The Evolution of Delayed Dispersal in Cooperative Breeders

Walter D. Koenig; Frank A. Pitelka; William J. Carmen; Ronald L. Mumme; Mark T. Stanback

The Quarterly Review of Biology, Vol. 67, No. 2. (Jun., 1992), pp. 111-150.

Stable URL: http://links.jstor.org/sici?sici=0033-5770%28199206%2967%3A2%3C111%3ATEODDI%3E2.0.CO%3B2-D

The Quarterly Review of Biology is currently published by The University of Chicago Press.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the 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 at http://www.jstor.org/journals/ucpress.html.

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

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals 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 take advantage of advances in technology. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.org Fri Jul 27 13:35:27 2007 VOLUME67, No. 2 JUNE 1992

THE EVOLUTION OF DELAYED DISPERSAL IN COOPERATIVE BREEDERS

Hustings Reservation and Museum of Vertebrate Zoology, University of California 38601 E. Carmel Valley Road, Carmel Valley, California 93924 USA

WILLIAMJ. CARMEN Hustings Reservation and the Department of Forestry and Conservation, University of California Berkeley, California 94720 USA

RONALDL. MUMME Department of Biology, Allegheny College Meaduille, Pennsylvania 16335 USA

Department of Zoology, University of Washington Seattle, Washington 98195 USA

ABSTRACT Why do the young of cooperative breeders-species in which more than two individuals help raise offspring at a single nest -delay dispersal and live in groups? Answering this deceptively simple question involves examining the costs and benefits of three alternative strategies: (I) dispersal and attempting to breed, (2) dispersal andfloatin~,and (3) delayed dispersal and helping. If, all other thinzs being equal, thefitness of individuals that delay dispersal is greater than thefitness of individuals that disperse and breed on their own, intrinsic benefits are paramount to the current maintenance of delayed dispersal. Intrinsic benefits are directly due to living with others and may include enhanced forazinz efficiency and reduced susceptibility to predation. However, if individuals that disfierse and attempt to breed in high-quality habitat achieve the highest fitness, extrinsic constraints on the ability of offspring to obtain such high-quality breeding opportunities force offsfiring to either delay dispersal or float. Therelevant constraint to independent reproduction hasfrequently been termed habitat saturation. This concept, of itseg fails to explain the evolution of delayed dispersal. Instead, we propose the

delayed-dispersal threshold model as a guide for organizing and eualuatinz the ecological factors potentially responsiblefor this phenomenon. We identify jive parameters critical to the probability of delayed dispersal: relative population density, the fitness differential between early dispersal/ breedinz and delayed dispersal, the observed or hypothetical fitness offloaters, the distribution of territory quality, and spatiotemporal environmental variability. A key conclusion from the model is that no one factor by itself causes delayed dispersal and cooperative breedinz. However, a dzfference in the dispersal patterns between two closely related species or populations (or between individuals in the same population in dzfferent years) may be attributable to one or a small set ~f~factors.Much remains to be done to pinpoint the relative importance of different ecolozical factors in promoting delayed dispersal. This is underscored by our current inability to explain satisfactorily several patterns including the relative significance of floating, geozraphic biases in the incidence of cooperative breedinz, sexual asymmetries in delayed dispersal, the relationship between delayed dispersal leading to helping behavior and cooperative polygamy, and the rarity of the co-occurrence of helpers andfloaters within the same population. Advances in this field remain to be made alonz several fronts. In particular, we advocate experimental tests of specific ecological factors affecting the parameters of the delayed-dispersal threshold model, studies of noncooperatively breedinz taxafocusinz on what constraints to independent reproduction exist and why they do not result in delayed dispersal, and studies of intraspecific variation in group size and composition of cooperative breeders in relation to local habitat gradients and patchiness.

INTRODUCTION tures justify this coupling. First, delayed dispersal and helping behavior usually co-occur. OOPERATIVE, or communal, breed- Second, the beneficial fitness consequences of C ing occurs when more than a pair of in- helping may also be a major selective benefit dividuals exhibit parent-like ("helping") be- for delayed dispersal (Rabenold, 1984; Stacey havior toward young of a single nest or brood. and Ligon, 1991). However, numerous ex- Numerous variations have been identified, ceptions to the first of these features are now including helping at the nest by offspring that known. Delayed dispersal, for example, is of- delay dispersal and remain with their parents, ten not associated with cooperative polyg- and various forms of cooperative polygamy amy, a phenomenon that may be more corn- or plural breeding in which more than a single mon than once thought (e.g., Rabenold et al., male or female share breeding status within 1990). Furthermore, even within systems in the same social unit. ~e~orted~~oradicall~in which nonbreeders delay dispersal, subse- the early 1900s, cooperative breeding was quent helping behavior is not universal [e.g., first reviewed by Skutch (1935, 1961), whose green jays Cyanocoraxyncas in southern Texas extensive experience in 'the neotropics ex- (Gayou, 1986) and Australian magpies Gym- posed him to avian social behavior found only norhina tibicen (Veltman, 1989)], and may rarely in temperate areas. The recent explo- frequently be characteristic of only some off- sion of work in this field stems largely from the spring [e. g., the northwestern crow Corvus cau- theoretical revolution inspired by Hamilton's rinus (Verbeek and Butler, 1981)l. It is there- (1964) theory of kin selection and the evolu- fore important to distinguish the questions of tion of social behavior combined with the ad- delayed dispersal and group living from why, vent of color-banding, which allowed workers once in groups, auxiliaries may subsequently to follow individuals throughout- their lives. help raise offspring not their own. Today, cooperative breeding in birds, mam- The latter of these questions has generated mals and even fishes garners" considerable at- considerable controversv over at least two is- tention in behavioral research. Several excel- sues: What is the relative importance of kin- lent reviews exist, most recently those of ship to the evolution of helping behavior? Brown (1987), Emlen (1991), and the collec- And, is helping behavior an unselected conse- tion of papers in Stacey and Koenig (1990). quence of group living? Valuable discussions Many early discussions of cooperative breed- of the first of these are provided in the reviews ing confound the evolution of delayed dis- cited above and in Emlen and Wrege (1989), persal and group living with the evolution of while the possibility that helping behavior is helping behavior per se. Historically, two fea- an unselected trait is discussed at length by JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 113

