European common frogs determine migratory direction by inclination magnetic compass and show diurnal variation in orientation

During migration, animals need a compass to choose and maintain correct direction; to date, we know of celestial (solar, stellar and polarized light) and magnetic compasses (Adler, 1982; Chernetsov, 2017; Mouritsen, 2018). For humans, the magnetic compass is the most mysterious, as we do not possess it naturally; at least, we have no conscious perception of the magnetic field (Chernetsov et al., 2021). Nevertheless, the magnetic compass has been demonstrated in animals by numerous behavioral experiments (Naisbett-Jones and Lohmann, 2022; Phillips and Diego-Rasilla, 2022; Wiltschko and Wiltschko, 2005, 2019). Animals can use two variants of the magnetic compass: the ‘polar compass’, which is based on the polarity of the horizontal component of Earth's magnetic field (i.e. north versus south), and the ‘inclination compass’, which indicates polewards or equatorwards and based on the direction with the smallest angle between the total field vector and the gravity vector (Wiltschko and Wiltschko, 2005). Among vertebrates, the polar compass has been shown in the sockeye salmon, Oncorhynchus nerka (Quinn et al., 1981), and mole rats, Cryptomys spp. (Marhold et al., 1997). A ‘fixed direction’ response was detected in the European robin, Erithacus rubecula, in total darkness, and this direction also depended exclusively on the direction of the horizontal field component, but neither coincided with the direction of migration or changed between seasons (Stapput et al., 2008). The ecological role of the darkness-elicited polar compass of birds remains enigmatic. The inclination compass has been reliably documented for different bird species (Beason, 1989; Bojarinova et al., 2023; Schwarze et al., 2016; Wiltschko and Wiltschko, 1972; Wiltschko et al., 1993) and loggerhead sea turtles, Caretta caretta (Light et al., 1993). Amphibians are a very interesting group in this regard. In experiments with the eastern red-spotted newt, Notophthalmus viridescens, in the case of Y-axis orientation (perpendicular to the shore, explained further below), the newts reversed their choice of direction in the circular arena following vertical component inversion, i.e. they used the inclination compass. However, the newts stimulated to breeding activity showed homeward orientation and responded only to the rotation of the horizontal component (polar compass) (Phillips, 1986). Thus, amphibians (at least, urodeles) appear to possess both compass variants, and such diversity is very interesting, considering the basal position of the group relative to all terrestrial vertebrates. However, these experiments have not been replicated, and no similar studies have been conducted on anurans.

Compass orientation could be used by anurans in three scenarios. (1) During homing (i.e. the process of navigating home), if we assume that amphibians orient according to the map-and-compass principle formulated by Kramer (1957). However, in order to reliably do so, frogs need to travel a distance much greater than 1 m (Shakhparonov and Ogurtsov, 2008), which creates difficulty in the confined space of laboratory experiments. (2) When a frog needs to quickly choose a direction in a body of water (e.g. escape to a deeper part) and during emergence onto land and dispersion of the metamorphs. For these cases, Ferguson and Landreth (1966) introduced the concept of a Y-axis: a learned direction which crosses a shoreline (the X-axis) at right angles. These authors emphasize that the Y-axis orientation is not related to homing, but is a stereotypic behavior when animals move along a previously learned axis, which, in nature, normally leads them to a desired location (Adler, 1970, 1980; Ferguson, 1971). (3) During seasonal migrations. Adult amphibians of the temperate climate migrate to the breeding ponds in spring, then to summer habitats, and to the hibernation sites in autumn (Sinsch, 1990b; Wells, 2007). The presence of compass orientation during the migrations has been shown in field experiments (i.e. could actually be used by frogs in nature). For example, in autumn, adult marsh frogs, Pelophylax ridibundus, moved to their hibernation site (a river) after being displaced by 450 m. At a distance of 1 km, the frogs moved in a stereotypic direction parallel to their natural autumn migration (Shakhparonov, 2012; Shakhparonov and Ogurtsov, 2008). In adult common frogs, Rana temporaria, behavior similar to compass heading was observed in experiments conducted 140 km from the hibernation site (Pasanen et al., 1994), although the authors did not interpret their results this way. Juvenile common frogs can use compass heading during the first wintering migration (Shakhparonov et al., 2022). In general, orientation towards the direction of migration is very similar to orientation along the Y-axis, and in some cases obviously represents the same process.

Orientation by the magnetic field in anurans has been tested for all three cases outlined above. (1) Attachment of magnets to the head of toads Rhinella spinulosa, Bufo bufo and Epidalea calamita in homing experiments performed during the breeding season influenced the initial orientation (Sinsch, 1987, 1988, 1990a); however, the effect of a permanent magnet is too non-specific to be used for identification of the compass type. Experiments on Bufo japonicus found no effect of magnets (Ishii et al., 1995). (2) Numerous studies demonstrate a wavelength-dependent effect of light on magnetic compass orientation in tadpoles of Alytes obstricans, Pelophylax perezi, R. temporaria and Lithobates catesbeianus (Diego-Rasilla and Luengo, 2020; Diego-Rasilla et al., 2010, 2013; Freake and Phillips, 2005). This is important evidence in favor of the inclination compass as, in theory, the light-dependent chemical compass is inclinational (Ritz et al., 2000), but not crucial proof, as only the horizontal component was varied in these experiments. (3) Experiments using a circular arena during spawning migration of B. bufo and experiments using a T-maze parallel to the migration axis for P. ridibundus stimulated to spawn. In both cases, the amphibians responded to rotation of the horizontal component; the vertical component was not altered, and the compass type could not be determined (Landler and Gollmann, 2011; Shakhparonov and Ogurtsov, 2017). Therefore, experiments that would allow for direct identification of the type of compass in anurans are warranted. To achieve this goal, it is first necessary to reliably control the parameters of a magnetic field and other possible orientation cues (odor, chorus sounds and familiar visual cues), which is possible using a coil system under laboratory conditions. Second, in order to correctly interpret the results obtained, we need to understand similar behaviors of frogs in nature and the underlying motivation.

