Calibration of magnetic and celestial compass cues in migratory birds - a review of cue-conflict experiments

Migrating birds have an inherited migratory program coding the general direction and distance to be travelled (reviewed by Berthold, 1991, 1996; Gwinner, 1996), and they use several different compass systems to determine the seasonally appropriate migratory direction. They use compass information derived from the geomagnetic field (reviewed by Wiltschko and Wiltschko, 1995), star patterns(Sauer, 1957; Emlen, 1975), the sun's position at sunset (Moore,1987; see also Schmidt-Koenig,1990) and patterns of skylight polarization(Able, 1982; Moore and Phillips, 1988; Helbig, 1991; Phillips and Moore, 1992). In order to maintain an accurate heading when changing weather conditions alter celestial cue availability and/or when directional information between compass systems diverges (e.g. because magnetic declination, the difference between magnetic and geographic north, varies globally), the compass systems used by birds must be `calibrated' with respect to a common reference system. This reference system may be one of the compass systems that has primacy over the others, or it may be an independent source of directional information. The simplest case would be a single reference system that is either (1) available prior to migration (if the initial calibration of the compass systems is retained throughout the migration) or (2) accessible on a regional or global scale (if the calibration of the compass systems is updated periodically during migration). Previous reviews of the compass calibration literature(Able, 1993; Åkesson, 1994; Wiltschko et al., 1997, 1998a; Wiltschko and Wiltschko, 1999)suggest that such schemes based on a single reference system may be overly simplistic, given the complex demands that long-distance migration places on the birds' orientation systems. Nevertheless, our reanalysis of the literature on cue calibration adds credence to the possibility of a single underlying reference system for calibration of the compass systems of migratory birds.

For decades, orientation researchers have been interested in the interrelationships among the different compass systems used by migratory birds. These interrelationships have typically been investigated in experiments in which birds were given conflicting directional information from two or more cues to determine which of the conflicting cues is given greater saliency by the migrants under different conditions, and whether the conflict results in a persistent change ('recalibration') of the directional information derived from one or more of the compass systems. Cue conflicts were created by (1) artificially shifting the horizontal component of the magnetic field relative to celestial cues, (2) shifting celestial cues (e.g. altering sun position with mirrors, creating an artificial skylight polarization pattern with polarizing filters that could be rotated to different alignments, or rotating an artificial starry sky) while exposing the birds to the natural geomagnetic field, or combinations of both treatments.

The findings of cue-conflict experiments have varied considerably, in some cases appearing to yield conflicting results (for reviews, see Åkesson, 1994; Wiltschko et al., 1997, 1998a; Wiltschko and Wiltschko,1999). In the present review we focus on the relationships between the magnetic compass and natural celestial cues (sun, polarized light patterns and stars). In the main analyses, we have only included experiments in which the direction of the horizontal component of the magnetic field vector was shifted relative to natural celestial cues. Thus, experiments manipulating the polarization pattern with artificial polarizing or depolarizing filters were not included in these analyses because polarization filters alter the natural polarization pattern considerably (e.g. changing the intensity and spectral properties of polarization, as well as artificially increasing the axial symmetry of the overall pattern), with unknown effects on the orientation systems of birds. Mirror experiments were also excluded because rotation of celestial cues was necessarily limited to the regions of the sky near the horizon, and because mirrors used in such experiments were made of non-ultraviolet transmitting glass (normal window glass) with the reflecting layer on the back surface, and thus are likely to have reduced the relative intensity of ultraviolet light, which has been suggested to play a role in polarized light detection by birds(Phillips and Waldvogel,1988). Also experiments performed under artificial star patterns were excluded, because of differences from the natural sky. Our aim was to find general patterns that could help to explain the wide variety of different(sometimes apparently conflicting) results in the literature concerning the use of magnetic and celestial cues.

We divided the studies/experiments into three groups:

1. After-effects of premigratory exposure to cue conflict (12 experiments). Juvenile birds raised in captivity or adult birds caught during the breeding period were exposed to a cue-conflict situation during the premigratory period. Later, once the birds had initiated seasonal migratory activity, the after-effects of the cue conflict were tested in orientation funnels to determine whether the use of a particular cue was affected, either alone or in combination with other cue(s).

2. Effects of exposure to cue conflict during migration (46 experiments). Birds (either inexperienced juveniles or experienced adults) caught during migration were exposed to a cue conflict while being tested in orientation funnels, to determine whether the birds preferentially use a particular cue or combination of cues.

3. After-effects of exposure to cue conflict during migration (18 experiments). Birds that were exposed to a cue conflict during the migratory season were subsequently tested in the absence of the cue conflict, either by being released after dark with light capsules that permitted visual observation of vanishing bearings and/or with transmitters that permitted radio tracking of the subsequent migratory paths, or by being tested in orientation funnels in some cases with a restricted subset of cues available.

