Archerfish coordinate fin maneuvers with their shots

Archerfish use water jets to down aerial prey from considerable distances (e.g. Lüling, 1963; Dill, 1977; Elshoud and Koomen, 1985; Rischawy et al., 2015; Schuster, 2018). The forces transferred to targets, the mass of water ejected and the speed of initial water release have been measured by Schlegel et al. (2006) for prey of various sizes, and the major result was that the fish adjust the force transfer to prey size so as to keep a considerable safety margin above the maximal attachment prey of that size can attain (Schlegel et al., 2006). Height-dependent changes in force transfer were reported by Burnette and Ross (2015), but even at large height a considerable safety margin was found. Vailati et al. (2012) showed that focusing of water at the jet tip can arise as a consequence of speed changes that occur because water is released from rest, so that water that leaves first is slower. A surprising later discovery was that the time at which this focus occurs is not fixed, but is adjusted to target distance, so that the later parts of the jet catch up with the front right before impact, maximizing force transfer (Gerullis and Schuster, 2014). Using fish that were trained to fire from fixed positions, it was possible to see that the distance-dependent adjustments made in the jets were linked to precisely controlled mouth-opening and -closing maneuvers during jet formation. Similar maneuvers have also been seen when archerfish use underwater ‘jets’ to uncover hidden prey from various substrates (Dewenter et al., 2017). During the analysis of high-speed videos of these trained fish (Gerullis and Schuster, 2014), we noted characteristic movement patterns of the fins of the shooter that appeared to be linked to shooting. Surprisingly, these movements appeared to set in just at a time when they would no longer seem to be required: they started just when the shooters had assumed and subsequently kept their stable positions during their shots. At this time, the recoil forces from the emerging jet start to act on the fish, and so a synchronized onset of fin movement would mean that the forces produced by these maneuvers might be needed to compensate recoil. We therefore analyzed the previous recordings with a focus specifically on the timing of the fin movements relative to jet release. For these recordings, the changes in the water jets with target height, as well as the stability of the shooter during the shot, had already been characterized (Gerullis and Schuster, 2014). After characterizing timing of the fin maneuvers in two trained fish, we next explored whether these findings extend to freely hunting archerfish of two species, Toxotes jaculatrix and Toxotes chatareus.

Over the past two decades, remarkable insight has been gained into the ways in which teleost fish operate smaller fins, such as their pectoral and pelvic fins, for braking maneuvers and for steering at high speed. By combining detailed three-dimensional kinematic analyses (e.g. Gibb et al., 1994; Lauder and Jayne, 1996; Drucker and Lauder, 1999; Standen and Lauder, 2005; Tytell, 2006; Lauder and Madden, 2007; Standen, 2008; Tytell et al., 2008; Flammang and Lauder, 2009), detailed measurement of the induced flow fields and force reconstruction (e.g. Drucker and Lauder, 2001a,b; Lauder and Drucker, 2002; Drucker and Lauder, 2003, 2005; Lauder and Madden, 2006; Peng et al., 2007; Han et al., 2020), as well as by critical verification with robotic models (Lauder et al., 2007; Lauder and Tangorra, 2015; Flammang et al., 2017; Gravish and Lauder, 2018; Mignano et al., 2019; Zhu et al., 2019), it has become possible to see that even minute details in the movement of these fins can play a role in generating precisely controlled forces at the right time. Particularly interesting are maneuvers with specific and demanding constraints; for instance, during prey capture it is essential that the attacking fish arrives at the right place at the right time and in an orientation that allows it to actually swallow its prey. Beautiful studies on this aspect include those of Standen and Lauder (2005), Higham et al. (2005) and Higham (2007). For a shooting archerfish, the situation appears to be far less dramatic than for a predator braking from a full-speed chase: when an archerfish releases its water jet, it has already moved to a favorable position and has oriented appropriately (e.g. Lüling, 1958; Dill, 1977; Timmermans, 2001; Gerullis and Schuster, 2014), and all that seems now required is to remain stable while the jet forms. This could be achieved simply by keeping the fins spread during the shot (as suggested e.g. by Lüling, 1958, 1963; Bekoff and Dorr, 1976), which might suffice to prevent any roll, pitch or yaw that might result from the recoil of jet release. Here, we demonstrate that, surprisingly, characteristic rapid fin-movement patterns are synchronized with shooting that even show similarities to those seen in braking fish (e.g. Higham et al., 2005; Higham, 2007).

