Partial Experience with the Arc of the Sun is Sufficient for all-Day Sun Compass Orientation in Homing Pigeons, Columba Livia

Beginning with the pioneering discoveries of Karl von Frisch with honeybees, Apis mellifera, and Gustav Kramer with starlings, Sturnus vulgaris, it has become clear that a wide variety of species use a time-compensated sun compass in navigation (von Frisch, 1950, 1965; Kramer, 1952a,b). To use the sun as a compass, an organism must compensate for the apparent changes in the azimuth of the sun throughout the day. These compensations are made by relying on an internal sense of time that is integrated with measurements of the changing position of the sun relative to some frame of reference (Braemer, 1960; Dyer and Dickinson, 1994; Hoffmann, 1960; Schmidt-Koenig, 1960, 1990; Schmidt-Koenig et al., 1991). The existence of sun compass orientation is well established in many species. However, in no species do we have a clear understanding of the behavioral strategies used in sun compass orientation (for a review, see Able, 1991).

One major unresolved question concerns how much experience with the course of the sun is needed for organisms to learn to use the sun for orientation. Insects do not need complete exposure to the arc of the sun to estimate its position at a time of day when they have never seen the sun (Dyer and Dickinson, 1994; Lindauer, 1957, 1960; Wehner and Müller, 1993). Similarly, Braemer (1960) reported that some species of fish (Lepomis cyanellus, Lepomis gibbosus and Aequidens portalegrensis) were able to compensate correctly for the apparent movement of an artificial sun at any time of day even with limited exposure to the natural sun. Both Schmidt-Koenig (1963a) and Hoffmann (1953, 1959, 1960) also provided anecdotal evidence that adult pigeons, Columba livia, and adult and young starlings were able to use the sun compass at a time of day at which they had never seen the sun. However, the prevailing hypothesis in birds, based on the research of Wiltschko and Wiltschko (1980) with young homing pigeons, is that they must have exposure to the entire arc of the sun to develop a sun compass that can be used at any time of day.

The conclusions of Wiltschko and Wiltschko (1980) and those of Schmidt-Koenig (1963a) and Hoffmann (1953, 1959, 1960) are contradictory. Thus, our study was designed to re-examine the development of the sun compass to determine whether homing pigeons with only partial experience with the course of the sun could use that experience to orient at a time of day when they had never seen the sun.

The subjects were 16 homing pigeons, Columba livia. They were bred at Bowling Green State University and did not see the sun until 5 weeks of age. During the subsequent 8 weeks, we provided the pigeons with food, water and grit ad libitum. During later experimental training, the pigeons were food-deprived to 85 % of their free-feeding body mass. We provided the pigeons with mixed grain during experimental training.

The pigeons were housed in groups at the Bowling Green State University Mercer Road Ecology and Ethology Research Station in Bowling Green, Ohio, USA (83°39′W, 41°23′N). Upon self-sufficiency at 5 weeks of age, the pigeons were randomly divided into two groups (for each group, N=8) and placed into two adjacent indoor lofts (2 m×2 m×3 m each) located within the research station building (ground area 9 m×9 m). The six windows (44 cm×30 cm) of the research station building were covered with two layers of translucent Plexiglas, the first 6 mm thick and the second 3 mm thick. The two layers of Plexiglas eliminated direct observation of the sun’s disk and reduced the penetration of ultraviolet light to less than 1 % of ambient light and the penetration of visible light to less than 10 % of ambient light. Inside the loft, the light levels remained constant throughout the day so it is unlikely that the experimental pigeons could have relied on any direct or indirect information about the path of the sun. Both lofts were connected to outdoor aviaries (3 m×2 m×2 m each). From the enclosed lofts, we could place the control and experimental pigeons in the outdoor aviaries. The control aviary was situated on the corner of the loft, which provided exposure to the sky from east-southeast to north. The experimental aviary was adjacent to the control aviary, which provided full exposure only to the western half of the sky.

The control and experimental groups received different exposure to the course of the sun during the 8 weeks before experimental training and during training. The control group lived continuously in the outdoor aviary except in extremely bad weather and could view the sky throughout the day. The experimental group, in contrast, could only view the sky when we placed them each day in the outdoor aviary from local noon (approximately 12:40 h Eastern Standard Time, EST), when the sun was due south, until sunset. During the mornings, the experimental pigeons were held in the indoor loft.

