BULLETIN OF MARINE SCIENCE, 55(1): 133-141, 1994 CORAL REEF PAPER

INTRASPECIFIC BY MALE RAZORFISHES (LABRIDAE) DURING BROADCAST SPAWNING: FILIAL OR INTRA-PAIR PARASITISM?

Simon C. Nemtzov and Eugenie Clark

ABSTRACT During field studies, we observed approximately 550 broadcast spawns (sometimes called pelagic spawns), in which pairs release their gametes high in the water column for external fertilization, by three congeneric wrasses: green (Xyrichtys splendens), rosy (X martinicensis) and pearly (X novacula) razorfishes. The released formed a cloud that remained visible for 3-5 s before dissipating. Male X. splendens and X. martinicensis ate ova from the cloud during approximately 40% of their spawns. Male X novacula, and females of all three spe- cies, never ate ova. Our observations show that in , parental care is not a prerequisite for filial cannibalism (the ingestion of one's own offspring), as was previously believed. Analysis of videotaped spawns showed that males were prepared to eat even before comple- tion of the spawning act. We propose that males may control the amount of sperm they release when engaging in spawning rushes. The eating behavior should be adaptive for ter- ritorial male fishes, particularly those with multiple daily spawns (we observed up to 23 spawns per day), because parasitism of nutrients from females (via the ova) may increase the males' chances for future reproduction, consistent with intraspecific predation theory.

Filial cannibalism (the ingestion by an of its own offspring) seems to conflict with conventional evolutionary theory in that it apparently reduces the parent's fitness. Rohwer (1978) explained that filial cannibalism could be adaptive if a parent achieved higher future reproductive success by lowering its immediate fitness. He presented his model in terms of species with brood cycles, in which males defend temporary breeding territories where they guard eggs, and then, after the eggs have hatched, they abandon the territories to feed. He reasoned that by sacrificing part of the current clutch, a starving male might improve his chances of bringing the remainder of the clutch (and future clutches) to maturity. Hoelzer (1988) showed that intraspecific can occur in fishes with parental care but without brood cycling, and that filial cannibalism is, therefore, probably more common than was previously thought. Rohwer's (1978) model was written concerning partial-clutch cannibalism in fishes, but it may also explain whole-clutch cannibalism which necessarily ends any parental care of that clutch (Hoelzer, 1992). Partial-clutch cannibalism can also occur as a form of custodial care by parents when removing diseased or moribund eggs from the nest (Perrone and Zaret, 1979; Clutton-Brock, 1991). Although Rohwer's (1978) model does not differentiate between the eating of viable versus non-viable eggs, it is likely that both behaviors can be maintained by similar selective forces. Parents manipulate their investment in offspring to maximize their own fitness (Alexander, 1974). In filial cannibalism, the male effectively parasitizes nutrients from the female that he can then invest in future reproduction (Rohwer, 1978). For such intra-pair parasitism to be adaptive for the male, there must be a rea- sonable certainty that there will be future opportunities for him to reproduce. For tropical fishes that spawn daily all year-round, there is a high probability of future reproduction because the next spawning period is less than 24 h away. Intraspecific oophagy is common in many species of fish (FitzGerald, 1991; Katano, 1992). Filial cannibalism is a subtype of intraspecific oophagy in which

133 134 BULLETIN OF MARINE SCIENCE, VOL. 55, NO. I, 1994 the ingested eggs represent potential offspring. Filial cannibalism and nutrient parasitism would seem to be unlikely in broadcast spawning fishes (in which individuals release their gametes high in the water column), because eggs are fertilized and disperse almost immediately (Petersen, 199]; Petersen et aI., 1992). Yet, before they disperse, the eggs and sperm often form a temporary cloud which is sometimes exploited for food by planktivorous fish (Clark et aI., 1988; Colin and Clavijo, 1988). The appearance of this concentration of eggs can provide a brief opportunity for adaptive intraspecific oophagy. Recently there has been a great deal of interest in the prevalence of filial can- nibalism in teleost fish (Dominey and Blumer, 1984; FitzGerald, 1992; FitzGerald and Whoriskey, 1992), and its importance as a selective force in a wide range of behaviors and reproductive patterns (Foster et aI., 1988; Petersen, ]990). In pre- vious discussions, parental care was assumed to be necessary for filial cannibalism in fishes, as it allowed parents easy access to offspring. Yet in other vertebrates, filial cannibalism is relatively common under a wide variety of conditions, in- cluding cases with and without parental care (Polis, 1981), To fully understand the importance of filial cannibalism as a selective force, we must also consider fish species without parental care. Here we present observations of cannibalism of eggs during broadcast spawn- ing by males of two congeners with no parental care.