The initial analyses used recordings from two archerfish, Toxotes jaculatrix (Pallas 1767), that had been trained to reliably score perfect hits over a wide range of distances from a fixed position and under heavy illumination so that various aspects of jet production could be monitored closely using digital high-speed video (Gerullis and Schuster, 2014). These recordings were particularly valuable, because a number of important issues had already been analyzed: (1) all shots at all target height levels were perfect hits, (2) the position and orientation of the shooter were stable during jet formation, and (3) the adjustments made in the jets and in the mouth-opening and -closing maneuvers for changing target height had already been characterized. The recordings allowed us to analyze temporal aspects of fin activation of the shooter, specifically, how tightly any fin movement was associated with shooting, and whether shooting-related fin movement patterns might correlate with changes in the characteristics of the jets. Recordings were taken from the two most successful fish in the earlier study (Gerullis and Schuster, 2014). These were 13–14 cm in length (snout to caudal peduncle) and were kept in a tank (1.2×0.5×0.5 m, length×width×height) filled to 25.5 cm with brackish water (temperature, 26±1°C; conductivity, 3.6–3.8 mS cm⁻¹; 12 h:12 h light:dark regime). To shoot, the fish had to swim into a channel (Fig. 1A) formed by suitable walls and covers above the water surface [polyvinyl chloride (PVC)] so that the fish could not jump or shoot successfully from positions outside the channel. Targets were small black spheres (polyethylene terephthalate; diameter, 5 mm, mass, 0.065 g) hung on filaments 20, 40 or 60 cm above the water surface. For each successful hit, the shooter was rewarded with a dead fly (Calliphora sp.).

A total of 28 additional fish (length, 8–17 cm) of two species, T.jaculatrix (eight individuals) and Toxotes chatareus (Hamilton 1822) (20 individuals), of eight groups were evaluated that were filmed mostly (in some instances from below and from the side) from above at 500 frames s^–1 (see Movie 3). The groups were housed in tanks sized 1.0×1.0×0.6 m and 1.3×1.3×0.6 m as described earlier (e.g. Reinel and Schuster, 2018), under the same conditions as the two spatially trained fish.

A Fastcam PCI R2 (Photron, San Diego, CA, USA; resolution, 480×512 pixels; Sony TV lens 1:1.8 f=16 mm) was used to record the fish from above and a Fastcam APX RS (Photron; resolution, 1024×1024 pixels; Micro-Nikkor-P.C. Auto 1:3.5 f=55 mm; Nikon GmbH, Düsseldorf, Germany) recorded the shooter from the side (see Movies 1 and 2). Sufficient illumination came from a 1000 W spotlight, placed behind a diffuser plate below the tank (Fig. 1A). The two cameras were synchronized and operated at 250 frames s^–1. A small mirror (7.3×4.7 cm) placed in the channel at an angle of 45 deg to the shooter's length axis allowed the monitoring of the hidden fins in the side views (Fig. 1A). The horizontal component of the pectoral fin angle was taken from dorsal views (resolution, 0.3 mm pixel⁻¹; Fig. 1B). An additional vertical component was analyzed using the synchronized side view (resolution, 0.2 mm pixel⁻¹; Fig. 1C). The movement of the pelvic fins was monitored from side views. Recordings were evaluated using ImageJ (version 1.41m). To monitor the timing of the fast-forward flap of the pectoral fin from the top views, we tracked the front end (leading edge) of the pectoral fins frame by frame for all frames from 10 frames before to 20 frames after the shot, unless stated otherwise. In each frame, three positions were taken that were easy to find and that led to a robust estimate of fin angle: (1) the approximate midpoint (b,b′ in Fig. 1B) on the front of the fin, (2) the base (a,a′ in Fig. 1B) of the fin, and (3) a point on the body surface located ∼1 cm caudally from the base of the pectoral fin (c,c′ in Fig. 1B). Using these points, we derived an effective angle (α,α′ in Fig. 1B) of the leading edge of the archerfish's pectoral fin. To estimate angular velocity, we simply considered the differences in angle from one frame to the next (i.e. in 4 ms), assigning it to the time of the later of the two frames. Because of the large distance from the lens (top view, 60 cm; side view, 75 cm), small vertical excursions could safely be ignored in the evaluation of the recordings from above. However, to get an impression of the relative coordination of the pectoral and pelvic fins we did also measure, as far as possible in the side views, two spots on the pectoral and the pelvic fins (d,e in Fig. 1C).

