Sun Navigation in Homing Pigeons—Attempts to Shift Sun Coordinates

It has been demonstrated repeatedly that homing pigeons can return to their loft when released at a place they have never been before. Since the pioneering work of Kramer & St Paul (1950, 1952), Matthews (1951, 1953; summarized in 1968), Wallraff (1967) and Schmidt-Koenig (1965), it has become clear that pigeons use some form of bi-coordinate navigation.

Matthews (1951), Pennycuick (1960) and Tunmore (1960) proposed that pigeons use the sun as the basis of their navigation. Kramer (1953), on the other hand, believed that the sun was used only as a compass and that the bi-coordinate navigation of ‘map’ was based on a different set of cues. This view has also been adopted by Schmidt-Koenig (1965) and Wallraff (1967). The continuing debate about the role of the sun in orientation is evident in two recent reviews: that of Matthews (1968) and of Schmidt-Koenig (1965).

Both Schmidt-Koenig (1958, 1961) and Matthews (1955) have exposed their birds to artificial light-dark cycles such that the pigeons’ internal clocks should be shifted out of phase with the real day. Matthews (1955) reports that birds given a 3 h shift orient as predicted by the sun navigation hypothesis; Schmidt-Koenig (1961) reports that birds with a 6 h shift orient about 90° to the left or right of home as predicted by Kramer’s (1953) hypothesis of a time-compensated sun compass. But, if a pigeon were, in fact, released at a point where the sun was 6 or 12 h ahead of or behind loft time, it would have to fly a distance of at least 4000 miles to reach home. Since the usual homing tests are all made at distances of less than 500 miles (805 km), it could be reasoned (see Pennycuick, 1961) that such a large clock shift was in some way ‘inconceivable’ to a pigeon and that the bird might not utilize sun navigation under these circumstances. For this reason, we have tried giving pigeons smaller shifts and assessing their effect on the pigeons’ orientation.

According to the theory of sun navigation, short time errors of less than an hour should move the apparent position of the loft in sun coordinates a few hundred miles on an east-west line. Such time errors should not seriously alter the birds’ perception of compass directions based on the sun, so the effects of the shift on the navigation should be separable by this technique from the effects on the sun compass alone. We have reported (Walcott & Michener, 1967) the initial results of a study of such short time shifts, and their effects on the homeward tracks of individual pigeons. It is the purpose of this paper to report on the conclusion of these experiments, conducted during the summers of 1965−9 inclusive. We also wish to report the results of experiments in which the image of the sun at the loft was shifted with planar mirrors, thereby providing an apparent north-south shift of the loft position in sun coordinates. Finally we will report the results of experiments in which the birds were exposed to a series of treatments designed to disturb their recollection of local solar time at the home loft.

The stock of pigeons used in these studies was obtained from the lofts of successful pigeon racers in the Boston area. The actual pigeons whose performance is reported in this paper were mainly raised in the loft at Lincoln, Massachusetts; a few were from the loft at Harvard University in Cambridge. Birds were trained by being released at increasing distances in one direction from the loft. There were four such ‘training lines’ used in these experiments; birds were released from the south, the south-west, the west and the north. In all cases the training direction is specified in the results. Details of the lofts, training and techniques used for aeroplane tracking of the pigeons are given in Michener & Walcott (1967).

As previously described (Walcott & Michener, 1967), birds to be treated with artificial light shifts were incarcerated singly or in pairs in plywood boxes measuring 2 × 2 × 3 ft (0·6 × 0·6 × 0·9 m). These boxes and their tops were made light-tight by the application of sealant and paint, and were provided with input and output air ducts and a continuously operating electric fan. Water and mixed grain were available in excess at all times. Lighting was provided by two 75 W tungsten incandescent bulbs, providing a light intensity of about 1000 lx (100 ft-candles) on the bottom of the box. The bulbs were switched on automatically at almanac sunrise plus the amount of shift. The almanac times and the accuracy of the automatic clock had a maximum error of 30 s.

The shift boxes were kept in the cellar of a large barn, with little or no daily activity around them, and the electric fans provided a continuous background noise which was found to exceed virtually all other incoming noises. The boxes were equipped with two perches so arranged that when the pigeon used either perch it made a mark on the chart of an event recorder. One of these perches was attached to the side of the box, the other was placed in front of the feeder in such a way that a pigeon eating grain must, of necessity, step on this perch.

The preliminary analysis of activity patterns indicates that the pigeons required between 3 and 6 days to entrain their activity as measured by perch-activity rhythms to an artificial day which was 2 h out of phase with the natural day. To ensure that the birds were given sufficient exposure to the shifted régime all pigeons were kept in the shift boxes for between 8 and 14 days. The radio transmitters were put on the pigeons some time in the afternoon before the day of release. On the day of release the birds were removed from the shift boxes in the darkened cellar and put into cloth-covered cages early in the morning. They were transported by car, then by aeroplane to the release point and they were not allowed a view of the sun or of their surroundings until the moment of release.

Another procedure, intended as a controlling treatment, was administered to some pigeons. They were put, singly, in a box, identical in basic construction to the shift boxes, which was kept in the centre of a large, open field. Instead of a plywood top, however, the box was covered only with a piece of oil cloth. This material is waterproof and forms a light-dispersing filter of about 10% transmission (density = 1·0), through which it is impossible to determine the direction of incident light since the light emerges scattered in all directions. The box had the same fans, ducts, food and water and recording perches as the shift boxes. In such a box the pigeon should be exposed to the natural daily light cycle, but without an opportunity to view the sun’s disk.