To properly understand the motivation, we must consider not only annual cycles but also the diurnal periodicity of certain processes (Adler, 1970). Moreover, the diurnal rhythm can vary by season, from night-active in summer, when days are hot and dry (Meek and Jolley, 2006; Sinsch, 1984; Terentiev, 1938; Wells, 2007), to day-active in autumn, when nights are cold (Bannikov, 1940; Koskela and Pasanen, 1974; Shakhparonov et al., 2022). In a number of cases, amphibians were proven to have an endogenous rhythm, potentially used to predict cyclic environmental changes (Adler, 1970, 1976; Jianguo et al., 2011; Kumar, 2017; Oishi et al., 2004). Consequently, motivation and orientation direction can change throughout the day. Therefore, laboratory experiments require a good understanding of the behavior of a given anuran species in nature and both its seasonal and diurnal characteristics.

The main goal of this study was to identify the type of magnetic compass used by adult frogs, whether polar or inclinational. We chose the European common frog (Rana temporaria Linnaeus 1758) as a model species: its biology is well studied, and the species itself is abundant enough to allow large-scale experiments. To resolve the problem posed, we studied the behavior of frogs during the breeding season, when in the absence of familiar cues they were motivated to choose a compass direction parallel to their breeding migration. To test whether European common frogs exhibit this behavior in nature, to identify the direction of the migration axis, and to test possible diurnal dynamics of orientation behavior, we first conducted field experiments in a large circular outdoor arena. The experiments aimed at identifying the compass type were performed under controlled laboratory conditions with changed horizontal and vertical components of the magnetic field.

The experiments were conducted at the Zvenigorod Biological Station of Lomonosov Moscow State University (Moscow Oblast, Russia; 55.700, 36.723) in 2019–2023 on the population of European common frog inhabiting the right bank of the Moskva River. Frogs leave the river (wintering site) in late March–early April and migrate south-southeast to a breeding pond. All studies were conducted on the frogs breeding in a single pond (dimensions approximately 70×30 m) located at the edge of the forest, 30 m from the river (Fig. 1; Fig. S1). The forest is located on the southern side of the pond, and there is a floodplain meadow on the northern side. As females are full of eggs, and the specimens are subjected to a great number of manipulations during laboratory experiments, we focused our study on males.

The aim of this series of experiments was to determine whether European common frogs exhibit compass orientation during spring migration in an unfamiliar area and whether there are diurnal dynamics to this behavior.

To determine the direction chosen by the frogs, we used a circular arena 20 m in diameter, enclosed by a translucent polyethylene wall. A 15 cm wide and 10 cm deep groove was dug around the inside perimeter of the arena. In the groove, 16 cylinder traps were buried at 22.5 deg intervals (for details, see Shakhparonov et al., 2022). The area of the arena construction had no constant slope towards any of the cardinal points (confirmed using a 3 m long board and a Bosh GLM 80 incline sensor) and had no local iron deposits that could distort the magnetic field [confirmed using HB0204.62A magnetometer, Magnitnie pribory, St Petersburg, Russia; the spatial variation of total intensity (F) is presented in Fig. S1A]. The arena was built in a meadow between the river and the forest, 950 m west-southwest (254 deg) from the breeding site. The distance from the center of the arena northwest to the river was 60 m and southeast to the edge of the forest was 40 m; individual trees closest to the arena were located northeast.

The frogs were transported in a closed opaque container to prevent the use of visual and odor cues. At the site of release, the container was rotated in different directions for approximately 1 min to disrupt kinesthetic orientation, described for amphibians by Endler (1970). Then, the frogs were placed in the center of the arena in a plastic basin (40 cm in diameter with 8 cm high walls), covered with the container, and allowed to calm down for a couple of minutes. After that, two researchers simultaneously removed the container and left in opposite directions, northward and southward. The frogs were released in groups of 8–33 animals. In series of studies conducted by other authors in outdoor arenas with groups of anurans, the animals demonstrated their independent choice of direction (Ferguson et al., 1965, 1968; Gorman and Ferguson, 1970; Landreth and Ferguson, 1967a). To monitor the weather conditions, we used the Vantage Vue weather station (Davis Instruments).

The frogs were checked 1 h after release (or 2 h, if the temperature at the time was low) and collected separately from each cone. Frogs that did not reach the groove were collected separately. Then, body length of each frog was measured with a ruler (1 mm accuracy), weighed on Massa k model bk 600.1 scales with 0.02 g accuracy, and photographed from the dorsal and ventral side for individual recognition and to prevent reuse. Each frog was tested only once. The average length of males in our experiments was 76.48 mm (from 58 to 96 mm) and average mass was 43.98 g (from 13.32 to 74.24 g). Analysis of movement only included individuals that reached the marginal groove.

The frogs were captured when they started to migrate from the river to the breeding pond. The migration usually began at approximately 12:00 h (the frogs emerged earlier on warm days, at approximately 09:30 h), and prior to that, the frogs accumulated in the water near the bank. We collected only individuals that crawled out of the river or were moving towards the pond, but not those that reached the breeding site. Frogs were collected only on a small section of the river and its bank (approximately 45 m long) directly opposite the spawning pond. After that, the frogs were transferred to the arena and tested. The experiments were conducted throughout the day from 10:30 h to 00:12 h, trying to cover the intervals of morning, daytime from noon to sunset, evening twilight (hereafter ‘twilight’), and night. These intervals were selected on the basis of their putative biological significance and available cues. We expected the frogs to exhibit different behaviors in the morning before 12:00 h compared with the afternoon, when migration usually starts. In the twilight, the frogs cannot see the sun and can only use polarized light; and it is suggested to be a very important part of the day for orientation using the magnetic field, when the latter is stable and homing occurs (Diego-Rasilla and Phillips, 2021). At night, instead of the no longer available polarized light, frogs might use the star compass or again the reasonably stable magnetic field.

As frogs are difficult to find in the morning, one group of animals for the morning experiments was collected during the day and released the next morning at 11:00 h. In total, we conducted 17 experiments with 336 male frogs. We performed eight experiments over the interval from noon to sunset, as it was the longest by far and frog behavior could change within this period.

