text.

After installation of the halogen illumination, we experimented with lighter wall covers in order to diminish these cues. Fig. 4B depicts the birds’ performance when intensity was varied by neutral-density filters, after the walls had been covered with yellowish-white coarse sandpaper. The main difference from Fig. 4A is a slight shift towards lower light levels. The performance, however, depended in part on the corridor that the birds were coming from: coming from corridor I resulted in a significantly steeper drop of the curve as a function of light intensity (𝒳2-test; P=0.0004; asterisks in Fig. 4B; no left/right preferences, 𝒳2-test; P⪢0.05), compared with birds coming from corridor II or III. Spurious intensity cues were still visible to us on the walls of the corridors.

Changing the sandpaper for white blotting paper affected the birds’ performances profoundly: discrimination dropped to chance levels, whereas we expected at least 80% correct choices at this luminance (100 cd m−2) (t-test; N=300; P<0.001). Performance was also affected (60% correct choices) when we mounted lights above the entrance of each corridor in an attempt to mask the intensity cues apparent on the sandpaper walls (t-test; N=245; P<0.001).

The following control experiments were performed in the sandpaper-covered apparatus. Removal of the polarizers above the corridors had no effect (t-test; N=116; P>0.999). Replacement of the central polarizer by a 0.3 log units neutral-density filter resulted in a drop to 50% correct choices (t-test; N=170; P<0.001). Performance was not influenced by insertion of broadband colour filters into the lightpath. Although one has to be careful using a photometer when working with non-human subjects, we estimated that the resulting luminances justified an expectation of 75% correct choices for the green filter (31 cd m-2) and better for the other two filters. We found 77% (green), 78% (blue) and 81% (orange) correct choices, indicating that the addition of these colour filters did not affect the behaviour of the pigeons (approximately 1000 trials for each colour filter).

Experiment 2

Experiment a

Experiments started in a sandpapered matt white box. After the birds had learned to peck the key for a food reward, they were trained daily for about 1 month to acquire a stable pecking rate. After this, pecking was rewarded only when the source of polarized light had the correct orientation (S+). These experiments proceeded first manually and eventually automatically for about 2 months with daily sessions of 1 h. We found no suppression of the pecking rate during S.

Experiment b

We decided to present S in combination with a conspicuous additional cue, a 10 Hz flicker, to three subjects. Pecking suppression was almost complete within one session. Next, the onset of this flicker was delayed by means of a staircase procedure. This delay could be increased or decreased in steps of 10% of the duration for which the stimulus was presented. When RS was below the criterion level, the delay of the flicker was increased, otherwise it was decreased. When the delay reached 100%, S remained steady and the additional cue was not presented.

After three sessions, one bird could discriminate between S+ and S- without the additional flicker cue. We noted that this bird walked around in the box before it pecked the key. To investigate the possibility that this bird scanned intensity differences on the box walls, we changed their reflective properties by covering them with white blotting paper. Following this procedure, discrimination between S+ and S- vanished (Fig. 5A). The results for the other two birds are presented in Fig. 5B. The suppression ratio depended entirely on the delay to the onset of the flicker. When the delay was maximal, the difference between the suppression rates following the presentation of S+ and S was insignificant (N=129; P=0.26).

Fig. 5.

Flicker was presented as an additional cue in experiment b. (A) The results of pigeon A. Bars are mean suppression ratios of consecutive sessions (number of trials ranged from 40 to 140 per session); the three sessions in which flicker was gradually removed are not shown. Left-hand arrow, introduction of flicker; middle arrow, flicker cues removed; right-hand arrow, box walls covered with white blotting paper. (B) Mean suppression ratio (±95% confidence interval) of the final five sessions (962 trials) as a function of the delay of the onset of flicker for pigeons B and C. The delay is presented as a fraction of the total stimulus presentation time. Error bars indicate 95% confidence intervals.

Fig. 5.

