Behavioral Ecology Vol. 11 No. 3: 260–264 Partner ﬁdelity and egg reciprocation in the simultaneously hermaphroditic polychaete worm Ophryotrocha diadema Gabriella Sellaa and M. Cristina Lorenzib aDepartment of Animal and Human Biology, University of Turin, via Accademia Albertina 17, 10123 Torino, Italy, and bDepartment of Veterinary Morphophysiology, University of Turin, viale Mattioli 25, 10125 Torino, Italy The mating system of the simultaneously hermaphroditic polychaete worm Ophryotrocha diadema consists of a regular egg exchange between partners of a pair. Such reciprocal egg exchange has been considered a form of cooperation, where one partner cooperates by offering its eggs to be fertilized and expects to receive partner’s eggs to fertilize. Frequency of cases in which hermaphrodites cheated (i.e., failed to give up eggs at their turn) and responses to cheating were estimated by analyzing the behavior of 38 triplets of ovigerous hermaphrodites over a 2-week period. The partner did not reciprocate 8% of egg layings. The cheated partner did not detect most cases of cheating (16 out of 25). Such a low frequency of cheating can explain why no retaliation mechanism evolved in this species. Sixty-eight percent of the individuals from the original pairs deserted even if their partners never cheated them; therefore, cheating cannot be considered the cause of desertion. Rather, desertion appeared to be a consequence of availability of a new partner whose oocytes were riper than those of the old partner. It occurred because the opportunity arose for an immediate reward, indicating that O. diadema egg exchange differs from that originally described in some serranid ﬁsh as egg trading. The relationship between costs of desertion and population size is discussed. Key words: cheating, cooperation, desertion, egg reciprocation, Ophryotrocha diadema, Polychaeta, simultaneous hermaphroditism. [Behav Ecol 11:260–264 (2000)] I t is generally assumed that there is a conﬂict of interests between sexual partners, as different selective pressures are associated with reproduction through eggs and reproduc- regular alternation of sexual roles within the mating pair. By means of such a reciprocal egg exchange, each individual fer- tilizes as many eggs as it lays. The egg-trading mechanism, tion through sperm, according to Bateman’s (1948) principle. however, severely restricts the potential advantages of the male Reproductive success of females is limited by the availability role: an individual cannot fertilize more cocoons than it re- of egg production resources, while male reproductive success leases. is determined by availability of females. In this mating system reciprocity in egg exchange is not In simultaneous hermaphrodites, both male and female simultaneous; it is delayed and therefore open to the risk of roles are present in the same individual. Therefore, if present, cheating by nonreciprocating partners (Axelrod and Hamil- sexual conﬂict is directly experienced by the same individual. ton, 1981; Trivers, 1971). A nonreciprocating partner is ex- If the male role is less costly in terms of energy expenditure, pected to prefer the male role because sperm are cheaper to mating time, and so on, then all the hermaphroditic individ- produce than eggs, and egg laying entails some costs (e.g., in uals of a population should prefer this role. When a pair is O. diadema, histolysis and production of the mucous cocoon). formed, partners of that pair should compete for the pre- Upon receiving a clutch of eggs to fertilize, a cheater should ferred sexual role. According to Leonard’s (1990, 1991) her- offer no eggs in return; in other words, it will try to play the maphrodite dilemma hypothesis, the best strategy to solve this male role instead of the female role. conﬂict should be based on reciprocation in assuming both Some aspects of the mating system of O. diadema (Premoli sexual roles during a mating encounter and a low level of and Sella, 1995; Sella, 1988) and O. gracilis (Sella et al., 1997) cheating in the preferred sexual role. The Leonard model could be considered mechanisms that have evolved to reduce suggests that mechanisms should exist to prevent or to punish vulnerability to cheating. First, at any reproductive bout, one cheaters. of the partners lays all its mature eggs. A clutch contains a Some pair mating, externally fertilizing, simultaneously her- few eggs and is laid approximately 2 days after each clutch maphroditic serranid ﬁsh (reviewed by Fischer and Petersen, laid by the other partner (Premoli and Sella, 1995). In con- 1987) have resolved the conﬂict for the preferred role by trast to serranid ﬁsh, which parcel available eggs, in O. dia- evolving a very peculiar mating system called egg trading (Fi- dema and O. gracilis no eggs are saved for successive matings scher, 1980). A similar mating system was also described by because each clutch contains