The testing apparatus was an eight-sided outdoor arena (1 m in diameter) of a design similar to those successfully employed in studies of sun compass learning in pigeons (Bingman and Jones, 1994; Chappell and Guilford, 1995; Gagliardo et al., 1996; Schmidt-Koenig, 1963a,b; see Fig. 1). Each wall of the arena had a 15 cm²opening through which a pigeon could reach a metal food cup surrounded by blinders. The ceiling of the arena was clear Plexiglas in which we cut a circular hole in the center. A cylindrical start chamber (30 cm in diameter, 27 cm in height) could be lifted through the hole. The experimenter manually lifted the start chamber with a pulley system to release a pigeon. The arena was level with respect to gravity and was placed on a rotating board so that its orientation could be changed. It was mounted on a table and placed in an open field that offered the pigeon in the arena no view of surrounding landmarks (Bingman and Jones, 1994).

We trained both groups between 12:40 h EST and 19:30 h EST from April to October 1998. Each pigeon was trained at different times in the afternoon for 10–15 consecutive days when the sun was visible. Once a pigeon had completed its training, we tested it on the morning immediately following its last day of training.

The following protocol was similar to the training procedures used by Bingman and Jones (1994). Once food-deprived, the pigeons were trained in the test arena. We initially placed each pigeon into the chamber for acclimation. Once they had learned to eat from the food cups accessible through the openings on the arena walls, they were trained towards a fixed compass direction. We trained four pigeons from each group to receive food reinforcement from the north and four pigeons from each group to receive food reinforcement from the south. During training, we placed food reinforcement into the training direction cup, ensuring that the food was not visible to the pigeon. For each trial, we placed a pigeon into the start chamber and then manually lifted the start chamber. We allowed a pigeon to search the cups until it found the food reinforcement. The experimenter recorded each cup in which the pigeon’s beak passed through an opening as a directional response. Once the pigeon had obtained reinforcement, it was returned to the start chamber to begin a new trial. We repeated this procedure 20 times per session. After each trial, the experimenter rotated the arena to a new orientation, selected at random, and randomly changed her viewing position. This ensured that the pigeons could not use either landmarks within the arena or the position of the experimenter to learn the training direction. An initial directional response of a trial to the cup in the training direction or to the two adjacent cups was recorded as correct (e.g. if north was the training direction, responses to the northwest, north and northeast cups were all recorded as a correct response). Although the start chamber was designed to minimize variability in the starting position, pigeons occasionally moved within the arena before orienting towards a cup. We recorded the adjacent cups as correct to take into account the variability in starting position from trial to trial. The pigeons were trained until they reached a criterion of 80 % correct responses for five consecutive sessions in which the first response of trial 1 was either towards the training direction cup or towards either of the two cups adjacent to the training cup. An experimental session generally lasted approximately 30 min. We trained the pigeons at different times throughout the afternoon to minimize the possibility of them learning to respond to the sun as a fixed landmark rather than as a true compass. The group range of the azimuth of the sun during training was 182–266 ° for control pigeons and 184–275 ° for experimental pigeons; hence, the direction of the rewarding food cup relative to the azimuth of the sun varied by as much as 90 ° from day to day.

After the pigeons had met the performance criterion, we tested them for their ability to use the morning sun to find the cup in the direction to which they had been trained in the afternoon. The pigeons were tested between 09:00 h EST and 10:00 h EST and received 10 trials without food reinforcement (the group range of the sun azimuth during test 1 was 93–124 ° for control pigeons and 94–111 ° for experimental pigeons). For the data analysis (see below), the first directional responses of the pigeons were the critical measure of behaviour because, during the non-reinforced test session, a pigeon might change its behaviour after the first trial because of the absence of reinforcement. However, the pigeons received 10 trials for two reasons. (i) We tried to keep the test session similar to the training sessions, which had multiple trials. (ii) We were curious to determine whether non-reinforced trials would affect the subsequent responses of the pigeons, i.e. would the pigeons change their search strategies because of lack of reinforcement?

All other variables were identical to those described for the training sessions. Immediately after a pigeon had chosen a food cup, or after the pigeon had made no response for 5 min, the experimenter returned the pigeon to the start chamber to begin the next trial.