Jamieson and Craig (1987) and Jamieson group living in cooperative breeders was Sel- (1989, 1991), who have advocated this con- ander (1964) in his monograph on Campylor- cept, and by Ligon and Stacey (1989, 1991), hynchus wrens, a largely tropical genus con- Koenig and Mumme (1990), and Emlen et al. taining several cooperative species. Selander (1991), who have argued against it on various (1964: 206) wrote: counts. Until recently, the question of why young Under [conditions of consistently high habitat of many cooperative breeders delay dispersal occupancy], a young, inexperienced bird dispersing from the parental territory might have and live in groups has generated considerably little chance of establishing an adequate terri- less controversy: with few exceptions, authors tory and breeding. Possibly there would be have endorsed the concepts of "habitat satu- greater advantage to the individual, in terms ration" or "ecological constraints" (Emlen, of total reproductive success, in delaying re- 1982a). In the years since Emlen's (1982a) production and remaining on the parental paper, however, this apparent consensus has home area on the chance that the parental been repeatedly challenged. Recently, for ex- territory or an adjacent area with which the ample, Stacey and Ligon (1987, 1991) and bird was familiar would become available. Zack (1990) criticized both the validity and Chances for survival presumably would be the logical basis of habitat saturation, propos- greater and, by serving as a helper, the young bird would gain experience in parental activi- ing in its stead their own hypotheses stressing ties which could enhance reproductive success the benefits of philopatry. later in life. Our purpose here is to present a new, inclusive framework for understanding the evolu- This passage is notable for several reasons. tion of delayed dispersal. ~ecauseithas domi- First, it clearly separates delayed dispersal nated much of the thinking in this field, we and group living from helping behavior. Most focus on nonbreeding helpers at the nest, indi- aspects of what has since become known as viduals that have delayed dispersal and inde- the habitat saturation hypothesis are formu- pendent reproduction and provide aid to off- lated to explain delayed dispersal. Selander spring not their own. Aid generally consists also provides a suggestion for the advantages of feeding nestlings or fledglings, but can also helpers might gain by helping. [Written con- include incubation and nest defense. Most of temporaneously with Hamilton's (1964) pa- our discussion will focus on birds, primarily per on kinship and inclusive fitness, and in because avian studies have played a preemi- the absence cf data on relatedness, it is not nent role in this field. The ideas discussed. surprising that the benefits proposed by Sel- however, are equally applicable to coopera- ander pertained to the direct (sensu Brown, tively breeding mammals and fishes, recently 1980), rather than indirect, fitness of helpers.] reviewed by Gittleman (1985) and Taborsky The possibility that territory quality may play and Limberger (1981), respectively, and in an important role in determining the costs some cases eusocial insects (Andersson, 1984). and benefits of delayed dispersal is presented, We begin with a historical review and cri- as is the potential importance of inexperience ticlue of habitat saturation. The review re- on the part of young individuals and the possi- flects our personal biases and is not compre- bility that nondispersers will someday inherit hensive. We present it here to aid in the their natal territory or at least breed nearby. understandhiof the current controversy and There is the implicit recognition that helpers to provide an alternative to the perspective of are better off obtaining a reproductive op- Brown (1987, 1989) on the origin and history portunity on their own, either on their natal of ideas in this field. We then present the territory or nearby, if and when they can. delayed-dispersal threshold model, which we Selander (1964), however, identifies few spe- offer as a conceptual successor to habitat satu- cifics as to the ecological conditions that might ration. favor delayed dispersal and group living other

HABITAT SATURATION than to suggest elsewhere that they might arise under conditions of relatively low, con- Historial Background stant adult mortality rates such as are presum- The first person to develop a concise hy- ably characteristic of "stable" tropical and sub- pothesis to explain delayed dispersal and tropical areas. 114 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67

