Journal of Comparative Physiology A

, Volume 197, Issue 2, pp 141–151

Hydrodynamic trail following in a California sea lion (Zalophus californianus)


  • Nele Gläser
    • Institute for BiosciencesRostock University
    • Marine Science Center
  • Sven Wieskotten
    • Institute for BiosciencesRostock University
    • Marine Science Center
  • Christian Otter
    • Institute for BiosciencesRostock University
    • Marine Science Center
  • Guido Dehnhardt
    • Institute for BiosciencesRostock University
    • Marine Science Center
    • Institute for BiosciencesRostock University
    • Marine Science Center
Original Paper

DOI: 10.1007/s00359-010-0594-5

Cite this article as:
Gläser, N., Wieskotten, S., Otter, C. et al. J Comp Physiol A (2011) 197: 141. doi:10.1007/s00359-010-0594-5


The mystacial vibrissae of pinnipeds constitute a sensory system for active touch and detection of hydrodynamic events. Harbour seals (Phoca vitulina) and California sea lions (Zalophus californianus) can both detect hydrodynamic stimuli caused by a small sphere vibrating in the water (hydrodynamic dipole stimuli). Hydrodynamic trail following has only been shown in harbour seals. Hydrodynamical and biomechanical studies of single vibrissae of the two species showed that the specialized undulated structure of harbour seal vibrissae, as opposed to the smooth structure of sea lion vibrissae, suppresses self-generated noise in the actively moving animal. Here we tested whether also sea lions were able to perform hydrodynamic trail following in spite of their non-specialized hair structure. Hydrodynamic trails were generated by a remote-controlled miniature submarine. Linear trails could be followed with high accuracy, comparable to the performance of harbour seals, but in contrast, increasing delay resulted in a reduced performance as compared to harbour seals. The results of this study are consistent with the hypothesis that structural differences in the vibrissal hair types of otariid compared to phocid pinnipeds lead to different sensitivity of the vibrissae during forward swimming, but still reveal a good performance even in the species with non-specialized hair type.


Sea lionMystacial vibrissaeHydrodynamic trail followingVibrissal hair types


The vibrissal system of pinnipeds, its special morphology and differences in structure between species aroused the interest of scientists many years ago. As in terrestrial species (Rice et al. 1986), the follicle at the base of each vibrissa is part of an intricately structured and densely innervated sensory organ, the follicle–sinus complex (FSC). The FSC of harbour seals and sea lions is similar to that of terrestrial mammals, but shows remarkable differences concerning size and structure of the sinus system, as well as the degree and pattern of innervation (Stephens et al. 1973; Hyvärinen and Katajisto 1984; Hyvärinen 1989). The FSCs of pinnipeds are up to three times bigger than those of terrestrial mammals (Dehnhardt et al. 1999) and they possess an additional cavernous sinus that serves as a thermic insulator for the receptor area below it (Mauck et al. 2000). Additionally, they contain up to ten times more nerve fibres than well-endowed terrestrial mammals such as rat and cat (Rice et al. 1986). These facts indicate a special importance of the vibrissal system for pinnipeds. Contrary to the similar structure of the FSCs in all pinnipeds, the hair shafts differ substantially in shape. The vibrissal hair shafts of the Otariidae (sea lions and fur seals) and Odobenidae (walruses), as well as those of some phocids (true seals) like the bearded seal and the monk seal, are oval in diameter and smooth in outline. In contrast, those of all other phocid species are flattened and have waved surfaces (Watkins and Wartzok 1985; Kastelein and van Gaalen 1988; Hyvärinen 1989; Dehnhardt and Kaminski 1995). So whilst the special morphology of the FSCs indicates a pronounced importance of the vibrissal system for pinnipeds in general, the structural differences of the hair shafts may represent either a different sensitivity in most of the phocids as compared to the otariids, or at least a different approach to the need to filter the relevant signals out of the background noise in otariids, maybe involving a higher degree of postprocessing than in phocids.