all the ripe eggs an individual Sella (1985) and Sella et al. (1997) in the simultaneously her- has in its coelom. This, however, might be considered a sort maphroditic polychaete worms Ophryotrocha diadema and O. of temporal parceling of ripe eggs. In this way the cheated gracilis. Egg trading consists of reciprocal egg fertilization by egg donor will detect a nonreciprocating partner and even- tually desert it to cut its losses. Second, in isolated pairs made up of an ovigerous her- Address correspondence to G. Sella. E-mail: firstname.lastname@example.org. M. maphrodite and an adolescent male (which cannot recipro- C. Lorenzi is now at the Department of Animal and Human Biology, cate egg exchange), time intervals between successive egg University of Turin, via Accademia Albertina 17, 10123 Torino, Italy. spawnings by the hermaphroditic partner are signiﬁcantly lon- Received 21 July 1998; revised 11 June 1999; accepted 21 July 1999. ger than time intervals between successive egg spawnings of a 2000 International Society for Behavioral Ecology hermaphrodite paired with another hermaphrodite (Sella, Sella and Lorenzi • Partner ﬁdelity in a polychaete worm 261 1988). It has thus been inferred, but not demonstrated, that Experimental design a nonreciprocated hermaphrodite should have the ability to From the offspring of 38 pairs of YY yy and 38 pairs of yy lower the reproductive success of a nonreciprocating partner. yy individuals, we selected 38 Yy and 38 yy individuals of Up to now cheating and safeguards that evolved against the same age, body length (15 segments), and degree of egg cheating were studied in isolated pairs only (Sella, 1988). We maturation. With then we set up 38 pairs, each in a separate do not know how these worms behave in situations in which 15-ml bowl, each pair consisting of a Yy and a yy ovigerous multiple partners are available, which leads to the possibility and virgin hermaphrodite. of deserting a nonreciprocating partner. To synchronize pair spawning, strengthen pair bond, and The aim of this study was to analyze the reciprocal egg- avoid interference of other individuals in establishing the pair exchange mechanism in triplets of worms of O. diadema to bond, each isolated pair was allowed to reciprocate egg ex- determine (1) how stable the pair bond is, (2) whether cheat- change four times. This takes an average of 8 days. Then both ers (i.e., individuals that play the male role instead of the partners were offered the option of a new partner when a female role) are present, and (3) how cheated individuals re- third white-egg ovigerous hermaphrodite (the intruder) of spond when they detect cheating. Moreover, we checked the same age was added. We allowed the triplet of worms to whether nonreciprocated individuals respond by deserting interact for 13 days. In this time interval we could observe the their partner. If this was not the case, we investigated the pos- stability of the pair bond. An individual was considered to be sible causes of desertion. a deserter only when, after leaving its previous partner and mating with the intruder, it spawned eggs with the latter. METHODS Yellow or white color of laid eggs allowed us to identify the Study animals female parent (which in addition could be identiﬁed because after spawning its coelom contains no more mature oocytes). Populations of O. diadema were found among the fouling fau- ˚ Generally, paternity was assessed by closeness of the male par- na of California harbors by Akesson (1976) and D.J. Reish ent to the female parent and egg cocoon. Some doubtful cases (personal communication). According to Reish (personal of paternity could be solved by rearing progeny to sexual ma- communication), such populations are at low density (1 ani- turity: if white eggs had been fertilized by a yy male, all the mal per 300 m2 of surface of fouling mussels). However, adults progeny was expected to have white eggs; if white eggs had produce a network of mucous trails that can be followed by been fertilized by a Yy male, half of the progeny was expected conspeciﬁcs, and thus the spatial distribution of animals is to mature yellow eggs. probably clumped. Because adults are 4–5 mm long (14–21 We observed and recorded egg reciprocation of worms dai- setigerous segments) and live in a heterogeneous environ- ly for 13 days; we also recorded body sizes by counting the ment, their behavior cannot be studied in natural conditions. ˚ number of setigerous segments (which also helped us to rec- All the O. diadema life cycle data (Akesson, 1976, 1982) and ognize the two individuals with the same color). Degree of the main features of its mating system (Premoli and Sella, oocyte maturation could be qualitatively assessed through the 1995a; Sella, 1985, 1988, 1990, 1991) have been obtained transparent body walls by looking at the oocyte size. When through laboratory observations. A brief male phase