A second notable ccntribution is the discus- ative breeder, it displays strong sociality in sion of the adaptive significance of territorial- both breeding and nonbreeding seasons. He ity in birds by Brown (1969a,b). Brown de- devotes a section to a comparative discussion vised a model that divided the potential effects of social systems in corvids and specifically of territorial behavior into three levels de- attempts to explain group living in gray- pending on population density. In level 1, the breasted (Mexican) jays (Aphelocorna ultrarna- density is low and all individuals are able to rina) and non-group-living in western scrub breed in their preferred habitat. In level 2, jays (A. coerulescens). Verbeek (1973 : 50) con- the density is higher and some individuals are cludes that the habitat reauirements of the forced to breed in less productive habitats. gray-breasted jay are narrower and that Finally, in level 3, "all habitats where breed- "whereas yearling Scrub Jays, because of the ing could possibly occur are occupied by ter- wider habitat range of the species, success- ritorial individuals, and a surplus of poten- fully exist in marginal habitats, the [Gray- tial breeders exists as non-breeding floaters" breasted] Jay presumably makes limited use (Brown, 1969a: 294, emphasis in the origi- of marginal habitats, because it probably does nal). Such surplus individuals may behave in not fa;or the survival of yearlings." He even a variety of ways, one of which is to remain presents (p. 48) a flow diagram of the possible in their natal groups as nonbreeding helpers. routes to sociality in corvids, an idea (and Brown proposes that this might occur if the figure) also developed by Brown (1974). population surplus is particularly large, in Whereas prior workers suggested only the which case "the probability of an individual's demographic conditions under which delayed eventually gaining the opportunity of success- dispersal in cooperative breeders might be ad- ful breeding . . . may be greater if it remains vantageous, Verbeek went beyond this by within the family group awaiting the demise proposing a set of ecological circumstances of its elders . . . than if it forsakes all claims that might lead to those demographic condi- to its old territory and attempts to establish a tions, specifically, a narrow ecological toler- new one in the face of uniformly fierce de- ance limiting the availability of marginal hab- fense" (p. 316). itat to young birds. In so doing, Verbeek's Brown's (1969a) model, which was not spe- model focused directly on the factors poten- cifically designed to represent cooperative tially constraining the alternative of floating breeding systems, differed subtly but im- as well as dispersal and independent breed- portantly from the earlier one of Selander ing, and thus allowed for the possibility of (1964). In particular, Brown was interested significant constraints to independent repro- in a simple system in which a species bred in duction in both cooperative and noncoopera- two distinct territory types and thus did not tive species. address the importance of territory quality to subsequent theoretical papers published in delayed dispersal. Rather, he suggested that the 1970s reiterated Selander (1964) and level-3 densities arise only when all habitats Brown's (1969a) ideas concerning the demo- where breeding could possibly occur are occu- graphic correlates of cooperative breeding. pied; in this way it is more extreme than Sel- Ricklefs (1975), for example, developed the ander's (1964) formulation of the same idea. idea that cooperative breeding occurs when a Like Selander (1964), however, Brown (1969a) high juvenile recruitment to adult mortality offered no indication of the specific conditions ratioleads to intense competition for space. that might make delayed dispersal a superior Brown (1974) summarized a variety of ideas route to eventual breeding other than the concerning the demographic causes and con- "uniformly fierce defense" offspring must sequences of cooperative breeding, many of overcome owing to the large population sur- which have held up well with subsequent work plus present under level-3 conditions. (see Brown, 1987). Brown (1974), however, did not elaborate on the ecolog.ical factors An hypothesis for what specific conditions .2 might apply to cooperative breeders was fi- leading to those demographic conditions nally provided by Verbeek (1973) in his other than repeating the hypothesis that coop- monograph on the yellow-billed magpie (Pica erative breeders should inhabit "stable, cli- nuttallz). Although this species is not a cooper- max vegetation forms" (p. 73). He did, how- JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 115 ever, suggest that increased productivity of tierizl,,, but failed to do so for Florida scrub groups with helpers is not likely to explain the jays, where the fitness payoff to independent evolution of cooperative breeding. Instead, breeding always outweighed those of helping. Brown proposed that helping behavior in- Emlen (1978: 257) concluded: "we should pre- creases fitness only under the specialized con- dict that non-breeding scrub jays will cease ditions of "intense competition in a popula- being helpers whenever they can successfully tion that is always at or near carrying capac- compete for a vacancy left by the disappear- ity" (p. 77). Although partly a repetition of ance of a previously breeding individual"(em- Selander (1964), an important feature of this phasis in the original). statement is the explicit recognition that in- This conclusion is a clear statement of the trinsic advantages ire unlikely to provide a hypothesis that, all other things being equal, general answer to why cooperative breeders the fitness of nonbreeders delaying dispersal live in groups. and helping is lower than what they would ~esidesthese largelv theoretical contribu- experience if they successfully dispersed to a .2, tions, initial publications from several impor- territory capable of supporting breeding and tant long-term studies of cooperative breeders bred independently. Why, then, do these became available in the period 1965- 1975. nonbreeders not disperse and breed indepen- This empirical work made possible major re- dently? The answer is that all other things are views by Brown (1978), Emlen (1978), and not equal; in particular, dispersal and inde- Gaston (1978). Gaston (1978) discussed mostly pendent breeding is apparently constrained the potential costs and benefits ofhelping, and due to the demographic conditions already like most other authors failed to provide any known by this time as habitat saturation (e. g., ecological context for delayed dispersal other Stacey, 1979a). than mentioning the apparent correlation The alternative to this habitat saturation with warm climates and, like Ricklefs (1975), hypothesis is that nonbreeders experience emphasizing the importance of a demographic higher lifetime fitness by delaying dispersal situation in which there is relatively high adult and helping than by attempting- - to breed inde- survivorship and thus few vacant territories. pendently at an early age because of the in- Brown (1978) discussed at length the life his- trinsic advantages of the former alternative. tory traits of cooperative breeders and reiter- This idea surfaces regularly in earlier works. ated the demographic correlates of this behav- For example, Rowley (1965 : 29 1) concluded ior, again providing no ecological context for superb blue fairy-wrens that "the presence other than to abandon the suggestion that it of supernumeraries in the family group short- is associated with stable and climax environ- ens the frequency of the breeding cycle to the ments. Emlen (1978), in contrast, maintained irreducible