The premigratory period was defined as the time during which the birds did not show migratory restlessness, and the migratory period as the period during which the birds increased their fat reserves and exhibited migratory restlessness. We discuss the following variables in our analyses: species,age, experimental season, testing location (place), direction of shift[clockwise (CW) or counterclockwise (CCW)], magnitude of shift, available cues during the cue conflict, type of experiment (i.e. cues available during testing of migratory response).

After exposure to the cue conflict during the premigratory periods, the birds in a majority of experiments were shown to have recalibrated their magnetic compass when subsequently tested during the migratory season(Bingman, 1983; Prinz and Wiltschko, 1992; Able and Able, 1993; Weindler and Liepa, 1999; Table 1)^†. That is,when subsequently tested with only magnetic cues available, birds having had access to celestial cues that included sunrise and sunset during the premigratory exposure were shown to have recalibrated their magnetic compass in 5 out of 6 cases (Table 1Ac,d, Table 1Ba,c-e). In all but one of the cases that showed recalibration of the magnetic compass, the birds were exposed to the cue conflict in cages that provided a full view of the sky (see below). An effect of the cue conflict was not evident when the birds were tested in a vertical magnetic field with access to the natural sky, indicating that the calibration of the birds'celestial compasses had not been affected(Bingman, 1984; Able and Able, 1997; Table 1Aa,b). In the one test in which birds were exposed to the cue conflict at times of day that did not include sunrise and sunset, subsequent tests carried out during the migratory season showed no evidence of recalibration of the magnetic compass(Table 1Bb).

The one inconsistency in this pattern can be found in the study by Able and Able (1990b; Table 1Ba-c) that exposed three groups of Savannah sparrows Passerculus sandwichensis with access to different celestial cues, to a shifted magnetic field. One group was exposed for continuous periods around the clock to all available celestial cues, i.e. sunrise, sun, sunset and stars (Table 1Ba). A second group was placed into the outdoor cage at night only, i.e. was allowed to experience the starry sky only, but no daytime and twilight cues (Table 1Bb). The third group was exposed to the cue conflict during daytime only, from prior to sunrise to after sunset (Table 1Bc). According to Able and Able(1990b), all groups recalibrated their magnetic compass when later tested during the migration season indoors with access to magnetic field cues only. However, they chose a control group that had never experienced a natural sky, while all the experimental groups had experienced natural celestial cues. We, therefore,considered Able and Able's `natural day sky group', which had experienced the natural day sky in an unshifted magnetic field, as a more appropriate control group. With this group as the control, only the experimental group experiencing all celestial cues (Table 1Ba) recalibrated the magnetic compass, while the other two groups(Table 1Bb,c) did not significantly shift direction, despite the fact that the third group(Table 1Bc) had access to both sunrise and sunset cues.

Interestingly, Prinz and Wiltschko(1992) observed recalibration of the magnetic compass only when the magnetic field was shifted CCW relative to the celestial cues (Table 1Ad). The birds were disoriented when the field shift was CW(Table 1Cc). This indicates a possible asymmetry in the birds' response depending on the direction of shift(see also evidence of an asymmetry in response to cue conflicts during migration).

At a first glance, the study by Alerstam and Högstedt(1983) also appears to be an exception to the general pattern discussed above, although the absence of significant orientation by controls prevents firm conclusions(Table 1Ca,b). They exposed pied flycatchers Ficedula hypoleuca during their nestling period to CW and CCW shifted magnetic fields in the nest boxes. The birds were then held in an enclosed room without access to celestial cues until the onset of autumn migratory restlessness. When the birds were tested outdoors under natural celestial and magnetic cues, they exhibited shifts in orientation that appeared to be consistent with celestial compass cues being recalibrated with respect to the rotated magnetic fields. Thus, they did not appear to recalibrate their magnetic compass as observed in the other experiments in which birds were exposed to a shifted magnetic field during the premigratory season. However, a closer look at the methods suggests a simple explanation for this apparent contradiction. The magnetic coils used to shift the magnetic field around the nest boxes were paired Helmholz coils attached to the lower part of the nest boxes (see picture published in Alerstam and Högstedt,1983). Due to the size and location of the coils, the nest box entrances were not exposed to the magnetic field inside the coils, but to the field outside of the radius of the coils looping back in the opposite direction. Nestlings sitting at the nest entrance were therefore exposed to a shift in the magnetic field that was smaller in magnitude (i.e. 45-50°)and in the opposite direction from the intended one. Consequently, the shifts in direction observed in these experiments were consistent with the nestlings recalibrating the magnetic compass at the nest box entrance where they had access to celestial cues.

In these experiments birds were tested during the migratory season while being exposed to a shifted magnetic field and natural celestial cues. In the majority of experiments that yielded meaningful results (see below), the birds shifted their orientation in accordance with the rotated magnetic field(Table 2A) indicating that they were using the magnetic field as the primary orientation cue.

