Kinship and Cannibalism

David W. Pfennig

BioScience, Vol. 47, No. 10, AIBS: The First 50 Years. (Nov., 1997), pp. 667-675.

Stable URL:http://links.jstor.org/sici?sici=0006-3568%28199711%2947%3A10%3C667%3AKAC%3E2.0.CO%3B2-U

BioScience is currently published by American Institute of Biological Sciences.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtainedprior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content inthe JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/journals/aibs.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academicjournals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community takeadvantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.

http://www.jstor.orgWed Oct 10 14:41:07 2007

---

Kinship and Cannibalism Understanding why animals avoid preying on relatives offers

insights into the evolution of social behavior

David W. Pfennig

Cannibalism-the ingestion of all or part of a conspecific- is common, occurr ing in

nearly every major vertebrate and invertebrate group (Elgar and Crespi 1992). In some s~ecies . certain indi- viduals even divelo; specialized weaponry to facilitate the capture and c o n s u m ~ t i o n of cons~ec i f ics (Figure 1).One consequence of can- nibalism is that cannibals may occa- sionally have t o choose between eat- ing a relative or a nonrelative. Do " cannibals discriminate among po- tential prey? In particular, do they avoid eating relatives? In this article. " I address these issues by reviewing both the theory of how kinship should affect prey choices and the evidence bearini on these ~redictions. I also explore the beha;ioral mechanisms used by animals to identify kin.

Interest in k i n s h i ~ and cannibal- ism stems from the advent of modern behavioral ecology, which has tended to view natural selection-the agent of most evolutionary change-as a struggle between genes rather than one between individuals or groups. Hamilton's (1964) theorv of inclu- sive fitness in'itiated such ah approach and provides a tool for understand- ing how behavior that benefits re-

David W. Pfennig (e-mail: dpfennig@email. unc.edu) is an assistant professor in the department of biology at the University of North Carolina, Chapel Hill, NC 27599-3280. His research interests are in behav- ioral ecology and evolutionary biology, especially the evolution of recognition sys- tems and the evolution and development of alternative phenotypes. O 1997Ameri-can Institute of Biological Sciences.

Often must decide between investing in personal reproduction

(by engaging in indis- criminate cannibalism) Or " a to (bynot

eating that relative)

cipients but decreases the fitness of donors can evolve. Investigations of kinship and cannibalism provide an excellent arena for testing Hamilto- nian ideas on the evolution of altru- istic behavior. A cannibal that avoids preying on kin thereby provides ben- efits t o its relatives, but the cannibal suffers the personal cost of dimin- ished growth or survival by forgoing a meal (Crump 1992). Such costs, however, may be only superficial. A cannibal that does not harm rela- tives may actually act in its own genetic self-interest by propagating genes shared with kin, including genes for kin recognition. I therefore begin my discussion of kinship and cannibalism by examining Hamil- ton's hypothesis.

Evolution of altruism Explaining the occurrence of altru- ism has perplexed biologists ever since Darwin first advanced his

theory of evolution by means of natu- ral selection (~ l exaAder 1974). Af- ter all. if natural selection favors those individuals that produce the most surviving young, how can a behavior evolve that entails helping others t o survive and reproduce, es- pecially if it reduces the helper's own survival and reproduction?

The resolution of this ~ r o b l e m comes from shifting the emphasis in evolutionary theory from individual organisms t o thei; genes (Dawkins 1976). Natural selection can be re- garde'd as the differential survival of alternative alleles, and individual organisms can be viewed as tempo- rary vehicles built by genes to enable them to survive and reproduce. The most successful alleles in this comDe- tition are those that most effectively promote their own transmission t o subsequent generations. Often this selection favors those individuals that produce the most surviving young. However, alleles can be shared not just by parents and offspring, but by all relatives, which can have similar alleles through descent from a com- mon ancestor. Natural selection does not distinguish alleles transmitted through descendants f rom those transmitted through collateral rela- tives, such as siblings. Thus, indi- viduals can propagate their alleles not only by reproducing themselves but also by helping relatives to re- produce. In essence, altruism-pro-moting alleles spread because they promote care for copies of them- selves (Dawkins 1976).

Hamilton (1964) formalized this idea and used it td develop a far-

Figure 1. Tiger salamander larvae (Ambystoma tigrinurn) occur in nature as two alternative morphs: a typical morph (upper left) that feeds mostly on invertebrate prey, and a cannibal morph (upper right) that has a wider gape, enlarged teeth, and modified oral bones to facilitate the ingestion of conspecifics (bottom). Cannibals are induced facultatively by high densities of conspecifics. Upper photos: James Collins.

reaching evolutionary principle, Hamilton's rule, which describes the spread of an altruism-promoting al- lele. This rule incorporates three terms: the coefficient of relatedness, r, between an altruist (the "actorn) and the individual receiving help (the "recipient"), where r is a measure of the genetic similarity between indi- viduals by virtue of descent from a common ancestor; the cost of the act, c, in terms of future offspring production that the actor loses by

behaving altruistically; and the ben- efit of the act, b, in terms of the extra offspring that the recipient gains. Hamilton's rule states that an allele for altruism will increase in frequency in a population if the following in- equality is satisfied:

In other words, an altruism-pro- moting allele will increase in fre- quency in a population if the genetic

profit from the altruistic behavior, which equals the recipient's extra offspring production multiplied by the relatedness of the altruist and the recipient, exceeds the genetic loss, which is the decrease in the altruist's offspring production resulting from its efforts to help other individuals.

Hamilton termed this measure of evolutionary success inclusive fitness, because it encompasses both trans- mission of alleles through offspring (the "direct" component of an actor's fitness) and transmission of copies of those alleles through collateral relatives (the "indirect" component of an actor's fitness). Inclusive fit- ness theory, or kin selection, helps to explain the evolution of nepotism, particularly in the unusual organ- isms, such as ants, many bees and wasps, and naked mole-rats, in which individuals have evolved to be subfertile or sterile and spend their lives nurturing collateral relatives (Hamilton 1964).

Kin recognition Although Hamilton's theory seemed to resolve an age-old problem-ex- plaining the evolution of altruism- it initially met, nevertheless, with skepticism, because it predicted that organisms would be able to identify relatives. Other than parent-off- spring recognition, such abilities were unknown for any organism. Soon, however, investigators began to look for-and to find-Hamilton's antici- pated kin recognition abilities in mul- tiple vertebrate and invertebrate taxa and across the spectrum of social and cognitive complexity (Fletcher and Michener 1987, Hepper 1991).