minimum occupied by the incuba- support for the hypothesis that cooperative tion and nestling stages and must therefore be breeding is associated with stable environ- regarded as a highly efficient adaptation to a ments. He also discussed the hypothesis that widely varying climate, by a multi-brooded cooperative breeding occurs when "suitable species." Similarly, Brown (1963), in an early habitat becomes filled or 'saturated"' (p. 250), publication on gray-breasted jays, stated "Te- a situation which he referred to as habitat sat- leologically, it would appear that the advan- uration. tages to the individual [Gray-breasted]Jay of The calculations made by Brown (1975, reserving a territory for himself are out- 1978) and Emlen (1978) comparing the rela- weighed-by the advantages gained through tive fitness contributions of a bird that delavs flock membership" (p. 151). He goes on to dispersal to one that disperses and attempts to list three advantages that might be gained by breed on its own were particularly notable. individuals in groups: the acquisition of abet- Despite including the effects of both group ter territory, increased foraging efficiency, living and helping behavior, these calcula- and reduced predation. tions proved inconsistent: they correctly pre- Although various benefits to delayed dis- dicted helping in the superb blue fairy-wren persal and group membership continued to (Malurus cyaneus) and among yearling Tasma- be identified between 1979 and 1984, con- nian native hens (Gallinula [Tribonyx] mor- siderable evidence began accumulating for 116 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67 the importance of the habitat saturation hy- the benefits accruing to grouped individuals pothesis (e.g., Stacey, 1979a; Atwood, 1980; lead to delayed dispersal, as intrinsic benefits. Koenig, 1981a; Woolfenden and Fitzpatrick, These and other papers helped forge a con- 1984). We note two theoretical developments sensus for the importance of ecological con- during this period. First, Koenig and Pitelka straints to the evolution of delayed dispersal. (1981) extending the hypothesis proposed by The consensus was short-lived. howev&. be- Verbeek (1973), developed a graphical model ing broken soon by the models of wiley'and of habitat saturation that we call the "marginal Rabenold (1984), Stacey and Ligon (1987, habitat hypothesis." This hypothesis, dis- 1991), Waser (1988), and Zack (1990), and by cussed in greater detail below, postulated a apparently contradictory data presented by high proportion of "optimal" versus "mar- Zack and Ligon (1985b), Stacey and Ligon ginal" habitats leading to a situation in which (1987), and McCaJlum et al. (in press). In gen- most marginal habitats would generally be eral, these alternative models stress the poten- occupied by individuals excluded from high- tial advantages to delayed dispersal (Hein- quality territories. Few habitats of marginal sohn et al., 1990). They have often been quality would then be available for occupancy advanced as alternatives to ecological" con- by either potential breeders or floaters, lead- straints, but the precise ways in which this is ing to relatively low fitness by individuals true have not always been clear, for reasons choosing either of those alternatives. discussed below. Currentlv we believe it is A second important development was that fair to say that there is more confusion than of Emlen (1982a; see also Orians et al., 1977), agreement concerning the role of ecological who pointed out the analogy between the constraints and habitat saturation to the evo- usualperception of habitat saturation leading lution of delayed dispersal and cooperative to restricted breeding options by offspring, breeding. and conditions where the cost of rearing young is prohibitive owing to adverse condi- Critique of Habitat Saturation tions. In both cases breeding options are re- The cause of this confusion lies squarely stricted, but in the latter situation indepen- with the use of the term habitat saturation. dent breeding is constrained not by the lack Habitat saturation has no generally accepted of a suitable territory, but by the inability definition and has been used in at least three of some individuals to acquire the requisite overlapping contexts. First, it is frequently resources for breeding other than space. With used to refer to population size relative to re- the "skill hypothesis* (see below) Brown (1985, sources, that is, to relative deiisity (e. g., Sta- 1987) expanded this idea to any situation in cey and Ligon, 1987; Jones et al., 1988). In which differential experience causes an en- this sense, habitat saturation is a continuous ergy threshold for successful breeding to be variable that is only meaningful when defined exceeded by some individuals and not others. relative to another population (Zack and Li- Species constrained by space include those gon, 1985a). Second, it is used in a general living under apparently stable ecological con- sense to describe a syndrome of intense com- ditions in which habitat saturation has gener- petition over territories that are rarely vacant ally been invoked to explain delayed disper- (Emlen, 1991; Walters et al., in press). In this sal, while species constrained by the high cost form, habitat saturation encompasses several of rearing young include those living under hypotheses concerned with the specific eco- highly variable conditions in which at least logical conditions leading to the demographic some pairs appear to be unable to breed by conditions that in turn select for delayed dis- themselves, independent of territory quality. persal and cooperative breeding (Walters et Emlen (1982a) subsumed both possibilities al., in press). Third, some authors have used under an ecological constraints model. In or- habitat saturation to refer to one specific set der to di~tin~guishthe two situations, we refer of possible ecological conditions leading to in- to the former. in which individuals do not tense competition over territories; that is, to directly benefit by the presence of others, as only a subset of the general syndrome just extrinsic constraints and the latter, in which referred to above. This more specific defini- JUNE 1992 DELA YED DISPERSAL IN (700PERATIVE BREEDERS 117 tion can be particularly troublesome, since habitat is so poor that delayed dispersal be- authors rejecting habitat saturation are often comes a preferable alternative. The view that doing so only for one possible form of the habitat saturation means that individuals are more general syndrome, not for the general forced to remain in their natal groups because syndrome itself. no unoccupied habitat of whatever quality ex- Consequently, we recommend that the ists (e.g., Zack, 1990; Ligon et al., 1991) has term habitat saturation be used only when it never been part of the concept of habitat satu- is explicitly defined by the user; however, this ration, at least as envisioned by Selander is not the only difficulty with this concept as an (1964), Verbeek (1973), or Koenig and Pi- explanation for delayed dispersal and helping telka (1981). behavior. As used by most authors, habitat Habitat heterogeneity and the scale of assessment. saturation is not a sufficient explanation for In any nondeclining population, the number delayed dispersal for at least the following of potential recruits produced over the long three general reasons. term must at least equal, and will usually ex- Ambiguity. Selander (1964) defined the con- ceed, the number of available vacancies, at ditions now known as habitat saturation as least in high-quality habitat. Thus, most terri- those in which young have little chance of torial populations will frequently exist, at least establishing an "adequate territory and breed- in some