A subset of experiments could not be included in the above analysis. In the experiments shown in Table 2Bthe orientation of both the control and experimental group was not significantly different from the position of sunset and was not directed into the species-specific migratory direction. It appears likely that phototactic behaviour towards the setting sun predominated over migratory orientation. As a consequence, these responses provide little, if any, information about the compass cues used for migratory orientation. We therefore excluded those experiments from further examination. We also excluded experiments in which either the control and/or experimental group was disoriented(Table 2C). In the case of disorientation, we could not distinguish an effect on the directional information provided by one or more of the compass systems from an effect on the birds' motivation. We included experiments in the analysis in which the experimental birds oriented towards the setting sun, but not the controls(Table 2Aa,b,d,e,o,p,r), since here the expected direction after the shift could coincide with the sunset position simply by chance, and the response of controls suggested that the birds were exhibiting migratory, rather than phototactic, orientation. We also included two experiments where the control birds oriented into the seasonally appropriate migratory direction, but where this direction happened to coincide with the position of sunset (Table 2Ac,q).

Interestingly, among experiments performed at sunset and resulting in significant shifts (Table 2A,references not in parentheses; see explanation in legend to Table 2), there was an asymmetry in the response of the birds exposed to CW and CCW shifts in the magnetic field. The variability in the responses exhibited by the birds exposed to the CCW rotated magnetic fields was significantly greater than that of birds exposed to the CW rotated fields (comparison of 50% inter-quartile ranges: CW, IQR=13.0°, N=10; CCW, IQR=39.0°, N=8; P=0.05; for statistical reference, see Muheim et al., 1999; Fig. 1A). Wiltschko and co-authors found that the response of Australian silvereyes Zosterops l. lateralis to CW and CCW shifted magnetic fields differed from that observed in species tested in the northern hemisphere, and suggested that the reason for this difference might lie in the reverse rotation of the sun in the southern compared to the northern hemisphere(Wiltschko et al., 1998b). To explore this possibility further, we grouped the data in `with the sun rotation' (WS) and `against the sun rotation' (AS). When we tested for variability in the expected shift, we also found a statistically significant difference between the two groups. The AS group was significantly more variable than the WS group (comparison of 50% inter-quartile ranges: WS,IQR=13.0°, N=12; AS, IQR=42.0°, N=6; P=0.04; Fig. 1B).

In addition, among the experiments not included in the above analysis, AS shifts were involved in 9 of the 11 experiments that reported sun attraction in the control and experimental group(Table 2B) and in 3 of the 4 experiments that showed directed orientation in the control group (indicating that the birds were motivated to exhibit migratory orientation), but disorientation in the experimental group(Table 2Cf-i). In Sandberg et al. (2000), the two species exposed to a CCW (AS) shift showed a delayed response to the treatment (Tables 2Bj,k, 3Af,g), while the two species exposed to a CW (WS) shift reacted immediately (Tables 2Ad,e, 3Ac,d)^‡. Overall, therefore, the birds reacted to AS shifts less readily, with lower accuracy and less consistency than to WS shifts.

In a number of recent experiments, birds were exposed to a cue conflict while being tested in orientation experiments or held in an outdoor cage and later tested in the absence of the cue conflict, either (1) under natural conditions (natural magnetic and celestial cues) in orientation funnels or release experiments, (2) in orientation funnels with access to celestial cues only (vertical magnetic field), or (3) in orientation funnels indoors with only magnetic cues available (Table 3). With four exceptions (see below), all birds continued to exhibit a shift consistent with the rotated magnetic field that they had experienced during the cue conflict. Birds tested in orientation funnels or in release experiments with access to natural magnetic and celestial cues (i.e. with the magnetic field returned to its original alignment) continued to exhibit shifted orientation (Sandberg et al., 2000; Åkesson et al., 2002; Table 3Ac,d,f-h), suggesting that information from the rotated magnetic field experienced during the cue conflict had been transferred to celestial cues (i.e. one or more of the celestial compass systems had been recalibrated). Consistent with recalibration of the celestial compass(es),birds tested in a vertical magnetic field with access only to celestial compass information, also exhibited shifted orientation (Wiltschko et al., 1998b, 1999, 2001; Table 3Ae,i,j,l;). Finally,birds tested after exposure to the cue conflict with only magnetic cues available exhibited orientation that was indistinguishable from controls(Wiltschko et al., 1999; Table 3Ak), confirming that the magnetic compass provided the primary reference system used to establish the migratory direction and was not recalibrated.

