Main

The magnetic compass of birds is light-dependent9,10, and exhibits strong lateralization with input coming primarily from the right eye11. However, the primary biophysical process underlying this compass remains unexplained. Magnetite3,4,5,12 as well as biochemical radical-pair reactions7,8 have been hypothesized to mediate sensitivity to Earth-strength magnetic fields through fundamentally different physical mechanisms. In the magnetite-based mechanism, magnetic fields exert mechanical forces3. In the radical-pair mechanism, the magnetic field alters the dynamics of transitions between spin states, after the creation of a radical pair through a light-induced electron transfer. These transitions in turn affect reaction rates and products7,8. Although in most radical-pair reactions the effects of Earth-strength magnetic fields are masked by stochastic fluctuations, model calculations13 show that such effects can be amplified beyond the level of stochastic fluctuations in specialized radical-pair receptor systems.

Exploiting the principles of magnetic resonance, we developed a diagnostic tool to identify a radical-pair process as the primary process for a physiological magnetic compass. No change in magnetic alignment of magnetite receptors is expected for weak oscillating fields with frequencies larger than 100 kHz (ref. 14). However, an oscillating magnetic field that is in resonance with the splitting between radical-pair spin states can perturb a radical-pair mechanism by directly driving singlet–triplet transitions. In typical biomolecules, many hyperfine splittings occur in the range of 0.1–10 MHz and a limited number may exist in the range of 10–25 MHz (ref. 15).

We used the migratory orientation of European robins, Erithacus rubecula, to detect the possible effects of oscillating magnetic fields on the underlying magnetoreception mechanism. Orientation tests were performed during spring migration under 565 nm light; conditions under which robins normally show excellent orientation using their inclination compass16,17. All birds were tested indoors, in the local geomagnetic field of 46 µT and 66° inclination. In addition to the control condition (geomagnetic field alone, no oscillating field), we used four experimental conditions in which an oscillating magnetic field was added to the geomagnetic field (Fig. 1).

Figure 1: Experimental set-up.
figure 1

Orientation of the 7.0-MHz oscillating magnetic fields (black arrows with sine curve) parallel, at a 24° (vertical) and at a 48° angle to the geomagnetic field (grey arrows; see Fig. 2c–e for results). In the parallel and 48° conditions, the oscillating fields have the same angle with respect to the birds in our experimental set-up.

Full size image

In the control condition, the robins exhibited seasonally appropriate northerly orientation (Fig. 2a), but in the presence of broadband (0.1–10 MHz, 0.085 µT) and single-frequency (7 MHz, 0.47 µT) oscillating fields, both vertically aligned (see Fig. 1), the birds were disoriented (Fig. 2b, d).

Figure 2: Effects of oscillating magnetic fields on magnetic orientation behaviour of European robins.
figure 2

Triangles indicate the mean headings of the 12 test birds, arrows represent the grand mean vectors (unit length = outer circle radius; see Table 1 for numerical values). The inner circles mark the 5% (dotted) and the 1% significance border of the Rayleigh test27. a, Tests in the geomagnetic field only. b, Tests in the geomagnetic field and a broadband (0.1–10 MHz) noise field of 0.085 µT. c–e, Tests in a 7.0-MHz field of 0.47 µT, oriented either parallel (c), at a 24° angle (d), or at a 48° angle (e) to the geomagnetic field.

Full size image

To confirm that the observed behavioural change was caused by a direct effect of the oscillating fields on the magnetic compass and not by nonspecific effects due to changes in motivation and so on, we varied the alignment of the 7.0-MHz field. The frequencies at which an oscillating field perturbs a radical-pair reaction depend not only on the chemical nature of the radical pair, but also on the alignment of the oscillating field with respect to the static field18. This implies that the responses of a magnetic compass system based on radical pairs in the presence of a weak, single-frequency oscillating field can depend on the alignment of the oscillating field, whereas nonspecific effects should occur independently of alignment. We tilted the oscillating field 24° to the north or 24° to the south, so that the two oscillating fields were aligned at the same angle relative to the vertical, but at different angles, parallel and 48°, relative to the geomagnetic field (Fig. 1).