subset of their range, under condi- ing." What, however, is an "adequate" terri- tions of habitat saturation in which breeding tory? territories are limited (Koenig and Mumme, A territory that is adequate (or other simi- 1987; Heinsohn et al., 1990; Emlen, 1991). larly vague labels, such as a suitable, success- At least two consequences follow. ful, optimal, or marginal) could be defined on First, habitat saturation is insufficient to an absolute, if arbitrary, basis. For example, explain delayed dispersal and helping behav- an optimal territory might reasonably be de- ior. Hunter (1987), for example, found that fined as one in which, on average, reproduc- limited availability of suitable breeding habi- tive success and survivorship are sufficient for tat in the purple gallinule Porphyrula martinica inhabitants to replace themselves, while a led to floating rather than helping by yearlings marginal territory might be one in which in- even though juveniles regularly acted as help- habitants cannot expect to do so. Applying ers, and Carmen (in press) found that the level such a criterion, however, tells us little about of demographic constraints may be identical the preferred dispersal strategy for individual among cooperative and noncooperative pop- offspring, which will depend on the quality ulations of scrub jays. It follows that habitat of the territories to which they can disperse saturation, although possibly necessary, is relative to that of their natal territory. For not sufficient to distinguish species in which example, because the benefits of remaining young delay dispersal from those in which on the natal territory and helping to raise re- they do not (Stacey and Ligon, 1987; Hein- lated offspring will correlate with the number sohn et al., 1990; Zack, 1990; Emlen, 1991). of offspring raised and hence most likely be Second, the scale at which habitat satura- related to territory quality, the minimally action is assessed is critical: within a single ho- ceptable territory to which an individual born mogeneous habitat of high quality, habitat on a high-quality territory might disperse will saturation is quite likely, whereas habitat sat- be of higher quality than that for an individual uration is unlikely at the level of a large popu- born on a low-quality territory (see below). lation spread out over many habitats of vari- Indeed, what constitutes an acceptable terri- able quality. Within any specified habitat, tory, and thus whether or not the habitat is saturation can be defined as occurring when perceived as being saturated, may very well the number of competitors is greater than the differ for every individual. number of suitable territories (Brown, 1987: Offspring can always disperse and occupy 71). By definition, high-quality habitats will space somewhere; the question is not whether generally fulfill these conditions while low- any unoccupied habitat exists, but under quality habitats will not. Unfortunately, we what conditions the available unoccupied are left with the nontrivial problem of distin- 118 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67 guishing different habitats. We are further sented by Koenig and Pitelka (1981) and is burdened with the even more difficult prob- analogous to the well-known polygyny thresh- lem of assessing available resources, which old model of Verner and Willson (1966) and in turn determines the potential "number Orians (1969). Various factors influencing of suitable territories" within any particular optimal group size in a similar model have habitat. been discussed by Brown (1982, 1987); other For these reasons, we believe that the em- authors who have presented models in the phasis on the demographic correlates of coop- context of delayed dispersal or joint-nesting erative breeding espoused by numerous au- include Gowaty (l981), Waser (1988), Powell thors, epitomized by the use of the term (1989), and S. Zack and B. J. Stutchbury habitat saturation in the general form dis- (pers. commun.). Our model focuses on the cussed above, has been misplaced. Many co- ecological factors leading to differing dis- operative breeders live at relatively high den- persal patterns of individuals compared either sities and under conditions in which young inter- or intraspecifically. birds have a difficult time obtaining reproduc- The basic model is presented in Figure 1, tive vacancies, but so do many noncoopera- where "habitat fitness" or "territory quality" tively breeding species. The critical differ- (Emlen, 1991) gradients are graphed. They ence, rather than being the existence of a are obtained by plotting the fitness state of population surplus and consequent competi- individuals pursuing a particular strategy on tion for space, is that this surplus consists of they-axis. Fitness or fitness states for an indi- offspring that delay dispersal and remain on vidual (ego) are functions of several parame- their natal territories rather than pursuing ters, including (1) ego's reproductive success some alternative strategy such as floating. and survivorship, (2) the reproductive success Circularity. The concept of habitat satura- and survivorship of kin due to ego's help (or tion entails an element of circularity: Does it hindrance), (3) the quality of the habitat or result in reduced dispersal or vice versa (Wi- territory on which ego lives, and (4) the con- ley and Rabenold, 1984; Austad and Rabe- tingent probability of switching to other fit- nold, 1986; Stacey and Ligon, 1987; Zack, ness states in the future as would happen 1990)? Stacey and Ligon (1 987), for example, when, for example, a floater or helper obtains argued that delayed dispersal has been se- a territory in high-quality breeding habitat. lected for by various intrinsic benefits and Parameters (1) and (2) correspond to the indi- that habitat saturation is a result, not a cause, vidual's future direct and indirect fitness ef- of reduced dispersal. Heinsohn et al. (1990) fects (Brown, 1980), and are generally easiest suggested a similar possibility based on the to envision as measured relative to individuals likelihood that most cooperative breeders of comparable age engaged in other fitness have been studied in preferred habitats. thus states. With respect to (3), the complete range biasing our view of the relationship between of available environments is ranked in order habitat saturation and cooperative breeding. from worst to best (based on their quality for This criticism highlights the difficulty that breeding) on the x-axis. This results in mono- there is generally no independent means of tonically increasing curves starting with "non- assessing habitat saturation other than mea- habitat," where the hypothetical fitness of an suring the extent to which individuals do not individual is zero, to the best available habi- disperse and establish independent territories tats, where fitness is relatively high. Parame- when we, as observers, think they should or ter (4) indicates that fitness is not measured could do so. instantaneously, but includes a time dimen- sion incorporating the probability of future THE DELAYED-DISPERSAL THRESHOLD MODEL successful reproduction, given that ego is cur- These difficulties with the concept of habi- rently engaged in a particular strategy. Thus, tat saturation prompt us to advocate a more at any point in time, an individual behaving general framework for the evolution of de- optimally will attempt to pursue the option layed dispersal in cooperative breeders, which yielding the highest fitness state. we call the "delayed-dispersalthreshold rnodeLn The first option, represented by Wd,is ego's The model is a generalization of that pre- fitness if it delays dispersal and remains on its JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 119