To characterize the efficiency of the vibrissal system of pinnipeds and to investigate if there are differences in sensitivity between the species, several experiments have been carried out in the past. Kastelein and van Gaalen (1988) demonstrated that a walrus (Odobenus rosmarus) can distinguish a circular disc from an equilateral triangle using only its mystacial vibrissae. In 1994, Dehnhardt et al. investigated the capability of a blindfolded California sea lion (Zalophus californianus) to discriminate diameter differences of circular discs by means of active touch (Dehnhardt 1994). The lowest relative difference threshold (Weber fraction, c) was c = 0.22, comparable to the results of the walrus (Kastelein and van Gaalen 1988). In an analogous study in harbour seals (Phoca vitulina), lowest Weber fractions were about half of those obtained for the sea lion (c = 0.08; c = 0.13; Dehnhardt and Kaminski 1995). Dehnhardt et al. (1997) showed that size discrimination capabilities of harbour seals remained essentially unaltered if tested in air and water.

Dehnhardt et al. (1998) considered that hydrodynamic events, i.e. water disturbances of biotic and abiotic origin, can provide another source of information to pinnipeds, and tested the function of the vibrissae as a hydrodynamic receptor system. By using dipole water movements generated by an oscillating sphere, a technique commonly used to test lateral line function in fish (Bleckmann 1994), Dehnhardt et al. (1998) demonstrated that harbour seals were able to detect water movements as low as 0.25 mm s−1 at 50 Hz, which was several orders of magnitude below the water particle velocities measured in the wake of a swimming fish (Bleckmann et al. 1991; Hanke and Bleckmann 2004). At 20 and 30 Hz, harbour seals can detect water velocities of 0.37 and 0.35 mm s−1, respectively. In comparison, thresholds of a California sea lion that were tested for hydrodynamic dipole stimuli at 20 and 30 Hz were even lower (Dehnhardt and Mauck 2008).

These results show that the whiskers of harbour seals and sea lions form a hydrodynamic receptor system with a spectral sensitivity that is well suited for the frequency range of fish-generated water movements. Given that particle velocities of an oscillating sphere and all other oscillating, but stationary, objects attenuate rapidly with distance from the source (Kalmijn 1989), Dehnhardt et al. (2001) conducted an experiment with a propeller-driven miniature submarine as a hydrodynamic trail generator. The submarine served as a model to mimic the hydrodynamic trails left by swimming fishes (Hanke et al. 2000). Its trail was a narrow street of turbulent water movements containing velocities in the same order of magnitude as the wake of a 30-cm fish, as calculated from measurements in smaller fish (Hanke et al. 2000). The blindfolded harbour seal was able to track hydrodynamic trails of the submarine as long as 40 m. However, in trials where the vibrissal system was covered by a stocking mask, the seal failed to follow any trail.

Schulte-Pelkum et al. (2007) demonstrated that harbour seals can also detect and follow biogenic hydrodynamic trails. They used two harbour seals, one as the trail generator and the other as the trail follower. The subcarangiform or thunniform swimming style of phocid seals (Williams and Kooyman 1985; Fish et al. 1988) generates a hydrodynamic trail similar in structure to that left by fishes, but at least 2 m wide (Schulte-Pelkum et al. 2007). The accuracy with which the following harbour seal could swim in the middle of the trail of its conspecific strongly indicates that it was able to analyse the inner structure of the trail.

Hydrodynamic and biomechanical studies of single vibrissae of harbour seals and sea lions (Hanke et al. 2010) showed that the undulated surface structure of harbour seal vibrissae serves to suppress self-generated noise during forward swimming. Given that the vibrissae of sea lions have been shown to be quite sensitive to hydrodynamic stimuli in stationary animals (Dehnhardt and Mauck 2008) but possess a different structure, the question arose whether hydrodynamic trail following is also found in otariids.

Here, the ability of a blindfolded California sea lion (Zalophus californianus) for tracking hydrodynamic trails generated by a miniature submarine was tested. We found a lower trail-following ability in the sea lion than previously found in harbour seals. This difference might be related to the different shape of the hair shafts in the two species.