Recently, several authors have suggested that the existence of kin recognition does not necessarily im- ply that natural selection has favored organisms to discriminate kin per se (e.g., Grafen 1990). In particular, the finding that certain organisms preferentially assist relatives rather) than nonkin cannot always be inter- preted as evidence of kin selection, because numerous adaptive and nonadaptive explanations could ac- count for such behavior (Emlen et al. 1991). For example, many frog and toad tadpoles associate prefer- entially with kin. Although this be- havior has been suggested to reflect

BioScience Vol. 47 No. 10

selection to recognize rela- No kin recognition Kin recognition formicivorus; Mumme et al. tives and to dispense nepo- 1983), and many species of tism (e.g., Blaustein and rodents (Elwood 1992). Waldman 1992), tadpoles may associate with any conspecifics that smell like the natal site as an inciden- tal effect of adaptive habitat selection (Pfennig 1990). That is, remaining near home, rather than helping

e \ A a-ar - <a rc--b+rz2za

A A A \.\ ;"\ 4 / ,q /I _qK

s 3 r a--O

-' ~--3cJL -r> L* -k - k d

Some species even avoid collateral, or non-descen-dant, kin. For example, when presented with their sisters' and less-related fe- males' broods, paper wasps (Polistes fuscatus) preferen- tially consume the latter

kin, may be the selective (Klahn and Gamboa 1983). Figure 2. Any gene that permits a cannibalistic bearer to recog- basis for this example of nize kin should spread because it promotes care for copies of Honeybees (Apis mellifera) apparent kin recognition. itself. Imagine two fish families, one of which does not carry such also avoid cannibalizing

therefore, the mere a gene and thus cannot identify its kin (left), and the other of kin. In particular, existence of kin recognition which is genetically predisposed to distinguish kin (right). par- workers discriminate be- does not demonstrate that ents produce two offspring each (only one parent is shown), but tween queen- and worker- the behavior evolved be- not all offspring survive, because they resort to cannibalism laid male eggs, preferen- cause of its beneficial ef- when faced with a food shortage. Thus, in the third generation, tially cannibalizing the fects on relatives. only half of the members of the family that cannot recognize kin latter (Ratnieks and Viss-

One way to determine if survive to reproduce; the others are eaten by siblings. By ,-her 1989). hi^ discrimi-a given instance of kin rec- contrast, three out of four fish survive in the family that can nation makes sense in light ognition has evolved by kin identify relatives because half of them ate fish from another ofHamilton,s rule. All hen-family. By the fifth generation, the family that is genetically

is to study kin dis- predisposed to distinguish kin predominates. eybee workers are female, crimination in contexts for and within any given hive, which the inclusive fitness these workers are mostly consequences are unambiguous. In als who prey on heterospecifics (Crump half sisters, because their queen other words, the best contexts for 1992). mother often mates with multiple applying Hamilton's rule are those Cannibalism may be costly, how- drones. Therefore, workers are often for which the associated costs and ever. Because conspecifics are often more highly related to sons of the benefits can be estimated. One such evenly matched in fighting ability, queen (their brothers) than to sons context is cannibalism. Inclusive fit- cannibalism may increase the can- of fellow workers. ness theory predicts that cannibals nibal's risk of injury. Moreover, as I The most dramatic examples of that can recognize and avoid de- discuss later, eating a conspecific may kinship altering prey preferences stroying kin should often be favored heighten the cannibal's chances of come from species in which some over individuals lacking this ability. acquiring deleterious parasites or individuals are at particularly high In essence, avoidance of kin canni- pathogens (Polis 1981). Finally, can- risk of harming relatives. Certain balism is a form of nepotism, and nibals may risk killing relatives and protozoans, rotifers, nematodes, and any gene encoding this behavior diminishing the indirect component wasp and amphibian larvae exist in should spread because it promotes of their inclusive fitness. nature as either of two structurally survival of copies of itself (Figure 2). Thus, the fitness consequences of and behaviorally distinctive mor-

cannibalism are relatively clear-cut, phs-they can be either cannibalistic of cannibalism making it an ideal behavioral con- or noncannibalistic (Polis 198 1).consequences text in which to apply Hamilton's Which form an individual assumes

Cannibalism involves fitness benefits rule. I therefore begin by asking depends primarily on the environ- and costs for the cannibal; the victim whether kinship affects prey choices ment in which it was raised. Thus, in usually incurs only costs, unless by in ways predicted by Hamilton's rule. these "cannibalistic polyphenisms" being eaten it can elevate its inclu- both types can occur within one sive fitness, a nonintuitive point to Kin recognition group of siblings. Inclusive fitness which I will return later. The pri- and cannibalism theory predicts that cannibalistic mary benefits of cannibalism are that morphs should have more refined the cannibal simultaneously obtains Numerous cannibalistic species iden- kin recognition abilities than non- a meal and kills a potential competi- tify and avoid eating kin. Parents cannibalistic morphs, because the tor. In terms of the first benefit, refrain from preying on their own former are at greater risk of harm- cannibalism can make the difference offspring but readily cannibalize less ing kin. between life and death to the canni- related young in desert isopods The strongest support for this pre- bal if alternative food sources are (Hemilepistus reaumuri; Linsenmair diction comes from spadefoot toads, scarce (Polis 198 1).Moreover, indi- 1987), milkweed leaf beetles (Labi- Spea, a desert-dwelling amphibian viduals who prey on conspecifics may domera clivicollis; Dickinson 1992), genus that breeds in rain-filled pools obtain a better balance of the nutri- waterstriders (Nummelin 198 9 ) , in western North America. Spadefoot ents that are necessary for growth poeciliid fishes (Loekle et al. 1982), toad tadpoles possess a unique means and body maintenance thanindividu- acorn woodpeckers (Melanerpes of acquiring extra nourishment,

November 1997 669

omnivores, feeding mostly on detritus and plankton. Occasionally, however, an individual chances to eat another tadpole or a freshwater shrimp. Eating such a meal can trigger changes in that individual's cranial musculature and

Figure 3. Larvae of polyembryonic wasps of the genus Copido- soma. (a) Precocious larvae. (b) Normal lar- vae. (The two figures are not drawn to scale.) Precocious larvae tend to be females that pref- erentially cannibalize their brothers. Drawing reprinted from Godfray 1992 and used with per- mission of Macmillan Magazines Ltd.

thereby hastening de- velopment (Pfennig 1992). All begin life as noncannibalistic

gut morphology, and, most impor- tant, in its dietary preference. These changed tadpoles start to feast ex- clusively on other animals-includ- ing conspecifics.