Sun navigation requires not only an accurate measurement of the sun’s altitude and a sense of time, but also an accurate comparison of the sun’s altitude at the release point with its position at the same time at the home loft. If a pigeon’s observations of the sun’s path at the loft were predictably distorted for the period immediately prior to its release in unknown territory, the bird’s choice of flight direction might also undergo predictable distortion.

Pigeons were confined, one at a time, in a wire cage in one corner of the south-facing wall of the loft (see Fig. 1). Besides being able to see around the compartment of the loft, their only view of the outside world was through a wire-covered port, 8 in (20 cm) square. The view through the port was divided into two portions: the lower 20° of the sky, the southerly horizon and the front yard could be seen directly through the larger bottom port; the sky between 30 and 80° of altitude could only be seen through a pair of mirrors, M1 and M2. Both were rear-surface plate-glass mirrors; M1 was anchored firmly to the wall of the loft, M2 was adjustable. Laterally, the pigeon could look straight south and about 20−30° either side of it. Thus the sun was visible for 2·7−3·5 h during midday, including local noon, through the nearly coplanar mirrors.

Adjustment of the mirrors was made as follows. Each day, during early morning, the mirrors were adjusted to be as parallel as possible, generally to within less than 1’ of arc of being coplanar. This meant adjusting the pairs of screws at three of the corners of M2 until the train of multiple reflexions of a pair of taut strings (one vertical, the other horizontal) appeared as straight lines, receding back into each mirror. (Multiple sets of reflexions add twice the error between the mirrors to the image, each time they are reflected. Thus, if more than ten reflexions are visible, it is easy to see an error of less than 1’ of arc.) After this setting, which corrected the apparatus each day for warp and other distortions of the wooden frame, the desired shift to be given to the sun was set into the adjusting screws. For instance, to raise the position of the sun’s image by 60’ of arc, it was only necessary to move the top of M2 toward the loft so as to tilt it by 30’ (since moving a mirror by 30’ changes both angle of incidence and angle of reflexion by 30’, the total angle from incidence to reflexion is changed 30 + 30 = 60’).

Frequent checks of the operation were made by measuring the apparent sun altitude through the mirrors with a bubble sextant placed in the pigeons’ viewing port. In all cases the measured sun altitude was found to be within ± 1·5’ of that calculated—an error typical of the sextant itself (equivalent to a shift of the loft position 1·7 miles (2·8 km) in a north-south line.) It might also be noted that, although the sunlight viewed at the port passed through two rear-surface mirrors, multiple or distorted reflexions were not evident, even when viewing the sun directly through the filters in the sextant. Thus, there were no obvious visual clues that the sun was not being viewed directly. Pigeons were confined to the mirror box for 10−14 days, and it was interesting that most birds spent much of their time in the area where the patch of sunlight was reflected on to the floor of the loft by the mirrors. When the sun was not visible the pigeons seemed to show no preference for any place in their wire enclosure.

The release points used in this study were all chosen so that they were unfamiliar to the bird being released there; in other words none of the birds’ previous tracks passed within 10 miles of this point. Yet for all such experimental release points we tried to have at least one control release of a pigeon from the same point. Fig. 2 shows all the releases treated here. The proportion of experimentals and controls at each point is clear from the symbols.

When released in unfamiliar territory a pigeon might first compute these two components, and then the resultant direction to the loft. Of course, no claim is intended that the pigeon manipulates symbols in the way the term ‘compute’ is normally used; all that the theory implies is that the behavioural equivalent of computation of the home direction would have to be based on the informative elements of the sun’s position which we conveniently represent by H and R.

As an example of the formal calculations, Fig. 3 shows the conditions existing on 5 September 1967, at 13.00 E.S.T., when pigeon 1534 with a 5 min clock shift was released from North Central Airport, 35 miles (56 km) SSW. of its loft. It should be emphasized that these sun coordinates apply only to the time of release, even though the entire track of the bird is shown. At the time of release the sun appeared at an altitude of 51·20° (H_rel); its rate of rise was of rotation of the earth (i.e. the sun was descending, early afternoon); the azimuth (Z) was 210·8° (approximately SW.). The sun-navigation hypothesis postulates that the pigeon compares H_rel and R_rel with the remembered-projected values at the loft at that time: H_L= 50·71°; R_L= − 0·377. Thus the sun is higher at release by ΔH = +0·49° and the pigeon should fly away from the sun (i.e. NE.) for S_H = 33·8 miles (54·5 km). The rate of increase in altitude is too negative at the release point by ΔR = 0·0034°/ degree of rotation, as compared with that at the loft, so the pigeon must fly right of the up-sun direction (i.e. NW.) for S_R = 9·3 miles (14·9 km). As Fig. 3 shows, the vector addition, assuming the earth’s surface to be a plane, produces a resultant 35·1 miles long (54·5 km), directed at the loft.

Analysis of cases where pigeons have been given a clock shift of time T by artificia light-dark cycles is quite straightforward. The only modification is that the bird should compare the sun coordinates at the release point with the sun at a ‘false home’, ‘T’ minutes west or east of the loft. Celestially, longitude and time are interconvertible, since the rate of rotation of the earth is constant at 15°/h (or 15’/minute of time). At this false home the sun’s course is identical with that seen at the loft, but each point on the course occurs T minutes before or after that point as seen at the loft.