Materials and methods

Experimental subject

A three-year-old female California sea lion (Zalophus californianus californianus), named Sunny, served as the experimental animal for this study. Sunny was born in captivity and housed with the sea lion group of Zoo Duisburg, Germany, since 2003. The test animal was experimentally naive, but participated in the daily show-training routine. During the time of the experiments, the animal was kept separated from the group in an additional enclosure (together with another female sea lion) and was not included in the daily shows. The training enclosure consisted of a pool with a total volume of about 230 m3 of freshwater and an adjacent land part. The daily diet consisted of 2.5 kg freshly thawed sprats and was fed mainly during the training sessions. Typically, experiments were conducted on workdays (5 days per week), in some cases also on weekends, depending on the motivation of the animal. The regular food deprivation between the last feeding in the afternoon and the first training session the next day was 12–18 h. To avoid any disturbances during the experimental trials, the second animal was gated into an indoor pen. The experiments were carried out under the guidelines established by the European Communities Council Directive of 24 November 1986 (86/609/EEC).

Animal training

The animal was trained using operant conditioning with food as the primary reinforcer, and short blows from a training whistle (approximately 8 kHz) as well as verbal praise as the secondary reinforcer. No negative reinforcement other than withdrawal of attention was used.

The animal was first trained to approach the trainer on a verbal and gestural command. It was trained to wait at a stationing target (a 10-cm plastic sphere at which it positioned its snout), and to leave it on a verbal command and retrieve the submarine using all sensory systems including vision. When the submarine had been found, the short blow from the training whistle (secondary reinforcer) was immediately given, and a food reward (primary reinforcer) was given after the animal had returned to the trainer. The secondary reinforcer was applied to achieve a close temporal proximity between the desired behaviour and the reinforcer. Verbal commands were used to terminate unsuccessful trials. In parallel, the animal was accustomed to the blindfolding mask using a shaping procedure. The transition from finding the submarine with and without vision was facilitated by an algal bloom that occurred in the experimental pool and prevented the animal from seeing the submarine when it started its search; it switched to the use of hydrodynamic information spontaneously.

Due to time constraints, testing trials began already after the animal had reliably found the submarine whilst blindfolded in one training session. The first 29 trials were conducted with linear trails (see below) of variable delay.

Stimuli, trail generating

The hydrodynamic trails were generated by a propeller-driven, radio-controlled miniature submarine and were visualized and measured using particle image velocimetry (PIV). The submarine represents an improved version of the one used by Dehnhardt et al. (2001). It is radio remote controlled, has a torpedo-like shape, and a single propeller drive to approximately mimic the hydrodynamic trails of fish (Fig. 1). Hydrodynamic trails could be generated that varied in direction, speed and shape. Additionally, the submarine’s remote control was equipped with a “dead man switch” to shut down all electrical systems after trail generation to prevent acoustical cues.
Fig. 1

The remote-controlled submarine used as the trail generator

Flow measurements (PIV)

We used particle image velocimetry (PIV; compare Westerweel 1997) to measure the trails of the submarine. Measurements were performed in a small outdoor pool with the dimensions 5 × 2 × 1 m (L × W × H). Neutrally buoyant polyamide seeding particles (Vestosint 1101, Degussa-Hüls AG, Marl, Germany) were added to the water and were illuminated in a horizontal plane by a fanned out diode pumped solid state laser (500-mW-DPSS-Laser, Entertainer 500, Quantum Physics, Newcastle, UK; optical fibre: Laserlight Showdesign, Berlin, Germany). The horizontal light sheet was about 2-mm thick. A CCD camera (DMK2001, The Imaging Source, Bremen, Germany) was mounted above the water surface filming the layer of illuminated particles. The water surface was smoothened by a glass plate (60 × 30 cm). The video signal was stored digitally by a DV camera (Canon XL1S, Canon Inc., Tokyo, Japan). The video recordings were digitally sequenced at a frequency of 50 half frames per second and analysed using custom-designed correlation programs in MatLab 6.5 (MathWorks, Inc., Natick, MA, USA) (Hanke and Bleckmann 2004), which followed the principles of digital particle image velocimetry (Willert and Gharib 1991; Hart 2000).

