BARBARA SCHMITZ* and JENS HERBERHOLZ /nstitut fur Zoologie, Technische Universitiit Munchen, Lichtenbergstr. 4, 85747 Garching, Germany *Corresponding author (Fax, 49 89 289/ 3674; Email, bschm@cip/.zoo.chemie.tu-muenchen.de).

During intraspecific agonistic encounters in snapping shrimp (Alpheus heterochaelis)the behaviour of the snapper,emitting a fast water jet by very rapid closure of the large modified snapperclaw, and the receiver was analysed by single frame video analysis before, during, and after the snap. During snapping the opponents usually face each other. Snapping is most frequently preceded by touch of frontal appendages.The snapping animal keeps its snapper claw slightly across the midline, shielding frontal body parts, and its tailfan bent downwards. The mean claw cocking duration (generating muscle tension) before snappingamounts to about 500 ms. In 58% of the snaps, the snapper claw pointed at the opponent, its claws, densely covered with sensory hairs, representingthe main target of the water jet. The mean distance for these directed snaps was 0.9 cm, while undirected snaps were emitted from larger distancesof on average 3.4 cm. The snapper usually withdraws immediately after snapping, the receiver approaches.Initial snaps are often answered by return snaps and both are emitted from smaller distances and hit more often than subsequentsnaps.

1. Introduction (Williams 1984; Gruner 1993). The animals used in this study originate from the Gulf of Mexico in northern Agonistic encounters in competition for space, shelter, Florida, where they live either individually or as and accessto mates or food are very common in crusta- intersexualpairs in the littoral zone within oyster banks. ceans and have been studied extensively in different They show a large, modified snapper claw on one (left orders and families (Dingle 1983; Hyatt 1983). Research or right) side, claw length reaching nearly half the body on agonistic behaviour in crustaceanshas been conducted length, and a small pincer claw on the other side in on mantis shrimp (Stomatopoda; Caldwell and Dingle both sexes.The morphological differences between sexes 1975; Caldwell 1987)and on severalfamilies of Decapoda include the broader pincer claw of the male with fringes including fresh water prawns (Palaemonidae;Barki et at of plumoserrate setae on dactyl (movable finger) and 1991), lobsters (Homaridae; Huber and Kravitz 1995), propus (Read and Govind 1991), and the slightly broader crayfish (Astacidae, Cambaridae and Parastacidae;Copp abdomenof the female with larger pleurites (Nolan and 1986; Soderback 1991; Pavey and Fielder 1996), and Salmon 1970). hermit crabs (Paguridae;Elwood and Neil 1992) to name In A. heterochaelis the dactyl of the snapper claw only a few reviews and recent studies. Snapping shrimp possessesa huge stopper-like tooth (the plunger), which differ from these crustaceansin the ability of producing fits into a socket in the propus (Brooks and Herrick a rapid water jet, which is used in interactions with 1891). Prior to snapping,the dactyl is cocked in a 1000 conspecifics and prey. position by cocontraction of a claw opener and closer Atpheus heterochaetis,the big-clawed snapping shrimp muscle, while the closer apodem is lifted over a pivot of the family Alpheidae (Decapoda, Caridea), which point, so that tension is generated until a second closer comprises about 425 species world-wide in subtropical muscle contracts (Ritzmann 1974). During the following and tropical oceans, reaches 5.5 cm in body length extremelyrapid closure of the snapperclaw (within about

Keywords. Snapping shrimp; Alpheus heterochaelis;behaviour; agonistic encounter; water jet

J. Biosci., 23, No.5, December 1998, pp 623-632. @ Indian Academy of Sciences 623 624 Barbara Schmitz and lens Herberho/z