able territory and attempts to breed (henceforth "dispersallbreeding"). The third, represented by W,, is its fitness if it disperses but becomes a floater that is not breeding for whatever reason (henceforth "dispersallfloating"). Note that while neither floaters nor helpers presumably reproduce while engaged in these options, both strategies may yield high fitness states by augmenting survivorship and future successful reproduction; in addition, helpers may facilitate the reproduction and survivorship of relatives. Also, although the model focuses on the options available to newly independent offspring, it is not limited to this stage and can also be applied to individuals at later stages in their lives, as long as comparisons are made with other individuals of similar age and status. For all three strategies, an individual's fitness is dependent on the quality of the territory on which it lives; this is represented by the parenthesized subscript. Thus, the fitness Worst Z Y x Best of an individual remaining on territory x as Environments ranked by a nonbreeder is W+) [Fig. l(a)]. This is a suitability for breeding probabilistic value that averages the fitness of all individuals engaged in this option relative FIG. 1. THEDELAYED DISPERSAL to those pursuing alternative options. The fit- THRESHOLDMODEL. ness of an individual that attempts to breed The habitat fitness or territory quality gradient on territory x is W4,,. Finally, the fitness of a for a population is obtained by ranking territories from worst to best on the abscissa, while graphing floater that occupies habitat x is W,(,). As with the fitness of occupants on those territories on the the other options, the fitness of floaters de- ordinate. Gradients can be derived for individuals pends on the quality of the habitat available pursuing alternative strategies. Territory quality to them; under certain circumstances, all gradients for individuals dispersing and breeding floaters may be able to occupy the best avail- (W,), delaying dispersal (Wd), and floating (Wf) are able areas [designated by Wf(,,)], perhaps by plotted here. The values for each of these depend sneaking in interstices of high-quality territo- on the quality of the environment (x) on which the ries, while under other circumstances, they individual lives. (a) Delayed dispersal yields higher may be excluded from high-quality areas and fitness than either dispersing and breeding on a forced to float in inferior habitats. Territories territory of comparable quality or floating, even in the best available habitat, i.e., Wd(,, > We(,)and are ranked according to their quality for Wd(x)> Wf(-), where W,(,,) is the fitness of floaters breeding. This ranking will usually, but not in the best available habitat for floating. (b) De- necessarily, be highly correlated with their layed dispersal yields lower fitness than dispersing quality as habitat for floaters. and breeding on a territory of comparable quality The distinction between intrinsic benefits but higher fitness than floating, i.e., We(,)> Wdw and extrinsic constraints currently main- > WJ(ml. In this case, individuals are forced to taining delayed dispersal and group living in delay dispersal because of extrinsic constraints. this model corresponds to the situations depicted in Figures l(a) and l(b), respectively. In Figure l(a), W,< Wd for all environments, natal territory, during which time it delays and thus an individual's fitness is greater if it breeding (henceforth "delayed dispersallhelp- delays dispersal and helps on its natal territory ing"). The second, represented by We,is ego's than if it disperses and breeds on a territory fitness if it disperses, settles on the best avail- of the same quality as that on which it was 120 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67 born; this corresponds to the situation in (Woolfenden and Fitzpatrick, 1984; Koenig which intrinsic benefits are sufficient to main- and Mumme, 1987; Rabenold, 1990), the fit- tain delayed dispersal (see below). In con- ness differential between W4,)and Wd(*)is es- strast, Figure l(b) shows the situation in sentially the cost of delayed reproduction. which extrinsic constraints are the proximate The fitness differential between W4,) and factor driving delayed dispersal and group liv- Wrcm,)is the cost of delayed reproduction and ing; in this case, W, > Wd for all environ- the additional cost (if any) owing to lower ments. survivorship of floaters compared to helpers. Consider an individual born on territory x [Fig. l(b)]. If it disperses and floats and is Assumptions of the Model able to occupy the best available habitat, it Here we discuss several assumptions rele- experiences fitness Wrc,,,). If it delays dis- vant to the delayed-dispersal threshold model. persal and remains a nonbreeder on its natal These assumptions are difficult to test and, territory, it experiences fitness Wd(+Finally, like those relevant to the polygyny threshold if it disperses and breeds, it chooses the best model (Davies, 1989), are unlikely to be met available territory from those in which vacan- in many,, . if not most, realistic situations. Most cies exist (sayy) and experiences fitness W*). of the assumptions, however, can be relaxed Which of these three options yields the without jeopardizing the model's usefulness; greatest fitness depends by definition on the this is done in the subsequent discussion. relative values of W,((,,), W+), and W*). In Relationship between territory quality andfitness. the example shown in Figure l(a), the intrin- Territories or habitats must vary in quality sic benefits of delayed dispersal and group in ways that directly influence the fitness df living are so great that, assuming the natal inhabitants. The most likely reason this might territory to be of reasonably high quality not be true is because of spatiotemporal envi- (which is quite likely, since high-quality terri- ronmental variation (discussed below; see tories by definition yield the majority of off- also Powell, 1989). If great enough, such vari- spring), this option is superior not only to ation may all but obliterate a correlation in dispersallfloating but to dispersallbreeding, quality of a given territory among years, even on the best available territory. In Figure thereby eliminating any correlation between l(b), territories x and y have been chosen so territory quality over the long term and fit- that Wdp)= W+)j thus, if the best territory ness. For example, this was apparently the available is of higher quality than y, an off- case for the banner-tailed kangiEoo rats (Di- spring should disperse and breed indepen- podomys spectabilis) studied by Waser (1988), dently on that territory, while if it is of lower in which young dispersed despite meeting all quality than y, it should delay dispersallhelp of the expected preconditions for philopatry. on its natal territory. Floating in this example Perfect knowledge. Individuals are assumed is unlikely for any offspring, since only in the to have perfect knowledge of both the exis- event that a bird is born on a territory of qual- tence of vacancies and the quality of territor- ity lower than z and there was no vacant terri- ies or habitats in which vacancies exist, even tory of higher quality, would a bird be better prior to their departure from the natal terri- off floating than delaying dispersal. For the tory, and thus are able to choose the best dis- offspring born on territory x, dispersinglfloat- persal alternative at any point in time. This ing, even in the best habitat, yields lower fit- assumption is dependent on the extent to ness than delaying dispersal and remaining which individuals examine available territor- on territorv x as a nonbreeder. This need not ies and are able to accurately discriminate be the case; indeed, the widespread existence among them prior to and during potential dis- of floaters indicates that Wrc,,) > Wdin many persal events. Some of the meager informa- cases. In most of the examples we discuss, tion currently available on these behaviors is however, Wrcm,)is set unrealistically low so as discussed on pages 124 and 129. to allow us to focus on the differences between Idealgree settlement. Individuals are free to Weand Wd. settle where their expected fitness is greatest. Since nonbreeding helpers are likely to Because of behavioral constraints such as in- have survivorship at least as high as breeders cest avoidance (e.g., Koenig and Pitelka, JUNE 1992 DELAYED DISPERSAL IN (7OOPERATIVE BREEDERS 121