Here, all recording was at 500 frames s^–1. In the analyses presented, recording was as described in Reinel and Schuster (2014) and involved a HotShot 1280M (NAC Image Technology, Simi Valley, CA, USA) that monitored the complete tank from above. To optimize contrast, four halogen lamps (200 W each) illuminated the tank's bottom through a diffuser plate (plexiglass, polymethyl methacrylate; transmission, 45%). Spatial resolution was 1.37 mm pixel^–1 (group of T.chatareus), 1.43 mm pixel^–1 (group of T.jaculatrix) and 1.48 mm pixel^–1 (recordings with three different target height levels). Wetted prey (dead Calliphora sp.) was stuck to the lower side of transparent disks, at fixed height above the water surface. For details of the arrangement, see Reinel and Schuster (2018). Qualitative aspects of any of the maneuvers could be seen in each of the recordings. In most, however, only parts of the complete maneuvers were clearly visible. To select maneuvers that were suitable for a quantitative analysis, we scrutinized a total of 5770 recordings and took those that fulfilled each of the following three criteria: (1) the onset of the shot could clearly be determined, (2) both pectoral fins were clearly visible throughout, and (3) recordings started early enough before the shot so that we could see the alternating beating of the pectoral fins of the later shooter. Additionally, two to 10 shooting events per individual were randomly selected and scrutinized for shooting-associated fin movement in eight more individuals (T.chatareus). Two of these were housed in a tank sized 1.2×0.5×0.5 m and filmed from the side, but with the same camera as described above. Six others were also T. chatareus and were members of three different groups. The other groups were kept in tanks sized 1.0×1.0×0.6 m (two groups) and 1.3×1.3×0.6 m and were filmed from below (also at 500 frames s^–1) using Photron Fastcam MC2 high-speed cameras (resolution, 512×512 pixel; Lensation 1/1.8″ F1.6/4.4–11 mm).

OriginPro (version 7.5) was used to estimate slopes and offsets of correlations between fin action of the left and right pectoral fin. All other tests (Kruskal–Wallis one-way ANOVA on Ranks and pairwise multiple comparisons using Dunn's Method; one-sample t-test, one-sample signed rank test and χ² test) were run using SigmaPlot (version 11.0) or custom written software (rank-based Brown–Forsythe and Fisher z-transformation).

In none of the 278 recordings made of the two trained (and pre-characterized; Gerullis and Schuster, 2014) fish were the pectoral fins already spread before the onset of the jet; also in none was their position stable during the release of the jet. Before the onset of the shot, the pectoral fins were beating slowly and asynchronously (as will be detailed below). However, slightly before the onset of the jet, we always noticed a characteristic rapid forward flapping of both pectoral fins (see example in Fig. 2A and Movies 1 and 2) that quickly gained speed, so that the horizontal angular speed was ∼1000 deg s⁻¹ by the time jet release was started (means±s.e.m., fish 1: 1256.0±51.0 deg s⁻¹, N=52; fish 2: 1028.5±153.6 deg s⁻¹, N=34). Our first objective was to examine the degree of variation in the timing of the forward flap across different shots; specifically, how well its timing was synchronized with the onset of water release. For this, we aligned each individual time course of the pectoral fin angle to the onset of the jet, i.e. we always set time to zero when the jet was first visible. After aligning a total of N=52 (fish 1) and N=34 (fish 2) courses of pectoral fin angle (recorded during successful shots fired at a target at 40 cm height), we averaged the individual courses. Should the onset of the fin flap not be synchronized to the onset of the jet or should the fin flaps vary from one shot to the next, then considerable scatter around the mean angle should occur at each point in time. In addition, our rather crude way of simply following the approximate midpoint of the leading edge of the pectoral fin (Fig. 1B) could also introduce errors that would increase the degree of scatter. On these notes, it is remarkable how little scatter actually did occur when we averaged the individually aligned time courses of horizontal pectoral fin angle (Fig. 2B) and of angular speed (Fig. 2C). This analysis already suggests that the forward flap is rather precisely coordinated with the onset of the water jet. In the trained fish, this allows averaging times relative to the onset of the shots from several recordings, yielding a considerable temporal resolution for the means. The forward flap appears to follow a robust time course in which it reaches a maximal horizontal angular speed slightly after the onset of the jet. The rapid forward movement is kept during approximately the first third of the water release, when shooters are stable (Gerullis and Schuster, 2014), but speed declines after this time and, by the time the jet is fully released, the fins reverse direction to then move toward the shooter's body.