There were four exceptions to the general pattern of findings from studies of the after-effects of cue-conflict exposures during the migratory period(i.e. four studies that did not show recalibration of celestial cues relative to magnetic cues). Two exceptions are from a study of Savannah sparrows by Able and Able (1995a; Table 3Ba,b), in which birds were exposed to the cue conflict in outdoor cages with a full view of the sky. Following round-the-clock exposure to 90° CW and CCW shifts of the magnetic field for 4 clear days and nights, a group of juvenile and a group of adult birds tested indoors with only access to magnetic cues exhibited shifts of approximately 90° in both conditions. Although the responses were bimodal, so the direction of shift could not be determined, the correspondence between the magnitude of the shift and the rotation of the magnetic field during the earlier cue conflict suggests that the magnetic compass was recalibrated. In the other two exceptions, Cochran and colleagues(2004) exposed grey-cheeked thrush Catharus minimus and Swainson's thrush Catharus ustulatus during migration to 70° to 90° CW rotated magnetic fields (Table 3Aa,b). The birds in Cochran et al.'s study were exposed to the cue conflict in outdoor cages in which they had a full view of the entire sky down to the horizon for variable periods up to about 3 h, including sunset. When subsequently released after the sunset period, radio-tracked birds followed tracks consistent with a recalibrated magnetic compass. On subsequent nights after the Swainson's thrushes had experienced the natural (unrotated) magnetic field through one sunrise and sunset period, the shifted orientation of radio-tracked birds was eliminated. Sandberg et al.(2000) carried out a study very similar to Cochran et al.(2004), but instead of using radio telemetry they followed the vanishing directions after releasing the birds (one of the four species used was Swainson's thrush, as in Cochran et al., 2004) with light capsules. Sandberg et al.'s results were comparable to the funnel experiments they carried out with the same individuals during exposure to the cue conflict prior to the release experiments(Table 3Aa,c,d,f,g), i.e. the birds did not recalibrate their magnetic compass. Cochran and colleagues reported that the directions of their thrushes were already deflected from the very beginning of the tracking, which would coincide with the visual observations of the initial orientation of birds carrying light capsules performed by Sandberg et al.(2000). The major difference between the two studies is unlikely to be due to the difference in the methods used to record the data (funnel experiments vs radio tracking/release experiment), as suggested by Cochran et al.(2004), but rather to the difference in exposure to the cue conflict (open cage with full view of sky at horizon vs view from funnel). Summarising, among the studies of after-effects of cue conflicts carried out during the migratory period, the experiments by Able and Able(1995a) and Cochran et al.(2004) were exceptional in two respects: (1) they were the only studies in which the birds experienced the cue conflict with a full view of the sky down to the horizon, and (2) they were the only studies to show recalibration of the magnetic compass.

The studies by Able and Able(1995a) and Cochran et al.(2004) suggest that recalibration of the magnetic compass may depend not on season (premigratory vs migratory), but on access to celestial cues(Table 4).

Cue-conflict exposures are typically carried out in two different ways: (1)by exposing the birds to the shifted magnetic field in outdoor cages with a full view of the surrounding, including the horizon or (2) by exposing them in orientation funnels (Emlen and Emlen,1966) or octagonal registration cages(Wiltschko et al., 1971; Wiltschko and Höck, 1972;Wiltschko and Wiltschko, 1975a,b)that restrict the view of the celestial cues near the horizon to different degrees. Most funnels allow the birds to view a 120-160° section of the sky centered around the zenith. A few studies have been carried out in funnels with a surrounding shield that allowed the birds to see a maximum of 90-102° around the zenith, as do the octagonal registration cages. Studies that did not explicitly give the view of sky in degrees and did not mention the presence of a shield surrounding the funnel, were categorized into the first group.

Ignoring season, birds that had a full view of sky during the cue-conflict exposure recalibrated their magnetic compass in 8 out of 9 studies(Table 4). Birds that did not have a view of the sunset cues down to the horizon (view limited to 95-160° centered around the zenith from a funnel or octagonal orientation cage) followed the shifted magnetic field and presumably recalibrated celestial compass cues (see studies with release experiments after cue-conflict exposure; Table 3A), or showed no magnetic compass calibration, in 28 out of 29 studies (Table 4). In only one of these 29 studies was there evidence of recalibration of the magnetic compass (Table 1Be). The four studies in which the surrounding shields restricted the view of the birds in the funnels to only 90° around the zenith showed no response whatsoever to the magnetic field shift (Sandberg et al., 1988a,b; Table 2Aa,b,o,p; see below).

It was previously argued that only one exposure to the cue conflict would lead to a dominance of magnetic cues, thus to a recalibration of the celestial compass(es), while repeated exposures to the cue conflict would lead to a dominance of celestial cues and to a recalibration of the magnetic compass(e.g. Wiltschko et al., 1997, 1998a; Wiltschko and Wiltschko,1999). In 10 out of the 22 studies in Table 2A tested at sunset the birds were exposed multiple times to the cue conflict(Table 2Ah-k,m,n,v-y), without recalibrating the magnetic compass. Some of the early studies(Wiltschko et al., 1971;Wiltschko and Wiltschko, 1975a,b; Table 2Ah-k) even exposed the birds to the cue conflict repeatedly during both sunset and sunrise cues, but in octagonal orientation cages that shielded all but 95-102° of the sky,and did not observe recalibration of the magnetic compass.