STUDY SPECIES

Green razorfish (Xyrichtys splendens), rosy razorfish (X. martinicensis) and pearly razorfish (X. novacula) are tropical wrasses (family Labridae) of the western Atlantic (Randall, 1965). They live on shallow grass or sand flats adjacent to coral reefs, where they forage on benthos and plankton. X. sp/endens and X. martinicensis generally live in single-male harems (Victor, 1987; Nemtzov, 1992). The mating system of X. novacula has not been described but it is apparently haremic (pers, obs.). All three species are protogynous hermaphrodites, i.e., they undergo sex change from female to male (Roede, 1972; Victor, 1987; Nemtzov, 1992). Like all tropical wrasses, razorfishes engage in external fertilization by broadcast spawning in which gametes are released when the fish are high in the water column. Razorfishes spawn only in pairs (Nemtzov, 1982, 1985, 1992; Clark, 1983; Thresher, 1984; Victor, 1987; Scarr, 1990), daily, during a late afternoon spawning period, all year-round (Clark, Shen and Scarr, pers. obs.).

METHODS

All three species of razorfishes were observed by stationary SCUBA divers in shallow (5-10 m deep) habitats around the southern Caribbean island of Bonaire, Netherlands Antilles (I2°N, 68°W) during February and June 1986 (by E.C.), and from June to August in 1989 and 1990 (by S.C,N.). Male X. sp/endens and X. martinicensis were identified by tags or natural markings, and were observed over the entire daily spawning period which began about 3 h before sunset and lasted 1.5-2 h. While watching focal males we noted the time each spawning occurred and whether or not the male ate eggs. Although we had observed many individual occurrences of egg eating, we present here data from 31 complete spawning periods of 15 male X. splendens, and II spawning periods of 7 male X. martinicensis (Table I). We also present data from 39 complete spawning periods of 6 male X. no- vacu/a (Table I) that were observed by Clark, Shen and Scarr (unpub!. data). Data from males with more than one spawning period are presented as individual means. An underwater videographic record (using ambient light) was made during some of the spawning periods of the razorfishes. Videotapes of 20 complete spawning rushes of the razorfishes were analyzed visually both in slow-motion and frame-by-frame. Sketches of behaviors (Fig. I) were traced from the screen of a television monitor in freeze-frame, and from single-frame print-outs. Figures ]a, and 1b were each drawn from a videotape of a single spawning rush, each with an elapsed time of about 5 s. The seven sketches in each of the two parts of Figure 1 were chosen as arbitrary representative views; they do not denote equal amounts of time between sketches, NEMTZOV AND CLARK: OOPHAGY BY RAZORFISH 135

Table 1. Intraspecific oophagy during broadcast spawning by Caribbean razorfishes (genus Xyrichtys)

X. .\'{Jlendens x. martinicen.si.s X. novacula Total number of spawns observed 295 55 126 Spawns with egg eating, N (%)* 107 (36.2) 23 (41.8) o Number of complete spawning periods observed 31 II 39 Spawning periods observed per malet 1.6 ± 0.4, 1-4 1.6 ± 0.7, 1-3 7.8 ± 1.7,3-14 Spawns observed per spawning period:!: 9.5 ± 1.0, 3-23 3.2 ± 0.3, 1-4 2.7 ± 0.6, 1-8 * There was no significant difference between X. splendens and X. martmicensis in the percentage of spawns with egg eating; G = 0.15, df = I. P > 0.3; 2 X 2 G-test for the significance of the difference between two percentages (Sokal and Rohlf. 1981). t Mean ± SE, range. 4:Menn of individual males' means ± SE, range.