This experiment was performed to determine whether the results of test 1 were based on sun compass learning or on some other compass mechanism (e.g. magnetic compass or some other directional cue, see Wiltschko and Wiltschko, 1988). After test 1, the pigeons’ training continued every 4 days during the afternoon as described above. On the first day of complete cloud cover, we tested the pigeons in the morning as described in test 1. If the orientation in the test arena was based on the use of the sun, then the cloud cover would prevent the birds from searching in the correct compass direction.

During training and testing, the experimenter recorded the direction of the first food cup chosen for each trial. The first trial from each session was used for the initial data analyses. During training, all trials were reinforced. Thus, only the first trial reflected learning across sessions and did not reflect within-session learning. Similarly, because the test trials were non-reinforced, the first trial during a test session was the critical measure of behaviour. During the test session, a pigeon might tend to change its behaviour after the first trial as a result of the lack of reinforcement. Three types of analysis were undertaken. (i) We examined the directional responses recorded on the first trial during the last 5 days of training. For each pigeon, a mean vector was computed from the five directional responses. We calculated a group mean vector for each group using the direction of the mean vector of the individual pigeons. (ii) For test 1, we analyzed the directional response of the first trial. We generated a mean vector for each group using the directional response of the first trial of the individual pigeons. (iii) For test 2, we analyzed the directional response for each bird on the first trial of the test session. We generated a mean vector for each group using the directional response from the first trial of the individual pigeons.

For all three analyses, we used a Rayleigh test to determine whether the distribution of angles for each group differed from uniform, and we also performed between-group comparisons of directional responses using a Watson U²-test (Batschelet, 1981).

Ninety-nine per cent confidence intervals were computed for each group (Batschelet, 1981) to determine whether the direction of the mean vector of each group differed significantly from either the training direction or a fixed angle direction based on the last day of training. Because a fixed angle direction is based on the angle between the training direction and the sun on the last training day (which was different for each bird), the first trial directional response of each bird was normalized so that confidence intervals could be computed (i.e. for each pigeon, the expected fixed angle response was normalized to 360 °).

Additional analyses on the first five test trials of each bird for test 1 and test 2 were computed to determine whether continued testing with non-reinforced trials would affect the response of the pigeons. However, one must be cautious in the interpretation of these results because of the absence of reinforcement in these trials. For these analyses, we again used a Rayleigh test to determine whether the distribution of angles for each group differed from uniform, and we also performed between-group comparisons of directional responses using a Watson U²-test (Batschelet, 1981).

On the last day of training and during test 1, using a stopwatch, we recorded the latency to the first directional response for each bird as an additional way of monitoring possible behavioural differences between the control and experimental pigeons during training and testing. With the latency to the directional response as the dependent variable, we used a two-factor within-group analysis of variance (ANOVA) with group (control and experimental) and day (last training day and test 1) as the independent variables to identify any possible differences.

The two groups learned the task equally well. The mean number of training sessions for the control group was 15.50±0.63 sessions and for the experimental group was 13.38±0.80 sessions (means ± S.E.M., t₁₄=2.09, not significant). The mean directional responses of the first trials during the last 5 days of training are presented in Fig. 2 and Fig. 3A,B. The first trial directional responses of the control and experimental pigeons on the last 5 training days were well oriented in the trained compass direction indicating that, from day to day, the birds changed their orientation relative to the sun and their orientation was significantly different from uniform. We noted no differences between groups in directional responses on the last 5 days of training.

During the morning test session, we expected the control pigeons to select the training compass direction because they had previously received exposure to the course of the sun in the morning. Hence, as shown originally by Kramer (1952a,b), the control pigeons should have been able to compensate for the change in the solar azimuth from afternoon to morning. However, for the experimental pigeons to orient towards the training direction, they would have to estimate the course of the sun in the morning based only on their knowledge of its apparent movement in the afternoon in a manner similar to that of insects (Dyer and Dickinson, 1994; Wehner and Müller, 1993).

The directional responses of each individual for the first trial are presented in Fig. 2 and Fig. 3C,D. Both the experimental and control groups were non-randomly oriented, and the distribution of mean vectors did not differ. To test whether the mean vector of each group was significantly different from either the normalized training direction or the normalized fixed-angle direction, a 95 % confidence interval for each group was computed. (i) The mean vectors of the control group and the experimental group did not differ significantly from the normalized training compass direction (360 °). The confidence intervals for the control group was 326±40 ° and that for the experimental group was 8±32 °. (ii) In contrast, the mean vectors of both groups differed significantly from the fixed angle direction (normalized to 360 °). The confidence intervals for the control group was 85±45 ° and that for the experimental group was 185±32 °.