750 ~s) a short, very intense sound is produced (Knowltonrostrum tip to telson) were used in our behavioural and Moulton 1963; Schmitz et al 1995) when both claw experiments. They were caught among other snapping surfaces hit each other. Furthermore, as first pointed out shrimp at the gulf coast of Florida in Panacea and at by Volz (1938) for different snapping shrimp species, a the Florida State University Marine Laboratory in St. rapid jet of water is formed, when the dactyl plunger Theresa (latitude: 29°55'N, longitude: 84°30'W) about 9 is driven into the propus socket, displacing water which months before the tests. Prior to the experiments the escapes through a narrow anterior groove. In A. hetero- animals were labelled witt. small bee queen numbers chaelis Schein (1975) evaluated water jet velocities of and kept individually for on average 53 days in perforated 1.5-4 m/s and a water jet width of 20-700 at 0.5-3 cm plastic containers (1 I x 11 x 15 cm, containing gravel distance from the tip of the snapper claw in a small and oyster shells for shelter) within a large tank glass petri dish using 100 frames/s film analysis. In (90 x 195 x 33 cm) with 330 I of seawater (salinity: 23- accordance, for A. heterochaelis snapping shrimp tethered28%0, temperature: 22-23°C). The water was permanently in an aquarium, a well focussed main flow of water filtered and proteins were removed; pH, carbonate, oxy- with a maximum velocity of 3-10 m/s in extension of gen, CO2, and NO3 were regularly controlled. The shrimp the long axis of the snapper claw was shown using high were exposed to a light/dark cycle of 12/12 h and fed speed video analysis (up to 4000 frames/s), flow visua- frozen shrimp,. fish, mussel meat, or Artemia salina three lization, and laser Doppler anemometry (B Schmitz, times a week. unpublished data). Experiments were conducted within a 25 x 15 x 16 cm Snapping can be used in offensive actions and was aquarium (water level: 8 cm, temperature: 22°C, floor: observed in successful attempts to stun or even kill covered with black cloth to facilitate walking) on a small prey such as worms and goby fish (MacGinitie vibration isolated platform (cf., Breithaupt et al 1995). 1937; MacGinitie and MacGinitie 1949; Suzuki 1986), The short sides of the aquarium were covered with shrimp (Downer 1989), and small mud flat crabs (Schultz transparent paper and two Schott KL 50 light guides on et al 1998), even though Hazlett (1962; Hazlett and each side were used to enhance contrast for video Winn 1962) concluded from his studies that snapping recordings (average intensity: 50 lux). Two animals of is not associated with the procurement of food. In the same or different sex were kept in the aquarium for addition, snapping may serve as a warning signal to 10 min for acclimatization, isolated by an opaque partition defend a shelter or a territory against (usually conspecific) in order to prevent visual and tactile contact. After intruders. Intraspecific agonistic interactions in A. hetero- removal of. the partition, all interactions between the chaelis were previously studied by Nolan and Salmon shrimp were videotaped (camera: Panasonic AG 455, (1970), Schein (1977), Conover and Miller (1978), and videorecorder: Panasonic AG 7355, monitor: Sony Tri- Hughes (1996a). Nolan and Salmon (1970) thoroughly nitron) for 20 min from above and from the side via a studied aggressive, submissive, and other acts in encoun- mirror angled in 45° from the base of the aquarium. ters of two snapping shrimp, established general sequencesThe shrimp moved freely about the interior of the of behaviours, and found e.g., that intrasexual interactions aquarium rather than solely along its edges, and often show more aggressive and less con tactual elements thanencountered each other in the middle of the aquarium. intersexual ones. In accordance with these authors, ScheinA snapping shrimp met the same conspecific only once (1977) found that larger animals usually win, that the and was usually tested at most once a day, on average eventually dominant shrimp displays more aggressive 4 experiments being conducted over a period of on acts (e.g., snaps) and also that the first animal to snap average 13 days per individual. In consecutive experi- is more likely to win in night encounters. Chemical ments with the same animal the number of snaps usually communication in A. heterochaelis was studied by Schein fluctuated and less often either increased or decreased (1975) and Hughes (1996b), and the use of the cocked continuously, i.e., multiple testing did not affect the snapper claw as a visual signal was shown by Hughes ability or readiness to snap. Thirty-two experiments with (1996a). a total of 2 I 8 snaps were subsequently characterized The present paper provides a detailed, quantitative using single frame analysis (50 frames/s; temporal reso- analysis of the behaviours of both the snapper and the lution: 20 ms) and statistical analysis [using Statgraphics receiver before, during, and after the snap in order to Plus 6.0 and SPSS 6.0. I for conventional statistics and understand the role of tactile and hydrodynamic stimuli evaluating the mean vector direction, the angular deviation in these interactions. and the length r of the mean vector for circular statistics (Batschelet 1981)]. For conventional statistics means and standard deviations were caiculated for each variable of 2. Materials and methods interest for each individual and only one value per Sixteen adult snapping shrimp (A. heterochaelis. 6 males individual was included in each statistical test (Friedman, and 10 females) 2.5 to 5.1 cm in size (length from Wilcoxon, and Mann-Whitney-U tests). Agonistic encounters in snapping shrimp 625