OBSERVATIONS

All three species of razorfish engaged in broadcast spawning, in which pairs rose quickly in the water column with their bodies touching along most of their lengths. At the apex of the spawning rush the two fish separated quickly ("snapped" apart), about 2-4 m above the substratum producing a small (diameter ca. 15 cm) white cloud in the water that dissipated in 3-5 s (Fig. I). We could sometimes discern individual ova within the cloud. After snapping apart at the apex of the spawning rush, both the male and female usually darted back down to the substratum (Fig. 1b). Occasionally, however, immediately after spawning, male X. splendens and X. martinicensis turned and ate from the cloud before it dissipated (Table I) and only then returned to the substratum (Fig. la). X. novacu- la were never seen exhibiting such intraspecific oophagy. Also, we have never seen this behavior by a female of any species. Slow motion analysis of videotapes of spawning rushes showed that those that occurred when males ate eggs differed from usual spawning rushes (Fig. Ib). In all spawning rushes in which oophagy occurred (n = 10), the male's body took on a pronounced S-cUfve (Fig. la) when nearing the apex of the rush. The S-curve was not observed in the (n = 10) spawning rushes without oophagy (Fig. Ib). When oophagy occurred, the male did not stop to look for the cloud before heading for it, but as the pair reached the apex of the spawning rush, the male circled back and went for the cloud even before it was visible on tape. The circling maneuver by which the male turned around to eat was one fluid motion, continuous from the S-curve during the rise (Fig. la). In the field, the actual cloud was not visible to the observers at each spawn. On one videotape of a spawning rush without oophagy (Fig. 1a), we could de- termine that the gametes were released in two clouds, first the eggs and then the sperm (Clark et aI., 1991). In the field we could always tell whether a male turned and ate, or whether he headed straight back down towards the substratum after spawning. Males that ate, stopped and took 1-6 bites at the cloud-the biting lasted for up to 3 s- and they then returned to the substratum. We did not analyze stomach contents because other behavioral studies of the males were in progress (Nemtzov, 1992). Yet it was obvious to us (and to our colleagues who viewed the videotapes) that the males were indeed eating eggs. During some spawning periods, eggs were eaten by male X. splendens and X. martinicensis each time they spawned, while during other spawning periods eggs were only eaten after some of the spawns, or none. Some males ate more often than others; there was a significant difference among the various males in the proportion of spawns in which they ate (Fig. 2, F[21. 18] = 18.8, P < 0.001); this 136 BULLETIN OF MARINE SCIENCE, VOL. 55, NO. I, 1994

b

Figure 1. Broadcast spawning by a single male X. martinicensis on one day with two different females, The sketches are sequences (from right to left) from seven single-frame print-outs of spawning rushes with (a) and without (b) intraspecific by the male. Figure la shows the male and female closely pairing before the beginning of the spawning rush, the pronounced $-curve of the male's body (arrow) before the apex of the spawning rush as he springs into a tight tum, the apex of the spawning rush in which the pair separate quickly ("snap" apart), the female heading back toward the substratum while the male completes his circling motion, and finally the male eating eggs from the cloud. Figure lb shows the same male making a gradual up-and-down "loop" (rather than a full tight circle) at the apex of the rush; after the female releases eggs the male releases milt (sperm) while the female heads immediately down towards the substratum.

difference was not attributable to a difference among species or days (by analysis of variance). Egg eating by male X. splendens was distributed randomly throughout the spawning period (Fig. 3. runs up-and-down test [Sokal and Rohlf, 1981] ts = 12, n = 17, P = 0.9). There was also no significant trend towards an increase in egg eating by male X. splendens in spawns later in the day (Fig. 3, Kendall's coeffi- NEMTZOV AND CL.ARK: OOPHAGY BY RAZORASH 137

10

8

III Cl) (ij E '0 •... Cl) .c E ::l Z

x. splendens X. martinicensis

20 40 60 80 100 Spawnswith egg eating (%)

Figure 2. Intraspecific variation in the frequency of egg eating during broadcast spawns by different male razorfishes, X. splendens (n = 15) and X. martinicensis (n = 7).

cient of rank correlation corrected for ties T = 0.059, n = 18, one-tailed P ~ 0.2). The amount of time that had elapsed from the previous spawn had no effect on the occurrence of oophagy (Fig. 4, D = 0.077, n. = 104, n2 = 140, two-tailed P > 0.9, Kolmogorov-Smirnov two-sample test).