The robust time course and alignment of the forward flap allowed us to explore whether it was kept or was adjusted when the fish changed their water jets to engage targets at different height h: the higher up the target, the longer is the interval of water released, the longer the lifetime of the jet and the later does water focus at the jet tip (Gerullis and Schuster, 2014). Jets aimed at different target distance could therefore produce different disturbances to the shooter's position and compensating them might require different fin activation. To examine this, we again determined jet-onset-aligned mean horizontal angles for a given individual (as in Fig. 2) but now also for a lower (h=20 cm) and for a larger (h=60 cm) target height (Fig. 3A,B). This analysis shows that the forward flap varies slightly but systematically with target height. Maximal horizontal angular speed was reached later when the fish engaged targets at larger height [Fig. 3C; means±s.e.m., 3.8±0.8 ms (20 cm; N=51); 8.9±0.8 ms (40 cm; N=52); 8.1±0.7 ms (60 cm; N=66); Kruskal–Wallis: P<0.001]. Also, the absolute value of maximum angular speed was larger for more distant targets [1853.1±40.9 deg s⁻¹ (20 cm; N=51); 1932.6±37.8 deg s⁻¹ (40 cm; N=52); 2017.3±25.9 deg s⁻¹ (60 cm; N=66); Kruskal–Wallis: P<0.001]. The maximal angular excursion of the pectoral fins was not significantly different [87.2±1.5 deg (20 cm; N=51); 85.7±1.1 deg (40 cm; N=52); 87.4±0.7 deg (60 cm; N=66); Kruskal–Wallis: P=0.110], but the maximal angle was reached earlier for more distant targets [Fig. 3E; 11.4±1.3 ms after the end of the shot (20 cm; N=51); −4.0±1.7 ms (40 cm; N=52); −7.8±1.5 ms (60 cm; N=66); Kruskal–Wallis: P<0.001]. For an estimate of the ‘start’ time of the forward flap, we used the time when both pectorals were moving in synchrony and at large angular forward speed (>500 deg s⁻¹). Using this estimate, we found that onset was significantly later when shots were fired at targets at larger height [Fig. 3D; −13.4±0.8 ms before the onset of the jet (20 cm; N=51); −3.3±0.8 ms (40 cm; N=52); −2.9±0.7 ms (60 cm; N=66); Kruskal–Wallis: P<0.001]. Also, the reversal of the synchronized pectoral fin movement occurred earlier for shots fired at more distant targets [Fig. 3F; 25.7±1.2 ms after the end of the jet (20 cm; N=51); 7.9±1.2 ms (40 cm; N=52); 2.9±0.9 ms (60 cm; N=66); Kruskal–Wallis: P<0.001]. Note that the precise coordination of fin maneuvers with jet release and the resulting small standard deviations (e.g. <6 ms in Fig. 3D) allowed mean timing to be determined quite accurately (see the s.e.m. values reported above).

The forward flap sets in after a phase in which the two pectoral fins were beating slowly and in opposing directions. This is illustrated by the example in Fig. 4A in which the fins had been monitored for much longer than in the majority of our recordings. The alternating low-angle beating can clearly be seen, as well as the switch to synchronous movement when the two fins are at similar angles. The example also illustrates that pectoral fin beating is variable during the pre-shooting phase. However, this changes dramatically just before shooting, when the forward flap sets in. Here, the two pectoral fins appear to move in synchrony, achieving similar angles at each point in time. Fig. 4A also illustrates another remarkable aspect: the forward flap sets in when the pectoral fins had arrived at similar angles and levels of forward speed. The example illustrates aspects that were seen in each of the 278 recordings analyzed in this study in the two specifically trained sharpshooters (fish 1: N=169; fish 2: N=109). Specifically, there were no recordings in which (1) only one pectoral fin was moving, (2) the two fins moved in the same direction before the forward flap, and (3) the forward flap started when the two pectorals had largely different angles. We next examined the actual degree of synchrony between the two pectoral fins during the forward flap. For a quantitative measure of synchrony, we plotted, at each instant, the angular speed of the right pectoral fin against that of the left fin at the same point in time. Fig. 4B reports this for the complete interval from 40 ms before onset until the end of water release. Correlations between angular speed of the left and right pectoral fin were strong at all target height levels (always R²>0.9; P<0.001; correlation for line through origin), confirming our impression of synchrony in the forward flap.