For each animal, sex, body length, snapperclaw side, shrimp) on averageamounted to I.I9:t0.I5 (range: 1.01- snapperclaw length, width and thickness were evaluated 1.57, N = 32) and the size difference to 0.60:t 0.45 cm in the living shrimp. Each snapping interaction was (range: 0.05-1.60 cm, N = 32). Regarding all 32 experi- characterized by the latency between a preceding tactile interaction and the snap (disregarding durations of 5 s and more), the distance between the opponentsand the body alignment (head-head,head-tail, tail-head, tail-tail) during the snap, the target of the snap, and the behaviour of both animals after the snap as well as its latency. Additionally, the following angles were measured: .. (i) The body axis angle fJ betweenthe longitudinal body axes of the interacting animals (ranging from 0° to :t 180°, positive when ipsilateral to the snapper claw of the snapping animal and negative when contralateralto it; see figure lA). (ii) The position angle () between the longitudinal body axis of the snapperand the line connectingthe midpoints of both shrimp (range and sign definition as for the angle fJ; see figure lA). (iii) The snapperclaw angle y betweenthe snapperclaw long axis and the longitudinal body axis in the snapping animal (negative when crossing the animal's midline; figure IB). (iv) The tailfan angle with respect to the abdomen in the snapper (0° -stretched out, 90° -bent downwards, 180° -fully folded) during the snap. Furthermore, to compare freely moving and tethered B shrimp, seven animals (five males and two females,two of which had also participated in the first set of experi- ments), were fixed for at most 20 min to a vertical holder with a screw by means of a plastic nut glued to the carapace.These tethered animals were submergedin a similar aquarium for at most 20 min, standing on a small platform of mesh net, and were stimulated to produce snaps 'with a thin hair attached to a glass rod. Using the same video system as before, 167 snaps were recorded and analysed frame by frame. For snaps emitted during agonistic encountersas well as for snaps of tethered animals the cocking duration, i.e., the time between completely opening the snapper claw, cocking the dactyl in a 100° position, and the beginning of claw closure was evaluated using single frame analysis.

Figure 1. Definitions of the body axis angle {3, the position 3. Results angle d, and the snapper claw angle y in snapping shrimp interactions during snapping. (A) The body axis angle {3 is given by the angle betweenthe longitudinal body axes of both 3, General characteristics animals,ranging from 00 to:t 1800; it is positive when ipsilateral to the snapper claw of the snapper and negative when con- In interaction experiments, the two snapping shrimp (A. tralateral to it. The position angle d is given by the angle heterochaelis) usually met within a few minutes after between the longitudinal body axis of the snapperand the line removal of the opaque partition, the first snapon average connecting the midpoints of both shrimp (range and sign occurring 276:t 306 s (range: 19-1098 s, N = 32) after definition as for {3). (B) The snapper claw angle y is given by the angle betweenthe snapperclaw long axis and the longitudinal the onset of the test. The size ratio of both animals body axis in the snappinganimal; it is negative when crossing (size of the larger animal divided by that of the smaller the animal's midline. 626 Barbara Schmitz and lens Herberholz

ments, the snapping frequency of the larger animal body axis of the snapper and the line connecting the (N = 116) exceeded that of the smaller one (N = 102). midpoints of both animals on average amounts to figure However, in 14 experiments the larger and in another 2.4 ° :t 42.8° (N = 215, r = 0.721, circular statistics; 14 tests the smaller animal snapped more often, with 3B) and the absolute position angle I c} I on average an equal number of snaps in the remaining four experi- amounts to 29.1 °:t ~4.0° (r = 0.824, circular statistics). ments. The mean body size of the snapping animal Thus, in 73% of the cases (N = 157) the receiver is (3.7 :t 0.8 cm, N = 218) and the receiving shrimp positioned in the rostral quadrant with respect to the (3.6:t 0-6 cm, N = 218) did not differ significantly (Mann- snapper during the snap (cf., figure lA). As illustrated Whitney-U test: P> 0-1). in figure 3C, small body axis angles in combination with small position angles predominate. In addition,both 3.2 Behaviour before and during snapping shrimp usually face each other during the snap (h-h, head of snapper oriented towards head of receiver: In order to obtain first informations about the stimuli N = 16, n = 187, 93%), while only rarelydoes the snapper eliciting snapping, we analysed tactile interactions face the tail of the receiver (h-t) or is its tailfan closest between the opponents before snapping occurred and to the receiver (t-h, t-t; figure 3D). measured the interval between the last contact and the As illustrated in figure 4A, the snapper keeps its snap, i.e., the moment when the cocked snapper claw snapper claw parallel to its longitudinal body axis or closed rapidly. As illustrated in figure 2, in both the slightly crossing the midline (negative angles),the snapper snapper and the receiver frontal appendages (antennae, claw angle r on average amounting to -9.6°:t 15.4° snapper and/or pincer claw) were touchedmost frequently, (N = 203, r = 0.964, circular statistics). Interestingly,the while legs, carapace, and tailfan were less oftencontacted. tailfan is usually kept at about right angle downwards The interval between touch and snap (disregarding from the abdomen (circular mean: 73.6°:t35.8°, N = 205, intervals of 5 s or more, which were rare) on average r = 0.805, figure 4B), which stabilizes the stand of the amounts to 1.5:!: 0.6 s (N = 16, n = 169). snapping animal, while on the other hand preparing for The body axis angle p between the longitudinal body escape after the snap (see below). axes of both animals during the snap on averageamounts With respect to the distance between the opponents to 7.8°:!: 58.9° (N = 2 15, r = 0.471, circular statistics; during snapping two situations were distinguished: (i) figure 3A). Thus, the distribution is rather symmetrical directed snaps: the snapper claw of the snapper points with respect to 0°, though positive angles slightlyprevail at the receiver, i.e., the main flow of water in extension « 0°: N = 86, 0°: N = 20, > 0°: N = 109). The absolute of the long axis of the claw (see § 1) hits the receiver body axis angle I pi on average amountsto 44.8°:!: 47.4° and (ii) undirected snaps: the snapper claw does not (r = 0.658, circular statistics); i.e., the longitudinal body point at the receiver. This distinction is based on the axes of the interacting animals usually encompasssmall result of the snap and does not imply intention in angles. The position angle ~ between the longitudinal the snapping animal, thougt this presumably is the case