To test whether the behavior of frogs changed after reaching the breeding pond, we conducted eight experiments with 245 males that reached the pond and were collected among the laid eggs. The experiments were conducted from 10:32 h to 00:04 h.

The experimental setup consisted of two mirrored T-shaped mazes placed in an artificial magnetic field with maze arms parallel to the axis of migration in the natural environment. A similar technique was used in the study by Shakhparonov and Ogurtsov (2017) on the marsh frog.

The maze was constructed from pipes with a diameter of 10 cm. Arms of the maze were connected with receiving cylinders (drop traps) of 55 cm depth (so the frogs could not climb out). The total length of the maze arms was 111 cm (distance between the openings of the receiving cylinders) (Fig. 2). We did not use a circular arena here, as it is a relatively wide empty space with well-defined borders. It has been shown in a number of studies that open space causes anxiety in animals, motivating them to hide near the wall (Eilam, 2003; Lebedev et al., 2012). Well-defined borders in an open-top apparatus can also stimulate frogs to escape by jumping. As a result, anurans tend to show escape behavior that interferes with realization of the orientation tests (Landler and Gollmann, 2011). Tubes provide a completely closed ‘cozy’ space to eliminate anxiety in the frogs, which is well established in our previous work (Shakhparonov and Ogurtsov, 2017). In contrast to our previous methodology (Shakhparonov and Ogurtsov, 2017), we added a longer start chamber (25 cm) that was level with the maze. This modification is due to the fact that European common frogs are more frightened when they enter the maze arms from above, and we have not observed them hiding in the internal corners of the maze. To prevent the frogs from escaping through the inlet compartment, it was equipped with a check valve. The labyrinth was installed on a table without metal parts. We controlled the horizontal position of the mazes using a Bosch GLM 80 digital inclinometer. The maze was positioned symmetrically in relation to sound sources, such as the power supply (50 dB noise level), and to light sources. The light sources included two LED lights (400–720 nm spectrum) and a full-spectrum LED lamp minifermer sunlike 4000K+660 nm (410–795 nm spectrum); the optical spectra were measured using a fiber optic spectrometer Ocean optics USB4000-UV-VIS. The light level outside the maze was 622 lx; inside in the center of the maze, it was 61.2 lx; at the edges of the maze arms, it was 10 lx (measured using Sonel LXP-10A). The temperature in the laboratory was 18–19°C, and the relative humidity was 34–55%.

The magnetic field was generated using the three axis Merritt three square coil system (Merritt et al., 1983) (Fig. 2A). The internal length of the side of the X, Y and Z coils was 200 cm, 191 cm and 182 cm, respectively. The coils were manufactured at the Faculty of Biology, Lomonosov Moscow State University, and the control unit and software were created by Oleg Korol ([email protected]). The field parameters were monitored with a Bartington Mag658 digital three-axis fluxgate magnetometer placed in the center of the system. The field could be set with an accuracy of 10 nT for each coordinate, and the system effectively stabilizes the field in a range of 5 nT from the set value (stabilization efficiency shown in Fig. S1B). The coils were leveled with the horizon and the vertical using calibrated level Bosh GLM 80+R60 Professional (grade error of ±0.1 deg). Thus, setting of the field inclination in our experiments was accurate to ±0.1 deg. The system simulated the magnetic field in the frog habitat area. Field measurements were taken directly in the laboratory prior to the start of the series of experiments. We used two laboratories. Lab 1 (2020) was located 1250 m in a 241 deg direction from the ponds, and Lab 2 (2021–2023) was 1180 m in a 249 deg direction. As the distance was fairly short, the differences in field parameters between the laboratories and the capture site were not significant (Fig. S1A).

The field parameters differed slightly between years of experiments (the measurements are given with the orientation of the X-axis to the magnetic pole). In 2020, inclination (I)=+72.89 deg, total intensity (F)=51.72; in 2021, I=+70.34 deg, F=52.44; in 2022, I=+70.61 deg, F=52.17; in 2023 I=+70.48 deg, F=52.24. The coil system and the T-maze were positioned in such a way that declination of the artificial field (hereafter ‘original’ magnetic field) was +11 deg in Lab 1 (2020) and +135 deg in Lab 2 (2021–2023) instead of +11 deg in the natural one.

The transfer from one laboratory to another was for purely practical reasons. Lab 1 was used while the adjacent rooms were not used by other research groups, and therefore their power supply could be switched off to lessen interference (however, this was difficult to achieve from 2021). Lab 2 was constructed in 2021 with our specifications: all wiring was at a distance from the magnetic coils, with no active electrical appliances in an 8.5 m radius (with the exception of lights and the power unit for the coils). The lights had common grounding (3.08 Ω resistivity), while frames of the power block, control console and coils had a separate laboratory grounding (0.67 Ω resistivity). Both laboratories were constructed of wood and had no massive iron deposits in their vicinity. Spatial variation of the magnetic field within laboratories is shown in Fig. S1A. The background magnetic field within the coil system and the stabilization we achieved for the working system are shown in Fig. S1B. The laboratories had no dedicated shielding; however, the background level of electromagnetic noise in the range 0.009–150 MHz was low (Fig. S1C,D, measured using Rigol DSA 815 tg spectrum analyzer and two active circular antennas manufactured and calibrated by Roman Cherbunin at St Petersburg State University; calibration was verified in the Russian Center for Testing and Certification (Rostest-Moskva).

Before release, the frogs were placed in a cloth bag and rotated in different directions to disrupt kinesthetic orientation. Frogs were individually released into the start chamber of the T-maze and tested only once. The test time was 20 min. The response was scored if the frog was in either one of the drop traps. If the frog did not drop into a trap, then at the end of the experiment the position of such individuals was determined visually by slowly opening both ends of the maze. If the frog was found at the very edge of the T-maze arm, its choice was also scored. Frogs that stayed in the center of the apparatus or moved half-way into the maze arms were considered ‘unresponsive’. We tested two frogs simultaneously in mirrored mazes to minimize the effect of possible right- or left-turn preference, described in some studies on amphibians (Adler, 1980). To remove odors and traces after each test, the maze was rinsed with water (in its entirety regardless of where the frog went), and remained wet, because preliminary experiments showed that European common frogs move much more actively in a wet maze than in a dry one.