1979), this may not always be true, although acceptable territory can be filled, thereby in- relatively few data are available to assess the cluding possible failure to settle and breed in importance of such constraints in any species. the first and perhaps additional breeding sea- Irrevocability of decision. In its simplist form, sons, and (3)delayed dispersal and remaining the model assumes that the choice of a breed- on the natal territory as a nonbreeder. ing territory is permanent. Territory switch- Two definitions are helpful. First, dispersal ing, however, is readily accommodated by en- is the process of permanently leaving the natal visioning the model to represent probabilistic territory. Once an individual disperses, it can "fitness states" rather than lifetime fitness. either settle on a territory where breeding can That is, once an individual chooses a terri- be attempted or become a floater. Second, tory, it can expect to accrue fitness at a certain floating designates the behavior of nonterrito- rate. This rate is based not only on its ex- rial and nonbreeding individuals that have pected survivorship and reproductive success left their natal territory. Floating is neither a on that territory, but also on the degree to simple nor uniform phenomenon; rather, it is which that territory provides opportunities part of a continuum of space-use behaviors for the owners to later move to a higher-qual- that grade into forays mad; by territorial indi- ity territory, or the degree to which it may viduals in search of reproductive opportuni- expose the owners to the possibility of a de- ties (see below). In particular, note that the cline in resource value forcing desertion and traditional view that floaters are necessarily dispersal in search of a new, more adequate individuals lacking site fidelity is wrong. In- territory elsewhere. stead, floaters frequently employ site-specific Several prior models (e.g., Fitzpatrick and tactics leading to future access to breeding Woolfenden, 1986; Stacey and Ligon, 1987; opportunities (Smith, 1978; Smith and Arc- S. Zack and B. J. Stutchbury, pers. com- ese, 1989; P. Arcese, pers. commun.; S. Zack mun.) explicitly assume that the choice of a and B. J. Stutchbury, pers. commun.; see breeding territory is permanent. Such inflexi- below). bility is certainly not general. For example, The model generally assumes that the three secondary dispersal by breeders, including dispersal options of delayed dispersallhelp- the return to their natal group as nonbreeding ing, dispersallbreeding, and dispersallfloat- helpers after having dispersed and bred else- ing, are qualitatively distinct. However, not where, is common among Californian acorn only may individuals switch back and forth woodpeckers (Koenig and Mumme, 1987), between options, but each may grade into the and extensive territory switching has been others in some species. For example, species documented in both the red-cockaded wood- such as Galdpagos mockingbirds Nesomimus pecker (Walters, 1990) and the noncoopera- parvulus (Curry and Grant, 1990) and western tively breeding western scrub jay (Carmen, bluebirds Sialia mexicana ('J. L. Dickinson, in press). Depending on the degree of such pers. commun.) include individuals that si- flexibility, the option pursued by an individ- multaneously help at one nest while breeding ual at any particular time will be a function of at another. Similarly, individuals that delay the relative quality of territories and expected dispersal and help may simultaneously spend payoffs of the relevant options at that time a significant amount of time foraying in and correspondingly less on long-term mean search of vacancies, at least at some seasons territory quality. (Woolfenden and Fitzpatrick, 1986; Hooge, Dispersal options are distinct. The model as- 1992). Such behavior is tantamount to float- sumes that the ultimate goal of a nonbreedcr ing, with the distinction that individuals have is to become a successful breeder, and that a home base to which they can return when both delayed dispersal and floating are poten- and if needed. Finally, although floaters are tial means to this end. Determination of the generally assumed to be nonbreeders, it is pos- preferred options therefore requires consider- sible, although as yet unsubstantiated, that ation of three alternatives: (1) dispersal to the they may occ~sionallyegg dump (females) or best available territory or habitat where inde- obtain extrapair copulations (EPCs, males), pendent breeding is at least a possibility, (2) thereby blurring the distinction between dis- dispersal and floating until a vacancy in an persallfloating ind dispersallbreeding. We 122 THE QUARTERLY REVZE W OF BIOLOGY VOLUME67 know relatively little about the extent, much Several authors have proposed an alterna- less the significance, of such potential com- tive view that conflict is usually absent be- plexities. tween parents and offspring and that delayed