In contrast to the pectoral fins, the pelvic fins were already fully spread before the shot and remained so until the start of the jet. This was observed in each of the 278 recordings with at least one pelvic fin clearly visible for at least 40 ms before onset of the shot. The recordings in which the pelvic fin was monitored 600 ms (mean±s.e.m. 635.6±14.4, N=10) before onset of the shot confirm this and show that the pelvic fins are spread long before the onset of the shot. The precise pattern of movement of the pelvic fins was much more difficult to assess from our recordings than that of the pectorals. However, it appears to be simply an adduction that starts with the release of the jet and that is not finished by the end of the shot (e.g. Movies 1 and 2). Synchrony between the two pelvic fins could be assessed in 236 of our 278 side-view recordings in which the movement of both pelvic fins could be clearly observed (i.e. in which the fin on the right side of the body could be seen in the mirror; Fig. 1A); only in 20 of these 236 recordings did one of the two fins start slightly later (one to maximally three frames, i.e. 4–12 ms) than the other. A simultaneous start of the two fins was clearly seen in 216 of the 236 recordings. The recordings sufficed to test for systematic changes in the timing of the adductive movement when the fish adjusted their jets for target distance. Onset of pelvic adduction started approximately at the onset of the shot, with no robust trend for the onset of motion to occur later when target distance was increased [Fig. 5A; Kruskal–Wallis; fish 1: P=0.004; fish 2: P=0.107; means±s.e.m., fish 1 (fish 2): −11.14±1.06 ms before the onset of the shot, N=51 at h=20 cm (0.1±1.7 ms, N=40); −6.8±1.1 ms, N=52 at h=40 cm (4.0±2.4 ms, N=34); −6.6±0.9 ms, N=66 at h=60 cm (4.9±2.1 ms, N=35)]. At all target distances, the movement of both pelvic fins continued throughout the shot. The pelvic fins stopped moving earlier after shots fired at more distant targets [Fig. 5B; Kruskal–Wallis; fish 1: P<0.001; fish 2: P<0.001; means±s.e.m., fish 1 (fish 2): 101.1±4.8 ms after the end of the shot, N=40 at h=20 cm (90.4±6.8, N=40); 60.6±3.3 ms, N=52 at h=40 cm (78.2±5.0, N=34); 58.5±1.9 ms, N=66 at h=60 cm (55.9±4.0, N=35)].

We next attempted to get a rough idea of the actual relative movement of the pectoral and pelvic fins. So far, we had only been looking at planar projections, which provided sufficient information to discover the temporal aspects reported this far. However, a closer analysis of the available recordings gave a good first impression of the spatial movement pattern. For this, we simply followed the trajectory of the midpoint on the trailing edge of the left-side pectoral and pelvic fins in a plane parallel to that of the shooter. In these analyses, we first aligned each individual recording to the onset of the shot and then calculated the mean positions measured at each point in time. Fig. 6 shows the mean position (±s.e.m.) every 4 ms after the onset of each shot for the pectoral (Fig. 6A) and pelvic (Fig. 6B) fins during shots fired by fish 1 at a target at 40 cm height. Given our crude way of obtaining these trajectories, it is remarkable how small the respective errors were. This suggests that the vertical component is synchronized to the shot and that it also follows a robust time course, at least for fixed target height.

We then tested for any systematic variation in the vertical component when target height was changed. The results of this analysis are shown in Fig. 6C and D, in which the trajectories obtained for target height 40 cm are shown as reference (gray shaded areas) and the means±s.e.m. for larger (upper graphs) and lower (lower graphs) target height are shown for the pectoral and the pelvic fins. The changes were clearly larger than the variations at any given target height, therefore suggesting that the fish adjust aspects of the relative coordination of their shooting-related pelvic and pectoral fin maneuvers.

In all 278 recordings, the anal fin was clearly visible and was always erected before the shot, comparable to the pelvic fins. However, in contrast to the pelvic fin, the anal fin appeared to remain fully erected throughout the shots. In 202 recordings, the spiny anterior part of the dorsal fin was clearly visible (in the side views). In all these recordings, it was erected before the shot and appeared to remain erected throughout the shot. The soft posterior part of the dorsal fin was visible in the side views of 185 recordings and in each of them started to move during the shot. Although our recordings did not allow us to characterize its apparently rather complex movement, they sufficed to show that it started at the onset of the shot with no apparent target height-dependent change in onset timing [means±s.e.m., 0.0±4.7 ms (N=24 at h=20 cm height), 3.3±1.8 ms (N=46 at h=40 cm), 2.8±2.6 ms (N=34 at h=60 cm); Kruskal–Wallis: P=0.169]. The caudal and dorsal fins also appeared to be active during the shots: both fins could clearly be seen in 104 recordings (dorsal views). In all these recordings, the upper half of the caudal fin and the flexible posterior part of the dorsal fin both moved during the shot, always in opposing directions (e.g. Jayne et al., 1996 for data on sunfish). We also found a clear laterality in the movement of the dorsal and caudal fins. The flexible part of the dorsal fin moved significantly (P<0.001, χ² test) more often (in 75 of 104 recordings) to the left side, and in all these recordings the upper part of the caudal fin then moved to the right side.