For the five test trial analysis, the mean vector for the control group was 312 ° and the mean vector length was 0.90 (not taking into account the mean vector length of each individual; P<0.01). The mean vector for the experimental group was 348 ° and the mean vector length was 0.62 (not taking into account the mean vector length of each individual; P<0.01).

Both the control and experimental groups were non-randomly oriented, and the distribution of mean vectors did not differ. These results are similar to the results of the analysis of the first trial directional responses.

The group results indicate that, when tested in the morning, control and experimental pigeons were oriented in the true compass direction in which they had been trained and were not simply adopting a fixed angle relative to the sun. It should be noted that examination of Fig. 2A indicates that at least one control pigeon may have been relying on a fixed angle response when tested in the morning. However, despite the response of this one control pigeon, the data critically demonstrate that the experimental pigeons were able to use the sun compass at a time of day when they had never seen the sun.

The results of the ANOVA for the latency to the first response between the last training day and test 1 are presented in Fig. 4. The ANOVA revealed a significant interaction between group and day (F_1,14=5.62, P<0.05). Fisher’s post-hoc analysis revealed no significant difference between the response latency for the control group and experimental group on the last day of training (P>0.05) and a significant difference between the response latency for the control group and experimental group on the test day (P<0.05). The difference in response latency between the control and experimental pigeons during test 1 indicates that the two groups responded differently to the morning sun, a difference that presumably reflects the lack of prior experience of the experimental birds with the morning sun.

Test 2 was a control test to determine whether the orientation of the pigeons in the arena was based on some environmental cue other than the sun compass (e.g. the magnetic field of the earth). The results from the first trials of test 2 are presented in Fig. 3E,F. The mean vectors of the control group and experimental group for the first trial data did not differ from uniform. Thus, the orientation of the pigeons recorded during test 2 was dependent on the availability of the sun.

Although the analysis of the first trial directional response was not significant, the apparent orientation towards the training direction of four of the eight control pigeons was curious. However, for the five trial test analysis, the mean vector for the control group was 225 ° and the mean vector length was 0.16 (not taking into account the mean vector length of each individual; P>0.05). Thus, the apparent orientation towards the training direction of these individual pigeons was not evident in the five trial test analysis. The mean vector for the experimental group was 263 ° and the mean vector length was 0.15 (not taking into account the mean vector length of each individual; P>0.05). Both the control and experimental groups were randomly oriented, and the distribution of mean vectors did not differ.

In this study, we examined whether young pigeons exposed only to the afternoon sun could generalize the information gained in the afternoon and use the morning sun for orientation. Insects and fish do not need complete exposure to the sun to use it as an orientation mechanism throughout the day (Braemer, 1960; Dyer and Dickinson, 1994; Lindauer, 1957, 1960; Wehner and Müller, 1993). However, there is conflicting evidence about the amount of experience birds need to use the sun as an orientation cue.

Schmidt-Koenig (1963a) tested sun compass orientation in adult pigeons under the arctic sun in summer using a cue-controlled arena similar to our own. Two of his three pigeons oriented in the training direction at a time of day when they had never experienced the sun (at midnight). Hoffmann (1959) obtained similar results with starlings displaced to the arctic and tested under the midnight sun. The results of Schmidt-Koenig (1963a) and Hoffmann (1953, 1959, 1960) indicate that birds might be able to use the sun compass at a time of day when they have never seen the sun. However, their research involved mature birds that had received extensive exposure to the diurnal course of the sun in lower latitudes. In addition, the starlings in the study of Hoffmann (1959) may also have seen a portion of the midnight sun before testing. Although Hoffmann (1953, 1960) did examine sun compass orientation in young starlings, they were 12 days old when removed from the nest and may have had some passive exposure to the sun. In a more substantial study, Wiltschko and Wiltschko (1980) raised young homing pigeons with exposure only to the afternoon sun. They flew a control group throughout the day and an experimental group only in the afternoon. They then released both groups in the morning from an unfamiliar location. During the morning, both groups were oriented towards home. Both groups were then subjected to a 6 h slow clock-shift and released in the afternoon (the subjective morning for the pigeons). The control group displayed a 90 ° clockwise shift in orientation, whereas the experimental pigeons continued to orient towards home. In a subsequent study, Wiltschko et al. (1981) concluded that the experimental pigeons were relying on a magnetic compass to orient, and they suggested that pigeons needed complete exposure to the daytime arc of the sun to use it for orientation throughout the day.

