The authors have declared that no competing interests exist.
Conceived and designed the experiments: SPO. Performed the experiments: SPO KG. Analyzed the data: SPO TDJ. Contributed reagents/materials/analysis tools: SPO JRT KG MS. Wrote the paper: SPO. Editorial input: JRT.
The hunting strategies of pelagic thresher sharks (
Dense aggregations of prey fishes, commonly termed ‘bait balls’, attract large marine predators to areas of high productivity across the globe
Reaching 365 cm in total length, approximately half of which comprises a scythe-like elongate tail fin,
Predation strategies employed by sharks are diverse and vary among species and individuals
Tail-slapping has been observed in a range of marine predators. Humpback and sperm whales (
When investigating how killer whales forage on schooling herring in Norway, Domenici et al. (1999) showed that tail-slapping enabled the predator to stun up to 33 prey fish with one strike alone. Since sardines school in dense aggregations
In this paper, evidence is provided to show that pelagic thresher sharks use their tails to prey upon sardines, and the kinematics associated with the behaviour are investigated. Hunting events were quantified from handheld video observations to address the following hypotheses: (1) thresher sharks execute a series of rapid body motions that drive tail-slaps during hunting events; (2) tail-slapping enables thresher sharks to stun several prey items at a time. Thresher shark hunting behaviour is discussed in relation to kinematics and hydrodynamics.
All of the research (including the handling of marine life, and the interruption of shark behaviour) was undertaken with the permission of the Governor of Cebu and adhered to the Philippine ‘Wildlife Resources Conservation and Protection Act’. The handling of marine life complied with Bangor University's
Pescador is a small coral island situated in the Tañon Strait (N 09° 55′ 44.2′, E 123° 20′ 61.2′), approximately five kilometers due west from Moalboal, Cebu, in the Philippines (
A dense aggregation of sardines
Fieldwork was undertaken over 70 days, spanning five months, from June to October 2010 (fieldwork was time restricted due to resource limitations). Handheld underwater video cameras were used by SCUBA divers to record thresher shark hunting behaviour during hour-long dives conducted between 09:00 and 16:00 hours. SCUBA divers used Sony Camcorders® FX-1 and HVR-Z1 housed in Gates Z1 underwater housings, fitted with dome ports, with their focal ranges locked to 0.4 m, and recorded their observations of thresher sharks onto MiniDVs in 1080i 50 (25 fps−1) and 1080i 60 (29.97 fps−1) HDV formats. Video records were captured opportunistically in the water column between 10 and 25 m depths with the camera recording when a thresher shark was present and observable in the viewfinder. Recordings were downloaded to a hard drive and screened for analysis.
On some occasions, divers interrupted the feeding behaviour of the sharks to collect stunned and dead sardines by hand from the water column. These were brought to the surface where they were inspected for injury, photographed, total length measured, and then released if they were alive. Observable injuries sustained by collected individuals were assumed to be associated with a thresher shark's predatory behaviour. Since no stunned or dead sardines were observed prior to thresher shark attacks, their presence in the water column was used as a proxy for a successful hunting event.
Video sequences documenting thresher sharks' hunting behaviour were classified into two main event types: those in which predation attempts were characterised by (1) an overhead tail-slap or (2) a sideways tail-slap. Overhead tail-slaps typically took place when the shark was positioned perpendicular to and facing the perimeter of the bait ball (
For analysis, hunting events were partitioned into ‘phases’ that were characterised by observable changes in a thresher shark's movement and behaviour during a tail-slap. Termed ‘preparation’, ‘strike’, ‘wind-down recovery’ and ‘prey item collection’, phases were analysed in 25 or 29.97 frames s−1 resolution using
The terms ‘motion’, ‘mechanics’ and ‘kinematics’ were adapted from their standard uses in the literature
Protocols developed by Slater for categorising behaviour
To estimate a thresher shark's length measurements (total length (TL); precaudal length (PCL); and dorsal caudal fin margin (CDM)), a still image was taken from its video record when the shark was planar to, or in contact with, one of the sardines it was hunting, and both were perpendicular to the axis of observation. Assumed to be equal to the mean (± SE) of the total lengths (cm) of the sardines collected by SCUBA divers
where (f) was shark, (p) pixels and (s) the referenced sardine. Shark sex was determined by the presence or absence of claspers.