As predicted by inclusive fitness theory, omnivorous and cannibalis- tic morphs have different kin prefer- ences (Pfennig et al. 1993). Omni- vores vreferentiallv school with

a

, 9

siblings, whereas their cannibalistic siblings preferentially associate with nonsiblings. Cannibals also nip at conspecifics, and, after tasting them, either eat them (if they are nonsiblings) or release them unharmed (if they are siblings).

~ a r v a l tiger salamanders, Am- bystoma tigrinum, also occur as small-headed "typical" morphs that eat ~rimarilv invertebrates but occa- sioially conspecifics and as large- headed cannibal morphs that feed mainly on conspecifics (Figure 1). All larvae are born with the typical morph phenotype. They remain typicals if they develop under low densities of consvecifics: however. under crowded cdnditioni, some in: dividuals transform into cannibals (Collins and Cheek 1983). When housed with smaller larvae'that dif- fer in relatedness, both morphs pref- erentially consume the more distant relatives (Pfennig et al. 1994). They even prefer to eat nonrelatives ( r = 0)

over first cousins ( r = 118). The two tiger salamander morphs do not differ in kin recognition abilities, however, probably because both are normally predacious; both would, therefore, need kin recognition abilities.

A remarkable example of nepotis- tic cannibalism is found in Copido- soma floridanum parasitic wasps, which infest the eggs of noctuid moths. The brood of this wasp are composed of two types of larvae: typi- cal parasitic larvae, which develop into adult wasps, and larger precocious larvae, which move about in the host haemocoel but die before vuvation (Figure 3). Precocious larvae tknd to be females and preferentially attack their brothers (Grbic et al. 1992). Cannibalism of brothers reflects a; asymmetry in relatedness between males and females of this species. These wasps are polyembryonic, meaning that each egg divides asexu- ally to produce a brood of geneti- cally identical siblings that feed to- gether on the same host individual. Brothers and sisters are related to their own sex by r = 1. However, because wasvs (as well as ants. bees. and certainlother animals) have a form of sex determination known as haplodiploidy-in which males develop from unfertilized eggs and are haploid, and females develop from normally fertilized eggs and are di~loid-brothers are related to sistlrs by 112, whereas sisters are related to brothers by 11.4. Thus, by preferentially killing their broth- ers rather than their sisters, preco- cious larvae benefit their more

closely related sisters and, indirectly, themselves.

Cannibalistic polyphenisms, therefore. vrovide remarkable illus- trations d ihow a cannibal's behav- ior is responsive to its kinship envi- ronment. Thev also reveal how kinship enviroiment might affect morphological development as well. Morphogenesis-like behavior-is often context devendent. and an abil- ity to alter moiphological develop- ment in response to kinship environ- ment would be especially valuable to organisms exhibiting cannibalis- tic polyphenism, in which individu- als develop structures that can be used as weaponry against conspecif- ics and, potentially, kin. In two re- cent tests of this hypothesis, tiger salamander larvae and spadefoot toad tadpoles were found to be sig- nificantly more likely to express the cannibal phenotype in mixed sibship groups than in pure sibship groups (Pfennig and Collins 1993, Pfennig and Frankino in press). Sibship-spe- cific olfactory signals, which are used in assessing whether to cannibalize more distait relatives in preference to close kin, appear to mediate larval development into a cannibal.

Although the feats of discrimina- tion outlined above are impressive, their existence is neither necessarv nor sufficient to demonstrate th& they are the result of kin selection. Indeed, not all cannibalistic animals avoid eating kin. For example, indis- criminate cannibalism of kin and nonkin has been observed in hatchling snails (Baur 1992), among nestmate ant queens (Leptothorax acervorum; Bourke 1994), in certain families of tiger salamanders (Pfenniget al. 1994), and, occasionally, in spadefoot toad tadpoles (Pfennig et al. 1993). A few species even ingest kin primarily; for example, in some sharks, only one embryo may survive in each oviduct after it engulfs related embryos and egg capsules (Wourms 1977); larval marbled salamanders (Ambvstoma opacum) prefer to eat siblings as opposed to nonsiblings (Walls and Blaustein 1995); acorn woodpeckers occasionally consume eggs of close kin (Mumme et al. 1983); and lactat- ing female black-tailed prairie dogs (Cynomys ludovicianus) cannibalize offspring of close relatives (Hoogland 1985). After examining how kin are

670 BioScience Vol. 47 No. 10

identified, I will return to the ques- tion of why some animals fail to avoid killing and eating kin.

How are kin recognized? The process of kin recognition can be conceptually partitioned into three steps: a recognition cue or label is ~roduced:an individual oerceives and interprets this label; consequently, it acts appropriately (Sherman et al. 1997). The labels may be produced by recivient organisms themselves

L,

("phenotypic recognition"), or, al- ternatively, they may lie in clues re- lated to time or vlace that do not involve recipient phenotypes ("non- phenotypic recognition"). The latter function only when relatives are en- countered at vredictable times or lo- cations. For example, acorn wood- peckers live in communal nests with several sisters. and females some-times destroy and eat their sisters' eggs. After a female begins to lay her own eggs, however, she will not dis- turb any eggs in the nest (Mumme et al. 1983), implying that these birds rely on timing clues to recognize their own eggs.

Timing cues are also implicated in parent-offspring recognition in sev- eral rodent svecies (Elwood 1992). Males and fimales bf such species show a temporal shift in infanticidal tendencies such that they cannibal- ize pups only when their own young are unlikely to be present. Thus, in female hamsters, gerbils, and wild house mice. cannibalism ceases with the onset of pregnancy. After their young are weaned, however, females typically resume pup cannibalism. This behavioral inhibition of canni- balism is probably mediated by the dramatic hormonal changes that oc- cur during gestation. Males also be- come less cannibalistic during times when their own young are likely to be encountered-namely, after they have mated and at approximately the time when their young are born.