Flicker was presented as an additional cue in experiment b. (A) The results of pigeon A. Bars are mean suppression ratios of consecutive sessions (number of trials ranged from 40 to 140 per session); the three sessions in which flicker was gradually removed are not shown. Left-hand arrow, introduction of flicker; middle arrow, flicker cues removed; right-hand arrow, box walls covered with white blotting paper. (B) Mean suppression ratio (±95% confidence interval) of the final five sessions (962 trials) as a function of the delay of the onset of flicker for pigeons B and C. The delay is presented as a fraction of the total stimulus presentation time. Error bars indicate 95% confidence intervals.

Experiment c

Experiments with several other cues demonstrated that these could be distinguished into two classes. One group consisted of cues that were conspicuous, even when the animal was looking towards the pecking key. We used flicker (as above), intensity differences between S and S+, and differently coloured spots accompanying the presentation of S+ and S projected next to the key. The pigeons were easily trained using all of these cues. The other group consisted of cues that could only be seen when the animal looked in a specific direction. These included coloured spots on the ceiling next to the polarizer, or the silhouette of a bar placed on the polarized ceiling. Pairing E-vector orientation with either of these cues did not enable the birds to peck successfully. As we supposed that the E-vector-sensitive part of the eye was located in its upper field of vision, we had, until these experiments, categorized an E-vector stimulus as belonging to the first category. In the next set of experiments, we decided to direct the birds attention towards the polarizer.

The birds were trained to discriminate between S+ and S when polarization direction was paired with differently coloured spots projected on either side of the pigeon key. These additional cues were projected at a higher position in subsequent sessions, until they were eventually next to the polaroid filter. After the birds had been trained in this way to discriminate between stimuli presented from above, the colour cues were successively reduced in intensity from session to session using neutral-density filters. The suppression values were found to be correlated with the intensity of the coloured spots (r=-0.95; P=0.0004). In some experiments, however, some discrimination remained when the suppression data were considered: the suppression ratio remained between 0.1 and 0.3 (P<0.05). This was not so for the reaction time data (Fig. 6A). These showed no significant difference when the additional cue was attenuated by more than 4 log units (P0.05). Examining the birds’ behaviour under these circumstances with a video camera revealed that when the stimulus was switched on, the birds looked towards the polarizer. After a while (e.g. 10 s), they pecked the key a few times. Next, they made scanning movements (they moved their heads continuously up and down) either in front of a wall or in a corner, occasionally giving a peck on the key. As we thought that the birds could be detecting reflection differences, we decided to change the reflective properties of the walls. After the old blotting paper had been replaced with fresh paper, the scanning behaviour did not alter but the performance after the presentation of S+ did not differ from that after the presentation of S-, when the help cue was attenuated by 4 log units or more (Fig. 6B; P⪢0.05). We conclude that ‘used’ blotting paper generates (weak) differential wall reflections because of its smutted surface. Comparing the reaction time data and the pecking data in this way allowed us to decide whether the pigeon was using information provided by a distinct cue, such as the coloured spots, or by ‘diffuse’ information in the neighbourhood of the pigeon key.

Fig. 6.

Coloured spots were projected as an additional cue next to the polarizer in experiment c. Psychophysical functions for pigeon B. Left ordinate and open symbols, suppression ratio (SR); right ordinate and filled symbols, reaction time difference (DRT). Error bars indicate the standard deviation. Each point is the mean of at least 60 trials. Discrimination performance is plotted as function of the intensity of the additional cue. Results obtained from two consecutive blocks of sessions. (A) In the first block, there was an apparent bias caused by differential wall reflections. (B) In the second block the bias disappeared after the blotting paper had been renewed.

Fig. 6.

Coloured spots were projected as an additional cue next to the polarizer in experiment c. Psychophysical functions for pigeon B. Left ordinate and open symbols, suppression ratio (SR); right ordinate and filled symbols, reaction time difference (DRT). Error bars indicate the standard deviation. Each point is the mean of at least 60 trials. Discrimination performance is plotted as function of the intensity of the additional cue. Results obtained from two consecutive blocks of sessions. (A) In the first block, there was an apparent bias caused by differential wall reflections. (B) In the second block the bias disappeared after the blotting paper had been renewed.