Only sagittal and transverse plane video observations were selected for kinematic analysis, in which all four phases of the tail-slaps occurred within full view of the camera, and where the shark was close enough to identify the key anatomical parts used for hunting. Of the 22 recordings of overhead tail-slaps, only six sagittal and two transverse plane events were considered suitable for analysis. None of the video records of the sideways tail-slaps met selection standards and were therefore only used to describe the behaviour.
For sagittal plane events, three key anatomical parts (i) the terminal caudal fin lobe (tip of the tail), (ii) the midpoint of the caudal peduncle, and (iii) the tip of the snout were tracked in two dimensions by analysing a sequence of video still images. Using the posterior base of the pectoral fin as a fixed reference point, the coordinates of the anatomical parts were plotted for each still frame (
For sagittal plane events, three key anatomical parts (a) the tip of the tail, (b) the midpoint of the caudal peduncle, and (c) the tip of the snout were tracked in two dimensions using the posterior base of the pectoral fin as a fixed reference point (x/y intercept =
During the peak accelerations of the strike phase, the speed with which the tip of the tail travelled exceeded the frame rate of the underwater cameras used to record it. As a result, some images of the terminal caudal fin lobe were blurred, for all of the selected recordings. To plot the coordinates of the blurred images, a still image taken from a point in the video sequence when the terminal caudal fin lobe was clearly observable was layered on top of the original still image. The leading edge of the caudal fin, the caudal notch and the lower caudal lobe for the two layered still images were aligned. Since all of the leading edges of the caudal fins aligned precisely and only the tips of the caudal fins were blurred for all occurrences, it was assumed that the position of the terminal caudal fin lobe would not alter, relative to its orientation in the strike phase, and its coordinates were plotted from the image layered on top.
For transverse plane events, the pectoral fins were the only anatomical features to be tracked. The ventral midpoint between the pectoral fins was used as a fixed reference point, and coordinates for both the tips of the pectoral fins and their posterior bases were plotted for each video still frame. The angles at which the pectoral fins protruded from a shark's body were measured and the angular velocity with which they adducted to initiate a tail-slap was calculated using trigonometry.
All statistical analyses were carried out in
Tail-slap arc lengths (AL) (
To generate conservative estimates of the speed with which the tip of the tail travelled, straight-line distances between the plotted coordinates were used for calculations, even if the path of the coordinates formed an arc-like motion. Time was expressed as the number of video frames it took for a motion to be completed divided by the frame rate used by the camera upon which the observation was recorded (25 or 29.97 frames s−1). Speed was calculated by dividing the straight-line distance between the coordinates of the motion by time (ms−1). Speed was only calculated for the motion of the tail, and not the forward locomotion of the shark.
Least squares linear regression analysis was used to examine if tail-slap speeds (max and mean) were related to a shark's size (PCL), and/or the length of its tail (CDM). A tail-slap's maximum speed could only be calculated from the six sagittal plane events for the regression, since frame-by-frame analysis of the arc, which formed the tip of the tail's travel path, was required. The times it took for the tip of the tail to reach maximum speed and arc height were standardised as a proportion of the duration of the strike phase, and compared using a paired
To assess the trajectory of a thresher shark's tail-slap, strike phases were standardised for all sharks by dividing the arc that formed the travel path of a thresher shark's tail-slap into nine equal time segments, with one trajectory angle measured for each (
To examine if a phase's duration varied by the size of a shark, a two-way analysis of variance (with duration as the response variable and PCL and phase type as treatments) was used. The variability between the duration of the different phase types (preparation, strike, or wind-down recovery) was also investigated (one-way ANOVA with duration as the response variable and phase type as treatment).