Other organisms use locational cues to identify kin. For example, before engaging in cannibalistic be- havior, cannibalistic female spiders, scorpions, and Tribolium beetles leave the area into which their young disperse (Polis 1984), thereby de- creasing the probability of killing their own offspring.

November 1997

Other votentiallv cannibalistic organisms use phenotypic cues to iden- tify kin. Such phenotypic recognition cues can be any aspect of an organism's phenotype that signifies kinship reli- ably (Pfennig and Sherman 1995, Sherman et al. 1997). Thus, Belding's ground squirrels (Spermophilus beldingi), paper wasps, desert iso- pods, honeybees, and tiger sala-mander larvae can recognize rela- tives by assessing their fakiliarity to odors of conspecifics. Most poten- tiallv cannibalistic animals ~ robab lv employ both phenotypic and non-phenotypic cues to identify relatives. For example, spadefoot toad tad- poles avoid eating relatives prima- rily by avoiding those areas of the pond where their relatives are found and secondarily by using taste to distinguish the phenotypes of kin and nonkin (Pfennig et al. 1993).

In the second stage of kin recogni- tion, a potential cannibal perceives the label associated with the recipi- ent and assesses its relatedness. Be- fore it can do so, an organism must learn these kin recognition labels from other relatives, its environment, or itself (Sherman et al. 1997). For instance, '~eldin~'sground sqiirrels learn recognition odors from nest- mates, which tend to be kin (Sherman and Holmes 1985). Paper wasps, by contrast, learn recognition odors from their nest, which serves as a rich repository of their nestmates' phenotypes (Gamboa et al. 1986). Some organisms, however, have no opportunity to learn about kin reli- ably from their relatives or their en- vironment (Sherman et al. 1997). These must resort to self-inspection through a phenomenon dubbed the "armpit effect" (Dawkins 1982). For example, when honeybees are reared in isolation. thev learn their own odor , , and then favor similarly smelling full sisters over paternal half sisters, whose slightly different genetic makeup re- sults in a different odor (Getz 1991).

Regardless of whether an indi-vidual learns kin recognition cues from its relatives, the environment, or itself, that individual forms a memory, or "template," of these la- bels. Recognition occurs when re-cipient phenotypes later match these templates closely enough (Sherman and Holmes 1985). Once recogni- tion has occurred, the outward mani-

festation of kin recognition-kin dis-- crimination-takes place only when the actor decides what action to take. This decision is often context depen- dent. In general, individuals are ex- pected to treat relatives and non-relatives differently only when the inclusive fitness benefits of discrimi- nation exceed the costs (Reeve 1989).

Hamilton's rule and cannibalism I now turn to the perplexing issue of why cannibalistic animals sometimes recognize kin and sometimes do not. We can address this issue by apply- ing Hamilton's rule to determine whether individuals facultatively adjust their behavior in ways that enhance their inclusive fitness. Such plasticity in expression of cannibal- istic behavior provides the best evi- dence for kin-selected kin recogni- tion in cannibalistic species. This flexibility is predicted by Hamilton's rule, which says that organisms should be sensitive to all three pa- rameters (r, b, and c), each of which can vary independently of the other two. Thus. indiscriminate cannibal- ism of kin and nonkin-indeed, even cannibalism of siblings-can be fa- vored if there is a sufficient decrease in b and/or increase in c to offset the increase in r. For example, the ben- efits of discriminating kin (6) will decrease if targets of cannibalism are

u

unlikely to reproduce anyway (e.g., when they are too small or mal-formed). and the versonal costs of ,, such discriminations to the cannibal (c) will increase if the benefits of feeding outweigh the fitness costs of eliminating even a close relative (e.g., when the cannibal risks death or reproductive failure by forgoing a meal: Eickwort 1973). Cannibalistic spadefoot toad tadpoles, for instance, are less likely to avoid eating siblings when they are hungry than when they have just eaten (Figure 4) . Pre- sumably, cannibals cease to discrimi- nate kin when their own survival is threatened-that is, when the cost of passing up a meal increases. This flexibility in kin preference makes sense, because a cannibal is always more closely related to itself than to its sibling.

Some species may cannibalize closely related individuals, includ-

671

24 h 48h 24 h Interval between meals

Figure 4. Cannibalistic animals are ex- pected to treat relatives and nonrelatives differently only when the inclusive fit- ness benefits of discrimination exceed the costs. Food was withheld from five carnivore morph spadefoot toad tad- poles for 24 li, 48 h, and then 2 4 h again (i.e., sequentially over a 96 h period). At the end of each fasting period, the canni- bals' consumption of siblings and nonkin was assessed. An asterisk indicates that the observed value was significantly less than random expectation (indicated by the horizontal line). This experiment shows that when satiated (i.e., when they have not eaten for only 2 4 h), can- nibalistic carnivore morph spadefoot toad tadpoles eat significantly fewer kin than nonkin. When hungry, however, (i.e., when they have not eaten for 48 h) these same animals engage in indiscrimi- nate cannibalism. Data are from Pfennig et al. 1993.

ing siblings, because close kin are often the only conspecifics available to a malnourished organism; again, even.cannibalism of close kin can be favored in such desperate situations because a cannibal is .more closely related to itself than to its kin. In these cases, the victim serves as a "food cache" of energy and nutri- ents for its kin. There may even be selection for sibling cannibalism in species in which the mother is unable to provide her offspring with the provisions they need to develop: Vic- tims essentially function as packages of live meat for their kin, a phenom- enon dubbed the icebox hypothesis (Alexander 19741. An extreme case bf this food-cache strategy can be seen in animals (e.g., certain insects and frogs) that produce infertile "trophic" eggs to nourish their new- born.

Although these ideas help to clarify why certain cannibals engage in in-

discriminate cannibalism, they do not explain why some species, as in some of the examples given above, actu- ally prefer to eat kin. One possibility is that such examples may reflect altruism by the victim. In other words, it may be in a victim's best genetic interest to sacrifice itself to a cannibalistic relative if the victim is unlikely to reproduce anyway (see box page 673).