Experimental setup

Experiments took place in a part of the pool that was 8 × 7 m in size and 1.2-m deep on average. It contained a platform big enough to store the experimental equipment (Fig. 2). The platform served as a start point for the submarine and the sea lion. Trials were recorded by a top camera with a wide angle lens 4.5 m above the pool that filmed the whole experimental area for offline analysis.
Fig. 2

Schematic drawing of the experimental platform with the animal stationed above the water surface and the submarine shortly before the start of a trial

Experimental procedure

At the beginning of every trial, the experimental subject was stationed at a target beside the experimenter who was kneeling on the platform. The platform was half flooded, so that the animal could be stationed in a shallow water area with its head above the water. From this position, the animal could glide easily into the deep water and start its search for the hydrodynamic trails. For visual masking, an opaque elastic mask was placed on the sea lion’s head, completely covering the eyes, but leaving the mystacial vibrissae uncovered. To achieve calm water conditions, the blindfolded test animal had to wait for approximately 30 s at its station until the submarine was started. Subsequent to this, the submarine was started directly in front of the platform at a depth of approximately 40 cm. It was remote controlled by a second person who stood on another platform outside the experimental area, having an overview of the complete test area. This person had to decide at the end of each trial if the test animal had performed a correct trail following or not. For a correct trail following, the animal was rewarded with fish. The submarine was steered by the human operator in a way that the paths were highly variable.

In the first part of the study, linear hydrodynamic trails were generated. The submarine was accelerated straight for approximately 2 m up to 6 m and then braked by shortly turning the propeller backwards and activating the “dead man switch” (see above). The change of swim speed was sufficient to identify the brake points on the video sequences. After a gliding phase of at most 2 m, the submarine came to a rest. The animal was allowed to submerse its head in the water only after the dead man switch had been activated and all sounds from the submarine were prevented, except for the trials with complex hydrodynamic trails (part 3 of the study, see below). The direction of the linear hydrodynamic trails was recorded as the angle of the submarine’s driving path in relation to the platform, with 0° defined as the direction straight ahead of the animal, perpendicular to one of the edges of the platform that was chosen as a reference. The angles of the trails were between +30 and −50°. In the second part of the study, trails including sharp curves to the left or to the right were introduced. In the last state of the study, complex trails with even more than one directional change were used. In all stages of the study, the test procedure remained the same as for the linear trails.

Criteria for successful trail following

For a valid trial, certain criteria had to be fulfilled. First of all, the trail had to be generated without breaking through the water surface to make all trails comparable with each other. Secondly, finding the submarine by means of acoustic cues, characterized by a straight and quick approach to the submarine, was not counted as a correct trail following, but separately noted as a shortcut. This could happen when the submarine hit the wall of the pool during the glide phase (with the motor switched off). It must be noted that linear trails were, without acoustic cues, not followed straight and quickly, but the sea lion showed deliberate movements and head undulations. A valid trial was counted as correct when the sea lion followed the submarine at least until the brake point where the engine was switched off to an accuracy of one boat length or better. This criterion was chosen because one boat length represents a good estimate of the trail width plus the spread of the vibrissae, and because after the brake point the submarine was just gliding, making the trail more difficult to follow.

Experiments to exclude secondary cues

Since aquatic animals may also use other sensory input such as chemosensory and/or acoustic cues for the perception of hydrodynamic stimuli (Enger et al. 1989; Weissburg 2000; Pohlmann et al. 2001; Ferner and Weissburg 2005), control experiments were carried out during which the sea lion’s mystacial vibrissae were covered by the opaque elastic eye mask. The sea lion was still able to open its mouth, allowing the animal to use chemosensory cues. The experimental procedure during control trials was identical to that of test trials. At this point, the animal was so familiar with the trainer that no special training was required for covering the vibrissae.

Data recording and analysis

For offline analysis, experimental sessions were recorded by a top camera (Sharp CCD outdoor camera) equipped with a wide angle lens and stored on a video recorder (FUNAI, Video Cassette Recorder, Model No 31A-650). By placing the camera at a height of 4.5 m, the complete experimental area could be recorded. Every sequence was digitized and transferred into bitmap frames by Main Actor for Windows (v. 3.65, Main Concept GmbH; frequency = 25 frames s−1). Sequences were viewed on an LCD monitor, the paths of the submarine and the animal were traced on transparency film and the relevant parameters were measured with a ruler. This procedure proved to be considerably less time consuming than digital techniques and is adequate for the goals of the present study. In addition, ten sequences with curved or complex hydrodynamic trails were evaluated digitally at a reduced frequency of 5 frames s−1. The single tracks of the submarine and the test animal were marked frame by frame to get the x and y coordinates for graphical description (Scion Image for Windows; Beta 4.02, Scion Corporation) and statistical evaluation.