The results of our initial analyses were consistent with the conclusions of earlier reviews, which suggested that the relative saliency of celestial and magnetic cues differs during the premigratory and migratory periods (e.g. Wiltschko et al., 1997, 1998a; Wiltschko and Wiltschko,1999). During the premigratory period, celestial cues provided the primary calibration reference; the magnetic compass was recalibrated when birds were exposed to a rotated magnetic field in the presence of natural celestial cues. In a majority of experiments carried out during migration,however, birds exposed to a rotated magnetic field during migration showed a corresponding shift in the direction of migratory orientation, indicating that they are giving precedence to magnetic cues, although the consistency of this response depended on the direction that the magnetic field was rotated (see further discussion below). When tested for after-effects under natural celestial cues in the absence of magnetic cues (vertical magnetic field), they maintain the shifted direction, suggesting that the celestial compass cues had been recalibrated relative to the magnetic field(Bingman and Wiltschko, 1988;Wiltschko et al., 1998b, 1999, 2001; Tables 3Ae,i,j,l,Bc).

Subsequent analyses suggested an alternative explanation for the differences in the response to the cue conflict. The response to the cue conflict during both the premigratory and migratory periods appears to be correlated with access to celestial cues during exposure to the cue conflict(Table 4). Birds exposed to the cue conflict during times of day that included sunset, recalibrated the magnetic compass if they had an unobstructed view including the region of sky near the horizon, regardless of whether they experienced the cue conflict during the premigratory (4 out of 5 experiments) or migratory (4 out of 4 experiments) season (Tables 1and 3). In contrast, birds that were exposed to the cue conflict at times of day that did not include sunset or were prevented from seeing the region of sky near the horizon, did not recalibrate the magnetic compass, but instead relied on the magnetic field as the primary source of compass information and used the magnetic field to calibrate the remaining celestial compass cues. Thus, the integration of the multiple compass systems used by migratory birds may involve a much simpler and more invariant cue hierarchy than suggested by previous reviews (cf. Able, 1993; Wiltschko et al., 1997).

Our analysis of cue-conflict experiments suggests that an unobstructed view of the sunset sky and, in particular, the region of sky near the horizon, is necessary for recalibration of the magnetic compass to occur. A number of studies suggest that polarized light patterns present at sunset (and, possibly also, sunrise) provide the reference information used to calibrate the magnetic compass. In contrast to controls, birds experiencing a shifted magnetic field under depolarizers that eliminated celestial polarized light cues, but permitted detection of both sun position and star patterns, did not show shifted orientation when subsequently tested with only magnetic field cues available (Able and Able,1993). Able and Able(1995b) also showed that birds exposed to a shifted artificial polarization pattern at dusk and dawn in the ambient magnetic field, without seeing the sun or stars, recalibrated their magnetic compass according to the shifted polarization pattern. Exposure to an artificial polarized light pattern at sunset was also found to produce recalibration of the sun compass in a migratory bird(Phillips and Moore, 1992). Evidence from homing pigeons also suggests that sunrise (and possibly also sunset) polarized light cues from the region of sky near the horizon may be used as a compass calibration reference(Phillips and Waldvogel,1988).

While there is considerable evidence for the role of cues available at sunset in calibrating other compass systems, the role of cues available at sunrise is less clear. Cue-conflict experiments by Sandberg et al.(2002) during sunrise showed results that were similar to their sunset results (the birds followed the magnetic compass), so sunrise factors alone, at least when the birds' view of the horizon sky is obstructed, do not provide the birds with the necessary information for magnetic compass recalibration. However, sunrise cues may be used as a calibration reference when birds have a full view of the sky down to the horizon, perhaps in combination with sunset cues (e.g. to provide an estimate of true geographic North; Phillips and Waldvogel, 1988),but no studies have been carried out yet to specifically address this possibility.

Interestingly, although birds tested in funnel experiments in which their view of the sky was restricted to between 95-160° centered around the zenith did not recalibrate the magnetic compass, they still responded to celestial polarized light cues at sunset (see references in Table 2 and Wiltschko and Wiltschko,1974). In the absence of magnetic cues, birds tested with depolarizers covering the funnels were disoriented, whereas birds given an undisturbed view of the natural sky from orientation funnels showed seasonally appropriate orientation (Helbig, 1990, 1991). Thus, a subset of the available skylight polarization patterns appears to be used by migratory birds at sunset even when a view of the horizon sky is blocked. Interestingly,however, birds whose view of the sky was restricted even further (i.e. restricted to 90° centered around the zenith; Sandberg, 1988, 1991; Sandberg et al., 1988a,b)showed aberrant orientation and/or did not respond normally to the shifted magnetic field or polarization patterns. Instead they showed: (1) no reaction to rotation of the magnetic field (Table 2Aa,b,o,p), (2) non-migratory orientation that was not significantly different from the sunset azimuth(Table 2Ba-e and Sandberg, 1988, 1991), or (3) no response to an artificially shifted polarization pattern(Sandberg, 1988). Thus birds followed the shift in magnetic direction instead of recalibrating the magnetic compass, and recalibrated the available celestial compass cues with respect to the rotated magnetic cues, only when they had access to a view of sky in the funnel experiments that was ≥95° but ≤160°; Table 2, and see below).