precedes oocytes are fully mature, they mask the underlying gut. the simultaneously hermaphroditic phase, but pairs are We measured frequency of cheating of the original part- formed preferentially between simultaneous hermaphrodites. ners, response to this behavior, frequency of desertion and of Courtship is on the average 4.5 h long. By mutual rubbing pair ﬁdelity, and individual reproductive success. Reproduc- during courtship display, both partners acquire information tive success was measured either as the mean number of co- on the degree of their egg maturation. Mating is achieved by coons per individual or as the total number of eggs per in- pseudocopulation, a process of external fertilization in which dividual and as the total number of free living larvae per in- partners maintain a close physical association before their syn- dividual produced over the 13-day period. The number of co- chronous release of gametes. Partners regularly alternate sex- coons per individuals is also a measure of the frequency of ual roles by reciprocally exchanging a cocoon of 20–25 eggs alternations in sexual roles. every 48 h on average. Immediately after mating the maternal For all statistical comparisons, sample sizes are the maxi- parent remains on the egg cocoon. Its mate stays nearby, and mum number of cases available for a particular data analysis, therefore both parents can be easily identiﬁed. Parents alter- so that, for example, in Table 1 reproductive success is mea- nate in egg caring, especially during the ﬁrst 3 days after egg sured in a subset of the original pairs. Test probabilities re- laying, then the parental care of embryos decreases. Cocoon ported are two-tailed. walls are transparent, and egg development can be easily fol- lowed with a stereoscopic microscope. Unfertilized eggs are generally eaten. At the ninth day after egg laying, offspring RESULTS are released from the cocoon as small four-segment individ- Frequency of cheating uals, soon ready to produce their ﬁrst sperm. When worms reach a body length of 14–15 segments, they become simul- Mean time interval between egg spawnings of a pair when egg taneous hermaphrodites. exchange was regularly reciprocated was 1.78 1.12 (SD) We used individuals of an O. diadema strain derived from days, and mean time interval between successive egg spawn- ˚ individuals collected by B. Akesson in 1976 and 1980 in the ings by the same individual was 3.66 1.13 (SD) days. In the Los Angeles and Long Beach harbors. To recognize animals, 13-day period, 23 out of the original 76 individuals (30%) that we took advantage of a genetically determined polymorphism we used to form the 38 pairs cheated (i.e., did not lay eggs at for a yellow or white egg color. The yellow coloration is de- their turn) at least once. Failure to supply eggs occurred in termined by a dominant allele that controls the presence of 25 out of 313 egg layings (8%). lutein in egg vitellum (Y); the white egg color is due to a recessive allele (y) (Sella and Marzona, 1983). Previously we Response to cheating ascertained that individuals homozygous for the yellow egg allele had the same reproductive rate (i.e., number of eggs/ In 9 out of the 25 episodes of cheating, partners that did not individual/day) as individuals homozygous for the white egg receive the expected eggs, waited on average 5 1.41 days allele and as heterozygous individuals. for an egg offering from their partner (i.e., signiﬁcantly lon- 262 Behavioral Ecology Vol. 11 No. 3 Table 1 Reproductive success of 10 faithful individuals in 10 stable pairs and of 20 deserting individuals in 20 divorced pairs (mean SD) Cocoons Spawned eggs Free-living larvae Faithful individuals 4.7 1.2 165.2 48.90 96.7 58.83 Deserting individuals 5.5 1.5 196.55 61.73 138.6 53.71 Test of faithful vs. deserting individuals, t (df)a 1.398 (28) 1.397 (28) 1.952 (28) a Student’s t test; all values nonsigniﬁcant. ger than reciprocated individuals, Student’s t 3.32; df size of 15 segments, growth differences during the experiment 102; p .01), then deserted and mated with the intruder. gave rise to differences of 1–3 segments in body size. Thus, This behavior occurred in 9 out of 313 egg exchanges (2.9%). we checked whether the old mate preferred its new partner In the remaining 16 episodes of cheating (performed by 14 because it had a larger body size at the moment of partner individuals), the cheated partners re-released their eggs at a desertion. In 11 out of 23 cases the new partner had a larger mean time interval of 3.47 1.12 days. This mean time in- body size; in the remaining 12 cases its body size was smaller terval is not signiﬁcantly different from the mean time inter- than that of the deserted partner. Therefore, data are consis- val between two successive egg layings by the same individual tent with the null hypothesis that worms mated randomly with (3.66 1.13) when reciprocation occurred regularly (Stu- respect to body size (G test; G 0.43; df 1; p .1). dent’s t 0.944; df 108; p .10). The 14 individuals be- (4) Was the old partner deserted because in egg exchanges haved as if they had not detected defection in egg recipro- before desertion it gave its partner signiﬁcantly fewer eggs cation. Frequency of this behavior (i.e., releasing more than than it had received? This hypothesis can be excluded because one parcel without egg reciprocation by its own partner) was no signiﬁcant difference was found between the mean num- 16 out of 313 egg exchanges (5%). ber of eggs spawned by the deserted partner (60.78 36.13) and by the deserting partner (66.15 27.49; paired t test; t 0.75; df 26; p .10). Frequency and causes of desertion (5) Was the intruder preferred by the deserting partner Among the 26 pairs that ‘‘divorced,’’ changes of partners oc- because its oocytes were riper than those of the old partner? curred either once or more than once for a total of 42 times. Under the null hypothesis desertion should occur randomly Only after day 7 was the frequency of pairs in which one of with respect to egg maturation. Out of 39 intruders, 32 were the two partners deserted no longer signiﬁcantly different preferred as mates because they had the ripest oocytes, where- from the frequency expected under the null hypothesis of as under the null hypothesis the expected frequency is 18.5. random matings. At the end of the experiment the frequency The observed frequencies are signiﬁcantly different from of divorced pairs was 26 out of 38 (Ho: random matings; G those expected under the null hypothesis (G test; G 17.23; test; G 0.099; df 1; p .1). In fact, 4 pairs divorced the df 1; p .001). day after the addition of the intruder and 12 between days 2 (6) Did the intruder routinely check out its options, or did and 6 after the addition. Pair bonds lasted more than 6 days it simply mate with whichever of the original pair deserted? in the other 10 pairs that later divorced. The expected frequency under the null hypothesis is 12.5, In only nine of the divorced pairs had the deserting partner whereas 19 out of 25 intruders were observed mating with the been cheated by its original partner. Therefore, cheating can- partner of the original pair which had the ripest oocytes. The not be considered the principal cause of desertion. The fol- observed frequencies are signiﬁcantly different from those ex- lowing hypotheses to explain desertion were tested: pected under the null hypothesis (G test; G 6.96; df 1; p (1) Was desertion contingent upon the previous partner’s .01). We can infer that the new pair was formed between behavior (i.e., failure of the partner to give up its eggs)? Un- the two worms that had the ripest oocytes. However, in the der the null hypothesis, among the deserted partners, individ- ﬁrst egg spawning of the new pair, the deserting partner re- uals that had not reciprocated should be deserted with the ceived no more eggs from the intruder (mean 21.5 2.3) same frequency as those that had reciprocated. In 13 out of than those (mean 22.9 2.1) it received from its old part- 36 cases desertion followed the partner’s failure to give up ner in its last spawning before divorce. eggs, while in the remaining 23 cases desertion was not con- Twelve pairs were faithful to each other from the very be- tingent upon failure to be reciprocated. The hypothesis must ginning of the experiment to its end and spawned or fertilized be rejected that desertion is a retaliation against a partner that their eggs on average 8.7 times. Among them, in seven cases does not give up its eggs (G test; G 2.78; df 1; p .10). either the intruder’s oocytes were less ripe than those of ei- (2) To save energy, a deserting partner who already played ther partner or resorbed; in three cases there was no differ- the male role should prefer to play the male role again with ence in egg ripeness among the triplet of worms; in two cases its new partner. Out of 28 deserting individuals, 19 played the the intruders’ oocytes varied in degree of maturation during male role after desertion, but most of them (13 out of 19) the 13-day observation period. behaved as females before deserting. Only 6 individuals played the male role twice. The other 9 played the female role (5 assumed the female role twice and 4 abandoned the Reproductive success of deserting and faithful individuals male role for the female role). Contrary to our expectations, no preference for the male role resulted from our data (G Estimations of reproductive success of deserting and faithful test; G 2.62; df 1, p .1), but the sample size is too small individuals over the 13-day period are listed in Table 1. Dif- to allow us to make a reliable conclusion. ferences between the three estimates of reproductive success (3) Although all worms started the experiment with a body of individuals that changed partners and individuals that did Sella