The wide angle lens that was necessary to film the whole experimental area and the fact that the camera could not be mounted exactly above the centre of the experimental area cause a distortion of the image. This distortion does not affect the results of the present study, because criteria for successful trail following were based on the length of the submarine at the respective position in the image. Images were scaled by measuring the distance between the experimental platform and the opposing wall on a satellite image (Google Maps, Google Inc., Mountain View, CA, USA). The scale bar provided on the satellite image was verified by measuring the track gauge of the nearby railway (1,432 mm) and was accurate to 2% or better. Slight changes in the field of view, which were unavoidable as the camera had to be removed after each session, were accounted for by aligning two corner points of the experimental platform in Microsoft Excel.

Statistical evaluation

The probability of finding the submarine trail’s end point by chance was calculated from the area that the sea lion had to reach (a circle of one boat length radius), divided by the area over which end points were distributed (35 m2), resulting in p = 2.8%. It must be noted that the sea lion was not allowed to perform an extended search after missing the end point, but trials were immediately aborted in that case. The percentage of unsuccessful trials as compared to successful trials was analysed using a binomial distribution to assess the probability that the number of successful trials was achieved by chance. Directions of linear trails and directions of right or left curves were chosen pseudorandomly (Gellermann 1933). The complex trails (curves and multiple direction changes) in parts two and three of the study are practically impossible to replicate by chance. To calculate the probability p that the sea lion follows a complex hydrodynamic trail by chance, the experimental area was divided into square fields of 0.5 submarine lengths. The probabilities qi that the animal entered the correct field by chance repeatedly whilst following the trail were multiplied and resulted in p ≪ 0.0001.


Figure 3 shows the spatial distribution of the braking points that were found and the braking points that were not found by the sea lion. The behaviour of the sea lion was similar in all trials. After the start signal, the test animal directly submerged from its waiting position into the start region of the submarine. Searching for the hydrodynamic trail was characterized by protracting the vibrissae to the most forward position and was sometimes accompanied by lateral head movements. When the trail was detected, the sea lion accelerated and then followed the course of the submarine, often mostly gliding. Frame-by-frame analysis of the video material showed that the test animal usually swam directly on the trail of the submarine, sometimes parallel to it, but never more than one boat length deviating from it. Sometimes the animal went on searching even after the point where the engine of the submarine had been switched off. In this phase, the search pattern was characterized by the sea lion additionally performing undulating head and body movements.
Fig. 3

The braking points of the 209 experimental trials. Green circles show the points found by the sea lion, red diamonds show the points not found. Point (0,0) is the starting point of the submarine and the approximate starting point of the sea lion

Perception of linear hydrodynamic trails

To test the perception of linear hydrodynamic trails, 104 trials were analysed. Trail lengths ranged from 2 to 6 m. The test animal could follow the submarine successfully in 84.6% of all linear trials (88 times out of 104). In the remaining trials, the hydrodynamic trail was lost far from the submarine’s braking point or even close to the beginning. Whilst the directions “left” or “right” relative to the zero (0°) direction were counterbalanced according to Gellermann (1933), the exact heading of the submarine and the length of the trails were varied at the discretion of the submarine pilot and were approximately evenly distributed.

Perception of curved hydrodynamic trails

In nature, following a hydrodynamic trail generated by a fish will often require the detection of one or several changes of swimming direction. Therefore, we conducted 84 trials with a change in direction, 42 curves to the right and 42 curves to the left. Trail lengths ranged from 3 to 11 m. An example is depicted in Fig. 4. The test animal performed successfully in 20 out of 42 right curves (47.6%) and 22 out of 42 left curves (52.4%). Contrary to the linear hydrodynamic trails, curved hydrodynamic trails were highly variable in shape, so the possibility to track them by chance was negligible. Although curved trails were essentially used throughout a session with only few linear trails interspersed, there was no indication that the animal expected a specific pattern, e.g. it did not follow a straight portion of the curve correctly and then expected a turn, but guessed the wrong direction. The animal appeared to analyse the trails independently of each other.
Fig. 4