Overall, these findings indicate (1) that access to the region of the sky immediately above the horizon is necessary for recalibration of the magnetic compass and (2) that access to a region of sky spanning at least 95°around the zenith appears to be necessary for the birds to be able to react to changes of the magnetic field (see below) or other celestial cues. These conclusions suggest that polarized light cues from the region of sky included in the view from 95° to 165° around the zenith, but not from the region of sky immediately around the zenith (i.e. within ±45°),functions differently (as a source of compass information) from those from the region of sky immediately above the horizon (as a compass calibration reference), as suggested by Phillips and Waldvogel(1988) for homing pigeons. And, secondly, that access to light from the celestial hemisphere may influence the birds' use of the light-dependent magnetic compass (see below).

The importance of sunset (and maybe also sunrise) cues for magnetic compass recalibration is further supported by two studies exposing birds to a full view of celestial cues down to the horizon that did not result in a magnetic compass recalibration. In both studies the cue-conflict exposure did not include sunset or sunrise.

Consistent with the conclusion that sunset (and possibly also sunrise)polarized light cues are necessary for recalibration of the magnetic compass,two studies found that exposure to a cue conflict with a full view of the sky did not result in recalibration of the magnetic compass, if the exposure did not include sunrise and sunset. In an experiment carried out during the premigratory season, Able and Able(1990b) exposed birds to the cue conflict only during the night, so that the birds could only use stars for calibration (Table 1Bb). Even though the birds were exposed to a full view of the starry sky in the shifted magnetic field, they did not recalibrate the magnetic compass. In an experiment carried out during migration, Åkesson et al.(2002) exposed birds to the cue conflict during the afternoon hours, thus the birds only had access to sun and other daytime polarized light cues and did not experience sunset cues during the exposure (Table 3Ah). When tested the same evening under natural conditions, the birds followed the magnetic field shift, thus had not recalibrated the magnetic compass. Although small in number, these studies support the theory that birds need a full view of celestial cues at sunset (or possibly sunrise)in order to recalibrate their magnetic compass.

If the birds given brief exposures to cue conflicts in Emlen funnels do not have access to celestial cues from the region of the sky near the horizon, but instead rely on the magnetic compass and use it to calibrate other celestial compass systems (see earlier discussion), what celestial cues are recalibrated relative to the magnetic field under these conditions? Clearly, other types of celestial cues must be involved, e.g. either star patterns or zenith polarized light patterns. Star patterns do not appear to be involved in recalibration of the magnetic compass. Experiments that exposed birds to a premigratory cue conflict between magnetic and stellar cues, by changing the direction of rotation of artificial star patterns, showed no calibration of the magnetic compass when the birds were tested with access to magnetic cues only(Wiltschko and Wiltschko,1976; Beason,1987; Weindler et al.,1998; but see Able and Able,1990a). Instead, the magnetic compass seems to be involved in calibrating the star compass, which by itself only provides information about the deviation of the migratory direction away from the center of celestial rotation, to establish the population-specific migratory direction (Weindler et al., 1996, 1997). Therefore, we envision a cue hierarchy in which celestial cues available at sunset/sunrise(presumably polarized patterns from the region of sky near the horizon)provide the primary reference system for calibration of the magnetic compass,while the magnetic compass in turn is used to calibrate the star compasses, as well as zenith polarized light patterns. The magnetic compass may play a similar role in calibrating the sun compass for daytime orientation, as suggested for homing pigeons (Phillips and Waldvogel, 1988; but see Wiltschko et al., 1976, 1983, 1984; Wiltschko and Wiltschko,1990).

Prinz and Wiltschko (1992)obtained evidence for an asymmetry in the response to premigratory exposure to shifted magnetic fields. They observed recalibration of the magnetic compass only when the magnetic field was shifted CCW/AS relative to the celestial cues. Their birds became disoriented when the field was shifted CW/WS. Unfortunately, in the only other studies involving cue conflicts during the premigratory period in which birds were exposed to a CW shift, the birds oriented bimodally (Able and Able,1990b), making it impossible to determine the direction of the shift, or they were not tested to determine whether recalibration of the magnetic compass had occurred (experiments in a vertical magnetic field; Able and Able, 1997). Our analyses of cue-conflict experiments carried out during migration also provided evidence for an asymmetry in the response to different directions of rotation of the magnetic field. Birds exposed to CW shifts (and even more to WS shifts) of the magnetic field generally followed the shift with remarkable accuracy compared to birds exposed to CCW/AS shifts, which were more variable in their responses (see also Åkesson, 1994). Moreover, CCW/AS shifts were also disproportionately represented in experiments reporting a phototactic response toward sunset (rather than migratory orientation) in both control and experimental groups and in experiments that showed directed orientation in the control groups, but disorientation in the experimental group^§. This asymmetry is analogous to the asymmetry reported by Prinz and Wiltschko(1992) during premigratory exposures, although they found the opposite tendency in the effects of the two treatments (i.e. a strong shift by birds exposed to a CC/AS shift and disorientation in birds exposed to a CW/WS shift).

