Animal Cognition

, Volume 11, Issue 4, pp 715–718

Harbour seals (Phoca vitulina) can steer by the stars


    • Institute of BiologyUniversity of Southern Denmark
  • Nele Gläser
    • General Zoology and NeurobiologyUniversity of Bochum
  • Wolfhard Schlosser
    • Astronomical InstituteUniversity of Bochum
  • Guido Dehnhardt
    • Institute for Biosciences, Sensory and Cognitive EcologyUniversity of Rostock
Short Communication

DOI: 10.1007/s10071-008-0156-1

Cite this article as:
Mauck, B., Gläser, N., Schlosser, W. et al. Anim Cogn (2008) 11: 715. doi:10.1007/s10071-008-0156-1


Offshore orientation in marine mammals is still a mystery. For visual orientation during night-time foraging and travelling in the open seas, seals cannot rely on distant terrestrial landmarks, and thus might use celestial cues as repeatedly shown for nocturnally migrating birds. Although seals detect enough stars to probably allow for astronavigation, it was unclear whether they can orient by the night sky. The widely accepted cognitive mechanism for bird night-time orientation by celestial cues is a time-independent star compass with learned geometrical star configurations used to pinpoint north as the rotational centre of the starry sky while there is no conclusive evidence for a time-compensated star compass or true star navigation. Here, we present results for two harbour seals orienting in a custom made swimming planetarium. Both seals learned to highly accurately identify a lodestar out of a pseudo-randomly oriented, realistic projection of the northern hemisphere night sky. Providing the first evidence for star orientation capability in a marine mammal, our seals’ outstanding directional precision would allow them to steer by following lodestars of learned star courses, a celestial orientation mechanism that has been known to be used by Polynesian navigators but has not been considered for animals yet.


Offshore orientationAstronavigationMarine mammalsHarbour seals


The mechanisms of animal large- and middle-scale orientation in rather featureless environments like the open seas are still largely unknown. Fixing one’s own position and pinpointing a goal under offshore conditions, without the availability of distant terrestrial or other local landmarks, most probably require animals to truly navigate without sensory contact to the goal. This would mean to either rely on precise idiothetic information or on global information not directly related to the goal. Path integration or “dead reckoning” gets by with log and compass, but would require the capability of precise vector addition to determine the present position using continuously updated distance and direction from an earlier known position (Müller and Wehner 1988). However, dead reckoning is prone to error accumulation even in terrestrial environments and for small- and middle-scale orientation; and open ocean drift due to wind, tides and currents makes it certainly unsuitable as a sole orientation mechanism, especially for long offshore trips. Charting, on the other hand, requires a map sense to incorporate reliable and regularly available global stimuli into a cognitive map or some other appropriate, possibly grid-like, map system. In a map system, both one’s present position and the goal is displayed, and from their respective spatial relationship in the map, course and distance can be determined. The course is set and kept by means of some kind of compass sense providing the map system with directional reference such as geographic north assumed to be analogous to magnetic north. However, although existence of cognitive maps in animals has been questioned (Bennet 1996), some evidence for true navigation in animals using global stimuli represented in a two-coordinate or grid-like map comes from the earth’s magnetic field, which is suggested to be used by sea turtles and lobsters in terms of magnetic inclination and field intensity (Boles and Lohmann 2003; Lohmann and Lohmann 1996).

Otherwise, global stimuli that are potentially useful for offshore navigation are mainly celestial objects. However, animal celestial orientation by the sun or corresponding skylight polarisation pattern, the moon, or the stars, has been found to serve for determination of compass direction only, either to deduce geographical north or to recalibrate a magnetic compass (Able and Able 1990; Emlen 1967; Mouritsen and Larsen 2001; Wehner 1984; Wiltschko and Wiltschko 1978). Contrary to that, technical astronavigation yields latitude and longitude for position fixing on a grid-like geographical map; however, the longitude component became available to human seafarers only when exact chronometers were invented.