Experiment d

From the above experiments we concluded that pigeons are not able to utilize the information provided by the orientation of the E-vector. We had shown that every other (faint) discrimination cue was detected and utilized. One further possible explanation of the apparent ability of pigeons to detect E-vector orientation under certain conditions remained to be examined. Therefore, because an axial cue (the orientation of the polarizer) might not readily be associated with a non-axial cue (flicker, colour), we decided to test the effect of another axial cue.

Two birds were trained to respond to a bar placed upon the Perspex diffusor, parallel to the E-vector direction. This was fairly easy as the birds were already trained to expect information from above (we gradually replaced the coloured spots with an elongated bar). Two experiments were conducted: (1) the polarizer was replaced by a neutral-density filter of 0.3 log units and (2) the polarizer remained in place. The suppression ratios and the reaction times of the birds were measured as a function of the length/width ratio of the bar. We found no significant differences (P⪢0.05) between the curves (Fig. 7), except for the differential reflection effects as described under experiment c (not shown).

Fig. 7.

Typical example of psychophysical functions for pigeon C, when the orientation of a bar was presented as an additional cue in experiment d. Discrimination performance (mean suppression ratio ± S.D.) is plotted as a function of the logarithm of the length/width ratio of the overhead bar. Filled symbols, the polarizer was in place; open symbols, the polarizer was replaced by a 0.3 log unit neutral-density filter. Each point is calculated from at least 100 trials.

Fig. 7.

Typical example of psychophysical functions for pigeon C, when the orientation of a bar was presented as an additional cue in experiment d. Discrimination performance (mean suppression ratio ± S.D.) is plotted as a function of the logarithm of the length/width ratio of the overhead bar. Filled symbols, the polarizer was in place; open symbols, the polarizer was replaced by a 0.3 log unit neutral-density filter. Each point is calculated from at least 100 trials.

When our results are compared with those of Delius et al. (1976), it could be concluded that we had replicated the earlier results (Fig. 4A). However, we are unwilling to attribute our findings to polarization perception, because the performance of the pigeons depended on the wall covering used, a variable not manipulated by Delius et al. (1976).

We realized that painting the inside of the box black was not a good choice. The amount of linearly polarized light which is reflected from a wall depends on the degree and angle of polarization. When the angle of polarization is parallel to the wall (horizontally polarized), the intensity of the reflected light will be maximal; when its orientation is perpendicular to it (vertically polarized), it will be minimal. These effects are ruled by physical laws (Born and Wolf, 1983) and therefore these intensity differences are inevitable. There are two ways to make them subliminal. A rough surface will depolarize the light and thus decrease the difference. A high albedo will increase the background intensity and can thus mask the remaining small differences of intensity caused by polarization. In our experiment with black surfaces, the birds could simply compare light reflected from differently oriented walls. Inhomogeneities in the sandpaper lining affect the maximum difference between those walls that are visible from each corridor. This might explain the somewhat different results for corridor I.

The conformity of the results of Delius et al. (1976) (stimulus: E-vector direction) and our results (stimulus: reflection differences) necessitated another approach to the study of the perception of polarized light. We decided to abandon the Y-maze because it was difficult to avoid reflection differences as there are too many surfaces that can be compared in this respect. We therefore concentrated on the very different experimental design of experiment 2.

The results of these experiments show that the pigeons utilized every cue that was present, except for the orientation of the E-vector. This cue must have been a prominent one, because it consisted of the difference in orientation between orthogonally oriented polarizers. Our confirmation of the negative results obtained by Montgomery and Heinemann (1952) was unexpected, as more recent investigations that claim to demonstrate polarization-detection appeared to be rigorous (Delius et al. 1976; Kreithen and Keeton, 1974). In order to explain the differences between our study and previous studies, the following should be noted.