and Lorenzi • Partner ﬁdelity in a polychaete worm 263 not are not statistically signiﬁcant but would suggest a higher Cooperation in a tit-for-tat strategy requires that a player level of ﬁtness in deserting individuals. recognize its partner among other individuals and recall the outcome of at least the last interaction with that partner (Ax- elrod and Hamilton, 1981; Dugatkin and Wilson, 1992). One DISCUSSION can imagine that such primitive animals lack the ability to In our laboratory population of O. diadema, approximately recognize the individual with which they had the last inter- 30% of individuals occasionally cheat: about 8% of egg layings action, but, according to Axelrod and Hamilton (1981), part- ˚ are not regularly reciprocated. Akesson (1976) estimated that ner recognition could be obtained simply by maintaining in the maximum fertility phase (corresponding to the ﬁrst 7 close contact. In O. diadema, when intruders enter the nesting weeks), an average of 28 4.2 egg masses per individual are sites of the old pairs, recognition (through close contact) of laid. Therefore, in that period, according to our estimate of the old partner would be interrupted, and individuals would the frequency of cheating, an individual runs the risk of being follow the rule of mating (and establishing bouts of recipro- cheated only about twice. Such a frequency of being cheated cation) with the individual that has the ripest oocytes. Our does not seem to be very relevant. In serranid ﬁsh cheating results indicate that 68% of the tested hermaphrodites de- reaches a frequency of 20% (Dugatkin and Reeve, 1998), and serted their partner due to the opportunity for an immediate cheated individuals delayed the next spawning if their part- reward. They deserted independently of what their partner ners failed to reciprocate. This behavior was considered a sort had done in its previous move, contrary to what would be of punishment, followed by a forgiving act (Fischer, 1988). expected under a tit-for-tat model. When cheating is present, some form of punishment is ex- Tit for tat is strictly a two-player game, but animals live in a pected to evolve (Leonard, 1990, 1991). Our results, however, population. This means that they have the option of leaving did not show this: most individuals ignored that they had been their partners and seeking better partners. In this case they cheated and spawned their eggs again without regard for the no longer play tit for tat (Connor, 1992). According to the behavior of their partner. Probably, because cheating is rare, complementarity model (Crowley et al., 1998), regular ex- no selective pressure was exerted for the evolution of retalia- change of roles can evolve in hermaphrodites when providing tory behaviors. In O. diadema both the advantages of cheaters eggs is substantially more costly than providing sperm. Con- and the losses by cheated individuals are sufﬁciently reduced trary to tit for tat, the complementarity model does not ex- by the fact that eggs are laid frequently and in small-sized clude the possibility of desertion. cocoons (Premoli and Sella, 1995). We advance the hypothesis In our experiment we tried to simulate what could happen that, by saving the energy necessary to produce a clutch of in a high-density situation where the opportunity for deserting eggs, cheaters may obtain only little gains, for example, in is maximized. An ovigerous hermaphrodite was present in a lifetime expectancy. Probably other factors than punishment very small bowl containing a reciprocating pair and so the cost keep the selective pressures for cheating low. of searching for a new partner was nil. In nature, populations We cannot consider cheating to be the principal cause of of O. diadema are expected to be at low density. From high- desertion, which had the following timing: 42% of pairs di- density laboratory populations we know that only about 20% vorced within the ﬁrst week and 26% during the second week. of individuals chosen at random are ovigerous at any time Rather, desertion seems to be correlated with attractiveness of (Sella, 1990). If in natural populations ovigerous hermaph- a new partner in terms of egg maturation. By choosing the rodites have the same frequency, costs of deserting should be partner who has the ripest eggs, the deserting partner prob- different from 0 even in crowded populations. Then potential ably increases its reproductive success because it obtains eggs beneﬁts of changing partners should be devaluated by the cost from the new partner sooner than from the old one. This of deserting (searching and courtship costs, risks of preda- supposed reproductive advantage could not be measured for tion), and beneﬁts from protracted mutual cooperation in re- two reasons. Either individuals always chose the best between ciprocal egg exchange should become larger than beneﬁts of the two available partners, be it the