Pigeons use both a navigational map and a compass to navigate home. Wiltschko and Wiltschko (1980) conducted experiments that involved homing releases. With respect to understanding the sun compass, homing releases are confounded because the sun compass functions together with the navigational map. Therefore, the apparent failure of their experimental pigeons to use the sun as a compass during homing releases may not be due to the absence of a developed sun compass. Instead, the pigeons may have possessed a functioning sun compass but simply have given priority to their magnetic compass when using compass information in conjunction with their navigational map. In many other contexts, animals will ignore an orientation cue when some other source of directional information is more reliable, but such a result is not evidence that the cue that is ignored is not used under different circumstances (see, for example, Gagliardo et al., 1996). A paradigm that isolates sun compass learning is necessary when examining the development of sun compass orientation.

In the present study, we trained and tested young pigeons in an outdoor arena that explicitly isolated sun compass learning from other navigational mechanisms. During training, there were no differences between the control and experimental groups in the mean number of sessions needed to reach the training criterion, and both groups learned to orient in the training direction. The results are consistent with previous studies: birds can use the sun as a compass for orientation when isolated in an experimental arena (Balda and Wiltschko, 1995; Bingman and Jones, 1994; Chappell and Guilford, 1995; Duff et al., 1998; Gagliardo et al., 1996; Hoffmann, 1960; Kramer, 1952a,b; Schmidt-Koenig, 1963a,b; Wiltschko and Balda, 1989).

More substantially, the experimental pigeons were able to learn to use the sun compass with only partial experience of the course of the sun. Thus, when the sun compass has been experimentally isolated from other spatial mechanisms, pigeons with exposure only to the afternoon sun are able to generalize this information and use the morning sun to orient. Consequently, pigeons resemble insects and fish (Braemer, 1960; Dyer and Dickinson, 1994; Lindauer, 1957, 1960; Wehner and Müller, 1993), which also develop a sun compass that works at all times of day even after only partial experience with the arc of the sun.

However, the results of the present study do not diminish the importance of the findings of Wiltschko and Wiltschko (1980). During the homing releases, their pigeons may have possessed a fully functioning sun compass. However, the pigeons may have been impaired in integrating the sun compass with the navigational map because they were deprived of flight experience at a time of day when they had not seen the sun. In contrast, our experiment isolated the sun compass from the navigational map and other homing mechanisms. The contrast between our results and those of Wiltschko and Wiltschko (1980) suggests that the development of sun compass orientation in the context of homing may be more complex than previously appreciated.

During test 2, both groups of pigeons displayed uniformly distributed directional choices under complete cloud cover. The results indicate that the orientation recorded during test 1 was dependent on the sun. Thus, in the experimental arena used in the present study, young pigeons with limited exposure to the arc of the sun do not rely on a magnetic compass or any other cue to orient when the sun is not visible.

Honeybees with only late afternoon exposure to the sun do not compensate for the apparent movement of the sun with perfect accuracy at other times of day. Rather, the bees place the morning sun in the direction of a 180 ° reverse angle of the azimuth of the sun during training and maintain this estimate for the entire period before noon (approximately east in the study of Dyer and Dickinson, 1994). Do young pigeons with exposure only to the afternoon sun compensate accurately for the movement of the sun during the morning or do they compensate only approximately in the way bees do? Superficially, the data from the experimental pigeons indicate that their compensation is quite accurate in the morning. However, we tested all the pigeons at approximately the same time during the morning (approximately 09:00 h EST) when the sun was roughly in the east. Therefore, if they relied on a rough approximation (e.g. 180 ° reverse angle of the sun at sunset), as bees do, similar results would have been recorded. Further experiments are necessary to determine whether the sun compass orientation of the experimental pigeons during the morning was fully time-compensated or simply an approximation similar to that used by partially experienced bees.

In summary, the data obtained from the experimental pigeons in this study indicate that birds resemble insects with respect to the development of sun compass orientation. As such, the data indicate that some properties of sun compass learning are shared across a broad array of taxonomic groups.