We used a blind control procedure. The field settings were coded with numbers, the decoding of which was not communicated to the people who placed the frogs in the maze and scored their choices. The choice was recorded from the absolute number of drop traps, which did not change and was tied to the geographical position of the maze arms.

As in the field experiments, we photographed each frog and measured its body length and mass. The average length was 78.6 mm (61–94 mm); the average weight was 46.88 g (21.28–80.26 g).

Frogs were caught both on their way to the breeding pond (N=40) and in the breeding pond among the eggs (N=296), from 09:30 h to 20:00 h. In our sample set, the majority of frogs were captured in the breeding pond, where they could be reliably caught at any time of the day and had a more stable motivation to travel towards the spawning pond, according to the results of field experiments. Frogs caught in the breeding pond were brought to the laboratory in pond water and kept in the same water until testing. Tests were conducted throughout the day from 10:09 h to 23:40 h. Experiments were performed in 2020–2023.

Based on observations in nature and field experiments (see also the results of part 1, field experiments), we tested the direction of frog movements: southeast (towards the ponds) or northwest (towards the river). For this purpose, we used four field settings: (1) original magnetic field; (2) reversed horizontal component: declination (D)+180 deg; (3) inverted vertical component [for example if original magnetic field inclination (I)=+70.48 deg, then inverted vertical component I=−70.48 deg]; and (4) reversed horizontal and inverted vertical components (Fig. 2C). We tested 84 frogs in each case (a total of 336 individuals).

These four settings were identical to those used by Wolfgang and Roswitha Wiltschko in their classic 1972 study in birds (Wiltschko and Wiltschko, 1972) and they allowed us to identify the type of compass used by the frogs: polar or inclination. If the compass is polar, then frogs should follow only the direction of the horizontal component, like a magnetic needle of a technical compass. If the compass is based on inclination, then the angle formed by the magnetic field lines and gravity is important (an example and explanation are given in Fig. 3). As under field settings 3 and 4, the two compass types would produce opposite results, we obtained the answer by combining the results of experiments based on these two models. The model that shows a statistically significant asymmetry in the choice of maze arms is the one that describes the mechanism used by the frogs. If the distribution across the maze arms is uniform both when combined according to the polar model and according to the inclination compass model, then either the animals do not use the magnetic field or they lack a clear motivation to choose only one direction.

To test for the absence of other cues and of potential uncontrolled stimuli affecting the symmetrical conditions of the maze, we performed control experiments in a vertical magnetic field (I=+90 deg), which provides no directional information. The vertical magnetic field strength was equal to the directional field strength in the same year. The deviation of the field from the center to the ends of the maze was less than 0.4 deg. A total of 114 frogs were tested.

As a number of researchers note that naturally occurring fluctuations in the Earth's magnetic field affect animal orientation (Keeton et al., 1974; Kowalski et al., 1988; Moore, 1977), we assessed their magnitude and influence during our research. We used the K-index to assess the state of the magnetic field at the time of the experiment; K₁₂ to assess the state of the magnetic field over the interval that also includes 9 h before that point; and the A_p-index to assess the general state of the magnetosphere on the day of the experiment. The data were obtained from the ‘Moskva’ magnetic observatory (Pushkov Institute of Terrestrial Magnetism, Ionosphere and Radio Wave Propagation of the Russian Academy of Sciences, Troitsk, Moscow) located 40 km southeast from the experimental site (http://serv.izmiran.ru/out/KindMOS/Kind-mos.html). Details on these indices are given in Supplementary Materials and Methods. The values obtained during field experiments are presented in Table S1, and those during laboratory experiments in Table S4. The values in nT that correspond to the K-index integers are presented in Table S1 legend, according to Shevnin and co-authors (2011).

The research was conducted in accordance with the laws of the Russian Federation and the requirements of the Committee for BioEthics of Lomonosov Moscow State University (GOST 33219-2014). The ‘Guidelines for accommodation and care of animals’ of the European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes (ETS No. 123) and the ‘Guidelines for the treatment of animals in behavioural research and teaching’ by the ASAB Ethical Committee/ABS Animal Care Committee (2012) were also followed.

To analyze the distribution of frogs in the arena, we used Rayleigh's uniformity test (Batschelet, 1981; Humphreys and Ruxton, 2017). In cases where we assumed a bimodal distribution, we used the angle doubling procedure (Batschelet, 1981; Landler et al., 2018). To analyze the preferred direction (if there was any), we used the mean vector with 95% confidence interval (CI) (Batschelet, 1981; Mardia and Jupp, 2000). The χ² test was used as a two-sample test (Batschelet, 1981). To ensure that most or all expected frequencies used in the calculation are greater than 5 when using the χ² test, we had to decrease the number of sectors to eight or four by merging them in pairs (N+NNE; NE+ENE; etc.) or tetrads (N+NNE+NE+ENE; E+ESE+SE+SSE; etc.). To counteract the multiple comparisons problem (25 testable hypotheses), we used the Holm–Bonferroni method (Holm, 1979). After these adjustments, α=0.0045. To understand what caused the differences, we compared the 95% CI, length of the mean vector, and the circular standard deviation of the datasets (Batschelet, 1981). If the 95% CIs of the two samples overlapped, the differences were due to scatter of the data (Batschelet, 1981). If the 95% CIs did not overlap, then the differences were in the direction of movement under a comparable vector length. The calculations were performed in Oriana 4.0 (1994–2011 Kovach Computing Services). Spearman rank order correlation was used to determine the effect of weather conditions and to study the relationship between the body length and mass of frogs and the angular deviation from the migratory direction in circular arena.