Example of a curved trail (right curve). Times in the upper right corner of each panel denote the time after the submarine was started. The experimental platform (compare Fig. 2) and the submarine are located in the lower left corner. Yellow lines show the path of the submarine until the engine was switched off. a–c Trail generation and start of the animal. d–f Trail following by the sea lion. Note the sharp turn in e

Perception of complex hydrodynamic trails

Additionally, complex hydrodynamic trails, defined by even more than one unpredictable change of the submarine’s course, were generated. In these trials, the increasing complexity of the hydrodynamic trail caused an increasing delay between the start of the submarine and the start of the sea lion. The increasing delay itself caused a decreasing trail-following success of the test animal. If it had to wait for more than 5 s after the start of the submarine, the sea lion mostly lost the trail at or before the first directional change. To study if this decreasing performance was related to the increasing delay, or rather to the complexity of the trail, 30 trials were conducted where the test animal could leave its station after 1–3 s, even when the engine of the submarine was still running (Fig. 5). Trail lengths ranged from 3 to 13 m. The animal showed successful trail following in 17 of 21 trials (81%). In nine trials, the animal made use of the acoustic cues given by the running motor of the submarine and reached it by taking a shortcut, i.e. a straight approach to the submarine without following its path. These trials were excluded from analysis. However, the trials in which the sea lion did follow the path of the submarine strongly indicate that hydrodynamic perception was used.
Fig. 5

Example of a complex trail (slight right curve in the beginning followed by a long left curve). The delay between the start of the submarine and the animal was 3 s. As in Fig. 4, yellow lines show the path of the submarine. ac Trail generation and start of the animal. d–f Trail following after the engine was switched off

Statistical analysis

Geometrical considerations (see “Material and methods”) show that the probability of finding the submarine’s braking point (within one submarine length) by chance was 2.8%. The cumulative binomial probability of finding the submarine’s braking point by chance in 147 out of 209 trials is therefore negligible (p  ≪ 0.0001). The exact following of the highly variable curved and complex trails that we observed (cf. Figs. 4, 5) was also extremely improbable to occur by chance. To obtain an upper bound for this probability, we divided the experimental area into square fields with an edge length of 0.5 submarine length. Passing through these fields, the sea lion has to decide repeatedly which of the three neighbouring fields ahead to choose. In a typical experiment with a trail length of 10 submarine lengths, the probabilities of p = 0.33 to pick the correct field by chance add up to p = 10−10. In reality, the sea lion’s performance was much more accurate than picking one of three fields ahead and included even sharper turns.

Dependence of the trail-following performance on the delay

Figure 7 shows the probability of successful trail following as a function of the delay between the start of the submarine and the start of the animal in linear, curved and complex trails. Numbers above each column depict the number of trials at the respective delay. In linear trails, the success rate was 75% or higher after up to 4.5 s. In curved trails, the success rate dropped below 75% already after 3.5 s. In complex trails, performance was still above 80% at 3.5 s, and the one trial performed at 4.5 s was successful. When complex trails were introduced towards the end of the study, the inability of the animal to follow the trails after more than 7 s was unambiguously observed by the experimenter and by the submarine pilot. It was however not recorded on video and is not represented in Fig. 7.

Experiments to exclude secondary cues

One test series containing 18 trials with linear and complex hydrodynamic trails was carried out: 10 trials with normal conditions interspersed with 8 trials where the mystacial vibrissae were covered by the stocking mask. Delays were short to keep the animal motivated. The animal performed successful trail following in all trials with exposed vibrissae, but always failed when the vibrissae were covered.

Swim speed

The mean swim speed of the test animal during the trail following was calculated from video recordings. For linear hydrodynamic trails, swim speeds of 1.77 ± 0.30 m s−1 were determined, 1.26 ± 0.21 m s−1 for trails with a left curve and 1.28 ± 0.24 m s−1 for trails with a right curve. Consequently, following of linear trails took the seal 1–4 s, and following of curved trails 2–6 s.