Certainly lacking a sufficiently precise internal clock and relevant astronomical chart information to use in position fixing calculations, there is no conclusive evidence for a time-compensated star compass or true star navigation in animals (Mouritsen and Larsen 2001). However, seals, and probably other marine mammals as well, are certainly able to detect enough stars to allow for some other kind of astronavigation (Mauck et al. 2005). Besides the possibility of using a time-independent star compass by pinpointing the geographical north as the azimuth of the rotational centre of the starry sky with the help of learned geometrical star configurations as has been suggested for birds (Emlen 1967; Mouritsen and Larsen 2001), marine mammals might follow singular lodestars to keep their course. Whales, seals and sea lions have been repeatedly shown to elevate themselves well above the water surface––the so-called spyhopping––to visually scan the surrounding environment for orientation. It might be during this spyhopping that marine mammals try to set their course by taking the bearing of suspicious stars shortly above the horizon. Therefore, we tested whether harbour seals can use the azimuth of a learned lodestar to determine their swimming direction.


Experiments were conducted with two male harbour seals (Phoca vitulina), Nick and Malte, performing a trained, free-choice, directional swimming task in a custom-made planetarium. The dome (5 m diameter, opaque Styrofoam/GRP sandwich construction spanning ~160° of a half sphere from ~10° above the horizon to zenith, inner surface painted white) was installed on a ring-like pontoon in the seals’ pool (Fig. 1). A black curtain hanging from the edges of the dome reduced background luminance to ~0.1 cd/m². A Zeiss KP1 projected ~6,000 stars of the northern hemisphere night sky (geographical position 51°N, 10°E; date and time 1 January 2007, 0000 hours CET, diameter of projected light points 1–10 mm for stars of stellar magnitude of 6 to −1.5, luminance of all projected light points 1.3 cd/m²); for each test trial, compass direction of the projection was changed by turning the star projector about the z-axis following pseudo-random schedules (Gellermann 1933). Eight infrared-sensitive video cameras with infrared LEDs and fish-eye lenses allowed observing and video recording the seals’ responses for offline analysis. In each trial, a seal started from the dome centre, searching and heading for a learned lodestar (Sirius, azimuth 170°25′; altitude +21°50′, diameter of projected light point 10 mm, same luminance as all other projected stars to avoid brightness orientation cues) and was rewarded for touching the inner wall of the dome with its snout at the lodestar’s respective azimuth position.
Fig. 1

Experimental setup. To enable the seal to perform a trained, free-choice, directional swimming task, a planetarium dome was built on a ring-like pontoon system. A Zeiss KP1 generated a realistic projection of ~6,000 stars of the northern hemisphere night sky


Both seals learned the basic experimental procedure with little difficulty. In an initial training phase with the experimenter inside the planetarium, we first guided the seals with a handheld laser pointer to the lodestars’ azimuth position. After that, the seals learned to indicate the azimuth position when the laser pointed directly to the lodestar. A seal’s directional response was counted as a hit when deviation from the lodestars’ azimuth was no greater than 30°, and the seals were differentially rewarded for being more precise. After abandoning the laser pointer and leaving the respective test seal alone in the planetarium, one of the seals (Malte) immediately reached the learning criterion of surpassing 80% correct choices in four subsequent sessions of 30 trials, while the other seal (Nick) reached the criterion after 11 sessions. Subsequent testing for the two seals’ precision was done in 11 (Nick) and 14 (Malte) sessions, respectively, during which both seals’ gradually increased their performance to 100% correct choices (Fig. 2). Plotting the deviation from the lodestars’ azimuth of a total of 332 (Nick) and 347 (Malte) choices against a chart of the projected starry sky showed that the directional responses of both seals clustered tightly around the azimuth position of the lodestar (Fig. 3) (Rayleigh test: Nick α = −2.22°, r = 0.944, P < 0.001, 95% confidence interval ±δ = 2.34°; Malte α = −0.66°, r = 0.991, P < 0.001, 95% confidence interval ±δ = 0.72°; V test: Nick u = 24.32, P < 0.0001; Malte: u = 26.1, P < 0.0001) (Batschelet 1981). Highly-trained response data of the last four sessions (n = 120 trials) yielded a mean deviation of α = −1.05° for Nick and of α = −0.56° for Malte, respectively (Rayleigh test: Nick r = 0.965, P < 0.001, 95% confidence interval ±δ = 2.09°; Malte r = 0.995, P < 0.001, 95% confidence interval ±δ = 0.62°; V test: Nick u = 14.95, P < 0.0001; Malte u = 15.41, P < 0.0001) (Batschelet 1981).
Fig. 2