In the present case, shown in Fig. 3, the pigeon actually had been given a 5 min shift: the lights in its box went on exactly 5 min after sunrise, and off 5 min after sunset. This corresponds to a point exactly 5 min (⁠$12$ or, at this latitude, 63·8 miles, 102·6 km) west of the loft on the same line of latitude. The false home is also shown on Fig. 3, along with the sun-navigation components. ΔH is very small, since the sun’s altitude at release is nearly the same as that at the false home point, amounting to S_H= 2·0 miles (3·2 km) displacement down-sun; R is large (−0·0239°/degree), giving a right of up-sun component of S_R= 64·6 miles (103·9 km).

However, as the track shows, the bird gave little evidence of flying towards the false home; it began flying SSE., in nearly its compass direction, then flew back past the release point and thence, semi-directly, to its loft. At no point in the track did it appear to orient toward the false home. This result is typical of all our releases; no clock-shifted pigeon ever flew to or investigated the false home area. Most were quite well oriented to the home loft when released in unfamiliar territory. Fig. 2 shows, as a composite map, the directions of the 10 mile bearings for all the tracks reported here. The 10 mile bearing is simply the bearing of that point at which a pigeon was 10 miles from the release point. At each release point, each 10 mile bearing is shown as a rectangle on the periphery of a circle. The rectangles contain symbols showing the sort of treatment given each bird, and are lumped to the nearest 10° category. Owing to a lack of space between adjacent release points, the illustration shows the bearing points at a scale radius of about 4 miles, but the miles scale at the right can be used to indicate accurately the points 10 miles from the release points. For each release point a fine line indicates the direction to the loft (L). The release points were new to the pigeons in every case, so that the pigeons could not have been orienting by landmark recognition. The map shows that, regardless of treatment, most of the pigeons started off in the homeward half of the circle for all release points around the loft.

In addition to the 10 mile vanishing points, we have examined the tracks produced by individual pigeons released at new places both after experimental treatment or after no treatment at all. An examination of such tracks shows no consistent differences between controls and experimentals. Rather than publish all the tracks here, we have given the results of various measurements made on them in Appendix table, p. 313. Table 1 is a comparison of these measurements for all the different experimental treatments and for the controls.

To analyse the 10 mile bearings with respect to treatment, we subtracted each of several reference directions from each bearing and recorded the difference. For each 10 mile track bearing (see Fig. 3), the angle between it and the direction to the home loft is called the ‘Home Error’; the difference between the 10 mile track bearing and the previous training direction is the ‘Compass Error’; and the difference from the predicted false-home direction is the ‘False Home Error’. Fig. 4 shows the same 10 mile bearings as in Fig. 2 but compiled with respect to the different treatments and with the ‘errors’ with respect to the three reference directions. The first column indicates, for that group of releases, the direction to home plotted as compass directions from the release point. This is included to show that we did not release the birds randomly around the loft for all treatments and thereby introduce a directional bias in some of the expected flight directions. The second column shows the 10 mile bearings, and, like the first, north (or o° of azimuth) is at the top of the diagram, as on a map. The next three columns to the right are distributions of ‘errors’, with zero error now at the top of the page. Thus, a 10 mile bearing directed in exactly the trained compass direction would be shown in column two as a rectangle pointing in that training direction (e.g. SE., 135°clockwise from north); in column three, it would show as a zero compass error, at the top of the diagram; in columns four and five it would appear at random with respect to the loft and the false home (each, o° of error at the top of the page), depending on the exact choice of release point and shift.

For each diagram the Rayleigh test was employed (Batschelet, 1965) and an empirical formula (see note) was used to estimate P, the probability that the observed samples were drawn from a randomly scattered parent population. The estimate of P is shown inside each diagram. Also, for all diagrams with a P of less than 0·1, the mean vector direction is shown as a heavy line projecting out of the circle.

Since for every treatment the same data are presented in each of several ways, it may help avoid confusion if we review the properties of a nearly perfect and conclusive experiment, in terms of the various diagrams. Ideally, the choice of release points should be such that the home directions for different birds released after a treatment should appear random (P large, nearly 1·0). In addition, if there is no tendency for the birds to fly in a fixed direction, the 10 mile bearings should also appear random with respect to north as a reference. Then if the data show consistent orientation to only one reference direction, e.g. the home loft, the scatter of errors in this category should be reduced, with the points grouped tightly at the top of the circle and P very small. The other two ‘error’ diagrams should show the same wide scatter as the first two diagrams.

The first row, where no treatment was given to the birds, is the largest. This represents tracks of only the birds used in one or more of the shift experiments which were used to make up this set of data, so that, as much as possible, the comparison from treatment to treatment was made on the same birds. In this case, the choice of release points, the 10 mile bearings and the compass errors look random, P in all cases being larger than 0·2. Only the home errors show significant reduction in P, being approximately equal to 0·00001, shown to the nearest three digits. Not only is the scatter low, but the mean direction is 006°, i.e. 6° right of home. Thus, the untreated birds were well oriented to the loft within i o miles of being released, and showed neither a tendency to fly in a fixed compass direction nor a tendency to orient in their trained compass direction as had pigeons studied previously (Michener & Walcott, 1967).