When the oscillating field was parallel to the geomagnetic field, the birds oriented in the migratory direction (Fig. 2c) and their response was indistinguishable from that of the control condition (Table 1). In contrast, when the same oscillating field was presented at 24° and 48° relative to the geomagnetic field, the birds were disoriented (Fig. 2d, e) and their response differed significantly from that of the control condition (P < 0.01). The intra-individual scatter in the distribution of nightly headings, as reflected by the length of the birds' mean vectors (rb), was indistinguishable from that of the control condition when the 7-MHz oscillating field was parallel to the geomagnetic field, but was significantly greater (lower rb) in the other three oscillating-field conditions (that is, broadband and 7 MHz at 24° and 48° angles) (see Table 1).

Table 1 Orientation of European robins in different oscillating magnetic field conditions
Full size table

Our findings show that it is unlikely that oscillating fields have an effect on magnetite-based receptors3,4,5,12, because the dampening effects of the cellular environment prevent magnetite particles from tracking weak radio-frequency magnetic fields. Even in very-low-viscosity physiological conditions (spherical single-domain magnetite in water) we can estimate, following ref. 14, that a 7-MHz field would require an intensity of 285 µT to produce a noticeable change in alignment, which is far stronger than the 0.47 µT fields used in our experiments. Likewise, frequencies used in these experiments of less than 10 MHz are far from the expected ferromagnetic resonance frequencies in the GHz range19, thus rendering thermal or lattice vibration effects of the oscillating fields on magnetite unlikely.

In contrast, resonance effects of oscillating magnetic fields in the frequency range of 0.1–10 MHz are expected in a radical-pair mechanism because hyperfine splittings occur in this range15. Resonance effects in this frequency range would also be expected in the context of Leask's optical pumping hypothesis6, although the lack of evidence for a biological source of energy in the radio-frequency range required by the optical-pumping process6 makes this mechanism unlikely.

By what physical mechanism could the remarkably weak oscillating fields used in our experiments (0.085 µT, 0.47 µT) affect a radical-pair reaction, and in turn, a radical-pair-based compass system? A simple model calculation (see Methods) suggests that at least in some radical-pair reactions (radical pairs with one dominant hyperfine interaction and a long lifetime), a resonant oscillating magnetic field of a thousandth of the geomagnetic field strength can produce a detectable effect. This remarkable sensitivity to very weak resonant oscillating fields is a noteworthy feature and further studies should analyse the limits of sensitivity in more realistic descriptions of radical pairs.

Our data, together with the above analysis, indicate that a magnetically sensitive radical-pair process in European robins is linked to the physiology of magnetic compass orientation. The most straightforward explanation for our findings is that the radical-pair process indicated by our experiments works as the primary process underlying magnetic compass orientation in European robins and probably in other birds10. Of course, we cannot exclude the possibility that a radical-pair reaction that is part of an unrelated biochemical pathway was affected. However, the fact that resonance effects are only expected in specialized radical-pair systems that can also detect the geomagnetic field7,13, makes it unlikely that a radical-pair process not associated with magnetoreception was affected. There is currently no evidence supporting the existence of such a magneto-sensitive radical-pair process outside the context of magnetoreception and even if one existed, it is uncertain whether it would affect orientation behaviour. In our study we observed no change in activity between birds in oscillating-field and control conditions; and food intake and the general appearance of the birds was normal, suggesting that their health and motivation were unaffected by the brief 75 min exposure to oscillating magnetic fields. In view of this, our findings probably reflect a direct effect of the oscillating fields on the compass mechanism.

This conclusion does not rule out the possibility that birds possess an additional magnetically sensitive system based on magnetite. Magnetite in the form of single domains and super-paramagnetic crystals embedded in specialized structures has been described in the ethmoid region and in the upper beak of migratory birds and pigeons20,21. However, behavioural evidence22,23,24 as well as electrophysiological recordings25,26 suggest that the magnetite discovered is not involved in magnetic compass orientation, but instead forms the basis of a magnetic-intensity sensor, potentially used in a magnetic ’map’ sense for determining geographic position.

Our study establishes the use of oscillating magnetic fields as a diagnostic tool that can indicate the involvement of a magneto-sensitive radical-pair reaction in birds. Extending this tool to determine the frequency range in which oscillating fields affect the radical-pair mechanism can reveal the chemical nature of the radical pairs involved. Finally, using oscillating magnetic fields as a diagnostic tool is not specific to birds and should be easily transferable to assays with other animal groups. The threshold intensity at which oscillating-field effects can be observed provides information about the underlying mechanism. Behavioural effects from oscillating fields of similar intensity to those used in the present study would suggest a radical-pair mechanism. The absence of behavioural effects from oscillating fields of intensities greater than 50 µT would make a radical-pair mechanism unlikely.