In general, the decision of whether to cannibalize or not (or whether to sacrifice oneself to a cannibal or not) is a problem of allocation. Cannibals often must decide between investing in personal reproduction (by engag- ing in indiscriminate cannibalism) or "helping" a relative to reproduce (by not eating that relative). Natural selection should strike a balance so that even cannibalism of close kin can evolve if the disadvantages of decreasing the indirect component of the cannibal's inclusive fitness are less than the consequences of starva- tion or reproductive failure owing to inadequate nutrition.

Viewing cannibalism as a way to reallocate resources can also be gen- eralized to include cases of redeploy- ment of tissues or cells within an individual organism. For example, many animals "cannibalize" their own muscle tissue as a means of acquiring needed protein when food intake is reduced. In this case, the cost of destroying such tissue is gen- erallv minimal. because the tissue bein; absorbed has no chance of getting into the germ line. Such in- tra-organismal cannibalism should involve greater "cooperation" be- tween cannibal and victim than in- ter-organismal cannibalism, because the different units (e.g., soma and germ line) of a single individual are closely related. By contrast, inter- organismal cannibalism typically takes dace between nonkin and is. therefbre, characterized by conflic; (see box page 673).

Another hypothesis for discriminatory cannibalism Kinship may influence the prey choices of cannibals for evolution- ary reasons other than, or in addi- tion to, kin selection. To see why, we return to an important cost of canni- balism: pathogen transmission.

Many pathogens are known to be transmitted via cannibalism (Polis 1981). Perhaps the most famous dis- ease whose spread is linked to canni- balism is kuru, a deadly nerve degen- eration disease afflicting humans. Although the exact cause of kuru remains elusive, the disease appears to be transmitted infectiouslv from victim to cannibal through contami- nation of hands during handling of infected tissue (Klitzman et al. 1984).

A more direct connection between disease transmission and cannibal- ism was established in natural popu- lations of tiger salamanders (Pfennig et al. 1991). In populations in which cannibals are rare, late-summer bac- terial blooms correlate with massive die-offs of salamanders (Pfennig et al. 1991). During epidemics, most salamanders in any given pond die, but cannibal morph larvae are dis- proportionately represented among the dead (Pfennig et al. 1991). The over-representation of cannibals among the dead may reflect lethal bioaccumulation of pathogenic bac- teria in tissues of cannibalistic larvae that feed on sublethally infected con- specifics. In support of this explana- tion, cannibals in the laboratory that were fed a single diseased conspe- cific were more likelv to die from disease than were typicals that were simply exposed to (but did not eat) a diseased animal, or cannibals that consumed only a single healthy con- specific (Pfennig et al. 1991).

Although these results show that pathogens can be transmitted via cannibalism, they do not necessarily show that intraspecific predation is more costly than interspecific preda- tion. Pathogens can be transmitted by interspecific predation as well as by intraspecific predation, and eat- ing a diseased heterospecific may be as dangerous as eating a diseased conspecific. However, individuals may be more likely to acquire patho- gens from conspecifics than from heterospecifics (Freeland 1983) be- cause conspecifics are more similar genetically than heterospecifics; thus, selection for pathogen host specific- ity and resistance to host immune defenses will render a pathogen more likely to infect a conspecific than a heterospecific. Indeed, cannibalistic tiger salamanders that were fed dis- eased conspecifics were significantly

672 BioScience Vol. 47 No. 10

Cooperation and conflict between cannibal and victim Are victims of cannibalism ever expected t o sacrifice themselves t o cannibals? Cannibalism of kin suggests a n intriguing possibility: Occasionally, it may be in an individual's best interest t o sacrifice itself t o a cannibalistic relative, if, by doing so, the victim can increase its inclusive fitness. T o identify the conditions favoring "cooperation" between cannibal and victim, I use a general form of Hamilton's rule (Grafen 1984) , which predicts that strategy x (e.g., cannibalism) will be favored over an alternative strategy y (e.g., no cannibalism) if

where r is the coefficient of relatedness between actor (e.g., cannibal) and recipient (e.g., victim of cannibal- ism); (Bx - By) is the expected change in number of offspring accruing to the recipient if the actor performs strategy x as opposed to strategy y; and (BIZ - BIy) is the expected change in number of offspring accruing to the actor if it performs strategy x as opposed to strategy y.

Compare the two strategies from the cannibal's perspective-engaging in cannibalism (cannibalism) versus not engaging in cannibalism ( n o cannibalism)-and from the victim's perspective-sacrificing oneself t o a cannibal (cannibalism) versus not sacrificing oneself t o a cannibal ( n o cannibalism). The expected number of offspring produced by the cannibal is C if it engages in cannibalism, and C ' if it does not engage in cannibalism. The expected number of offspring produced by the victim is 0 if it sacrifices itself, and V if it does not sacrifice itself.

First, from the cannibal's perspective, what conditions favor cannibalism? The cannibal is the actor, and the quantity (BIx - B'y) in the equation above therefore equals ( C - C ' ) , which is the change in the expected number of offspring produced by the cannibal when it engages in cannibalism versus when it does not. The quantity (Bx - By) equals ( 0 - V), which is the change in the expected number of offspring produced by the victim when it is cannibalized versus when it is not. Substituting these values into the equation above yields

Y(-V)+ ( C- C ' ) > 0

which simplifies to

r c ( C- C') /V

Next, under what conditions is a victim favored to sacrifice itself t o a cannibal? In this case, the victim is the actor, and the quantity (B', - B',) equals (0- V), which is the change in the expected number of offspring produced by the victim when it is cannibalized versus when it is not. The quantity (B, - B ) equals ( C - C' ) , which is the change in the expected number of offspring produced by the cannibal w&en it engages in cannibalism versus when it does not. Substituting these values into the first equation yields

Y ( C - C ' ) + (-V) > 0

which simplifies t o

Thus, if V < ( C - C ' ) , then ( C - C ' ) / V > 1, and the cannibal will always be expected to engage in cannibalism, because the third equation, r c ( C - C ' ) /V, can never be satisfied (because r cannot exceed 1). Cannibalism will also be favored from the victim's perspective (i.e., it is expected to sacrifice itself), so long as the fifth equation, r > V / ( C - C ' ) , is satisfied. In other words, we expect cooperation between cannibal and victim.