We next examined to which extent the patterns monitored in the two highly trained fish that fired from a controlled position were typical for shooting archerfish. We surveyed 5820 additional digital high-speed video recordings made of a total of 28 archerfish, mostly during projects (Reinel and Schuster, 2014, 2018) that examined the predictive C-starts of the fish towards the later landing point of their prey (Movie 3). These fish were members of eight groups of archerfish that could hunt freely. All had extensive experience in downing prey from at least 30 cm height but were not especially trained for high hit rates. Most importantly, they were free to select the position from which they shot at prey. Three groups were filmed from above at 500 frames s^–1 and the movement of the pectoral fins could regularly be seen (see Materials and Methods). In eight additional fish (members of five further groups), 40 additional recordings were made from below (through the transparent bottom of the respective experimental tank) and 10 recordings from the side. The existence of the forward flap of the pectoral fins could be confirmed in all recordings. The pelvic fins could be seen clearly in at least 80 recordings. All these confirmed that the pelvic fins are extended before the shot and adducted during it. A number of recordings fulfilled all of the following criteria: the onset of both (1) the forward and (2) the subsequent backward flap could be determined accurately, (3) the prior transition from alternate beating of the pectoral fins to synchrony could be checked, and, most importantly, (4) the onset of the jet could be determined accurately. Fig. 7 presents the results of a quantitative analysis of such recordings in two groups of archerfish (Fig. 7A) that downed targets from 30 cm height. One group consisted of six T. jaculatrix, the other of six T. chatareus. For both species, the onset timing of the forward flap and of the backward flap (determined at target height 30 cm; Fig. 7C,D) lay in the ranges we had found in the trained fish at 20 cm and 40 cm (Fig. 3D,F), with no difference between the two species (Brown–Forsythe: P>0.084). Estimates of the durations of the forward (Fig. 7E) and backward (Fig. 7F) flap also showed consistent timing, with no difference between species (Brown–Forsythe: P>0.375). Another remarkable finding is that the stability of the shooter's mouth during jet release (Gerullis and Schuster, 2014) is also seen in the freely hunting archerfish (Fig. 7G), again with no apparent difference between the two species (Brown–Forsythe: P=0.490). The variation in shooting positions selected in the freely hunting fish is shown in Fig. 7H. The distributions of horizontal distances d from prey showed considerable scatter but did not differ between the two species (Brown–Forsythe: P=0.103), and the same conclusion holds for the shooting angle ε (Brown–Forsythe: P=0.150; not shown). The variations in shooting position allowed testing of whether the timing (Fig. 7I) and duration (Fig. 7J) of the forward flap depended on horizontal distance from the target, but no correlations could be found (linear regression: R²<0.046; F<3.826; P>0.054; no differences between both species, Fisher z-transformation: P>0.333). The same held for timing and duration of the backward flap (not shown; linear regression: R²<0.069; F<5.911; P>0.071; no differences between both species, Fisher z-transformation: P>0.423) and also for the respective correlations with shooting angle (not shown; timing and duration of the forward flap: linear regression: R²<0.046; F<3.803; P>0.055; no differences between both species, Fisher z-transformation: P>0.291; timing and duration of the backward flap: linear regression: R²<0.075; F<6.393; P>0.128; no differences between both species, Fisher z-transformation: P>0.450; no differences between correlations with distance or shooting angle, Fisher z-transformation: P>0.140).