The seals’ performance. Percentage of correct choices in the test sessions for Nick (black squares) and Malte (grey circles)
Fig. 3

The seals’ directional responses. A total of 332 directional responses of Nick (a) and 347 directional responses of Malte (b) were plotted against the projected starry sky. The trained lodestar (Sirius) is indicated by a grey circle and the correspondingly correct compass bearing is indicated by a grey line


Our results demonstrate that seals are able to reliably identify singular lodestars in the night sky and indicate the respective azimuth with an outstanding precision. This suggests that they can steer by the stars while travelling at open seas and provides the first evidence for star orientation capability in a marine mammal. Visual pattern recognition and identity concept formation capability of seals (Mauck and Dehnhardt 2005) can certainly be assumed to be sufficiently efficient to identify lodestars within salient star constellations of the night sky. Due to celestial rotation and azimuthal change in higher latitudes, lodestars can serve as salient landmarks for taking a reliable bearing only when elevation is rather small near rising or setting, and visible star constellations of the starry sky gradually change over the year. However, stars have the advantage over the sun that, although they rise each night 4 min earlier than on the previous night, they do so always at the same point on the horizon relative to a stationary observer. With the stars rising at a given point on the horizon during dusk or dawn changing gradually over the year, seals have the chance to learn the relationship between relevant terrestrial landmarks in their home range to various lodestars during these short periods of the day when both are visible. Once learned to be relevant for keeping course from a given starting point to a certain goal, seals could recognize lodestars despite rotation of the starry sky during the night in a pattern recognition process such as mental rotation (Mauck and Dehnhardt 1997). Polynesian and Micronesian navigators are known to use a type of star navigation making use of zenith stars pointing down to certain islands and of lodestars, either belonging to the so-called siderial compass or to a so-called kaveinga (the “carrier”, star route or star path) (Lewis 1970). While steering their sailing vessels from island to island, prominent stars setting or rising at the horizon are used to set and keep course, either until the respective star has disappeared behind the horizon or seems to be too high in the sky to provide a good bearing, and is, therefore, replaced by another star of the same kaveinga. When the course has to be changed for the next leg of the journey, depending on the goal and the estimated logged distance, another kaveinga with respective lodestars is chosen to be followed. Experienced navigators know various “star courses” by heart depending on the season, the weather, local currents and their sailing vessel (Gladwin 1970; Lewis 1975). Polynesian and Micronesian navigators are reported to obtain course accuracy of up to 1–3° of azimuth, although between 5 and 10° are assumed to be the safe limit of accuracy for voyages to expanded target such as archipelagos (Lewis 1970). Thus, our seals’ course accuracy would be clearly sufficient to reliably reach foraging areas or hauling out sites, respectively, orienting by the night sky.

The cognitive mechanism of storing a sequence of landmarks encountered while moving from one place to another has been called steeple-chasing. If the landmarks are sufficiently close together––or visible from a distance as in the case of stars––the animal can make direct sensory contact from one to the next, so that no calculations are required. This kind of pilotage and other forms of route learning have been suggested to be employed by animals as diverse as birds (e.g. Biro et al. 2004), insects (e.g. Collett et al. 2003) and horses (Nicol 2002).

We suggest that marine mammals might learn to identify lodestars in the pattern of the night sky and to use these lodestars as distant landmarks or even complete kaveingas (star routes) to steer by in the open seas. This might be at least one possible mechanism of offshore orientation until an expanded target like a coastal region is reached and goal-related terrestrial orientation mechanisms can correct their swimming direction. However, field work would be needed to determine whether this orientation mechanism is in fact used by seals in the wild.


This work was supported by a grant of VolkswagenStiftung to G. D. We thank the Schulmuseum Leipzig for kindly providing the star projector. The experiments comply with the German animal protection legislation.

Copyright information

© Springer-Verlag 2008