By contrast, if V > ( C - C' ) , then the cannibal will be expected to engage in cannibalism only if the third equation, r c ( C- C ' ) /V, is satisfied. However, the victim is never expected to give itself up willingly, because V / ( C- C ' ) > 1, and the fifth equation, r > V / ( C- C' ) , can never be satisfied because, again, r cannot exceed 1. In other words, there should be conflict between cannibal and victim.

T o summarize, cannibal and victim are expected to cooperate if V < ( C - C ' ) and r > V / ( C - C ' ) . By contrast, they are expected to conflict if V > ( C- C' ) and r c ( C- C')/V. Empirical da ta clearly demonstrate that conflict is more common than cooperation. This observation makes sense, because the condition for conflict, V > ( C - C') , will almost always be met. In other words, the total number of offspring produced by individuals tha t escape cannibalism (V) will almost always be greater than the change in number of offspring accruing to cannibals as a result of engaging in cannibalism ( C - C' ) . Note, however, tha t self-sacrificial behavior will be favored if the victim has little chance of reproducing on its own (i.e., if V = 0 ) .

November 1997 673

more likely to die from disease than those that were fed similarly dis- eased heteros~ecifics.'

Pathogen transmission is clearly an important cost of cannibalism, and it may provide a general expla- nation for whv cannibalism is infre- quent in most species. Selection against pathogen transmission may also help to explain unusual cases in which food-limited animals kill con- specifics but do not consume them. For instance, lactat ing female Belding's ground squirrels kill in- fants of nearby mothers but rarely ingest their victims (Sherman 1981); the larvae of two species of mosqui- toes (genus Toxorhynchites) attempt to kill, but do not eat, all other larvae in their tree hole or bamboo hollow (Hamilton 1970); and cer- tain nestling birds kill their siblings but do not eat them (Stanback and Koenig 1992). Presumably, these animals forgo consuming a freshly killed prey item, even though there would be no risk of retaliation or of diminishing the indirect component of the cannibal's inclusive fitness (be- cause the victim is already dead), as a conseauence of the enhanced risk of acquiring deleterious pathogens.

The observation that pathogens may be especially transmissible among conspecifics has important implications for understanding the evolution of kin recognition in canni- balistic species. Just as conspecifics may be more likely than heterospecifics to infect each other, close relatives may be more likely than nonrelatives to exchange pathogens. Indeed, some pathogens exhibit such a high degree of host specificity that they only in- fect hosts with a single genotype (Burdon and Jarosz 1988). More- over, there is mounting evidence that contagious, debilitating pathogens may be more highly transmissible among genetically similar conspecif- ics (i.e., kin) than among less geneti- cally similar conspecifics (i.e., nonkin; Shykoff and Schmid-Hempel 1991). Again, this difference may reflect the fact that individuals that have diverged from one another more recently (e.g., conspecifics and kin, as compared with heterospecifics and nonkin) have more similar immune

ID. W. Pfennig, S. G. Ho, and E. A. Hoffman, 19 9 7, submitted manuscript.

systems and are, thus, afflicted by similar varieties of pathogens.

If individuals are indeed more likely to acquire pathogens from rela- tives than from nonrelatives, then natural selection should favor indi- viduals that avoid those forms of contact with close relatives, such as cannibalism, that facilitate disease transmission. Whether selection to avoid harmful diseases can explain kin recognition behavior in any can- nibalistic species, however, is cur- rently unknown. Future investiga- tion is required to resolve this important issue.

Conclusions The fundamental premise of inclu- sive fitness theory is that individuals will behave in ways that maximize their genetic representation in subse- quent generations. Hamilton's rule- the mathematical embodiment of inclusive fitness theory-provides us with a powerful theoretical tool for examining the evolution of nepo-tism. Investigations of kinship and cannibalism are ideal for studying the evolution of nepotistic behavior because the fitness consequences of cannibalism are relatively unambigu- ous. Consequently, the parameters in Hamilton's rule can be estimated, making it possible to test inclusive fitness theory. Cannibalism also rep- resents a clear manifestation of the conflict between individuals and is, therefore, an excellent context in which to study the evolutionary reso- lution of such strife (see box page 673).

A number of issues regarding kin- ship and cannibalism require further clarification. Three in particular promise to be fruitful areas for fu- ture research. One issue is the adap- tive significance of kin recognition in natural populations: Does kin rec- ognition enhance the indirect com- ponent of a cannibal's inclusive fit- ness? Another unresolved issue is the role of kinship in facilitating disease transmission. Enhanced risk of ac- quiring disease may explain why can- nibalism is not more common; can disease transmission also explain why some cannibalistic animals avoid eating kin (especially in those in- stances in which a relative is killed but not eaten)? Finally, do victims

sometimes sacrifice themselves to cannibals and thereby enhance the indirect component of their inclu- sive fitness? The answer to this ques- tion may help to further illuminate the conditions favoring cooperation.

Acknowledgments I thank Rebecca Chasan, Karin Pfennig, Kern Reeve, Paul Sherman, and four anonymous referees for helpful comments on this manuscript and Benjamin Pierce for recommend- ing that I prepare this paper. My research was supported by National Science Foundation grants (BSR-9001828 and IBN-9512110).

References cited Alexander RD. 1974. The evolution of social

behavior. Annual Review of Ecology and Systematics 5: 325-383.

Baur B. 1992. Cannibalism in gastropods. Pages 102-127 in Elgar MA, Crespi BJ, eds. Cannibalism: ecology and evolution among diverse taxa. Oxford: Oxford Uni- versity Press.

Blaustein AR, Waldman B. 1992. Kin recog- nition in anuran amphibians. Animal Behaviour 44: 207-221.

Bourke AFG. 1994. Indiscriminate egg canni- balism and reproductive skew in a mul- tiple-queen ant. Proceedings of the Royal Society of London B Biological Sciences 255: 55-59.

Burdon JJ, Jarosz AM. 1988. The ecological genetics of plant-pathogen interactions in natural communities. Philosophical Trans- actions of the Royal Society of London B Biological Sciences 321: 349-363.

Collins JP, Cheek JE. 1983. Effect of food and density on development of typical and cannibalistic salamander larvae in Ambystoma tigrinum nebulosum. Ameri- can Zoologist 23: 77-84.

Crump ML. 1992. Cannibalism in amphib- ians. Pages 256-276 in Elgar MA, Crespi BJ, eds. Cannibalism: ecology and evolu- tion among diverse taxa. Oxford: Oxford University Press.