8--A 10 7 70

~ 10 5 5 :60 -Cl 11 c: :0::as 4 8 6 Cl) 50 15 Cl 1 1 1 7 Cl CI) 40 11 .s;;; ~ 20 13 ~ 30 7 (I) 29 27 26 c: ~ as 2Q a. C/) 9 10-

A ,I 1st 5th 10th 15th Order of spawn in the spawning period

Figure 3. The percentage of spawns in which egg eating occurred for the nth spawn in spawning periods with at least n spawns, by male X. splendens (black bars) and X. martinicensis (white bars). The number above each bar is the number of spawning periods with n spawns. 138 BULLETIN OF MARINE SCIENCE, VOL. 55, NO. J, 1994

35

30

l!! 25 ~ fE 20 '0 CD 15 .c § 10 Z 5

o o 234 5 8 7 B 9 10 Elapsed time from previous spawn (min)

Figure 4. Relationship of the occurrence of egg eating by male X. splendens (.) and X. martinicensis (0), with the elapsed time from one spawn to the next. The plot shows the distribution of the number of spawns that occurred after various elapsed times after a spawn, to a spawn with (solid lines) and without (dashed lines) egg eating. An elapsed time of zero minutes represents spawns that occurred within the same minute (seconds were not recorded). Not shown in the figure are spawns that occurred after more than 10 min had elapsed. There was no significant difference between the curves for eating and not eating for either species (Kolmogorov-Smirnov tests, P > 0.5).

Egg eating did not occur more frequently when males had a high number of spawns that day. There was no significant correlation between the proportion of spawns in a spawning period at which a male ate, and the total number of spawns the male had during that day's spawning period (r = 0.03, df = 41, P > 0.9, square-root transformation).

DISCUSSION Field observations of X. splendens and X. martinicensis showed a high fre- quency of instances of intraspecific oophagy during spawning by some males. We observed complete spawning periods for only a limited number of males (Table 1), and they exhibited variable rates of oophagy (Fig. 2). It is therefore, difficult for us to determine the exact prevalence of this behavior in these species. Al- though the behavior may be entirely non-adaptive and may cause a real reduction in fitness, its high rate of occurrence, at least in some males, demands an expla- nation. Both X. splendens and X. martinicensis males are planktivorous and engaged in intraspecific oophagy, whereas male X. novacula, which are benthic feeders, did not. This dichotomy suggests that planktivory is necessary for this type of cannibalism, or that benthic feeding precludes it. Similarly, planktivorous male threespine stickleback (Gasterosteus aculeatus) are not cannibalistic on benthic eggs, whereas benthic feeding sticklebacks are (Foster, 1988), Although the stick- lebacks show the exact opposite dichotomy to that of the razorfishes, both sets of observations suggest that feeding in the same place where eggs are laid is associated with egg cannibalism. More study is needed in order to determine what specific factors caused this behavior to occur only in these two razorfish species. One possible explanation for this phenomenon is that males may have eaten eggs only during rushes when they were unable to fertilize them. An inability to NEMTZOV AND CLARK: OOPHAGY BY RAZORFISH 139 fertilize could occur either if a spawning rush took place during a refractory period (when sperm release was physically impossible), or if the male's sperm supply was depleted. First, we could find no published studies of refractory periods of fish with multiple daily spawns. We have seen male razorfishes spawn with different fe- males as quickly as three times in <20 s without egg eating, but we could not determine if sperm were released each time. We found that razorfish oophagy was not associated with the elapsed time between spawns (Fig. 4). In addition, males often ate eggs at the first spawn of the day (Fig. 3) when there could be no refractory period. Therefore, an inability to release sperm is an unlikely expla- nation for the occurrence of razorfish oophagy. Second, we believe that sperm depletion (Nakatsuru and Kramer, 1982) is an unlikely explanation for razorfish oophagy based on three observations: (1) males did not eat eggs more often on days when they had many spawns, (2) males often ate eggs at the first spawn of the day when that day's sperm supply could not yet be depleted, and (3) males were no more likely to eat eggs at early or late spawns (Fig. 3). Recent studies have shown that male bluehead wrasses (Thalassoma bifascia- tum) can control the amount of sperm they release at each spawning rush in relation to the size of the female (Shapiro, Marcanato and Yoshikawa, pers. comm.). If male razorfishes have a similar ability for precise control of sperm release, then they may not have released any sperm during some spawns. Ac- cording to this hypothesis the oophagy we observed would represent intra-pair parasitism and not filial cannibalism, because the eggs the male ate were not his offspring. While incurring almost no long-term costs, the male could successfully steal the nutrients that the female had invested in the eggs. More study is needed to determine whether male razorfishes release less sperm (or any) during spawns when they eat, and what clues the males may use to make the choice of when to eat. Once they are released, the ova represent potential food for both fish, yet only males engaged in egg eating. If a male were in an energy deficit he would be expected to eat the eggs and thereby increase the probability of remaining on the territory for future reproduction. In contrast, a female that ate her own eggs would simply be producing a portion of tomorrow's clutch at the expense of today's, a non-adaptive tactic. Because opportunities for future reproduction are almost cer- tain for these razorfishes, a territorial male that ate eggs during even some of his spawns would increase his own future reproductive success. In conclusion, the cloud of eggs created during broadcast spawning provides opportunity for site-restricted males to supplement their energy intake at the ex- pense of their mates, whether the eggs are fertilized or not. Male X. splendens and X. martinicensis have apparently adopted a tactic to maximize their fitness, similar to behaviors previously observed only in male fishes that guard eggs.