In a similar way it is possible to analyse the two treatments originally regarded as controls. These were designed to assess the effect of the 1−2 week incarceration in a shift box. The ‘cloth box’ results are unclear. Only seven releases have been made and neither compass error nor home error show a scatter reduction that is significant, although six of the seven home errors are less than 90°, whereas only three of the compass errors are less than 90°. That is, the 10 mile bearings were closer to being homeward orientated than to the direction of previous training.

The ‘zero’ shift refers to pigeons put in a clock-shift box in which the on and off times of the lights exactly matched almanac sunrise and sunset for the loft. There were only four birds which actually were released after this treatment, and the choice of release points was restricted to north and NW. of the loft. Thus, the 10 mile bearings show a significant tendency to head SE., yet there is no reason to believe that this treatment would favour a fixed direction tendency over the pigeons’ already established tendency to head homeward. But here again the home errors are not significantly clumped; they are all less than 90°, and more closely grouped than the compass errors. It can be said, then, that these two sets of data, small though they be in number, show that birds held in boxes for 8−14 days do orient to the loft almost as well as they do after merely having been released with no treatment other than the displacement.

Clock shifts of 5−120 min can be considered together on the basis of the scatter diagrams, as an examination of their home errors will indicate. Several points stand out: (1) the majority of 10 mile bearings were within 90° of the home direction, with the scatter of home errors, measured as a probability of randomness, all less than 0·06. (2) The mean directions of the home errors progress from 11° left of home (349°) with 5 min shifts, to 37°right of home with 2 h shifts. (3) In no case do the false home errors show a reduction in scatter over the raw data (10 mile bearings with respect to north) nor are they grouped in the false-home half of the circle. (4) The home errors are, in three of the four categories, less scattered than the compass errors, the exception being the bearings of birds given a 10 min shift. These points will be more fully discussed below.

The 6 h shifts, five in number, seem random with respect to all possible explanatory directions: neither compass, nor home, nor false-home errors show any scatter reduction.

The birds released after being given mirror-box shift were well oriented toward home. The calculation of the false-home direction for such shifts is as straightforward as that for the time-shifted birds. A pigeon watching the sun through the mirrors, which shift the noon position (and, proportionally, the rest of the sun’s path as well) to an altitude 31’ higher than normal at the loft, should home to a point on the earth’s surface where noon occurs at the same time as at the loft (on the same north-south line) but where the sun appears 31’ higher than that at the loft. This is exactly equivalent to tilting the horizon by 31’, or to moving to a point 31’ of latitude to the south of the loft (31’ of latitude = 31 nautical miles = 36 statute miles = 57 km). The following shifts were given: five of 31’ (three higher, two lower altitude); five of 70’ (two higher, three lower); and one with no net shift up or down. As the last row of Fig. 4 shows, neither compass errors nor false-home errors were significantly clumped, while the home errors showed very little scatter (P = 0·002) and a mean of 332° or 28° left of home.

The homing speeds and accuracy of the straight track segments shown in Appendix table (portions of tracks in unfamiliar territory which are at least 20 miles long yet deviate from a straight line by less than ± 1 mile along their entire length) of all the groups were comparable to those published previously (Michener & Walcott, 1967) except for the 6 h shifted birds; only two of the five ever homed, and they averaged 1·05miles/h, as compared to 6·86miles/h for the ‘zero’ shifted birds. The poor returns made us reluctant to increase the number of 6 h shifts in our sample.

The results reported here do not agree with the tentative summary we gave in our previous paper (Walcott & Michener, 1967). The numbers of tracks and experiments were, at that time, quite small, and evidently the sample was not representative. Taken as a whole, the results reported here, like those of the longer shifting experiments of Schmidt-Koenig (1961), argue strongly against the sun-navigation hypothesis. There is no evidence that shifts of 5−120 min were disorienting to pigeons released in unknown territory. The only treatment that affected homing speeds was the 6 h shift treatment It could be argued that, with so large a shift in the sun compass, the birds’ courses did not converge on the loft; and it is well known from the work of Schmidt-Koenig (1961), Keeton (1969) and others that birds with 6 h shifts home poorly.

While the 6 h shifts seem to disrupt both initial orientation and homing speed, the only other treatment that seemed to seriously increase the scatter of the initial orientation was the cloth box, even though the other measures of homing performance (see Table 1) were comparable to the control groups. It is somewhat puzzling why this treatment should have had an effect only on the birds’ 10 mile orientation. One possible explanation which we explored is that the dimming of the effective daylight within the box might have been interpreted by the pigeons as a shorter day length. An analysis of the activity rhythms of two pigeons confined for months to the cloth box by Miselis & Walcott (1970) showed that the pigeons actually spent an average of 37 min less time being active each day than under the conditions of the normal loft. The sun-navigation hypothesis would suggest that a pigeon confined under conditions of shorter days during the north temperate summer time would conclude it was at a point well south of the loft (if the bird held the season to be a constant), or that the incarceration at the loft was further south (in sun coordinates) than normal, and refer the sun coordinates of the release point to that southerly location of the projected loft (if the bird held the place of incarceration to be constant). In the first case the pigeon might fly nearly due north from any release point within 100 miles of the true loft; in the second case the bird would orient in a predominantly southerly direction, calculated to bring it back to an area to the south of the real loft, where the summer days are shorter and the nights are longer. These two interpretations are mutually exclusive, but, as can be seen from Fig. 4, second row, second column, the 10 mile track bearings were neither all northerly nor all southerly. Here again, the results are in conflict with the various possible interpretations of the sun-navigation hypothesis. We think it probable that the effect on the initial orientation may be the spurious effect of so small a sample, especially since six of the seven home errors lie at less than 90° from zero, indicating a homeward tendency, even if rather scattered.