We next tested whether we could also detect height-dependent changes in the temporal characteristics of the pectoral flap in fish that downed targets from one of three possible heights, and whether such changes would also be independent from the shooter's horizontal position from the target. An extensive analysis was made in a group that fired at targets at 35, 55 or 65 cm height (Fig. 8A). First, we analyzed the horizontal distances assumed by shooters. They varied (as in Fig. 7H), but with an interesting trend: the mean horizontal distance chosen systematically increased with target height (Kruskal–Wallis: P<0.001; Fig. 8B), whereas shooting angles were approximately equal (only difference lowest versus medium level, Dunn's method: P<0.05; Fig. 8C). Interestingly, shooting durations also increased systematically (Kruskal–Wallis: P<0.001; Fig. 8D) with target height, as we had described previously in specifically trained fish (Gerullis and Schuster, 2014). Furthermore, also in this group, the mouth position of the shooters was quite stable during jet release (Fig. 8E), particularly when they fired at targets at great height (change in position not significantly different from zero, one-sample signed rank test: P>0.475), whereas small changes away from the target occurred at the smallest target height (one-sample t-test: P<0.001). Start of the forward flap and of the backward flap were again tightly linked to the onset or end of the jet, respectively (Fig. 8F,G). Target-height-dependent changes could be seen at the onset of the backward flap (Kruskal–Wallis: P=0.007; Dunn's method: 65 cm: P<0.05). No changes in the time of onset of either the forward flap (Fig. 8H) or backward flap (Fig. 8I) versus the horizontal distance of shooter or versus the shooting angle (not shown) could be found at any height level [either horizontal distance or shooting angle; forward flap (onset and duration): linear regression: R²<0.044, F<3.903, P>0.051 (R independent of target height; Fisher z-transformation: P>0.131); backward flap (onset and duration): linear regression: R²<0.040, F<2.327, P>0.131 (R independent of target height; Fisher z-transformation: P>0.289)]. Durations of the forward flap (Fig. 8J) and backward flap (Fig. 8K) were comparable to the findings reported above, and significant changes in the duration of the backward flap with target height could be seen (Fig. 8K; Kruskal–Wallis: P=0.030; Dunn's method: 35 cm versus 65 cm, P<0.05). Finally, we checked whether we could detect any differences between shots that were perfect hits and shots that missed their target. However, the start of the forward flap, for instance, did not differ between hits and failures (Fig. 8L; Kruskal–Wallis: P=0.269) and the same held for duration of the forward flap, as well as onset and duration of the backward flap, along with changes in mouth positions during jet release [Kruskal–Wallis: P<0.002 but all pairwise comparisons (Dunn's method) P>0.05; data not shown].

In this study, we show that a number of fin maneuvers are tightly associated with shooting in archerfish. Most notable is the rapid and synchronized forward flap of the two pectoral fins that is precisely coordinated with the onset of the water jet. The fin maneuvers could conveniently be studied in fish that had been specifically trained to shoot from a fixed position. These fish could be filmed from two angles, and aspects of their jet production and stability had already been characterized, as these fish reliably scored perfect hits from a fixed position at targets at different distances (Gerullis and Schuster, 2014). It is striking that some of these fin maneuvers, most notably a rapid forward flap of the two pectoral fins, set in just when the shooter had already achieved its shooting position and perfectly maintained it. During this time, the beginning recoil from the onset of the jet starts to act, and so forces are needed to stabilize the shooter. The synchronized forward flap could equally well be seen in all freely moving 28 individuals of the two archerfish species – T. jaculatrix and T. chatareus – that we analyzed additionally. Moreover, the onset and duration of the forward flap and of the subsequent backward flap were similar across all fish. Aspects of the timing depended on target height (but not horizontal distance) (Figs 3, 5, 6 and 8) and could therefore potentially be needed to account for the systematic changes in the water jets (Gerullis and Schuster, 2014). Our findings suggest that coordinated and precisely timed fin maneuvers are an important aspect of the archerfish's unique ability to fire powerful long-distance water jets and that they contribute to stabilizing the shooter against recoil during the release of its water jet.

Using trained fish that reliably fired from a fixed spot made it convenient to monitor the timing of all fin maneuvers relative to the onset of the shots. The tight synchrony with jet onset made it possible to measure mean timings with great accuracy (despite a rather low frame rate used in the trained fish). Height-dependent changes occurred in the timing of the pectoral (Fig. 3) and pelvic (Figs 5 and 6) fins. During an extensive survey of 28 additional archerfish, filmed during their natural hunting behavior as they positioned themselves freely below their targets, we found that the rapid forward flap was seen in each shooting maneuver. The most noteworthy aspects found are as follows: (1) the mouth position of the shooters was stable during the shot (Fig. 7G and Fig. 8E), suggesting that this might indeed be important for all fish that manage to hit distant targets (Gerullis and Schuster, 2014); (2) the increase in the duration of jet release with increasing target height (Gerullis and Schuster, 2014) could also be confirmed (Fig. 8D); (3) the shooters fired from a horizontal distance from prey as described previously by many authors (e.g. Timmermans, 2001). By varying target height we show that the fish seem to favour a specific angular range but not a specifc distance. (4) The considerable variation around the preferred shooting position (or angle) allowed us to rule out that changes in the temporal characteristics of the forward and backward flap of the pectoral fins changed as a result of changed position relative to the target (Fig. 8H,I). Most importantly, however, the survey showed that the synchronized rapid forward flap and backward flap can be seen in every shooting maneuver of 28 freely moving archerfish of two species and, where results could be analyzed quantitatively, showed the basic characteristics observed in the two highly trained fish. On this basis (Figs 7 and 8), we conclude that the fin movements we describe here appear to be representative for shooting archerfish, at least for individuals that have learned to hit distant targets.