To compare the number of frogs that moved in two alternative directions in the T-maze, we used the χ² test. ‘Unresponsive’ individuals were not included in the analysis, but the number of such animals was clearly indicated. The same data were analyzed in three ways: (1) according to the magnetic axis of migration; (2) according to left–right turn preference; and (3) according to any uncontrolled stimuli associated with the position of the maze in a room (‘topographic’ bearings: geographic south or geographic north in Lab 1 and geographic west or geographic east in Lab 2). As each data set was subject to four different tests, we applied simple Bonferroni correction and lowered the alpha level to 0.0125. To compare the differences in the number of ‘unresponsive’ individuals between the experiment and control groups, we used the two proportions Z-test. To study the effect of various factors on the distribution of frogs in the maze, we used a GLM with binomial error structure. The constructed model was then tested using a likelihood ratio test (drop1 function in R) to identify statistically insignificant factors; the model was then reduced. This operation was repeated until the model contained only significant predictors (Faraway, 2016). The calculations were performed in R 4.3.1 (http://www.R-project.org/).

To test the relationship between body length and mass and the choice in the T-maze, we used one-way MANOVA. The Kolmogorov–Smirnov test was used to test for normality, and the Levene test was used to test for equality of variance (P>>0.05 in both cases).

The calculations were performed in Statistica 8.0 (Statsoft Inc. 1984–2007).

The raw data are given in Table S1.

Male frogs captured on the migration in the morning were oriented ESE, so parallel to the migratory direction (Fig. 4A), except the frogs held in captivity overnight, which had a uniform distribution (Table S1). In the experiments performed during the rest of the day before sunset, the frogs unexpectedly showed heterogeneous behavior. In the early afternoon from 12:00 h to 16:35 h, the frogs oriented SE in the migratory direction (Fig. 4B), but in the late afternoon and early evening (between 16:35 h and 19:22 h) approximately half of the individuals went NW, towards the Moskva River, resulting in a bimodal distribution (Fig. 4C). At twilight, in one experiment the frogs oriented predominantly N towards the river, and ESE in another experiment held in rainy weather; the combination of these sample sets resulted in a bimodal distribution, similar to the previous interval (Fig. 4D). At night, most of frogs oriented north, towards the river (Fig. 4E).

It is worth noting that in releases made after 16:35 h (late afternoon and early evening, twilight, and night), the majority of frogs oriented similar to the migratory direction in rainy weather. When we combined these three samples (Table S1), we obtained the following distribution: N=48, Rayleigh test statistic Z=18.99, P<0.001, length of mean vector r=0.63, circular standard deviation (CSD)=55.2 deg, mean vector µ=101 deg, 95% CI=85–117 deg. The combination of the other seven experiments conducted in the rainless weather showed a general direction towards the river: N=110, Z=20.1, P<0.001, r=0.43, CSD=74.7 deg, µ=347 deg, 95% CI=331–5 deg. The differences were statistically significant (χ²₇=45.51; P<0.001).

As larger individuals may be more experienced as a result of their age or less affected by a decrease in temperature because of their greater mass, we investigated the relationship between the length and mass of frogs and the angular deviation from the migratory direction (135 deg). For this purpose, we chose the experiments conducted between 16:35 h and sunset, as their distribution remained bimodal even after the exclusion of the experiments performed in the rain (N=43, Z=10.96, P<0.001, Rayleigh test applied after doubling the angle), i.e. there was a great variety of chosen directions. The analysis showed no statistically significant correlation (N=43, R=0.024, P=0.88 for length and R=0.058, P=0.71 for mass).

Male frogs collected in the breeding pond in the morning and early afternoon were oriented SE and SSE, the same as the frogs collected during migration (Fig. 4F,G). Frogs released in the late afternoon and early evening, unlike the males caught during migration, generally oriented south (Fig. 4H). Interestingly, the rain that occurred during one of the releases had a slightly different effect on these frogs. They were actively moving around the arena, but less than half reached the groove and most of them oriented towards the northeastern part of the arena (Table S1). Frogs captured in the breeding pond and released at twilight oriented NE, towards the trees nearest to the arena (Fig. 4I). At night, frogs oriented NNE with the 95% CI of the mean vector covering a wide range of directions: to the river, to the nearest trees, and even to the capture site (Fig. J).

As ectotherms are highly sensitive to temperature changes, we measured the temperature both during the experiments and over the entire breeding season from 2019 to 2023. Air temperature during the experiments ranged from 2.2 to 17.4°C and overlapped greatly between the allotted time intervals (Fig. 5A). However, the temperature increased from the beginning to the end of the experiments performed before 16:35 h, while at a later time of the day, it decreased. Relative humidity (RH) ranged from 29% to 96%, and changed synchronously with temperature, as expected (Fig. 5C) [RH difference (ΔRH) highly correlated with temperature difference (ΔT): N=25, R=−0.917, P<0.001].

We tested for correlation of parameters characterizing frog orientation [r, c.s.d. and value of deviation of mean vector (µ) from the migratory direction (135 deg)] with weather conditions during the experiment [minimum and maximum temperature, minimum and maximum RH, minimum and maximum atmospheric pressure (AP), and ΔT, ΔRH, ΔAP] in experiments where distribution of the frogs was unimodal (N=21). We only found a strong negative correlation of the deviation of µ from migratory direction with ΔT (R=−0.872, P<0.001) and ΔRH (R=0.751, P<0.001), as the latter is correlated with ΔT. In other cases, R ranged from −0.33 to 0.34; P>0.12. It is also worth noting that the frogs were not disoriented under overcast skies. The effect of the wind could not be tested because of the generally calm weather. There were no magnetic storms during the experimental days (Ap-index ≤29, K-index ≤4, K₁₂ ≤12; Table S1); however, because of the strong dependence of the orientation parameters of frogs on the time of day and temperature, a simple correlation analysis was not applicable here, and the sample size was not sufficient for analysis of variance.

Regarding the daily temperature variation over the spawning season (Fig. 5B), we observed the following: after sunrise until 10:00 h the temperature increase accelerated, then slowed, and the average temperature started to lower between 16:00 h and 16:30 h (as clearly seen in Fig. 5D), with the most rapid decrease occurring during sunset and civil twilight. At dawn twilight, night and in the early morning hours, there may be frosts below zero. Fig. 5B shows that the confidence interval of average temperature crosses the zero mark. The minimum temperature during our experiments was −10.2°С on 1 April 2020 from 05:00 h to 05:30 h. The duration of such frosts can be short, 1–2 h, or long, 9–12 h. In 2020, frosts lasted for 2 days, interrupting the spawning, and the breeding pond was covered with ice.