There is a suggestion from the data that the clock shifts only affected the initial orientation via predictable disruption of the sun compass. If one looks (Fig. 4) from the ‘zero’ shift down the home error column to the 120 min shift, there is a noticeable progression from the left of the home direction to the right, as predicted from the direction of the shift. For the shifts of less than 20 min this compass shift would not be noticeable in the data, with the orientation diagrams as scattered as they are, since the change in the sun’s azimuth during 20 min varies with the time of day, but usually amounts to approximately 5−12° clockwise. In order to analyse this problem accurately the sun azimuth was calculated for all shifts for each release point and time. This was subtracted from a similar calculation made for the sun at the loft at the effective shifted time. This difference (ΔZ) gives the actual azimuthal shift expected for each 10 mile bearing, since it compares the birds’ best estimate of the sun azimuth at the true loft at the exact shifted time, with the actual sun azimuth as seen at the release point. The 10 mile bearing errors for the 120 min and 6 h shifts are shown in Fig. 5, corrected in each case for this shift in sun azimuth (ΔZ). If the shifts affected only the sun-azimuth compass, such a correction applied to each home error datum should both narrow the scatter of the home errors and pull the mean vector closer to the true home direction. As Fig. 5 shows, the results of these individual corrections produce exactly this effect for the 2 h shifts (compare with Fig. 4, 120 min shift row). The compass errors and false-home errors still appear largely random, but the home errors are less scattered (P = 0·004 instead of 0·012) and directed more nearly toward home (mean direction 5°left instead of 37°right). The 6 h shifts still, after the subtraction of ΔZ from the initial bearings, appear random in every respect. Here again, the 6 h shifts appear to be disruptive in their effects on our birds.

It may well be argued that there exists an ambiguity in the direction of the ‘predicted false home’ with our time shifts. Matthews (1968, p. 138) pointed out that a bird might conceive of the shifting container as being at the release point instead of being near the loft. Thus, a pigeon, upon finding itself with evidence of its chronometer being, say, 2 h fast when it is first put into a shifted light–dark régime, might well regard the experience as evidence of its having been transported several miles in a westerly direction. During its stay in the shift box the bird might remember this apparent transportation, and, if and when it was able to escape from the box, it should fly east. The predicted false home, then, might well be to the east of the release point, rather than the west.

The whole issue hinges upon the pigeon’s believing that the shift box is a place close to the loft. This consideration would lead to an entirely different set of false homes than are shown in Fig. 4. We have repeated the analysis with our short-time shifts considering this point, and we find that the orientation toward the home loft still shows the best orientation; there is no statistical difference between the errors to the two false homes.

The mirror-box results seem less subject to the ambiguity of the two false homes since the pigeons were, in this case, confined in one corner of the actual loft, in full sight and hearing of the other birds. Nevertheless, we have also checked each mirrorbox result, and reversing the direction of the false home leads, if anything, to more widely scattered false-home errors.

It may also be possible that, as Pennycuick (1960) points out, pigeons may selectively use one set of sun coordinates at a time (e.g. R_H, S_H) because they are easier to evaluate more accurately at a particular time of day than the others (S_R). Thus some birds might preferentially orient by one set and seldom by the other. To check this, we calculated the initial sun-coordinate conditions for release point, loft and false home for all the releases reported here, and laid them out in the manner of Fig. 3 along with each track. Here again, we find no indication that any pigeon tended to make use of any part of the information in any consistent way. There was no correlation of any aspect of the sun’s coordinates with any aspect of any pigeon’s track.

Matthews (1955) has suggested that pigeons’ chronometers might be unaffected by regular light-dark régimes, while their activity rhythms and sun-compass clocks might show entrainment. Our time-shifting results, even though involving longer periods of exposure to the shifted régime than normally employed (8−14 days instead of, e.g., Schmidt-Koenig, 1961, 4−7 days) are still open to this criticism. A similar argument might be found against the results with the mirrors—perhaps birds as old as the ones we used (1−3 years) had already worked out the sun’s course at the loft so well that the mirrors were simply not believed. In an attempt to clarify these remaining difficulties, we performed one more shifting experiment which combined several treatments at the same time.

If a pigeon is navigating by using the sun, it must have a memory of the solar coordinates at a given time of day at the home loft. In the shift experiments reported above, the pigeon was isolated in an enclosed box (for as long as 14 days) under artificial lights, without an opportunity to see the sun. Yet, such birds homed as well as controls. It has been commonly reported in the pigeon-racing literature, that birds displaced from the home loft and held for substantial periods of time, even up to a year or more, home rapidly to the original loft when released. That such birds could have a clock accurately entrained to solar time at the home loft seems unlikely. Yet, it would be helpful in evaluating the theory of sun navigation if one could unequivocally destroy a bird’s recollection of time at the home loft and then assay its ability to navigate. While there is no way of being certain that the bird’s clock is completely uncoupled from time at the home loft, there are techniques that should make such coupling less probable. We have used two techniques: random light-dark intervals and giving pigeons 30% D₂O in their drinking water.