The relative amplitudes of the movements (defined as the vertical distance between the lowest and highest points attained by a tracked anatomical part) of the tip of a thresher shark's tail, its snout and its caudal peduncle were compared among preparation, strike and wind-down recovery phases for the sagittal plane events, by using a two-way ANOVA (with amplitude as the response variable and phase type, anatomical part and associated interactions as treatments).
A total of 25 thresher shark hunting events were recorded at all times of day (09:00 to 16:00 hours; June – October 2010), 22 of which were overhead tail-slaps. Although divers observed sideways tail-slaps
Sex | PCL | CDM | FL | TL |
Unknown | 97.20±8.93 | 90.85±8.35 | 106.16±9.75 | 188.05±17.27 |
Female | 136.14±12.51 | 118.61±10.90 | 147.96±13.59 | 254.75±23.40 |
Female | 142.12±13.05 | 136.68±12.56 | 158.25±14.54 | 278.80±25.61 |
Female | 153.77±14.13 | 167.06±15.35 | 167.93±15.43 | 320.82±29.47 |
Unknown | 157.37±14.46 | 158.23±14.53 | 170.87±15.70 | 315.60±28.99 |
Male | 173.51±15.94 | 179.38±16.48 | 189.78±17.43 | 352.89±32.42 |
Male | 184.95±16.99 | 161.77±14.86 | 202.82±18.63 | 346.72±31.85 |
Female | 186.82±17.16 | 219.03±20.12 | 197.37±18.13 | 405.86±37.28 |
Male | 195.61±17.97 | 218.99±20.12 | 215.90±19.83 | 414.60±38.08 |
Male | 197.52±18.14 | 138.59±12.73 | 218.20±20.04 | 336.11±30.87 |
Lengths were calculated from video images by counting the number of times a referenced sardine could fit lengthwise into the selected length measurements of a thresher shark. Precaudal (PCL), dorsal caudal fin margin (CDM), fork (FL) and total (TL) lengths are presented±their standard deviations. The six sagittal plane events, which were selected for kinematic analysis, are in italics.
Event phases were defined by observable changes in the speed, vertical motion and directional orientation of the tracked anatomical parts, in particular, the positioning of the tip of the tail. A thresher shark accelerating in a lunge towards the bait ball characterised preparation phases. Strike phases were characterised by a slap of the tail. Strikes began with a shark adducting its pectoral fins, a manoeuvre that changed the shark's pitch promoting its posterior region to lift rapidly, and stall its lunge approach. The shark's tail then accelerated in a whip as it travelled overhead the length of its body to the tip of its snout (
A thresher shark lunged at the bait ball in the horizontal plane (1–3). It then adducted its pectoral fins in a manoeuvre that changed its pitch, promoting its posterior region to lift rapidly and stall its approach (4–6). After adducting its pectoral fins, the shark rotated them laterally in a surge to counter the momentum of its body from precipitating forward (7–10). A rapid and powerful ventro-dorsal peduncular motion drove its tail from its base in a trebuchet catapult motion that terminated overhead in a slap (7–10). The tail-slap occurred with such force that it caused dissolved gas to diffuse out of the water column forming small bubbles that entrained and grew in size (circled in 9–14). The shark returned its pitch to the horizontal plane in a wind-down recovery (11–14), turned 180°, and proceeded to collect the five sardines it had stunned (15). The center of mass about which the movements associated with the shark's overhead tail-slap occurred, changes in camber and time stamps are shown in white.
A motion animation (top) represents 1.08 s−1 of an event which was recorded by handheld underwater video camera on 17 June, 2010. Center inserts profile the key characteristics of the behaviour, while inserts shown in the transvers plane (bottom), were interpreted from other video sequences.
A motion animation (top) represents 3.16 s−1 of an event that was recorded by handheld underwater video camera 19 August 2010. Inserts show a thresher shark circling and collecting three sardines that were stunned during the strike phase of a successful hunting event.