A B

A s SIP P L I T body part touched

Figure 2. Frequency histograms of body parts of the snapping animal (A, N = 16, n = 187) and of the receiving animal (B, N = 16, n = 188) touched by the opponentbefore snapping occurred. Grand meansof percentages across individuals and standard deviations are shown. A, antenna; S, snapperclaw; SIP, snapper and pincer claw;P, pincer claw; L, leg; C, carapace; T, tailfan.

c Agonistic encounters in snapping shrimp 627

for hits. For these directed snaps (N= 15, n= 117,58%) 3.3 Behaviour after snapping the distance between the tip of the snapper claw and the nearest body part of the receiver in extensionof the The behaviour after the snap differs remarkably between snapper claw long axis was evaluated and on average the snapping animal and the receiver: the snapper(figure amounts to 0.9:t0.7cm. For undirected snaps (N=14, 5A, open bars) usually retreats (i.e., moves backwards n = 100, 42%) the distance between snapper claw tip by walking at least one step; 47.5% of the cases) or and the nearest body part of receiver in any direction even tailflips (23.2%), and less often does not change was evaluated and on average amounts to 3.4:t 1.4cm. its behaviour (18.4%) or even approachesthe receiver Thus, directed snaps are more frequent than undirected (i.e., moves forwards by walking at least one step; snapsand occur at significantly smallerdistances between 10-9%). In contrast, the receiver (figure 5B, open bars) the opponents (Wilcoxon test; P < 0.01, N = 13). The rarely retreats (12-2% of the cases) or tailflips (10-7%), target of directed snaps in most cases is either the but usually does not change its behaviour (32-7%) or snapperclaw or the pincer claw (N = 16, n = 90, 79.8%), approaches the snapper (44-4%). In both the snapping which are densely covered with sensoryhairs (Read and and the receiving shrimp the behavioura1changes occur Govind 1991), while the area between the claws, the more frequently following directed snaps(snapper: 58.3% legs, the carapace, or the tailfan (summarized as B in retreat, 14.1% tailflip, 15.5% approach, filled bars in figure 4C) are hit less often. figure 5A; receiver: 13.2% retreat, 14.4% tailflip, 53.6%

A B

50 50 r

-90 0 90 180 -180 90 0 90 180 body axis angle (O) position angle (O)

c D

180T I.

Figure 3. Frequency histograms of the body axis angle fJ between the longitudinal body axes of the opponents (A, N = 215) and of the position angle" between the longitudinal body axis of the snapper and the line connecting the midpoints of both shrimp (B, N = 215) during snapping as well as scatter plot for both parameters (C). See figure 1 and text for definitions and details. (D) Frequency histogram of body alignments of both shrimp during the snap (N = 16, n = 215). Grand means of percentages across individuals and standard deviations are shown. h, head; t, tail; first letter for the snapping animal and second one for the receiver; i.e., h-t means that the head of the snapper is closest to the tail of the receiver.

~ 628 Barbara Schmitz and lens Herberholz

approach, filled bars in figure 5B) than following all snaps; in other words the animals less often do not A change their behaviour after directed snaps. For these directed snaps as well as for all snaps the percentages of the four behavioural response types in the snapper as well as in the receiver differ significantly (Friedman tests: P < 0.01 in each case). Especially, Wilcoxon tests showed that retreats in snappers and approaches in receivers occur significantly more often than all other response types (P < 0.01 in 5 cases and P < 0.05 in 1 case). The latency betweenthe snap and the behavioural responseon averageamounts to 33.6:!: 37.8 ms (N = 16, n = 166) for the snapping animal and is significantly smaller than that for the receiver(61.1 :!:36.3 ms, N= 15, n = 137; Wilcoxon test: P < 0.05). The retreat of the snapper (mean latency: 25.2:!: 38.7 ms, N = 14, n = 105) usually starts significantly earlier after the snap B than the approach of the receiver (mean latency: 56.6:!: 22.8 ms, N = 14, n = 88; Wilcoxon test: P < 0.05).