In the experimental group, 45 of 336 frogs chose neither direction; the distribution of the remaining frogs was random relative to the direction of magnetic field when analyzed both according to the inclination compass model (153 frogs oriented towards the breeding pond and 138 towards the river, χ²=0.77; P=0.38) and according to the polar compass model (155 towards the river and 136 towards the pond, χ²=1.24; P=0.27).

However, in field experiments, we found that male frogs most reliably oriented towards the breeding pond in the afternoon if the temperature was rising. Analysis of temperature variation during the day showed that the temperature began to drop after 16:00 h. Therefore, we divided the experiments conducted into several time intervals: morning until 12:00 h, early afternoon until 16:00 h, late afternoon and early evening from 16:00 h until sunset (19:20 h), and after sunset (twilight and night were combined, as in the field experiments, frog behavior did not differ between these periods). The distribution of frogs between the maze arms in relation to the direction of the horizontal component of the magnetic field (polar compass) did not differ from random during all four time intervals (Fig. 6B, Table 1). However, according to the inclination compass model, in the morning, the distribution did not differ from random; from 12:00 h to 16:00 h, significantly more frogs moved towards the breeding pond; and after 16:00 h, the distribution again became uniform (Fig. 6A). The fact that the frogs made a choice based on the magnetic field is confirmed by the random distribution relative to the two ‘topographic’ directions of the experimental apparatus in all cases (Table 1).

Because in our experiments there were differences in the place of capture (during migration or in the pond) and the place of testing (Lab 1 or Lab 2) of the frogs, and also in the degree of geomagnetic field disturbance (Ap-index [8, 45], K-index [0, 4], in one experiment 5, K₁₂ [3, 15]; Table S4), we investigated their influence (both independently and taking into account the time of day) on the distribution of frogs in the T-maze according to the inclination compass model. Binomial GLM showed that all these factors were not statistically significant (P>0.05 in all cases; Table S5), and the result depended only on the time of day (P<0.001), as previously found.

No significant differences were found in the length and mass of frogs that oriented using the magnetic field (inclination compass model) in the direction of spawning migration, compared with the frogs that chose the opposite direction and with the unresponsive individuals between 12:00 h and 16:00 h (one-way MANOVA, F_4,232=0.167, P=0.95; Table S3). For other time intervals, when frogs made choices at random (combined data obtained in the morning and after 16:00 h), no differences were found either (one-way MANOVA, F_4,232=1.398, P=0.23; Table S3).

In the control group (vertical magnetic field), the distribution of frogs between the maze arms was random (Table 2). However, 28 of 114 frogs did not make a choice, which is a statistically significantly higher portion than in the experimental group (P=0.005, two proportions Z-test). Binomial GLM also showed that the type of magnetic field (vertical or directional) is the only statistically significant predictor of the number of active individuals, while time of day, location of capture, laboratory, or K, K₁₂ and A_p indexes were not (Table S5).

There was no preference for turning right or left in either the control or the experimental group (Tables 1 and 2).

We found that during the breeding period, motivation of male frogs depended on the time of day, as indicated by differences in orientation within the circular arena. Frogs caught on their way to the pond oriented southeast in the migratory direction in the morning and early afternoon experiments; in the experiments conducted from 16:45 h until sunset (when air temperature decreases rapidly), the distribution became bimodal, and at twilight and on nights without rain, most frogs oriented towards the river. Frogs captured in the breeding pond were more stable in choosing the southern direction in the late afternoon and early evening, and chose a general NE direction at twilight and at night.

During spawning, European common frogs exhibit homing from a distance of at least 100 m (Elmberg and Lundberg, 1988) and prefer the odor of the breeding pond water in T-maze tests (Lukianov et al., 1985). However, only in the night experiments with frogs captured in the breeding pond did the 95% CI of the mean vector cross the home direction (ENE), most likely because of a wide scatter, as only 3 frogs out of 50 oriented directly ENE. Thus, we can conclude that at 950 m distance, the frogs probably could not reliably locate their breeding pond. In a number of experiments, however, the frogs oriented in parallel to the migration direction, i.e. a direction that would lead them to the correct body of water in their habitat. This is a case of Y-axis orientation, often found in amphibians in the absence of familiar cues (Ferguson and Landreth, 1966; Landler and Gollmann, 2011; Landreth and Ferguson, 1967b; Shakhparonov, 2012; Shakhparonov and Ogurtsov, 2008). Interestingly, frogs in the arena experiments positioned themselves on the Y-axis north relative to the pond, whether they had been caught at the river bank or in the breeding pond. The latter group could have used the arena's location in a meadow as a cue, as the only meadow in their habitat was to the north from the pond. This explanation is consistent with Adler (1970), who wrote that frogs orienting along the Y-axis do not attempt to return home, but behave as though they are already in their native habitat.

The most interesting effect found in the frogs caught during migration, switching from orientation in the migratory direction towards the ponds to orientation towards the river, can be explained by avoidance behavior of possible night frosts, which are common in late March–early April. Frosts pose a great danger to European common frogs, which lack any specific cryoprotectors (Ludwig et al., 2015). At a temperature of −1.5 to −2.5°C, frogs begin to die after just 1 day (Pasanen and Karhapää, 1997), at a temperature of −4.5°C, most frogs die after 5 h (Berman et al., 2017). During our studies, we recorded even lower night temperatures and found dead frogs on land. Hiding in the water is a guaranteed way to avoid freezing during night frosts in spring or during winter (Ludwig et al., 2015). Another type of avoidance in frogs is, apparently, burrowing in the ground. This happened in one of the night experiments, where we could not locate five frogs after the first check, but found them the next day oriented in the migratory direction (Table S1). However, a particular temperature is not the main trigger of freeze avoidance in spring, as there was no correlation between the chosen direction and the minimum temperature. In our opinion, it is instead the change in temperature (ΔT) that has a significant role, with which we found a strong correlation. After 16:00 h, the temperature difference was negative, and the maximum rate of decrease was 0.8°C per 30 min. This means that under natural conditions, frogs on their way to a pond are at risk of experiencing a sudden temperature drop. At temperatures close to 0°C, frogs would lose mobility (Berman et al., 2017) and would no longer be able to hide. As a result, the motivation to find a breeding pond is replaced by freeze avoidance in all cases except for when it rains.