When considering only preparation, strike and wind-down recovery phases, the mean (± SE) duration for overhead tail-slaps was 1.91±0.19 seconds (95% CI: 1.54–2.28 seconds). The preparation phase lasted significantly longer than the strike and wind-down recovery phases (f2,45 = 11.53, p < 0.0001) (
Of the 16 events considered suitable for analysis, 15 involved a thresher shark turning 180
Preparation phases were characterised by thresher sharks lunging at the bait ball in the horizontal plane (mean 10.36±4.80° SD). Lunging never resulted in prey capture but was usually followed by a tail-slap. As a thresher shark accelerated into a lunge, there was little vertical movement of the tracked anatomical parts (
A) The movements of the tip of the tail (solid), the caudal peduncle (dotted) and the snout (dashed) were tracked in relation to their relative distances (cm) from the posterior base of the pectoral fin (x/y intercept =
Angles for transverse plane events (n = 2) were measured for preparation (blue); strike (red); and wind-down recovery (green) phases, and are aligned from the point at which the tips of the tail reached their maximum height (x/y intercept =
Strikes were always preceded by lunges and often resulted in prey capture. The mean (± SE) duration of the strike phase was 0.39±0.01 seconds (95% CI: 0.36–0.41 seconds). The duration of a strike phase and size of shark were not correlated (f1,14 = 0.07, p = 0.795).
To initiate a strike, a thresher shark first lowered its snout and flexed its body dorso-ventrally, causing the caudal-peduncle and tail to dip and tension (
A two-way analysis of variance found no significant differences between the shapes of the trajectory that formed the path of a thresher shark's tail-slap among the six events that were recorded in the sagittal plane (f5,40 = 1.04, p = 0.409). Figure ten shows the extent to which the trajectory paths were similar when standardised for precaudal length (
The relative distances (cm) the tip of the tail was from the posterior base of the pectoral fin (x/y intercept =
During a strike, the mean (± SE) speed with which the tip of the tail travelled over a thresher shark's body was 14.03±1.01 ms−1 (95% CI: 12.05–16.01 ms−1) for all sharks combined (n = 16). Tail-slap speeds were biphasic and would accelerate to their maxima, and then decelerate until the tip of the tail reached its terminal point above the snout (
During three of the successful events, bubbles were observed to form where the tip of the tail reached its maximum speed and height (
Speeds are expressed in meters per second (ms−1), and are regressed against the caudal fin (CDM) lengths of individual sharks that were observed in the sagittal plane (n = 6).
Wind-down recovery phases, during which the snout, caudal peduncle and tail were returned to their original condition, always began when the tip of the tail reached the terminal point of its trajectory above the snout (
The relative amplitudes of the tracked anatomical parts varied among preparation, strike and wind-down recovery phases (anatomical part: f2,45 = 28.89, p<0.001; phase: f2,45 = 20.99, p<0.001; part*phase interaction: f4,45 = 8.01, p<0.001) for all events (
Although SCUBA divers observed sideways tail-slaps
Sideways tail-slaps were always preceded by a successful overhead tail-slap and took place during an ongoing prey item collection phase of a thresher shark's predation attempt. However, not all overhead tail-slaps were followed by sideways tail-slaps. Preparation, strike and wind-down recovery phases could not be delineated from the video records, even though they were observed
During sideways tail-slaps, there was little vertical movement of the snout, caudal peduncle or tail. First a thresher shark positioned itself alongside the schooling prey fish. Then the pectoral fins were adducted and a strike was initiated by flexing the trunk and caudal peduncle laterally. The tail was then whipped laterally to one side of a thresher shark's body. The trajectory of the tip of the tail followed a horizontal path, which terminated in line with the first dorsal fin. The snout, caudal peduncle and tail were then returned to their original condition (
A motion animation (top) represents 4.86 s−1 of an event which was recorded by handheld underwater video camera on 14 June, 2010. Center inserts profile the key characteristics of the behaviour, while inserts shown in the transvers plane (bottom), were interpreted from
While it has long been suspected that thresher sharks hunt with their tails, little was previously known about the behaviour in the wild. This study represents the first attempt to quantify the kinematic patterns associated with alopiid predatory behaviours in their natural environment, and implies that adaptive foraging techniques play an important role in the hunting strategies of large marine predators.