3.4 Comparison of different snap types >- ~(,)Q 60 ~ In our experiments30 snaps of a snapping shrimp (initial g. L_~ snaps) were immediately (i.e., within 1 s) followed by Q 40 c .,t= a return snap of the opponent. In 3 of the 30 cases the exchangeof snaps was followed by a second snap from 20t the initial snapper.As shown in figure 6A, initial snaps were more ~ften directed towards the receiver, i.e., hit 0 it more often (72.2%, N = 9) than return snaps (62.8%, 0 30 60 90 120 150 180 N = 9) and other snaps (48.4%, N = 9) but the percentage tailfan angle (O) of hits does not differ significantly for the three snap c types (Friedman test: P > 0.1). In each case, i.e., for initial, return and other snaps, the distances betweenthe opponents for directed snaps are smaller than those for undirectedsnaps. The combination of both distancemeas- ures reveals that the meandistance progressivelyincreases '-"~ from initial snaps (1.0::!:0.9 cm) to return snaps (1.5 ::!:0.9 cm) and to the other snaps (2.9:!: 1.0 cm) with ~ Q) significant differences between all snap types (Friedman 5- test: P < 0.05, N = 9) and betweeninitial and return snaps ~ versus the other snaps (Wilcoxon test: P< 0.05, N = 9; figure 6B). The cocking duration of the snapper claw before snapping (see § 2) in agonistic encounters on average amounts to 499:!: 114ms (N = 16, n = 193) and does not differ significantly between initial snaps (492:!: 125 ms, N = 8), return snaps (484:!: 176ms, N = 8) and the other Figure 4. Frequency histogmms of the snapper claw angle snaps(499:!: 82 ms, N = 8) (Friedmantest: P> 0.5; figure 'Y betweenthe snapper claw long axis and the longitudinal body axis in the snapping animal (A, cf. figure lb, N = 203) and of 6C). Cocking duration is not correlatedwith snapperclaw the tailfan angle relative to the abdomen in the snapper (B, length or body length. Snaps elicited in tethered shrimp N = 205). (C) Frequency histogram of the targets [i.e., those by slightly touching the snapper claw show a cocking receiver body parts first hit by the water jet of the snapper duration of on average472:!: III ms (N = 7; figure 6C), (N = 16, n = 215)]. Grand meansof percentagesacross individuals and standard deviations are shown. N, none; S, snapperclaw; which does not differ significantly from the meancocking SIP, between snapperand pincer claw; P, pincer claw; B, body, duration of all animals tested in agonistic encounters i.e., leg, carapace,or tailfan. For more details see text. (Mann-Whitney-U test: P > 0.5, N = 23). The comparison Agonistic encounters in snapping shrimp 629

of cocking durations in two snapping shrimp, which this point the shrimp might be aware of each other by were tested both in agonistic encounters as well as when visual, hydrodynamic and/or chemical informations. After tethered, reveals only minor differences of the mean the first tactile contact, the opponents actively align their values (496:t 170 ms vs. 489:t 65 rns and 720:t 178 ms body axes and prepare for snapping. In many cases they vs. 692:t 110 ms), which leads to the idea that each change from a side by side or head-tail position to a individual might have its characteristic cocking duration. head-head position during this phase of the encounter. The shrimp thus are positioned as shown in figure lA, in most cases facing each other, the receiver usually Discussion situated in the rostral quadrant with respect to the snapper, and the body axes of both mostly encompassing small By rapidly closing their large modified snapper claw angles (figure 3). snapping shrimp produce a fast, well focussed water jet, which is used in defensive and offensive interactions with conspecifics and prey, as well as a short intense 4.2 Tactile interactions before snapping sound signal. We quantitatively analysed intraspecific agonistic interactions between intact male and female A. The reduced distance between the animals allows addi- heterochaelis snapping shrimp to gain first insights in tional tactile interactions, which occurred within 5 s the mechanisms of detection, identification and localiza- before the snap in most cases. The last contact before tion of the opponent, the preparations for snapping (like snapping predominantly involves frontal appendages- the approach and alignment of both shrimp, the posi- antennae and claws (figure 2). Here, cases of active tioning and cocking of the snapper claw), the snap itself antennal probing [especially of the snapping animal, see (including the target of the water jet), and finally the also the mutual antennulation mentioned by Nolan and behaviours of both animals after the snap (approach and Salmon (1970)] and active claw probing are included. return snap, retreat, etc.). The analysis of the behaviour The mean interval between touch and snap amounts to in intact snapping shrimp provides the basis for and will about 1.5 s, which strongly indicates that most snaps are be complemented by studies in animals deprived of actually elicited by the tactile stimulus, especially in visual, mechanosensory or chemosensory input. view of snapper claw cocking durations of about 0.5 s, which are included in these latencies.