old or the new one, or cheating and deserting. Although our study does not give such the short duration of the experiment did not allow us to high- evidence, we may advance the hypothesis that cooperation in light differences between the reproductive success of stable egg-trading persists only as long as the opportunity of pairing pairs and divorced pairs. In ﬁsh, egg trading is constrained by with a more attractive partner is lacking. the necessity of laying all mature eggs within each spawning The results of this study suggest that O. diadema mating period to avoid their decay. In contrast, in O. diadema oocytes system differs in some way from that of serranid ﬁsh, which are preserved from decay because their maturation stops at persist in cooperative behavior and stable pair bonds even in the end of prophase of the ﬁrst meiotic division. This frees high-density conditions. The effect of mating-group size on O. diadema hermaphrodites from the constraint of perform- sex allocation could be analyzed in O. diadema for new in- ing rigorous egg trading and allows them to abandon old part- sights in this kind of egg exchange. ners for new mate, with only minor risks compared to those incurred by egg-trading ﬁsh. Fischer (1988) described egg trading in serranid ﬁsh as a This research was supported by funds from the Italian MURST to G.S. We thank Phil Crowley, Professor Allasia, and two anonymous referees special case of the tit-for-tat strategy (Axelrod and Hamilton, for their helpful comments on a previous version of the manuscript. 1981). In this context the tit-for-tat strategy would be provid- ing eggs to a partner (cooperation) unless the partner failed to provide eggs in the previous egg exchange. In the classical REFERENCES tit-for-tat strategy, rewards for cooperation or for defection are ﬁxed, while retaliation against defection is expected and ani- ˚ Akesson B, 1976. Morphology and life cycle of Ophryotrocha diadema, mals should act accordingly without variation or error. How- a new polychaete species from California. Ophelia 15:25–35. ˚ Akesson B, 1982. A life table study on three genetic strains of Ophry- ever, in O. diadema retaliation mechanisms have never otrocha diadema (Polychaeta, Dorvilleidae). Int J Invert Reprod 5: evolved. According to Stephens et al. (1995), when recipro- 59–69. cation by the partner is not certain and when there is the Axelrod R, Hamilton WD, 1981. The evolution of cooperation. Sci- possibility of choosing between more than one partner, defec- ence 221:1390–1396. tion can be more proﬁtable than continuing the game or re- Bateman AJ, 1948. Intrasexual selection in Drosophila. Heredity 2:349– taliating. 368. 264 Behavioral Ecology Vol. 11 No. 3 Connor RC, 1992. Egg trading in simultaneous hermaphrodites: an Leonard JL, 1991. Sexual conﬂict and the mating systems of simul- alternative to tit-for-tat. J Evol Biol 5:523–528. taneously hermaphroditic gastropods. Am Malacol Bull 9:45–58. Crowley PH, 1998. Solving the complementarity dilemma: evolving Premoli MC, Sella G, 1995. Sex economy in benthic polychaetes. Eth- strategies for simultaneous hermaphroditism. J Theor Biol 195:13– ol Ecol Evol 7:27–48. 26. Sella G, 1985. Reciprocal egg trading and brood care in a hermphrod- Dugatkin LA, Reeve HK, 1998. Game theory and cooperation. In: itic polychaete worm. Anim Behav 33:938–944. Game theory and animal behavior, 1st ed. (Dugatkin LA, Reeve HK, Sella G, 1988. Reciprocation, reproductive success and safeguards eds). New York: Oxford University Press; 18–63. against cheating in the mating system of a hermaphroditic poly- Dugatkin LA, Wilson DS, 1992. The prerequisites for strategic behav- chaete worm, Ophryotrocha diadema. Biol Bull 175:212–217. iour in bluegill sunﬁsh Lepomis macrochirus. Anim Behav 44:223– Sella G, 1991. Evolution of biparental care in the hermaphroditic 230. polychaete worm Ophryotrocha diadema. Evolution 45:63–68. Sella G, Marzona M, 1983. Inheritance, maternal inﬂuence and bio- Fischer EA, 1980. The relationship between mating system and si- chemical analysis of an egg color polymorphism in Ophryotrocha multaneous hermaphroditism in the coral reef ﬁsh Hypoplectrus ni- diadema. Experientia 39:97–98. gricans (Serranidae). Anim Behav 28:620–633. Sella G, Premoli MC, Turri F, 1997. Egg trading in the simultaneously Fischer EA, 1988. Simultaneous hermaphroditism, tit-for-tat and the hemaphroditic polychaete worm Ophryotrocha gracilis (Huth). Be- evolutionary stability of social systems. Ethol Sociobiol 9:119–136. hav Ecol 8:83–86. Fischer EA, Petersen CW, 1987. The evolution of sexual patterns in Stephens DW, Nishimura K, Toyer KB, 1995. Error and discounting the seabasses. Bioscience 37:482–489. in the iterated prisoner’s dilemma. J Theor Biol 176:457–469. Leonard JL, 1990. The hermaphrodite’s dilemma. J Theor Biol 147: Trivers TL, 1971. The evolution of reciprocal altruism. Q Rev Biol 46: 361–372. 35–57.