ACKNOWLEDGMENTS

We wish to express our sincere thanks to T. Boettcher, K. Grief, P. Pemberton, H. Roffey, D. Scarr, and D. Shen who provided expert field assistance, to R. Korth for long hours shooting underwater videotapes, to S. Kogge and A. Snowhite for analysis of the videotapes, and to M. Chen and B. Rodgerson for preparing Figure 1. For helpful discussions on earlier versions of this paper we thank M. Bell, R. Cowen, C. Janson, S. Foster, and G. C. Williams. S. C. N. was supported by the Smith- sonian Tropical Research Institute, the Lerner-Gray Fund for Marine Research, the Explorers Club, the Raney Fund, and Sigma Xi. E.C. was supported by the University of Maryland Foundation and the National Geographic Society. This is contribution no. 912 from the Graduate Program in Ecology 140 BULLETIN OF MARINE SCIENCE. VOL. 55. NO. I. 1994 and Evolution, State University of New York at Stony Brook, and represents a portion of a dissertation submitted by S.C.N. to that program in partial fulfillment of the requirements for a Ph.D. degree.

LITERA TORE CITED

Alexander, R. D. 1974. The evolution of social behavior. Ann. Rev. Ecol. System. 5: 325-383. Clark, E. 1983. Sand-diving behavior and territoriality of the Red Sea razorfish Xyrichtys pentadac- tylus. Bull. Inst. Oceanogr. Fisher., Cairo 9: 225-242. ---, M. Pohle and J. Rabin. 1991. Stability and flexibility through community dynamics of the spotted sandperch. Nat. Geogr. Res. Expl. 7: 138-155. ---, J. S. Rabin and S. Holderman. 1988. Reproductive behavior and social organization in the sand tilefish, Malacanthus plumieri. Envir. BioI. Fishes 22: 273-286. Clutton-Brock, T. H. 1991. The evolution of parental care. Princeton University Press, Princeton. Colin, P. L. and I. E. Clavijo. 1988. Spawning activity of fishes producing pelagic eggs on a shelf edge coral reef, southwestern Puerto Rico. Bull. Mar. Sci. 43: 249-279. Dominey, W. J. and L. S. Blumer. 1984. Cannibalism of early life stages in fishes. Pages 43-64 in G. Hausfater and S. B. Hrdy, eds. Infanticide: comparative and evolutionary perspectives. Aldine, New York. FitzGerald, G. J. 1991. The role of cannibalism in the reproductive ecology of the threes pine stick- leback. Ethology 89: 177-194. ---. 1992. Filial cannibalism in fishes: why do parents eat their own offspring? Trends Ecol. Evol. 7: 7-10. --- and F. G. Whoriskey. 1992. Empirical studies of cannibalism in fishes. Pages 239-256 in M. A. Elgar and B. J. Crepsi, eds. Cannibalism: ecology and evolution among diverse taxa. Oxford University Press, Oxford. Foster, S. A. 1988. Diversionary displays of paternal stickleback: defenses against cannibalistic groups. Behav. Eco!' Sociobiol. 22: 335-340. ---, V. B. Garcia and M. Y. Town. 1988. Cannibalism as the cause of an ontogenetic shift in habitat use by fry of the threes pine stickleback. Oecologia 74: 577-585. Hoelzer, G. A. 1988. Filial cannibalism in a non-brood cycling marine fish. Envir. BioI. Fishes 21: 309-313. ---. 