Body alignment before snapping 4.3 Cocking of the snapper claw The sequence of events during agonistic interactions of two snapping shrimp starts with the approach of the While cocking its snapper claw the snapping animal animals as described by Nolan and Salmon (1970) for usually shields its frontal body parts by keeping the intra- and intersexual encounters in A. heterochaeli.\". At claw slightly across the midline (figures IB and 4A), a A

~ ~ ~'-" ~'-' >. >- ~u ~Co) v Q) 6-~ 6- .J::

tailflip retreat nc approach tailflip retreat nc approach behaviour after snap behaviour after snap

Figure 5. Frequency histograms of behaviours after the snap in the snapper (A) and the receiver (B). Open bars show the frequency of the respective behaviour following all snaps (snapper: N = 16, n = 208; receiver: N = 16, n = 209) and filled bars show that following directed snaps (snapper: N=15, n =115; receiver: N=15, n=117). Grand means of percentages across individuals and standard deviations are shown. nc, no change.

B4. 630 Barbara Schmitz and lens Herberho/z

posture which actually does not differ for directed snaps (the claw pointing at the receiver) and undirected snaps. Thus, apart from being a weapon the snapper claw represents a protective device. Cocking of this claw serves two functions. On the one hand, tension is built up by cocontraction of a claw closer and a claw opener ] muscle (Ritzmann 1974), which when released allows 21] the extremely fast and powerful claw closure. On the :E other hand, the cocked snapperclaw, exposing the white """' 0 80 plunger of the dactyl, can serve as a visual signal. As Q) 60 shown by Hazlett and Winn (1962) for Alpheus armillatus ~, and Synalpheushemphilli the presentationof an isolated 40 snapperclaw induces snapping and the number of snaps 20 is larger for open claws than for closed claws. Hughes (1 996a)performed similar experimentsin A. heterochaelis 0 and showed that the presentation of an open snapper initial return other claw elicited open claw displays in females and males, snaps the number of displays in the latter increasing with the size ratio betweenthe own claw and the presentedclaw"- Since snapper claw size is correlated with body size (see also Schein 1975), Hughes (1996a) concluded that male snapping shrimp use the visual open claw display for size assessmentof the opponent. In the same species Schein (1977) found that the open claw display is used more frequently during day encounterswhereas the snap is used more often during interactions at night. Conover and Miller (1978) immobilized the snapper claw dactyl (in the open or closed position) in A. heterochaelisand showed that these shrimp -no longer able to snap-could not compete successfully with intact individuals in acquiring or holding a shelter. The snapperclaw can be opened and closed again slowly, so that the duration of the visual signal may be controlled by the shrimp. Once the claw is cocked, however, it can only be closed rapidly by snapping. In our experiments with A. heterochaelis cocking 1000 """' duration does not depend on the size or behavioural '" situation of the shrimp (figure 6C), but shows individual 5800 ~ characteristics in two animals. Since the cocking of the 0 "": 600 snapperclaw for about 0.5 s is immediately followed by ~ the water jet, we assumethat this duration is not analysed ~400 visually by the receiver, but that it predominantly effects ~ the strength(velocity) of the hydrodynamicsignal. Longer :.g 200 cocking durations should result in the generationof more 0 u muscle tension and thus faster claw closure and higher water jet velocities. We are currently using laser Doppler anemometry in tethered snapping shrimp to test this hypothesis.

Figure 6. Comparison of different snap types. (A) Percentage of hits, i.e., of snaps directed towards the opponent for initial 4.4 Target and purpose of water jetj snaps, return snaps, and the remaining snaps. (B) Distance between the tip of the snapper claw of the snapping animal As alreadymentioned by Volz (1938) for Alpheus dentipes and the receiver for the three snap types. (C) Comparison of cocking durations for the three snaptypes in agonisticencounters and Synalpheus laevimanus, the snapping sound may and for snaps of tethered snappingshrimp. Grand meansacross merely representa side effect of the rapid claw closure. individuals and standard deviations are shown. Furthermore, auditory organs have not been detected in