First, our pigeons clearly did not spontaneously pay attention to cues emanating from a part of their environment that was not covered by their frontal visual field. In experiment 1, we added a cardboard platform. Jumping on and off this platform presumably forced the pigeons to attend to their whole environment. In experiment 2, it was necessary to guide the birds’ attention, by means of help cues, towards the cue being tested. Second, a source of polarized light above the pigeons always produces intensity differences between adjacent walls at eye level. Pigeons are very good at detecting minute luminance differences (Hodos et al. 1985). An obvious way to overcome this is to make the differences subliminal. Roughening the surface has a depolarizing effect, which decreases the maximal difference. As pointed out by Umov (1905), the behaviour of dark surfaces is very different from that of white surfaces when reflecting linearly polarized light. On a dark surface, there are pronounced differences in intensity of reflection depending on whether the E-vector of linearly polarized light is parallel or orthogonal to the reflecting surface. We have measured these maximum intensity differences after reflection (angle of incidence was 45°) on the materials we used. We found that matt black paint produces differences which exceed 0.2 log units, whereas (fresh) white blotting paper yields values less than 0.03 log units. Hodos et al. (1985) found a mean luminance difference threshold of 0.11 log units at 300 cd m−2. Most pigeons should, therefore, easily be able to detect the differences produced by differential reflection of polarized light on a black surface. To eliminate spurious reflections as much as possible, our experiments have shown that the surface’s albedo must be as high as possible.

These points, combined with the remark of Delius and Emmerton (1979) that they had difficulties in training pigeons to discriminate the orientation of an overhead bar, make it plausible that these authors actually trained their pigeons to discriminate intensity differences at eye level. Their experimental apparatus had matt black surfaces. Moreover, every orientation of the polarizer (0°, 45° and 90°) that they presented had a counterpart in a parallel wall of their octagonal Skinner box. This ensures that remaining intensity differences could have been large and were presented optimally for the pigeon. This supposition is supported by an examination of the data presented in Fig. 4A.

Our results emphasize the technical difficulties involved in psychophysical experiments concerning polarization perception. Consider the process of refraction, when light passes non-normally through curved surfaces, such as those of the lamp bulb or lenses of a slide projector. These surfaces polarize a light beam by a few per cent. When we take this into account, an experiment requiring discrimination between a stationary and a rotating polarizer could be performed as a flicker discrimination task. As pigeons are very sensitive to intensity differences, this could be an alternative explanation for the results of Kreithen and Keeton (1974), who employed heart rate conditioning to determine whether pigeons were able to discriminate between a stationary and a rotating polarized light source.

Replacing the polarizer with a neutral-density filter is a control experiment that proves that the polarizer is crucial. However, it does not disentangle E-vector information proper from unwanted reflections (or flicker), so it does not prove that polarization perception is possible. To do this, the E-vector component must be left intact while the unwanted components are varied. Changing the wall covering is such an experiment, as is the application of background illumination.

It could be argued that we often required the birds to couple an on/off cue with an axial cue. It might have been difficult for pigeons to achieve this association. However, in experiment 2 the birds proved to be able to couple the orientation of an overhead bar (an axial cue) with a colour cue (an on/off cue) within a few sessions. This is strong evidence against this proposed argument.

Our experiments and studies that were executed under a natural sky (for instance: Able, 1982; Helbig and Wiltschko, 1989) differ in that the nature of our stimulus is exactly known, whereas the colour and intensity distributions of the natural sky provide additional directional cues, which might explain the positive results in the open-air experiments.

We thank R. J. Loots and W. Maasse for developing the electronic and mechanical equipment and E. M. Brenner for helpful discussion. This work was supported by the Netherlands Organization for Scientific Research (N.W.O.).