1992. The ecology and evolution of partial-clutch cannibalism by paternal Cortez damsel fish. Oikos 65: 113-120. Katano, O. 1992. Cannibalism on eggs by dark chub, Zacco temmincki (Temminck and Schlegel) (Cyprinidae). J. Fish BioI. 41: 655-661. Nakatsuru, K. and D. L. Kramer. 1982. Is sperm cheap? Limited male fertility and female choice in the lemon tetra (Pisces, Characidae). Science 216: 753-755. Nemtzov, S. C. 1982. Monandric protogynous hermaphroditism in the Red Sea razorfish, Xyrichtys pentadactylus (Teleostei, Labridae). M.Sc. Thesis, Depart. Zool. Univ. Md., College Park, Mary- land. ---. 1985. Social control of sex change in the Red Sea razorfish, Xyrichtys pentadactylus (Teleos- tei, Labridae). Envir. BioI. Fishes 14: 199-211. ---. 1992. Intraspecific variation in the social and mating behavior of Caribbean razorfishes (Teleostei, Labridae). Ph.D. Dissertation, Depart. Ecol. Evo!., State Univ. N.Y., Stony Brook, New York. Perrone, M. and T. M. Zaret. 1979. Parental care patterns of fishes. Amer. Nat. 113: 351-361. Petersen, C. W. 1990. The occurrence and dynamics of clutch loss and filial cannibalism in two Caribbean damselfishes. J. Exp. Mar. BioI. Ecol. 135: 117-133. ---. 1991. Variation in fertilization rate in the tropical reef fish, Halichoeres bivittatus: correlates and implications. BioI. Bull. 181: 232-237. ---, R. R. Warner, S. Cohen, H. C. Hess and A. T. Sewell. 1992. Variation in pelagic fertilization success: implications for production estimates, mate choice, and the spatial and temporal distri- bution of mating. Ecology 73: 391-401. Polis, G. A. 1981. The evolution and dynamics of intraspecific predation. Ann. Rev. Eco!' System. 12: 225-251. Randall, J. E. 1965. A review of the razorfish genus Hemipteronotus (Labridae) of the Atlantic Ocean. Copeia 1965: 487-501. Roede, M. J. 1972. Color as related to size, sex, and behavior in seven Caribbean labrid species (genera Thalassoma, Halichoeres and Hemipteronotus). Stud. Fauna Curac;:ao Caribb. lsI. 42: 1- 166. Rohwer, S. 1978. Parent cannibalism of offspring and egg raiding as a courtship strategy. Amer. Nat. 112: 429-440. Scarr, D. 1990. The gentle sea. Professional Association of Diving Instructors, Santa Anna. 147 pp. NEMTZOVAND CLARK:OOPHAGYBY RAzORFISH ]41

Sokal, R. R. and F. J. Rohlf. ] 981. Biometry, 2nd ed. W. H. Freeman, San Francisco. 859 pp. Thresher, R. E. ]984. Reproduction in reef fishes. T.F.H. Pub]ications, Neptune City. 399 pp. Victor, B. C. ]987. The mating system of the rosy razorfish, Xyrichtys martinicensis. Bull. Mar. Sci. 40: 152-160.

DATE ACCEPTED: October 6, ]993.

ADDRESSES: (S.c.N.) Department of Ecology and Evolution, State University of New York at Stony Brook, Stony Brook, New York 1i 794-5245; PREsENT ADDRESS: Department of Zoology, Tel Aviv University, Ramat Aviv, israel; (E.c.) Department of Zoology, University of Maryland, College Park. Maryland 20742.