The rapid synchronized forward flap of the pectoral fins (and its being accompanied by pelvic fin action) is strongly reminiscent of the rapid forward movement of the pectoral fins seen in braking fish as they prepare for a catch or avoid colliding with an obstacle (e.g. Drucker and Lauder, 2002; Standen and Lauder, 2005; Lauder and Madden, 2006; Lauder et al., 2007; Higham et al., 2005; Higham, 2007; Kane and Higham, 2011; Higham et al., 2016). A large body of detailed evidence on braking maneuvers has been obtained in the sunfish, which, also being a perciform fish, shares with archerfish similar shape and location of the pectoral fins. However, in contrast to braking, the rapid forward flap of shooting archerfish sets in just when the shooter is already at a stable shooting position and then holds it. Because the rapid fin action thus does not cause any changes in the movement of the fish, the question obviously is which other forces balance their effect. Clearly, and based on the synchrony with onset of the jet, the recoil forces of the jet would be chief among them. Or, reversing the argument, the ‘braking-like’ maneuvers would seem useful in stabilizing the shooter against the recoil of its jet (∼100 mN for prey of the sizes presented here; Schlegel et al., 2006). The forces produced during the forward flap must produce sufficient lift to keep the shooter at the surface, which would clearly be possible (e.g. Breder, 1926; Harris, 1937; Webb, 1973; Blake, 1983; Walker and Westneat, 1997). None of our recordings allowed us to reconstruct the complexity of the changes in shape, orientation and rigidity of the fins that was clearly evident (e.g. see Movies 1 and 2); however, a number of observations could be made. During the forward flap, the shape and orientation of the pectoral fins underwent massive changes. In 236 of the 278 recordings, the plane of the pectoral fins could be seen both from the side and frontal views (i.e. also using the mirror that was placed in front of the shooter, see Fig. 1A). In all these events, the area of the pectoral fin was not constant. The images shown in Fig. 2A provide a good illustration: the visible area of the pectoral fins changes during the fin flap so that, at the end of the shot, its major area is actually orthogonal to the water surface. There is thus a clear rotatory component in the movement of the more trailing parts of the pectoral fins. Our finding that timing and duration of the pectoral flap did not change with shooting angle implies that these aspects (and not timing) might be important to provide the appropriate amount of lift required for the various shooting angles.

All fins are activated to prepare the shooter, some long before the jet actually starts and with different degrees of synchrony to jet release (Fig. 9). About half a second before onset of the jet, when the shooter takes up water, the anal fin, the soft parts of the dorsal fin and the spiny parts of the dorsal fin get erected (Fig. 9A). The exact timing and temporal order of these maneuvers varies considerably, particularly when compared with later maneuvers (Fig. 9B). Here, the pelvic fins first start to be spread, allowing then the precisely timed onsets of pelvic and pectoral fin maneuvers. The forward flap also involves preparation, i.e. the pectoral fins must change from an alternating pattern to synchrony (Fig. 4), allowing the forward flap to then set in. Hence, coordination of the pectoral fins with the shot must set in already before the shot starts.

All 6098 shooting maneuvers scrutinized in this study in a total of 30 archerfish showed the characteristic rapid shooting-associated forward flap of the pectoral fins. Wherever other fins could be seen during shooting, they followed the pattern that we had discovered in the two specifically trained fish. On this basis, we conclude that shooting-associated fin maneuvers are an integral aspect of shooting in archerfish and their timing relative to the expected recoil forces from the jet suggests that they are needed to keep the shooter stable.

We thank Martin Krause and Nora Voigt for sharing their unpublished videos (yielding additional fin movement patterns in eight individuals), Wolfram Schulze for his help during all stages of the project and our reviewers for many helpful comments.