Dawkins R. 1976. The selfish gene. Oxford: Oxford University Press.

. 1982. The extended phenotype. San Francisco: W. H. Freeman.

Dickinson JL. 1992. Egg cannibalism by lar- vae and adults of the milkweed leaf beetle (Labidomera clivicollis, Coleoptera: Chrysomelidae). Ecological Entomology 17: 209-218.

Eickwort KR. 1973. Cannibalism and kin selection in Labidomera clivicollis (Co- leoptera: Chrysomelidae). The American Naturalist 107: 452-453.

Elgar MA, Crespi BJ. 1992. Cannibalism in amphibians. Pages 256-276 in Elgar MA, Crespi BJ, eds. Cannibalism: ecology and evolution among diverse taxa. Oxford: Oxford University Press.

Elwood R. 1992. Pup-cannibalism in rodents:

BioScience Vol. 47 No. 10 674

causes and consequences. Pages 299-322 in Elgar MA, Crespi BJ, eds. Cannibalism: ecology and evolution among diverse taxa. Oxford: Oxford University Press.

Emlen ST, Reeve HK, Sherman PW, Wrege PH, Ratnieks FLW, Shellman-Reeve J. 1991. Adaptive versus nonadaptive ex-planations of behavior: the case of alloparental helping. The American Natu- ralist 138: 259-270.

Fletcher DJC, Michener CD. 1987. Kin rec- ognition in animals. New York: John Wiley & Sons.

Freeland WJ. 1983. Parasites and the coexist- ence of animal host species. The American Naturalist 121: 223-236.

Gamboa GJ, Reeve HK, Pfennig DW. 1986. The evolution and ontogeny of nestmate recognition in social wasps. Annual Re- view of Entomology 31: 431-454.

Getz WM. 1991. The honeybee as a model kin recognition system. Pages 358-412 in Hepper PG, ed. Kin recognition. Cam-bridge (UK): Cambridge University Press.

Godfray HCJ. 1992. Strife among siblings. Nature 360: 213-214.

Grafen A. 1984. Natural selection, kin selec- tion and group selection. Pages 62-84 in Krebs JR, Davies NB, eds. Behavioural ecology: an evolutionary approach. 2nd ed. Sunderland (MA): Sinauer Associates.

. 1990. Do animals really recognize kin? Animal Behaviour 39: 42-54.

Grbic M, Ode PJ, Strand MR. 1992. Sibling rivalry and brood sex ratios in polyem- bryonic wasps. Nature 360: 254-256.

Hamilton WD. 1964. The genetical evolution of social behaviour. I & 11. Journal of Theoretical Biology 7: 1-52.

. 1970. Selfish and spiteful behaviour in an evolutionary model. Nature 228: 1218-1220.

Hepper PG. 1991. Kin recognition. Cam-bridge (UK): Cambridge University Press.

Hoogland JL. 1985. Infanticide in prairie dogs: lactating females kill offspring of close kin. Science 230: 1037-1040.

Klahn JE, Gamboa GJ. 1983. Social wasps: discrimination between kin and nonkin

brood. Science 221: 482-484. Klitzman RL, Alpers MP, Gajdusek DC. 1984.

The natural incubation period of kuru and the episodes of transmission in three clusters of patients. Neuroepidemiology 3: 3-20.

Linsenmair KE. 1987. Kin recognition in sub- social arthropods, in particular in the desert isopod Hemilepistus reaumuri. Pages 121-208 in Fletcher DJC, Michener CD, eds. Kin recognition in animals. New York: John Wiley & Sons.

Loekle DM, Madison DM, Christian JJ. 1982. Time dependency and kin recognition of cannibalistic behavior among Poeciliid fishes. Behavioral and Neural Biology 35: 315-318.

Mumme RL, Koenig WD, Pitelka FA. 1983. Reproductive competition in the commu- nal acorn woodpecker: sisters destroy each other's eggs. Nature 306: 583-584.

Nummelin M. 1989. Cannibalism in water- striders (Heteroptera: Gerridae): is there kin recognition? Oikos 56: 87-90.

Pfennig DW. 1990. "Kin recognition" among spadefoot toad tadpoles: a side-effect of habitat selection? Evolution 44: 785-798.

. 1 9 9 2 . Polyphenism in spadefoot toad tadpoles as a locally-adjusted evolution- arily stable strategy. Evolution 46: 1408- 1420.

Pfennig DW, Collins JP. 1993. Kinship af- fects morphogenesis in cannibalistic sala- manders. Nature 362: 836-838.

Pfennig DW, Frankino WA. In press. Kin mediated morphogenesis in facultatively cannibalistic tadpoles. Evolution.

Pfennig DW, Sherman PW. 1995. Kin recog- nition. Scientific American 272: 98-103.

Pfennig DW, Loeb MLG, Collins JP. 1991. Pathogens as a factor limiting the spread of cannibalism in tiger salamanders. Oecologia 88: 161-166.

Pfennig DW, Reeve HK, Sherman PW. 1993. Kin recognition and cannibalism in spadefoot toad tadpoles. Animal Behav- ior 46: 87-94.

Pfennig DW, Sherman PW, Collins JP. 1994. Kin recognition and cannibalism in

polyphenic salamanders. Behavioral Ecol- ogy 5: 225-232.

Polis GA. 1981. The evolution and dynamics of intraspecific predation. Annual Review of Ecology and Systematics. 12: 225-251.

, 1 9 8 4 . Intraspecific predation and "in- fant killing" among invertebrates. Pages 87-104 in Hausfater G, Hrdy SB, eds. Infanticide in animals and man: compara- tive and evolutionary perspectives. New York: Aldine Press.

Ratnieks FLW, Visscher PK. 1989. Worker policing in the honey-bee. Nature 342: 796-797.

Reeve HK. 1989. The evolution of conspe- cific acceptance thresholds. The Ameri- can Naturalist 133: 407-435.

Sherman PW. 1981. Reproductive competi- tion and infanticide in Belding's ground squirrels and other organisms. Pages 311- 331 in Alexander RD, Tinkle DW, eds. Natural selection and social behavior. New York: Chiron Press.