Interestingly, frogs captured in the breeding pond continued to orient southward in the late afternoon and early evening, as they apparently position themselves on the Y-axis closer to the breeding pond than frogs caught migrating near the river. After sunset, the general direction shifted to NE, where the trees closest to the arena are located. The trees are attractive because they provide a place to burrow into the rotting leaf litter during the cold. This may well reflect a tendency to switch to orientation by local cues. We did not find any relationship between the length and mass of the frogs and the avoidance behavior. However, according to laboratory experiments (Laitinen and Pasanen, 1998), the number of frogs attempting to go into the water at low temperatures decreased in the series: frogs less than 1 year old→older immature frogs→mature frogs, which can be interpreted as dependence on age and, therefore, size. Future research may clarify this.

The reason for the disorientation of frogs held in captivity overnight in the morning experiments is as yet unclear. We doubt that it could be a loss of motivation due to the captivity. For example, in other arena experiments, overnight captivity did not result in a loss of motivation (Ferguson and Landreth, 1966; Ferguson et al., 1968). In a T-maze, European common frogs began to reject the breeding pond odor after more than a day in captivity (exact time not specified) (Lukianov et al., 1985); in a similar experiment on toads, Incilius valliceps, demotivation was observed only after 2 days (Grubb, 1973). In addition, as we noted above, frogs that hid in the arena overnight did not lose their motivation to move in the migratory direction the next day. Therefore, it is more likely that the aforementioned behavior is related to unstable motivation in the morning, as we usually observed the migration of frogs in the afternoon, when the shallow river water and soil are sufficiently warm. Our laboratory experiments also agree with this (see below).

For the first migrants from the studied population, the process of finding a breeding pond without the cue of a chorus can be divided into three stages: (1) choosing the correct bank to emerge from the river, (2) choosing the correct direction on the bank, and (3) short-range orientation near the pond.

For the first stage, frogs can potentially use rheotaxis to choose the correct river bank, although the presence of rheotaxis has been studied mainly in animals with a lateral line: tadpoles or aquatic amphibians (Simmons et al., 2015). Presumably, in this case, the need to turn left when choosing the right bank could lead to a population-level turn preference in frogs, which was absent in the laboratory experiments. It is also possible that frogs use global cues: magnetic (studied in the present work) and astronomical compasses (Gorman and Ferguson, 1970; Taylor and Ferguson, 1969). The presence of the two systems is important in terms of mutual redundancy, as the solar compass is ineffective under overcast skies, and the magnetic compass is disturbed by strong magnetic storms.

For the second stage, in addition to the aforementioned global cues, the frogs may use their kinesthetic sense to maintain the chosen direction (Endler, 1970) and use some available local cues. To move up a river bank, frogs can use the surface slope. This form of orientation is known for newts and turtles (Omland, 1998; Salmon et al., 1992). When the frogs sufficiently ascend the slope, the silhouette of the forest behind the pond becomes visible, which can also be a guiding landmark. A fundamentally similar way of orientation has been described for juvenile sea turtles (Salmon et al., 1992). During this period, frogs could use olfaction to navigate (Lukianov et al., 1985; Nakazawa and Ishii, 2000; Ogurtsov, 2004; Sinsch, 2006), as well as the humidity gradient (Brändle and Lázár, 1994; Reshetnikov, 1996). However, in our opinion, the odor and humidity gradient from the breeding pond would be concealed by the river as a stronger source of both signals because of its greater surface area. Therefore, the frogs first need to exit its area of influence. Our field experiment simulated this second stage; however, orientation mechanisms based on the slope (the arena was constructed on a relatively flat surface), the kinesthetic sense (frogs were rotated before release) and the breeding pond odor (because of the long distance) were not available. Note that the frogs were not disoriented under overcast skies; thus, the magnetic compass or the silhouette of the forest compensated for the impossibility of using a celestial compass.

For the final stage associated with finding the breeding pond and the spawning site, the main cues would be local: the odor and possibly some visual landmarks.

In all our experiments in the T-maze (for the experimental and control groups, regardless of the time), distribution relative to the geographical cardinal directions was random. Therefore, with the exception of the magnetic field, all other cues were compensated for or located symmetrically in relation to the maze.

If we analyze the direction relative to the magnetic field, frogs from the experimental group chose the direction towards the breeding pond in the early afternoon from 12:00 h to 16:00 h, the same time when we observed the most stable choice of migratory direction towards ponds in field experiments. However, inversion of the vertical component caused an identical change in the chosen direction in the maze to the rotation of the horizontal component of the magnetic field. Therefore, when grouping the results according to the inclination compass model, we obtained a statistically significant preference for the direction towards the breeding pond, and when grouping according to the direction of the horizontal component only (polar compass model), the choice was random. Thus, our data directly confirm that European common frogs can use the inclination compass for orientation along the Y-axis, as does N. viridescens (Phillips, 1986). This is also in good agreement with the wavelength-dependent effect of light on magnetic compass orientation in anuran tadpoles (Diego-Rasilla and Luengo, 2020; Diego-Rasilla et al., 2010, 2013; Freake and Phillips, 2005). However, we wish to avoid unnecessary speculation regarding the physiological mechanisms of the frog compass, as our study does not address them, e.g. light dependence may not be a necessary attribute of the inclination compass, as shown in sea turtles (Lohmann and Lohmann, 1993); in blind cave salamanders, Proteus anguinus, the direction of spontaneous magnetic alignment changed in response to rotation of the magnetic field under infrared light (Schlegel, 2008; Schlegel et al., 2009) (although the type of Proteus compass is not specified, it would be light independent). The only detail we can currently confirm is the irrelevance of ultraviolet light for the inclination compass in frogs, as this part of the light spectrum was absent in our experiments.