Thresher sharks are physiologically adapted for thermo-tolerance and demonstrate distinct crepuscular vertical migrations
Anecdotal evidence suggests that pelagic thresher sharks circumnavigate the surface waters of the Philippines at all times of the day. The sharks regularly visit cleaning stations by day
During overhead tail-slaps, preparations were significantly longer in duration than strike and wind-down recovery phases. The lunge acceleration, which characterises a preparation phase, may position a thresher shark within reach of the sardines it is preying upon to facilitate stunning them
After the wind-down recovery phase, almost all of the observed thresher sharks turned 180°, presumably to search for and collect dead and/or stunned sardines. One third of the overhead tail-slaps resulted in prey items being collected, and when successful, a thresher shark consumed more than one sardine. Carnivorous oceanic sharks generally pursue one prey item at a time
While the manoeuverability of small schooling fish can make individuals elusive to predators chasing them individually
Bubbles were observed to form at the apex of an overhead tail-slap's trajectory during most of the observed successful thresher shark hunting events (
While the sample size of the hunting events that met the selection standards for analysis was too small to quantify stereotypy, tail-slaps were remarkably invariable among trials for different sharks. To launch a strike, a thresher shark first adducted its pectoral fins, a manoeuvre that changed its pitch promoting its posterior region to lift rapidly (
Immediately after adducting its pectoral fins, a thresher shark rotated them laterally to counter the momentum of its posterior region from precipitating forward. Wilga and Lauder (2000, 2001) described the vertical movements of leopard (
The analyses of the video records showed that the dorsal margin of a thresher shark's caudal fin, which is flexible and narrows to the tip
Although a tail-slap's rotational speed was invariable among different sharks, the mean and maximum speeds of a tail-slap were directly related to a thresher shark's size. In physics, the ‘kinetic link’ principle implies that during a catapult launch motion, energy and angular momentum are transferred from one body segment to another, in a sequential manner, all the way to the distal segment
From their investigations of common thresher sharks Aalbers et al. (2010) described a second predominant strike behaviour during which a shark “positioned itself in close proximity and parallel to the prey item before initiating a lateral strike of the dorsal lobe”. Yet at Pescador Island, pelagic thresher sharks' sideways tail-slaps were rare, only occurring after successful overhead tail-slaps, when sardines had already been stunned (
During sideways tail-slaps, thresher sharks were observed to target single prey fish, which swam erratically and were not elusive, presumably because they had been previously stunned and/or injured. It is possible that sideways tail-slaps are a specialist technique to debilitate maimed prey further for consumption.
There were occasions during
The evidence is now clear; thresher sharks really do hunt with their tails. Tail-slaps comprise four distinct phases that sequentially function to windup, strike, and recover the tail, and if successful, collect stunned prey items. Tail-slapping is an efficient strategy for hunting schooling prey since thresher sharks are able to consume more than one prey item at a time. Larger thresher sharks tail-slap faster than smaller ones because their tails are longer. A thresher shark's pectoral and caudal fins appear to have evolved, at least in part, to deploy tail-slaps. Analyses of the peduncular, radialis and axial musculature, which are likely to be recruited for the execution of a tail-slap, as well as the changes in kinetic energy of the fins and body relative to a thresher shark's centre of mass, would help to elucidate the motor patterns driving the kinematics of this unusual hunting behaviour in future studies.
(M4V)
(M4V)
(M4V)
(M4V)
We are very grateful to James Monnington, Gary Avenido, Junior Regalado Deniega, Hermann Pauli, Lorenzo Schweizer and the Blue Abyss Dive Shop for their field and technical support. We thank Jan Acosta, Jaime Sabanate, and Tony Exall for contributing material, Peter Klimley, Nigel Hussey and Michel Kaiser for feedback, Alison Beckett, Gary Cases, and Ian Prior for their general support and guidance.