Methods

Test birds

European robins are small passerines that migrate at night. The test birds were mist-netted as transmigrants in early September 2002 in the Botanical Garden near the Zoological Institute in Frankfurt am Main (50° 08′ N, 8° 40′ E). The birds were kept indoors in individual cages over the winter on a photoperiod that simulated local conditions until December, but was then reduced to 8:16 h light:dark. In the beginning of January 2003, the photoperiod was increased to 13:11 h light:dark. This induced premature Zugunruhe (nocturnal migratory restlessness); the experiments took place between 13 January and 17 February 2003.

Test conditions

The tests took place in wooden huts in the garden of the Zoological Institute within the local geomagnetic field of 46 µT and 66° inclination. To produce the oscillating fields, we modified a test design developed by J.B.P. for similar tests (J.B.P., unpublished), consisting of a coil antenna (210 cm diameter) mounted on a rotatable wooden frame surrounding the test arena. Oscillating currents from a high frequency (HF) generator (Stanford Research Systems DS 34) were amplified by a HF amplifier (Research AF Model 25 W 1,000) and fed into the coil through a resistance of 51 Ω. The coil consisted of a single winding of coaxial cable (RG62A/U, 93 Ω) with 2 cm of the screening removed opposite the feed. The HF field was measured daily, before each test session, using a spectrum analyser (HP89410A) and a magnetic field probe (Rohde & Schwarz, HZ-11816.2770.0, 3 cm probe). Within the test arena, the inhomogeneity of the field was less than 15%. Variations in field intensities between tests were less than 20% of the average value. The HF generator and amplifier were placed outside the huts in varying positions with respect to the test arena. They were switched on during the majority of control tests, but with the power generator turned to zero; comparisons with control tests without this arrangement revealed no observable effect of this procedure.

Test apparatus and procedure

Testing followed standard procedures16: birds were tested individually in funnel-shaped PVC cages (35 cm upper diameter; 20 cm high) lined with coated paper (BIC Germany, formerly TippEx); the birds left scratches in the coating as they moved. The cages were covered with an opaque plexiglass cover and placed in PVC cylinders, the top of which consisted of a plastic disk carrying the same green light-emitting diodes as those used in earlier studies9,16 (peak frequency at wavelength λ = 565 nm, with λ/2 at 533 and 583 nm, respectively). The light passed through two diffusers before reaching the bird with an intensity of 2.1 mW m-2.

The birds were tested once per day. Tests began when the light went out in the housing cages and lasted about 75 min. Each bird was tested three times in each condition. The three tests were arranged in sets; the set of second and third tests began after the set of first and second tests respectively was completed. Within each set, the tests in the various conditions were performed in a pseudo-random order, with the sequence differing between birds.

Data analysis and statistics

For the data analysis, the coated paper was divided into 24 sectors, and the scratches per sector were counted by experimenters that were blind to the test condition. The heading of the respective test was determined by vector addition. From the three headings per test condition for each bird, the mean vector with heading αb, and length rb, was calculated. The twelve αb values were combined to a grand mean vector, which was tested for directional significance using the Rayleigh test27. The distributions of the birds' αb values in different conditions were compared using the Mardia–Watson–Wheeler test27. The rb values, representing the intra-individual variance in locating the migratory direction, are not normally distributed; and so medians are given for each test condition. The rb values were compared with those obtained under the control conditions using the Wilcoxon test for matched pairs of data.

Model calculations

We used a one-proton radical-pair model28 with an isotropic hyperfine coupling, a, of 0.5 mT, an anisotropy, α of 0.3, and a lifetime of 20 µs (corresponding to the observed lifetime of flavin-tryptophan radical pairs15). We solved the stochastic Liouville equation to determine the triplet yield in the presence of a static magnetic field of 46 µT. We then calculated, by direct numerical integration of the stochastic Liouville equation, the change in triplet yield, ΔΦOMF, caused by an additional 1.3 MHz oscillating magnetic field in resonance with the splitting due to the 46 µT static field. For comparison, we also calculated the triplet yield change, ΔΦstatic, resulting from a decrease of 12 µT in static field, noting that such a change led to disorientation in the magnetic compass orientation responses of robins29. The intensity of the oscillating field required for ΔΦOMF to equal ΔΦstatic is 0.033 µT, that is, less than any of the intensities employed in our experiments.