l20lOO Agonistic encounters in snapping shrimp 631 snapping shrimp. Thus, we suppose that the water jet an evasive strategy. while the receiver by approaching is analysed by the receiving shrimp and that mechano- increasesits chancesto hit the opponent more strongly. sensory hairs are used to analyse these hydrodynamic The immediate retreat of the snapper and the approach or tactile signals. Directed snaps usually hit the receiver of the receiver after the snap representa general strategy at its snapper claw or pincer claw (figure 4C). Interes- in intraspecific encountersof snapping shrimp and help tingly, the claws are densely covered with at least 6 retaining a defined distance between the opponents different hair types (long serrulate setae, short serrulate throughout the interaction. This distance obviously is setae,simple short setae,plumose and plumoserratesetae well suited for hydrodynamic and possibly chemical as weII as tubercles; Read and Govind 1991). According information transfer and at the same time reduces the to their morphology and distribution long serrulate setae risk of physical injury. are good candidates for the analysis of water jet velocity and direction as well as of tactile stimuli (Sullivan and 4.6 Size-specific differences Schmitz 1997). The amplitude of the hydrodynamic signal at the Many studies of crustaceanagonistic encounters have receptors of the receiver depends on the intensity of shown that the larger animal usually wins a fight (table 5 the signal as weII as on the angle and distance between in Hyatt ]983). This is also true for intrasexualencounters the sender and the receiver. In 58% of all cases and in A. heterochaelis snapping shrimp, where the larger most initial and return snaps (figure 6A) the snapper animal usually becomes dominant and snaps more fre- claw of the sender was pointing at the opponent during quently than its smaller opponent,so that the likelihood snappingand for thesedirected snapsthe distancebetween of winning is increasedfor the larger animal (Nolan and the animals on average amounted to less than I cm. At Salmon 1970; Schein ]977). In these two studies the this distance from the tip of the snapper claw water jet winner was either defined as the animal gaining the velocities can stilI amount to about 1 mis, however, possessionof a shelter or (if no shelter was offered) as obvious damagesin the receiver have not beenobserved. the shrimp performing the least number of withdrawal In addition, the receiving shrimp usualIy does not retreat, behaviours. In the presentstudy no shelter was provided but rather approachesthe sna~per after being hit by the and retreats were often observed after snapping. Here water jet (figure 5B) and may even launch a return the snapping frequencyof the larger shrimp only slightly snap. Thus, we conclude that snapping in intraspecific exceeded that of the smaller animal when regarding all encounters in A. heterochaelis is not used to damage experiments (]]6 vs. ]02), and was similar for both the opponent but that water jets can be viewed as threat shrimp (57 vs. 63) in intrasexual encounters. These displays alIowing an assessment of the opponent's results can be attributed to the low average size ratio strength,fighting ability and possibly other characteristic and the lo~ average size difference and conform to features of the snapping animal. In good agreementwith results by Hughes (1996a) who found a significant our results Knowlton and Keller (1982) showed that in correlation between relative size and the likelihood of intraspecific fights in the snappingshrimp Alpheusarmatus winning a competitive interaction in A. heterochaelis. physical damage does not result from water jets during Future studies, involving the deprivation of visual, snapping but from direct contact betweenthe animals. mechanosensory,and chemosensory input will further elucidate the mechanisms underlying detection, recog- 4.5 Behaviour after snapping nition, as well as localization of the opponent in snapping The snapperusually retreats within 25 ms after the snap, shrimp agonistic interactions. while the receiver approaches with a longer latency of about 57 ms. Thus, the retreat of the snapping animal Acknowledgements obviously rather presents a spontaneous action, being We would like to thank the Florida State University prepared by bending the tailfan downwards alreadybefore Marine Laboratory and especially Cheryl Morrison for and during the snap (figure 4B), than a responseto the help in collecting the animals. We also thank Georg approachof the receiver. Retreatsof the snappinganimal Heine for writing a circular statistics program, Daniel following directed snapsare more frequent than following Gehr, Raffael lturrizaga, Lars Miindermann, and Eva all (directed and undirected) snaps (figure 5A), which Tiedge for help with video recordings and data analysis, might indicate that the shrimp receives either visual or and Peter Fraser for critically reading an earlier draft mechanosensoryfeedback about the successof its attack. of the manuscript. Supported by grants of the Deutsche The retreat of the snapperincreases the distance between Forschungsgemeinschaft(Schm 693/5-1 and 5-2). the opponents,which is compensatedfor by the approach of the receiver. About 24% of these approacheswere References followed by a return snap of the receiver within 1 sand Barki A, Karplus I and Goren M 1991 The agonistic behaviour for these casesthe retreat of the initial snapperrepresents of the three male morphotypes of the freshwater prawn 632 Barbara Schmitz and lens Herberholz