Matthews (1953) has shown that the chronometers of pigeons could be thoroughly upset by treating them for 4 or 5 days with very irregular light-dark cycles. In addition Suter & Rawson (1968) have shown that D₂O slows down the circadian activity rhythm in the deer mouse (Peromyscus) and L. R. G. Snyder (in preparation) has shown the same effect in homing pigeons. Snyder reports that pigeons, in constant dim light and given 30% D₂O in their drinking water, slow their free-running activity rhythms by 3−5%. This amounts to a change of 45−75 min/day. The full effect appears within 5−7 days after the pigeons have begun drinking the D₂O.

Birds, whose previous homing performance was well known, were exposed to irregular light-dark cycles while being given 30% D₂O in their drinking water. The homing performance of such birds was compared with that of birds taken directly from the loft. Mere time disruption, however, would affect both sun compass and sun navigation leading to a scattered pattern of initial orientation. It was somehow necessary to allow the pigeons to re-synchronize their clocks to a time reference sufficiently close to loft time that the sun compass would not show any appreciable error, but with enough time difference so that sun navigation would be predictably upset.

The actual experiment was performed as follows. Nine birds were chosen with nearly identical past training from the south (New Bedford, 52 miles south). On 3 August 1969, six were put into two shift boxes with a random light schedule identical to that used by Matthews (1953) to disrupt their chronometers. Food and 30% D₂O drinking water were available at all times. The treatment was maintained for 9 days. The remaining three were left in the loft.

On 12 August the six pigeons were put into a covered cage and flown to Ithaca, New York, where they were held for us by Dr William T. Keeton and Mr Andre Gobert in an outdoor flight cage. This allowed them full view of the terrain, sky and sun for an additional 6−9 days. Food and normal drinking water were available for this time. Eachbird was then provided with a transmitter, placed again in a covered cage and flown east to Orange, Massachusetts, and together with the three taken on that date directly from the home loft were released and individually tracked by ground stations and aeroplane. This release point is 52 miles (84 km) west of the loft and 213 miles (343 km) east of the Ithaca holding cage. In a north-south direction, the holding point at Ithaca lies 2·8 miles (4·5 km) north of the line of latitude of the home loft, and the release point at Orange Airport lies 11·0 miles (17·7 km) north of this same line. In terms of sun navigation, the progress of the sun at the holding point and release point appears identical to that at the home loft within 10’ of arc (about one-third the diameter of the sun’s disk) except that they occur at different times of day. The difference in local time, therefore, is the only reliable basis by which an observer using celestial navigation could tell apart these three points. Pigeons transported to Ithaca after 9 days of disrupting lights and D₂O treatment should, at the very least, have no accurate evaluation of the time difference between the loft and the holding point (amounting to about a 20 min time shift).

An exact prediction of the pigeons’ behaviour after this treatment is impossible to make because so many variables have been introduced. We can imagine three possibilities: (1) the birds might fly west and return to the Ithaca loft. This might be the outcome if the navigational clock had been entrained to Ithaca time. (2) The pigeons might fly randomly in any direction. This might be the result for a number of reasons, including some specific effect of D₂O, or the effect of irregular photoperiod. This seemed the most likely result since so many factors might combine to produce it. (3) The pigeons might simply return to the home loft indicating that their navigation had been unaffected by our treatments.

Fig. 6 shows the results of the experiment. It give the 10 mile track bearings of six birds exposed to the disruption and holding, as well as the bearings of the three birds with similar training taken directly from the loft to Orange Airport and tracked on the same days as the experimentals. To these three tracks were added three ‘no treatment’ track bearings from Orange taken from the first part of this paper (see Fig. 2, release point ‘ORG’, the three black rectangles), bringing the number of birds in each group to six. All 12 were unfamiliar with Orange Airport. It is clear that the initial orientation of the treated birds was, if anything, slightly better than that of the controls: P = 0·004 as opposed to P = 0·008 for the controls, mean vector 9° right of home as opposed to 29° right of home for the controls. Since complete tracks were made of the six treated birds and of four of the control birds, a further analysis was made. The two groups differed in only one noticeable respect: the treated birds tended to perch longer and fly less than the controls, and thus averaged 10·9 m.p.h. homing speed as compared to 22·8 m.p.h. for the controls. This might well be expected for pigeons carrying transmitters which had not flown freely for 15−18 days. It also might have been due to a weakening effect from the dosage of D₂O, but this point was not investigated further. It should be said, however, that all the birds treated with D₂O are still alive and well in our loft at the time of writing.

The fact that both the control and experimental pigeons homed equally well from Orange is surprising. It suggests that either (1) the disruptive treatment was without effect upon their navigational clock or that (2) if the clock had been upset it was in some way reset at Ithaca to loft time or that (3) the pigeon’s navigation does not depend upon an accurate sense of time.