Flow measurements

Flow measurements using digital particle image velocimetry (Westerweel 1997) showed that the submarine left a narrow (20–30 cm) trail of turbulent water movements which contained water velocities of approximately 10 cm s−1 after 3 s, and 8 cm s−1 after 5 s. Contrary to fish trails, the turbulent eddies were small compared to the length of the trail generator (up to approximately 3-cm diameter). Figure 6 shows the decay of water velocity over time. At each point in time (25 images per second), water velocities were averaged over the swimming direction of the submarine, resulting in 21 velocity values along the direction perpendicular to the swimming path. The mean of the highest five velocity values was taken as representative of the water velocity in the submarine’s trail. This procedure was chosen to reduce measurement noise caused by suboptimal particle density and limited camera speed. Examples of vector fields can be found in Wieskotten et al. (2010b), who used the same miniature submarine in a trail-following study with a harbour seal (Phoca vitulina).
Fig. 6

Time course of the water velocity in the submarine’s trail as measured with particle image velocimetry. At each point in time (25 images per second), water velocities were averaged over the swimming direction of the submarine, resulting in 21 velocity values along the direction perpendicular to the swimming path. The mean of the highest five velocity values was taken as representative for the water velocity in the submarine’s trail


Significance of results from a single experimental subject

The performance of single experimental animals can in principle not be taken as representative for the species. “Sunny” is a comparatively young animal with little experimental experience. However, “Sunny” learnt the task quickly and was well motivated on the days chosen for data collection. It should further be mentioned that our first sets of trail following experiments in harbour seals (Phoca vitulina) were conducted with an even younger animal (“Henry”) which nevertheless showed a better performance than “Sunny”. Therefore we believe that the present results are indicative of a reduced trail-following ability in sea lions as compared to harbour seals. To safely confirm the idea of differential sensitivity of the vibrissal systems between harbour seals and sea lions, further studies with different individuals and extended training are needed.

Dependence of the trail-following performance on the delay

Figure 7 indicates that trail-following performance was best in linear trails, but was only slightly reduced in complex trails for the delays depicted here (where the motor was still running). Curved trails were more difficult to follow. The good performance of the sea lion in linear trails as compared to the other trail types is not surprising; they were certainly the least demanding task. The relatively good performance in complex trails may be due to the fact that these trials were conducted towards the end of the study where the animal was most experienced, or due to the fact that the motor was still running, potentially enabling the animal to use both hydrodynamic and auditory cues. However, from the very precise following of the submarine’s path, we conclude that acoustic cues played a minor role at most. Curved trails (Fig. 7b) may be considered most representative for the sea lion’s performance, as the change in direction is demanding, whilst acoustic cues are excluded. These data show that the sea lion’s trail-following performance drops drastically after delays of approximately 5 s.
Fig. 7

Probability of successful trail following as a function of the delay between the start of the submarine and the start of the animal, a in linear trails, b in curved trails, and c in complex trails. Numbers above each column depict the number of trials at this delay

The performance of our sea lion compared to that of harbour seals

The results of the present study show that sea lions are, like harbour seals, able to detect and track hydrodynamic trails. Control experiments with the vibrissae covered showed that the vibrissae were the sensory system mediating this ability in both harbour seals and sea lions. The successful trail following of the sea lion during linear trials (84.6%) is nearly identical to that of the second tested harbour seal in the study of Dehnhardt et al. (2001; Nick: 82.2%). The undulating head and body movements that were observed in the sea lion whilst trying to find the submarine in its glide phase have also been described for harbour seals by Schulte-Pelkum et al. (2007). This search type might help to track weak hydrodynamic trails by comparing the surrounding water movements with the possible trail.

Hydrodynamic trails that contained a directional change to the left or to the right could be followed successfully in 50.0% of all trials. This performance level is much lower than that observed in harbour seals (Dehnhardt et al. 2001; 86.6%), but is still significant because the trails varied in shape in a way unpredictable to the sea lion and were nevertheless accurately followed.

An obvious difference between the performance of our sea lion and harbour seals was found when delays longer than 5 s were introduced. Whereas the performance of harbour seals was hardly affected by increasing delays between the start of the miniature submarine and the start of the seals’ search (Dehnhardt et al. 2001; Wieskotten et al. 2010a, b), the sea lion was unable to perform correct trail following after waiting for more than a few seconds. By shortening the delay and giving the sea lion the possibility to start searching whilst the submarine was still running, it was shown that complex hydrodynamic trails could be detected and tracked if the trail was not older than 4–5 s. Although acoustic cues were present in this situation, it appears reasonable to assume that hydrodynamic information was the dominant factor in those trials where the animal did not approach the submarine directly (i.e. used no ‘shortcut’), especially because ‘shortcuts’ were observed in several cases where acoustic information was sufficient and clearly within the behavioural repertoire of the animal.