Sherman PW, Holmes WG. 1985. Kin recogni- tion: issues and evidence. Pages 437-460 in Holldobler B, Lindauer M, eds. Experimen- tal behavioral ecology and sociobiology. Sunderland (MA): Sinauer Associates.

Sherman PW, Reeve HK, Pfennig DW. 1997. Recognition systems. Pages 69-96 in Krebs JR, Davies NB, eds. Behavioural ecology: an evolutionary approach. 4th ed. Ox- ford: Blackwell Scientific Publications.

Shykoff JA, Schmid-Hempel PS. 1991. Ge- netic relatedness and eusociality: para- site-mediated selection on the genetic com- position of groups. Behavioral Ecology and Sociobiology 28: 371-376.

Stanback MT, Koenig WD, 1992. Cannibal- Ism in birds. Pages 277-298 in Elgar MA, Crespi BJ, eds. Cannibalism: ecology and evolution among diverse taxa. Oxford: Oxford University Press.

Walls SC, Blaustein AR. 1995. Larval marbled salamanders, Ambystoma opacum, eat their kin. Animal Behaviour 50: 537-545.

Wourms J. 1977. Reproduction and develop- ment in chondrichthyan fishes. American Zoologist 17: 379-410.

November 1997

You have printed the following article:

Kinship and CannibalismDavid W. PfennigBioScience, Vol. 47, No. 10, AIBS: The First 50 Years. (Nov., 1997), pp. 667-675.Stable URL:http://links.jstor.org/sici?sici=0006-3568%28199711%2947%3A10%3C667%3AKAC%3E2.0.CO%3B2-U

This article references the following linked citations. If you are trying to access articles from anoff-campus location, you may be required to first logon via your library web site to access JSTOR. Pleasevisit your library's website or contact a librarian to learn about options for remote access to JSTOR.

References cited

The Evolution of Social BehaviorRichard D. AlexanderAnnual Review of Ecology and Systematics, Vol. 5. (1974), pp. 325-383.Stable URL:http://links.jstor.org/sici?sici=0066-4162%281974%295%3C325%3ATEOSB%3E2.0.CO%3B2-D

The Ecological Genetics of Plant-Pathogen Interactions in Natural Communities [andDiscussion]J. J. Burdon; A. M. Jarosz; B. D. HarrisonPhilosophical Transactions of the Royal Society of London. Series B, Biological Sciences, Vol. 321,No. 1207, The Epidemiology and Exology of Infectious Disease Agents. (Oct. 31, 1988), pp.349-363.Stable URL:http://links.jstor.org/sici?sici=0080-4622%2819881031%29321%3A1207%3C349%3ATEGOPI%3E2.0.CO%3B2-8

Cannibalism and Kin Selection in Labidomera clivicollis (Coleoptera: Chrysomelidae)Kathleen R. EickwortThe American Naturalist, Vol. 107, No. 955. (May - Jun., 1973), pp. 452-453.Stable URL:http://links.jstor.org/sici?sici=0003-0147%28197305%2F06%29107%3A955%3C452%3ACAKSIL%3E2.0.CO%3B2-N

http://www.jstor.org

LINKED CITATIONS- Page 1 of 3 -

Adaptive Versus Nonadaptive Explanations of Behavior: The Case of Alloparental HelpingStephen T. Emlen; Hudson K. Reeve; Paul W. Sherman; Peter H. Wrege; Francis L. W. Ratnieks;Janet Shellman-ReeveThe American Naturalist, Vol. 138, No. 1. (Jul., 1991), pp. 259-270.Stable URL:http://links.jstor.org/sici?sici=0003-0147%28199107%29138%3A1%3C259%3AAVNEOB%3E2.0.CO%3B2-G

Parasites and the Coexistence of Animal Host SpeciesW. J. FreelandThe American Naturalist, Vol. 121, No. 2. (Feb., 1983), pp. 223-236.Stable URL:http://links.jstor.org/sici?sici=0003-0147%28198302%29121%3A2%3C223%3APATCOA%3E2.0.CO%3B2-N

Infanticide in Prairie Dogs: Lactating Females Kill Offspring of Close KinJohn L. HooglandScience, New Series, Vol. 230, No. 4729. (Nov. 29, 1985), pp. 1037-1040.Stable URL:http://links.jstor.org/sici?sici=0036-8075%2819851129%293%3A230%3A4729%3C1037%3AIIPDLF%3E2.0.CO%3B2-8

Social Wasps: Discrimination between Kin and Nonkin BroodJeff E. Klahn; George J. GamboaScience, New Series, Vol. 221, No. 4609. (Jul. 29, 1983), pp. 482-484.Stable URL:http://links.jstor.org/sici?sici=0036-8075%2819830729%293%3A221%3A4609%3C482%3ASWDBKA%3E2.0.CO%3B2-9

"Kin Recognition" Among Spadefoot Toad Tadpoles: A Side-Effect of Habitat Selection?David W. PfennigEvolution, Vol. 44, No. 4. (Jul., 1990), pp. 785-798.Stable URL:http://links.jstor.org/sici?sici=0014-3820%28199007%2944%3A4%3C785%3A%22RASTT%3E2.0.CO%3B2-X

Polyphenism in Spadefoot Toad Tadpoles as a Logically Adjusted Evolutionarily StableStrategyDavid W. PfennigEvolution, Vol. 46, No. 5. (Oct., 1992), pp. 1408-1420.Stable URL:http://links.jstor.org/sici?sici=0014-3820%28199210%2946%3A5%3C1408%3APISTTA%3E2.0.CO%3B2-C

http://www.jstor.org

LINKED CITATIONS- Page 2 of 3 -

The Evolution and Dynamics of Intraspecific PredationGary A. PolisAnnual Review of Ecology and Systematics, Vol. 12. (1981), pp. 225-251.Stable URL:http://links.jstor.org/sici?sici=0066-4162%281981%2912%3C225%3ATEADOI%3E2.0.CO%3B2-A

The Evolution of Conspecific Acceptance ThresholdsHudson Kern ReeveThe American Naturalist, Vol. 133, No. 3. (Mar., 1989), pp. 407-435.Stable URL:http://links.jstor.org/sici?sici=0003-0147%28198903%29133%3A3%3C407%3ATEOCAT%3E2.0.CO%3B2-U

http://www.jstor.org

LINKED CITATIONS- Page 3 of 3 -