Able
,
K. P.
(
1982
).
Skylight polarization patterns at dusk influence migratory orientation in birds
.
Nature
299
,
550
551
.
Able
,
K. P.
(
1989
).
Skylight polarization patterns and the orientation of migratory birds
.
J. exp. Biol.
141
,
241
256
.
Born
,
M.
and
Wolf
,
E.
(
1983
).
Principles of Optics, Sixth (corrected) edition
.
Oxford
:
Pergamon Press
.
Bowmaker
,
J. K.
(
1977
).
The visual pigments, oil droplets and spectral sensitivity of the pigeon
.
Vision Res
.
17
,
1129
1138
.
Coemans
,
M. A. J. M.
and
Vos Hzn
,
J. J.
(
1989
).
On the perception of polarized light by the homing pigeon, Columba livia L
. In
Seeing Contour and Colour
(ed.
J. J.
Kulikowski
,
C. M.
Dickonson
and
I. J.
Murray
), pp.
632
638
.
Oxford
:
Pergamon Press
.
Coemans
,
M. A. J. M.
,
Vos Hzn
,
J. J.
and
Nuboer
,
J. F. W.
(
1990
).
No evidence for polarization sensitivity in the pigeon
.
Naturwissenschaften
77
,
138
142
.
Delius
,
J. D.
and
Emmerton
,
J.
(
1979
).
Visual performance in pigeons
. In
Neural Mechanisms of Behavior in the Pigeon
(ed.
A. M.
Granda
and
J. H.
Maxwell
), pp.
51
70
. New York, London: Plenum Press.
Delius
,
J. D.
,
Perchard
,
R. J.
and
Emmerton
,
J.
(
1976
).
Polarized light discrimination by pigeons and an electroretinographic correlate
.
J. comp. Physiol. Psychol.
90
,
560
571
.
Helbig
,
A. J.
and
Wiltschko
,
W.
(
1989
).
The skylight polarization patterns at dusk affect the orientation behavior of blackcaps, Sylvia atricapilla
.
Naturwissenschaften
76
,
227
229
.
Hodos
,
W.
,
Bessette
,
B. B.
,
Macko
,
K. A.
and
Weiss
,
S. R. B.
(
1985
).
Normative data for pigeon vision
.
Vision Res.
25
,
1525
1527
.
Kreithen
,
M. L.
and
Keeton
,
W. T.
(
1974
).
Detection of polarized light by the homing pigeon, Columba livia
.
J. comp. Physiol.
89
,
83
92
.
Montgomery
,
K. C.
and
Heinemann
,
E. G.
(
1952
).
Concerning the ability of homing pigeons to discriminate patterns of polarized light
.
Science
116
,
454
456
.
Neuss
,
M.
and
Wallraff
,
H. G.
(
1988
).
Orientation of displaced homing pigeons with shifted circadian clocks, prediction vs observation
.
Naturwissenschaften
75
,
363
365
.
Schietecat
,
G.
(
1988a
).
De weerkundige en geofysische invloeden op wedvluchten
.
NPO orgaan
15
,
492
498
(in Dutch).
Schietecat
,
G.
(
1988b
).
De weerkundige en geofysische invloeden op wedvluchten (Slot)
.
NPO orgaan
16
,
521
526
(in Dutch).
Schmidt-Koenig
,
K.
(
1979
).
Avian Orientation and Navigation.
London
:
Academic Press
.
Smith
,
J.
(
1970
).
Conditioned suppression as an animal psychophysical technique
. In
Animal Psychophysics, the Design and Conduct of Sensory Experiments
(ed.
W. C.
Stebbins
), pp.
125
156
.
New York
:
Appleton-Century-Crofts
.
Umov
,
N.
(
1905
).
Chromatische Depolarisation durch Lichtzerstreuung
.
Physik. Z.
6
,
674
676
.
von Frisch
,
K.
(
1949
).
Die Polarisation des Himmelslichtes als orientierender Faktor bei den Tänzen der Bienen
.
Experientia
5
,
142
148
.
Waterman
,
T. H.
(
1981
).
Polarization sensitivity
. In
Invertebrate Visual Centers and Behavior. I. Handbook of Sensory Physiology
, vol.
VII/6B
(ed.
H.
Autrum
), pp.
281
469
. Berlin, Heidelberg, New York: Springer.
Wehner
,
R.
(
1989
).
Neurobiology of polarization vision
.
Trends Neurosci.
12
,
353
359
.
Wiltschko
,
W.
and
Balda
,
R. P.
(
1989
).
Sun compass orientation in seed-caching scrub jays (Aphelocoma coerulescens)
.
J. comp. Physiol. A
164
,
717
721
.
Young
,
S. R.
and
Martin
,
G. R.
(
1984
).
Optics of retinal oil droplets, a model of light collection and polarization detection in the avian retina
.
Vision Res.
24
,
129
137
.