Macrobrachium rosenbergii (Crustacea, Palaemonidae); of escalation potential in a symbiotic, territorial snapping Behaviour 116 252-277 shrimp; Behav. Ecol. Sociobiol. 10 289-292 Batschelet E 1981 Circular statistics in biology (London: Knowlton R E and Moulton J M 1963 Sound production in the Academic Press) snapping shrimps Alpheus (Crangon) and Synalpheus;Bioi. Breithaupt T, Schmitz B and Tautz J 1995 Hydrodynamic Bull. 125 311-331 orientation of crayfish (Procambarus clarki/) to swimming MacGinitie G E 1937 Notes on the natuml history of several fish prey; J. CompoPhysiol. A177 481-491 marine Crustacea;Am. Midi. Nat. 18 1031-1036 Brooks W K and Herrick F H 1891 The embryology and MacGinitie G E and MacGinitie N 1949 Natural history of metamorphosis of the Macroura; Mem. Natl. Akad. Sci. marine animals (New York: McGmw Hill) Washington5 319-576 Nolan A N and Salmon M 1970 The behaviour and ecology Caldwell R L 1987 Assessmentstrategies in stomatopods;Bull. of snapping shrimp (Crustacea: Alpheus heterochelis and Mar. Sci. 41 135-150 Alpheus IWrmannl);Forma et Functio 2 289-335 Caldwell R L and Dingle H 1975 Ecology and evolution of Pavey C R and Fielder D R 1996 The influence of size agonistic behavior in Stomatopods;Naturwissenschaften 62 differential on agonistic behaviour in the freshwater crayfish, 214-222 Cherax cuspidatus(Decapod".; Pamstacidae); J. Zool. London Conover M R and Miller D E 1978 The importance of the 238 445-457 large chela in the territorial and pairing behaviour of the Read A T and Govind C K 1991 Composition of external setae snapping shrimp, Alpheus heterochaelis;Mar. Behav. Physiol. during regeneration and transformation of the bilaterally 5 185-192 asymmetricclaws of the snappingshrimp, Alpheus heterochelis; Copp N H 1986 Dominance hierarchiesin the crayfish Procam- J. Morphol. 207 103-111 barus clarkii (Girard, 1852) and the question of learned Ritzmann R E 1974 Mechanismsfor the snapping behavior of individual recognition (Decapoda, Astacidea); Crustaceana51 two Alpheid shrimp, Alpheus californiensis and Alpheus 9-24 heterochelis;J. CompoPhysiol. 95 217-236 Dingle H 1983 Strategies of agonistic behavior in Crustacea;in Schein H 1975 Aspects of the aggressiveand sexual behaviour Studies in adaptation: The behavior of higher Crustacea(eds) of Alpheus heterochaelisSay; Mar. Behav. Physiol. 3 83-96 S Rebach and D W Dunham (New York: Wiley) pp 85-111 ScheinH 1977The role of snappingin AlpheusJzeterochaelis Say, Downer J 1989 Mit den Augen der Tiere. Teil: Laute der Natur. 1818,the big-<:lawedsnapping shrimp; Crustaceana33 182-188 Universum, BBC and WDR (WestdeutscherRundfunk) Schmitz B, van der Wall M and Simon B 1995 Underwater Elwood R W and Neil S J 1992 Assessmentand decisions. A acoustics of snapping in the pistol shrimp (Alpheus hetero- study of information gathering by hermit crabs (London: chaelis); Verh. Dtsch. mol. Ges. 88 215 Chapman and Hall) Schultz S, WuppermannK and Schmitz B 1998 Behavioural Gruner H-E 1993 Lehrbuch der speziellen Zoologie. Band I: intemctions of snapping shrimp (Alpheus heterochaelis)with WirbeIJose,4. Teil: Arthropoda (Jena: Gustav Fischer Verlag) conspecifics and sympatric cmbs (Eurypanopeusdepressus); Hazlett B A 1962 Aspects of the biology of snapping shrimp Zool. -Anal. ComplexSyst. (Suppl. I) 101 85 (Alpheus and Synalpheus);Crustaceana 4 82-83 SOderbackB- 1991 Interspecific dominance relationship and Hazlett B A and Winn H E 1962 Sound productionand associated aggressiveintemctions in the freshwater crayfishes Astacus behavior of Bermuda Crustaceans(Panulirus, Gonodactylus, astacus(L.) and Pacifastacusleniusculus (Dana); Can.J. mol. Alpheus, and Synalpheus);Crustaceana 4 25-38 69 1321-1325 Huber R and Kravitz E A 1995 A quantitative analysis of Sullivan J and Schmitz B 1997 The mechanosensorysystem of agonistic behavior in juvenile American lobsters (Homarus snapperand pincer claw in snapping shrimp (Alpheus hetero- americanus L.); Brain Behav. Evol. 46 72-83 chaelis); in Proceedings of the 25th Gb'ttingenNeurobiology Hughes M 1996aSize assessmentvia a visual signal in snapping Conference1997 (eds) N P.lsnerand H Wiissle (Stuttgart: shrimp; Behav. Ecol. Sociobiol. 38 51-57 Thieme) Vol. 2, p 250 Hughes M 1996b The function of concurrentsignals: visual and Suzuki D 1986 Stunning sounds. Series: The nature of things; chemical communication in snapping shrimp; Anim. Behav. CBC (CanadianBroadcasting Corporation) 52247-257 Volz P 1938 Studien tiber das "Knallen" der Alpheiden. Nach Hyatt G W 1983 Qualitative and quantitative dimensions of Untersuchungenan Alpheus dentipes Guerin und Synalpheus crustaceanaggression; in Studies in adaptation: The behavior laevimanus(Heller); Z. Morphol. akol. Tiere 34 272-316 of higher Crustacea (eds) S Rebachand D W Dunham (New Williams A B 1984 Shrimps,lobsters, and crabs of the Atlantic York: Wiley) pp 113-139 coast of the Eastern United States, Maine to Florida Knowlton N and Keller B D 1982Symmetric fights as a measure (WashingtonDC: SmithsonianInstitution Press)

MS received 3 June 1998; accepted 15 October 1998

Corresponding editor: VIDY ANAND NANJUNDIAH