It is possible, but unlikely, that neither the D₂O nor the irregular light-dark cycles had an effect upon the pigeon’s navigational clock. It has been shown by Miselis & Walcott (1970) that the activity rhythms of pigeons behave like the activity rhythms that have been described for a variety of other animals. Both the onset and cessation of activity seem to be keyed to the absolute light intensity rather than to either sunrise or sunset, leading to a system whose accuracy is probably closer to ± h than it is to ± 1 min. Such an accuracy is perfectly appropriate for a sun compass but would lead to disastrous errors in a sun-navigation system. The hypothesis tested here is whether there might be another clock which was much more accurate and was rigidly fixed to solar time at the home loft. This suggestion was made by Hoffman (1965), but he cautioned that there was no direct evidence for the existence of such a stable, loft-time clock. If such a stable clock did exist, it might also be affected by the D₂O since D₂O has such a profound effect upon a wide variety of biochemical reactions (see Thompson, 1963). Whether the effects of D₂O are due to a slowing down of the basic clock mechanism or to a delay in the expression of the clock is not clear from the literature, but it is important to note the widespread nature of its effects. It seems to us unlikely that any biological time-keeping mechanism would be unaffected by D₂O, especially one required to maintain an accuracy of 1 or 2 min over a period of 2−3 weeks.

Assuming that the navigational clock was upset by our treatment, is there any chance that the pigeon could have reset it while being held at Ithaca ? It is quite clear that the pigeons could and presumably did reset their compass clocks to Ithaca time. But, if they had reset their navigational clock to local Ithaca time, they should have flown west from Orange rather than east. The only remaining possibility is for them to have set their navigational clock to some sort of universal Zeitgeber. But, as Snyder (1970) has pointed out, even a pigeon with access to a universal time cue will have trouble using a clock set in this way for sun navigation, because the information necessary to navigate by the sun must be referred to the constantly changing solar day at the home loft. Taking all of these factors into account then, it seems unlikely that the demonstrable system of bi-coordinate navigation used by our pigeons depends upon an accurate knowledge of time at the home loft. This conclusion is supported, in part, by Keeton’s (1969) demonstration of accurate initial orientation in pigeons after 6 h clock shifts as long as the sky was overcast and the sun not visible.

Snyder (1970) has also used D₂O in a pigeon-navigation experiment. He argued that a pigeon, held in a loft, given D₂O for several days, then released at a new release point, should be, at best, unable to accurately measure the sun’s rate of change of altitude at the release point. If the bird then compared this erroneous value of R with that remembered at the loft, it would orient in the wrong direction. The experiment was made at a place and time when the sun’s altitude alone held little or no information about the direction to the loft (loft and release point lay on approximately the same Sumner circle of position). Snyder showed that the pigeons watered with 30% D₂O homed as well as the controls from the unfamiliar territory. Here again, the exact nature of the predicted errors could take a variety of different forms, depending on how, exactly, the D₂O affected the presumed chronometer; but, once again, the D₂O seemed to have no effect on the pigeons’ navigation.

Since the early nineteen-fifties it has been clear that many pigeons return to their lofts from unfamiliar territory too quickly to be explained completely by their having used landmarks to pilot home. Matthews (1953) and Kramer & St Paul (1952) at first, then Wallraff (1959) and Schmidt-Koenig (1963) showed conclusively that pigeons oriented to the loft itself, as a goal, without reference to a fixed compass direction, previous training, or known landmarks. Michener & Walcott (1967) showed that individual pigeons could orient toward the loft under sunny conditions with a high degree of accuracy, again apparently without reference to known landmarks, from distances of up to 100 miles. In this study no pigeon was ever observed to fly in unfamiliar territory for more than a few miles when the sun was not visible as a disk in the sky, even though other authors have reported homing under total overcast from unfamiliar territory (e.g., most recently, Keeton, 1969).

There has been, therefore, a widely agreed-upon need for a workable theory of goal orientation (bi-coordinate navigation), but, to date, the sun-navigation theory (Matthews, 1951, 1953; Pennycuick, 1960,1961) is the only complete theory proposed that has been based exclusively upon known sense organs (see Matthews, 1968, for an extensive review). Unfortunately there has been no direct, undisputed evidence that the sun is actually used as a source of positional information (a map sense), as opposed to an azimuthal compass reference. Most of the evidence has been that under some conditions the sun is essential for some aspect of the homing performance. Here it is notoriously hard to argue conclusively, since what appears to be a total overcast at the release point can, unknown to the observer, have breaks large enough for pigeons to get a sight of the sun somewhere along the homeward track. Instead, the sun navigation theory has been extensively considered and argued because it seemed not only the most plausible theory, but one that was predictive enough to test.

The only direct evidence for the use of the sun as a navigational reference comes from two reports by Matthews (1953, 1955)-He performed two sorts of experiments: 3 h clock shifts and preventing the birds from seeing the sun during the autumnal equinox. These experiments have been extensively repeated and discussed by several authors (Rawson & Rawson, 1955; Kramer, 1955, 1957; Hoffman, 1958; Schmidt-Koenig, 1961) who obtained results which conflict with those of Matthews. The clockshifting results of Schmidt-Koenig (1961) and those reported here do not seem to have affected the birds’ navigation. Their effect can be explained solely in terms of a sun compass. The experiments depriving the pigeon of a view of the sun during the equinox have not been successfully repeated.