One interesting possibility is that the asymmetry in the birds' responses to CW/WS and CCW/AS rotations of the magnetic field may be related in some way to the strong asymmetry in the role of the right and left eyes in mediating the light-dependent magnetic compass response of birds(Wiltschko et al., 2002). Evidence that the magnetic compass of birds is light dependent(Wiltschko et al., 1993)suggests that birds perceive the magnetic field as a pattern of light intensity and/or color, either superimposed on the normal visual input or processed by a separate (i.e. non-visual) photoreception system. For example,a recent theoretical model by Ritz et al.(2000) suggests that the biophysical process that underlies the light-dependent magnetic compass will cause a decrease in sensitivity to light in magnetically sensitive photoreceptors aligned either parallel or antiparallel to the axis of the magnetic field. If so, the magnetic field might be perceived as two darker areas (or areas of a particular color, if the magnetically sensitive receptors provide inputs to a color-opponent mechanism) superimposed on the visual field at either end of the magnetic field axis^¶ (see Fig. 6 in Ritz et al., 2000). In either case, the inclination of the magnetic field in the northern and southern temperate zones will cause this pattern to be `tilted' at a steep angle, with one end of the pattern viewed against the sky and the other viewed against the substrate; in orientation studies, the lower end of the pattern will typically be viewed against the interior of an Emlen funnel or octagonal registration cage (cf. Fig. 2).

If birds only respond to the magnetic field in the upper (or lower) visual field, then the problem of the axial symmetry of the pattern is eliminated(see earlier discussion; Fig. 2).However, the lateralization of the avian magnetic compass could make perception of this pattern in only the upper (or lower) hemisphere strongly dependent on the horizontal alignment of the bird's head. This dependence might account for the tendency of birds to respond more readily to∼90° rotations of the magnetic field in one direction (e.g. CW),rather than the other (e.g. CCW), but only if birds tend to align their heads non-randomly with respect to absolute (i.e. geographic) coordinates. So, for example, if birds use celestial cues to preferentially align their heads toward a particular geographic direction (e.g. geographic North) when checking the magnetic compass, the dark area produced by the magnetic compass in the upper visual field would be more easily detected in alignments of the magnetic field that caused this portion of the pattern to fall in the visual field of the right (magnetically sensitive) eye, rather than in the visual field of the left (non-magnetically sensitive) eye. Because of the differences in the inclination of the magnetic field in the two hemispheres (inclined down to the north in the northern hemisphere and up to the north in the southern hemisphere; Fig. 2), the asymmetry in the birds' response to CW and CCW rotations of the magnetic field should be reversed (see below).

At present, there is no evidence that we know of to suggest that birds align themselves with respect to a geographically fixed reference, such as celestial light cues, when consulting the magnetic compass. However, the well-known asymmetry in the distribution of light intensity and spectral content of the evening sky may produce the same result. At sunset, there is a strong gradient in the intensity (brightest toward the sun) and spectral content (proportion of short wavelengths greatest away from the sun; McFarland and Munz, 1975) of natural skylight. If the magnetic field is perceived as a pattern of light intensity and/or color, and therefore is sensitive to the wavelength and intensity of light (as suggested by recent studies on the magnetic compass orientation of migratory birds; Wiltschko et al., 2000; Wiltschko and Wiltschko, 2001; Muheim et al., 2002), then a ∼90° rotation of the magnetic field in one direction (e.g. CW) may result in a superposition of the magnetically derived pattern on the natural distribution of light intensity and color that makes the pattern more easily detectable, while the opposite rotation of the magnetic field (e.g. CCW) may result in the pattern being much less detectable(Fig. 2). The crucial point,however, is that the asymmetry in the birds' response to CW and CCW rotations of the magnetic field should be opposite in the northern and southern hemispheres. This is independent of whether the asymmetry in the response to CW and CCW rotations of the magnetic field is due to the lateralization of the magnetic compass and a tendency for birds to align themselves non-randomly when consulting the magnetic compass, or to the relationship of the light-dependent response of the magnetic compass to the rotated magnetic field and the natural distribution of light intensity and wavelength in the evening sky. Unfortunately, relatively few cue-conflict experiments under unmanipulated celestial cues have been carried out in the southern hemisphere,but nevertheless, data both from the cue-conflict experiments carried out during migration (see Tables 2and 3), and from Prinz and Wiltschko's (1992)cue-conflict experiment carried out during the premigratory period, are compatible with this prediction, suggesting that further work is clearly justified.

It has been previously suggested that there is a difference in the relative saliency of celestial and magnetic cues between the premigratory and migratory season. The magnetic compass was shown to be calibrated with respect to celestial cues during the premigratory season, but was suggested to act as the primary reference for calibration of other compass systems (including polarized light compasses) during migration. Able and Able(1996) found that recalibration of the magnetic compass in autumn was no longer evident in spring, which suggests that recalibration of the magnetic compass occurs prior to the onset of each migration. In addition, Weindler and Liepa(1999) showed that adult pied flycatchers recalibrate their magnetic compass when exposed to a cue conflict before the onset of migration, suggesting that calibration of the magnetic compass prior to migration occurs in adult, as well as juvenile, birds. As noted previously, however, the studies by Able and Able(1995a) and Cochran et al.(2004) that reported recalibration of the magnetic compass during the migratory season, suggest that magnetic compass calibration is not necessarily limited to the premigratory season.