The observed ‘shortcuts’, the straight approaches to the submarine by making use of acoustic cues, show that orientation and hunting under water are multimodal processes and the animals are able to use the cues that are most prominent in a special situation. A lot of fish species, including typical prey fish of pinnipeds, are known to produce a variety of sounds. Cods produce click sounds (Vester et al. 2004) and herrings produce broadband pulses (Wilson et al. 2004). One possibility for the sea lion whilst hunting could be to localize the approximate position of the prey fish by making use of acoustic cues, turn onto the hydrodynamic trail to get closer and finally catch the fish when it is close enough to be seen. The behaviour of harbour seals in the study of Dehnhardt et al. (2001) regarding the use of ‘shortcuts’ was comparable to that of our sea lion.

Swim speed

The mean swim speed determined for the sea lion during the linear trail-following trials was 1.77 m s−1. Feldkamp (1987) showed that the energetically ideal swim speed (minimum cost of transport) for two sea lions was 1.4 body lengths per second (BL/s) at a water temperature of 26°C. A third animal studied at a water temperature of 19°C showed no distinctive minimum, but cost of transport continued to decline over the range of swim speeds examined (0.4–2.1 BL/s). In our study, water temperature was not monitored on a daily basis, but it was probably in a similarly moderate range in the shallow pool during May through June, and lower than 26°C in most cases. Further, it should be noted that water depth, which was not specified in Feldkamp (1987), may influence the cost of transport. Assuming that a value of 1.4 BL/s applies to our test animal, the energetically ideal swim speed would be 1.82 m s−1or higher. The speed in our linear trail-following experiments (1.77 m s−1) is very close, tending to be lower than energetically optimal speeds. Swim speeds in the trails with one change of direction were significantly lower (1.26 m s−1 or 1.28 m s−1, respectively). These reduced swim speeds in the more difficult tasks indicate that the sea lion’s ability to follow hydrodynamic trails decreases with increasing swim speed. Mean gliding speeds of 1.6–2.4 m s−1 determined for the new Zealand sea lion (Phocarctos hookeri) (Crocker et al. 2001) or 2.6–3.6 m s−1 for the Steller sea lion (Eumetopias jubatus) (Stelle et al. 2000) also indicate that voluntarily chosen swim speeds are slightly higher than the swim speeds in our trail-following experiments. Dehnhardt et al. (2001) estimated a swim speed of 2.0 m s−1 for the harbour seal following linear hydrodynamic trails, a value comparable to the results of the present study.

Do the different hair types lead to different trail-following abilities in sea lions and harbour seals?

The vibrissae of sea lions and harbour seals differ remarkably in the structure of the hair shaft: whilst the vibrissae of sea lions and other otariids have an elliptical cross section and are smooth in outline, those of harbour seals and most other phocids have a flattened cross section and an undulated surface structure (Watkins and Wartzok 1985; Dehnhardt and Kaminski 1995; Fish et al. 2008; Ginter et al. 2010). It has recently been shown that the structure of harbour seal vibrissae serves to reduce vortex-induced vibrations, i.e. the vibrations that are caused by vortices being shed from the vibrissae as the animal moves forward (Hanke et al. 2010). The present study supports the hypothesis that the specialized shape of harbour seal vibrissae helps the seal to avoid vortex-induced vibrations and keep its vibrissae still whilst swimming forward, thus increasing sensitivity in the swimming animal. This explanation appears especially likely in the light of the fact that sensitivity to hydrodynamic stimuli in stationary, not moving, animals was higher in the sea lion than in harbour seals (Dehnhardt and Mauck 2008), and thus the processing of mechanical stimuli once they reach the mechanoreceptors within the follicle–sinus complex appears not to be a limiting factor.


We thank the directors and staff of Zoo Duisburg for their great cooperation. This study was supported by grants of the Volkswagenstiftung to GD, and the German Research Foundation (DFG) to GD and WH (SPP 1207). The experiments were carried out under the guidelines established by the European Communities Council Directive of 24 November 1986 (86/609/EEC).

Copyright information

© Springer-Verlag 2010