The most conclusive evidence that some pigeons do not need the sun to exhibit apparent goal-directed orientation in unknown territory comes from the recent report of Keeton (1969), where pigeons oriented initially very quickly and relatively accurately under heavy overcast. Keeton also reports that pigeons with 6 h clock shifts oriented homeward more accurately under overcast than under sunny conditions. The use of pigeons with 6 h clock shifts effectively disposes of the possibility that his pigeons were in some way able to view the sun’s disk through the clouds. These data all suggest that Keeton’s pigeons have some alternative compass to use in place of the sun compass, and that it is unaffected by time errors introduced by the clock shifts. It also suggests, at the very least, that the bi-coordinate navigation, which must precede the compass orientation in new territory, is not only time independent, but also sun independent. These data, of course, show that Keeton’s pigeons could and did use something other than the sun for navigation, when the sun was not visible. The clock-shifting results, including those presented here, as well as the D₂O disruption experiments show that pigeons apparently did not use the sun for navigational purposes when it was visible to them, except as azimuthal compass reference. To show that pigeons cannot use the sun as a bi-coordinate reference is clearly impossible, but the evidence against the sun’s providing a coordinate system now heavily outweighs that adduced for the system. Kramer (1953) showed quite rapid homeward orientation (apparently) by pigeons when S_H was nearly zero, and the choice of homeward direction based on sun coordinates could only be made by an accurate evaluation of S_R, which would take some time to observe. Time-shifting procedures have been used widely without any report of a pigeon ever having oriented toward a false home. Our present results extend these general findings, examining in particular the entire homeward tracks of the experimental birds and cover situations where false sun coordinates in both longitudinal and latitudinal directions were available to the pigeons for extended periods during the holding treatments. Again, we report the same result, that none of the possible expected directions from such treatments were evident in the pigeons’ behaviour, either initially or at any point in their tracks. Probably the most conclusive tests are the disruption experiments, where it is hard to imagine how an accurate biological chronometer could have survived the treatment given our six birds which none the less oriented toward the loft slightly more accurately than the untreated pigeons.

All points considered, sun navigation probably does not account for any of the observed goal orientation frequently reported in homing pigeons. What guidance system, then, could provide navigational information to these artificially transported birds, to allow them to orient routinely within ± 45° of the loft direction ? This question is made the more difficult because other theories based on inertial summation by the birds or geophysical measurements have been widely discounted (see Schmidt-Koenig, 1965; Matthews, 1968). Whatever is found to provide the ‘map sense’ for pigeons must, apparently, have the following qualifications, as a partial list: (1) it must not depend on a view of any celestial body; (2) it must not depend on an accurate time reference to local loft time; (3) it must be independent of local or distant landmarks (since it works over the open ocean, see Fig. 2); (4) it must be operative and accurate after a period of many days of imprisonment away from the home loft, either in a darkened box or a widely removed flight cage (as in our disruption experiment). We know of no theory, yet proposed, based on known sensory processes which fully meets these qualifications.

We are grateful to the many students and colleagues who have assisted in this work. Particularly we thank Norman Budnitz, Alexander Davis, Jerome Hunsaker III, Donald Maclver, Richard Miselis, Douglas Smith, Lee Snyder and Neil Tractman who assisted with the training and tracking and all other aspects of the study. Professor William T. Keeton, Jr., and Mr Andre Gobert generously cared for our experimental pigeons at Cornell. Professor Donald Griffin kindly read and criticized this paper. The research was supported in part by grants from the Committee on Research and Exploration, the National Geographic Society, the National Science Foundation, Grant GB 6777 and the National Institutes of Health, Division of Neurological Diseases and Stroke, Grant Number 5 ROI NS 08708–01.

All the releases discussed in the first portion of the paper are tabulated here. Each bird is identified by a band number and colour of the band. The track number refers to the number of tracks after training. The date of release is given next with the last digit of the year, i.e. 5 refers to 1965. Release time is given in Eastern Standard Time. The 10 mile bearing is the direction from the release point when the pigeon was at 10 miles from the release point. The trained compass bearing is the direction from the training point to the home loft, and the compass error is the difference between this bearing and the 10 mile bearing from the release. The home bearing is the direction of the home loft from the release point, and the home error is the difference between this bearing and the 10 mile bearing from the release point. The next column indicates whether the pigeon at 10 miles from release was more nearly oriented toward the home loft (H) or in its trained compass direction (C). The homing speed is given in statute miles/h, and is based on the time between release and arrival at the loft. The length ratio is the length of the pigeon’s track divided by the straight line distance from release to home. The deviation is a measure of how far the pigeon’s track deviates from a straight line. It is measured by determining the distance of the track from a straight line at ten equally spaced points. These distances are then averaged and expressed as a percentage of the straight line distance from the release point to the loft. The weather is given in standard symbols: ○, clear; ⦶, scattered clouds; CD, broken clouds; ⦷, overcast. The Ci refers to thin cirriform overcast through which the sun was visible. The visibility is in statute miles from the release point. The wind direction is the direction from which the wind was blowing, and the velocity is in statute miles/h. The number of straight track segments (STS) refers to the number of segments of track, 20 miles (32 km) or more long, that would fit within a rectangle only 2 miles wide made by each pigeon in territory which was at least 10 miles from all previous tracks. The length of each STS is given in statute miles, the bearing of its long axis and the bearing to the home loft at the start of each segment. The difference between the bearing of the STS and the direction to the home loft at the beginning of the STS is given as a ‘home error’; the compass error compares the bearing of the STS with the pigeon’s training direction. Finally for shifted birds, the false-home error expresses the difference between the observed bearing from the release point at to miles and the bearing from the release point to the false home. The letter a after a figure in the table means that the figure is an estimate based on only a partial track.