Does it make sense that the magnetic compass would only be recalibrated before the start of the migratory journey and not during migration?Recalibration of the magnetic compass enables the birds to correct for differences in magnetic declination (i.e. the difference in the direction of the magnetic and geographic poles at a given location on the Earth's surface)that would otherwise cause a discrepancy between the directional information provided by the magnetic and celestial compasses. Therefore, calibrating the magnetic compass with respect to celestial cues before the onset of migration would eliminate the discrepancy between magnetic and celestial bearings in areas with large declinations (Able and Able, 1993, 1999). As a bird migrates to lower latitudes, changes in declination are less pronounced and the magnetic compass becomes more reliable, while at the same time, the course of the sun and the alignment of the polarized light pattern at sunset change with latitude, and star patterns learned during ontogeny are replaced by new configurations. This combination of factors has been suggested to explain why the magnetic compass appears to become the primary reference for maintenance of the migratory direction (and calibration of other cue systems) once migration is under way (reviewed by Alerstam, 1990; Muheim et al., 2003). Based on theoretical considerations, Alerstam(2001) concluded that it would not be efficient for bird migrants to follow magnetic loxodrome routes(constant magnetic courses) on the American continent, when the movement has a relatively strong east-west component. There, migrants following a constant magnetic course need to travel a longer distance than when following a geographic loxodrome (constant geographic route). In these cases, at least,repeated recalibrations of all of the compass systems, including the magnetic compass, with respect to a reference system that is tied to true geographic coordinates (e.g. celestial cues) would be adaptive. This, however, reverses on the European continent where strong east-west traveling migrants may travel shorter distances by following a magnetic loxodrome(Alerstam, 2001; Bingman et al., 2003). Therefore, regular recalibration of the magnetic compass may be a more efficient strategy on the American continent than in Europe and Asia. In conclusion, selection may have favored different orientation strategies in different regions of the world (Alerstam,2001). It should be noted, however, that there are a number of alternatives to using the magnetic compass as a calibration reference during migration that would help to minimize the distance traveled by migratory species in Europe and Asia. Birds that use a single calibration reference derived from celestial cues both prior to and during migration could: (1)exhibit programmed change(s) in compass heading at appropriate points along the migration route (Gwinner and Wiltschko, 1978; Beck and Wiltschko, 1988; Helbig et al., 1989; Munro et al.,1993), or (2) exploit seasonal changes in celestial polarized light patterns at sunrise or sunset to produce adaptive deviations from the geographic loxodrome routes (i.e. more closely approximating magnetic loxodrome routes).

A large number of cue-conflict experiments have been performed over the last 35 years, and the aim of this review was to determine whether a reanalysis of the findings of these experiments, including those carried out during the past 5 years since the last comprehensive review of this literature, can provide any new insights into the integration of celestial and magnetic cues by migratory birds. Our findings suggest that accessibility to celestial (polarized light) cues at sunset from the region of sky immediately above the horizon can account for the differences in the response to the cue conflict observed between cue-conflict experiments carried out during the migratory season and those carried out prior to migration, rather than seasonal differences in the relative saliency of magnetic and celestial cues. If this interpretation is correct, the hierarchy of compass cues used by migratory birds may be much less plastic than previously thought. Based on these analyses, we propose that under natural conditions a single reference system is ultimately responsible for calibrating all of the compass systems of migratory birds. The proposed reference system involves polarized light cues visible at sunset from the region of sky near the horizon, as suggested previously for homing pigeons (Phillips and Waldvogel, 1988).

An interesting question for future research is whether sunset polarized light patterns alone are responsible for calibration of avian (or at least migratory bird) compass systems. A reference system based on sunset polarization patterns alone will shift considerably over the course of the year, due to the seasonal changes in the position of the setting sun. While in some instances a shifting compass course may be adaptive (see previous discussion) in cases where a fixed geographic loxodrome route is more efficient, averaging information from the polarization patterns present at sunrise and sunset can be used to derive an estimate of geographic north that is independent of time of year. An interesting question for future research,therefore, is whether polarized light cues present at sunrise play any role in the calibration of avian compass systems.

We want to thank Thomas Alerstam, Susanne Åkesson and Johan Bäckman and two anonymous referees for valuable comments on an earlier version of the manuscript. This study was supported by traveling fellowships by the Royal Physiographical Society in Lund, Sweden and the Journal of Experimental Biology, as well as the Swiss National Science Foundation(postdoctoral fellowship; all to R.M.), and by the National Science Foundation grant IBN04-25712 (to J.B.P.). Part of this study was financed by a grant from the Swedish Science Research Council (to S. Åkesson).