Some authors note that natural disturbances in the Earth's magnetic field may affect orientation in animals. A positive correlation was found between the dispersion in flight directions of passerine nocturnal migrants and current K-index (Moore, 1977) or some lasting effect of prior magnetic disturbance (correlation with K₁₂) on the initial orientation of homing pigeons (Keeton et al., 1974; Kowalski et al., 1988; Larkin and Keeton, 1976). Other researchers found no such K-effects (Able, 1987; Richardson, 1974). We observed no significant effect of small disturbances on either activity or choice in frogs in the T-maze. Of course, our results do not indicate a total lack of effects of natural disturbances on the frogs' behavior: no large magnetic storms occurred, a T-maze with only two directions might not have accounted for all reactions of the frogs, and the field in experiments was stabilized. However, the lack of correlation with the indices that describe larger time intervals (K₁₂ and Ap) indicates that disturbances within 45 nT do not demotivate frogs from using magnetic compass.

In experiments conducted in the morning, late afternoon and early evening, and after sunset, the choice of maze arms in relation to the magnetic field was random. Unfortunately, the T-maze we used cannot differentiate between disorientation and bimodal orientation along the axis. Therefore, the uniform distribution obtained in the laboratory can be interpreted in three ways: (1) the frogs lose the ability to use the magnetic compass; (2) the frogs do not use the magnetic compass in these time intervals; or (3) there are diurnal changes in motivation, which result in either no motivation for migration at all, or mixed motivation of individuals in the experimental group, where half of them oriented towards the river and half towards the ponds (bimodal orientation), as we observed in evening experiments with migrating frogs and as described in many studies (Freake et al., 2002; Landreth and Ferguson, 1967a; Phillips, 1987; Shakhparonov et al., 2022).

Hypothesis 1 is unlikely: it is difficult to imagine that the magnetic compass is only available over narrow time intervals during the day, considering the stable conditions (magnetic field, light spectrum and luminous flux, temperature, and humidity) in our laboratory. We also know that other amphibians can orient by the magnetic field at different times of the day (Phillips and Diego-Rasilla, 2022).

Hypothesis 2 is based on the idea that in the field experiments, some frogs caught in a breeding pond and released at twilight and at night switched to the use of local cues. However, local cues are not available in the laboratory experiments, so it is unlikely that frogs would not use the magnetic field in this case. In a vertical magnetic field, more frogs remained in the release chamber, as they were unable to choose direction using the only available cue. The portion of such frogs remained approximately the same between the time intervals. Daily variation in the magnetic field or the light spectrum (if we assume a light-dependent magnetic compass) also cannot be a reason to avoid using the magnetic compass at certain times of the day. The solar quiet daily variations of the horizontal and vertical component for 50 deg N latitude are approximately 40 nT and 20 nT, respectively (Matsushita, 1967), with the greatest changes occurring from 08:00 h to 16:00 h, and smaller ones at twilight and night. Frogs in our experiments oriented exactly during the day, and they were not demotivated by irregular disturbances of similar amplitude. Although in nature, short-wavelength light grows stronger in the spectrum with decreasing solar elevation, from daylight to nautical twilight (Spitschan et al., 2016), no part of the light spectrum disappears as a result of changes in their proportion (by factors and not orders); moreover, it is suggested that short-wavelength light is necessary for the correct functioning of the magnetic compass, according to the studies on tadpoles and newts (Phillips and Diego-Rasilla, 2022).

Hypothesis 3 generally explains the results well. In the evening and at night, the mechanism of freeze avoidance is activated in frogs on land and forces them to seek shelter. As there are no external reference points in the laboratory that would help frogs to determine where they are on the Y-axis (closer to the river or ponds), neither of the two directions has a significant advantage, which makes the distribution in the maze uniform. However, while frogs in the field experiments could change their behavior by responding directly to external temperature changes or the presence of precipitation, the temperature in the laboratory experiments was constant and always higher than outdoors. Therefore, nothing could signal to the frogs that the air temperature was beginning to drop in the evening. The change in motivation must thus be supported by some internal rhythm. Otherwise, the frogs transferred to the laboratory would always demonstrate orientation towards the ponds, as this transfer to the laboratory is itself accompanied by a rise in temperature. In the morning, the frogs seem to have a generally weaker motivation – in nature, there is often no active migration occurring at this time, so the distribution is also random.

European common frogs possess an inclination magnetic compass, as do newts, birds and sea turtles, and can potentially use it during the spring migration to orient themselves along the Y-axis after leaving a body of water, to choose a correct direction at the bank. A vertical magnetic field confuses the frogs, probably because of an inability to choose direction.

The relationship between temperature and diurnal rhythm appears to be complex. Cyclical changes in temperature, in particular the presence of night frosts, led to the emergence of an internal diurnal rhythm controlling the change in motivation from migratory to freeze avoidance. This certainly allows the frogs to effectively predict environmental changes and prepare for them in advance. The study of these processes would help to shed light on the high adaptability of the European common frog, which allows it to have a wide distribution range, and to better control the behavior of these frogs in experiments.

We would like to thank Prof. Valery Gavrilov (director of the MSU Zvenigorod Biological Station) for the opportunity to conduct experiments and Prof. Alexey Katrukha for sponsoring the construction of a new laboratory. We express our deep gratitude to the late Dr Vladimir Mikhailovich Shakhparonov from the Faculty of Physics of MSU for his invaluable help in designing, manufacturing and testing the Merritt coil system, to Oleg Korol for manufacturing the control unit and testing the entire coil system, to Igor Mukhin for his help with magnetic field calculations, to Dr Roman Cherbunin for his help in measurements of the local radio-frequency magnetic fields to Dr Ivan Konyukhov for light spectra measurements, and to Prof. Eldar Rakhimberdiev for his consultation in statistics; to our colleagues Dr Sergey Ogurtsov, Alexander Golovlev and Anna Dubrovskaya for their help in winding, transporting and mounting the coil system, and Yaroslav Vyatkin for his help in circular arena construction. We are also thankful to Varvara Bogatyreva for invaluable help in translating the article and to Nikita Chernetsov and the anonymous reviewers who made very useful and important remarks.