Furthermore, ours is not the only, way, in dispersal represents a form of "parental facili- which to structure the alternatives available tation" (Brown and Brown, 1984) or extended to offspring. For example, Brown (1987: 103- parental care (Ligon, 1981 ; Fitzpatrick and 104), Zack (1990), and Walters et al. (1992) Woolfenden, 1986; Ligon and Ligon, 1988). consider two dispersal alternatives instead of Overt conflict between parents and offspring three and contrast delayed dispersal ("stay and involving persistence of the latter in groups, foray") with dispersal ("depart and search"). however, can sometimes be directly observed One rationale for this is that floating and inde- (e.g., Hannon et al., 1987; Koenig et al., pendent reproduction are simply different 1988), indicating that conflict does exist, at outcomes of the single strategy of early dis- least under some circumstances. Further- persal; in other words, an individual disperses more, the fact that yearlings and older off- and then either finds an acceptable vacant spring have higher reproductive value than territory, settles, and subsequently attempts eggs or nestlings, cited by Fitzpatrick and to breed, or continues to float. Although this Woolfenden (1986: 154) as a reason for a lack may be true, we prefer to separate these latter of parent-offspring conflict and offspring re- two alternatives in order to emphasize the tention, is a demographic fact true for all spe- range of options potentially available to an cies and cannot, in and of itself, be sufficient individual during a particular breeding sea- to explain these phenomena. Instead, the son (see also Brown, 1987: 96-97). costs and benefits accruing to both breeders Point of view and potential conflict. The dis- and offspring as a consequence of differing persal strategy of an offspring can be exam- dispersal strategies are important in deter- ined from either the point of view of the off- mining which option is pursued (e.g., Orians spring itself or that of the breeding group et al., 1977; Emlen, 1982b). members (usually parents) to whose group it belongs. The interests of these two sets of Delayed Dispersal: Intrinsic Benefits individuals are not always concordant, and or Extrinsic Constraints? conflicts can be expected between them (e.g., As shown by the contrast of Figures l(a) Emlen, 1982b; Vehrencamp, 1983; Koenig and l(b), there is a dichotomy in the potential et al., 1988). Following prior treatments, we causes for delayed dispersal and group living. assume that parents or other group members First, there may be intrinsic benefits to de- cannot force offspring to delay dispersal, but layed dispersal and group living such that in- that they can evict offspring if it is in their dividuals achieve higher lifetime fitness by own best interest. Thus, just as there may be delaying. - dispersal compared to either at- constraints on the ability of offspring to obtain tempting to breed independently or floating, high-quality breeding territories, there may all other things being equal. Two major eco- be constraints on their ability to delay dis- logical reasons why this might be true are low- persal and help, depending on the real or hy- ered predation and increased foraging effi- pothetical consequences of such an action on ciency (Alexander, 1974). Alternatively, the the remaining members. Parental tol- intrinsic benefits promoting delayed dispersal erance of nondispersal is likely, however, be- and group living may not outweigh the costs, cause parents will usually benefit from the in which case the lifetime fitness of individuals retention of offspring beyond the point at that delay dispersal will be lower than that of which delayed dispersal is beneficial to the those dispersing early and attempting to offspring themselves (Emlen, 198213). This breed independently, all other things being prompts us here to take the point of view of the equal. In this latter situation, grouped indi- offspring rather than the parents. Constraints viduals do not gain because of the presence of that may be imposed by conflicts of interest others, but rather gain "solely from the pres- with other group members are subsumed un- ence of some other resource in the environ- der the conditions. discussed below, influenc- ment" (Alexander, 1974: 329); in other words, ing the dispersal options of offspring. they are ecologically forced to live in groups JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 123 in order to have access to a critical resource major types of ecological constraints identi- (see also Altmann, 1974; Walters et al., in fied by Emlen (1982a): Habitat saturation press). Such individuals are "making the best and a biased sex ratio are extrinsic constraints of a bad job" and gain by group living only to independent breeding, whereas if indepen- in comparison to living without access to the dent breeding is restricted because the cost of critical resource, not by the presence of other rearing young is prohibitive for some pairs, individuals per se. In principle, at least, iden- then delayed dispersal is a consequence of in- tification of the critical resource is straightfor- trinsic benefits to group living. Cooperative ward: If the constraint is removed, young will breeders that are classically territorial are disperse. Mates and breeding space (territo- most likely to fit into the first of these categories) are likely candidates for resources that ries, while those that are not are likely to fit might be limited and act in this way. into the second (Emlen, 1982a). Note that the The distinction between intrinsic benefits term "intrinsic benefits" stresses the advan- and extrinsic constraints can only be applied tages to group living accruing via this route, at the level of current functional utility and but we could equally as well use the term "in- not at the level of evolutionary origin (sensu trinsic constraints" to stress the limitation on Tinbergen, 1963; see also Sherman, 1988). independent breeding resulting from the high For example, even if a population is forced cost of rearing young by pairs or small groups. into group living due to extrinsic constraints, Species in which intrinsic benefits appear individuals may still reap secondary benefits to be particularly important include stripe- from living in groups (see below). These sec- backed wrens Campylorhynchus nuchalis, where ondary benefits can be expected to coevolve the ability of pairs and trios to deter predators with delayed dispersal and group living (Creel and thus breed successfully is extremely low and Creel, 1991), eventually leading to a situ- compared to larger groups (Rabenold, 1984), ation in which secondary benefits of group Harris's hawks Parabuteo unicinctus, in which living might exceed the fitness loss due to de- cooperative hunting appears to be a crucial layed dispersal. At that point, individuals in foraging strategy (Bednarz, 1988), and white- the population will appear, on a proximate winged choughs Corcorax melanorhamphos, where level, to delay dispersal and live in groups the skills needed for independent reproduc- because of intrinsic benefits, even though de- tion apparently require several years to delayed dispersal originated as a consequence velop, and group living is virtually obligatory of extrinsic constraints. Thus, the distinction (Heinsohn et al., 1988; Heinsohn, 1991; see between these two routes to delayed dispersal also below). Intrinsic benefits also are likely and cooperative breeding may, at least in to be an important component of helping in some cases, be primarily historical. cases in which helpers do not necessarily delay Because of this problem, it will almost al- independent breeding, since helping in such ways be more profitable to ask whether intrin- species is a behavioral option chosen by indi- sic benefits or extrinsic constraints are cur- viduals that might otherwise be putting addi- rently exercising greater control over delayed tional time and effort into raising their own dispersal rather than to attempt to reject one offspring. For example, as mentioned earlier, or the other hypothesis as playing some role in GalLpagos mockingbirds (Curry and Grant, in any particular system. Documentation of 1990) and in western bluebirds (J. L. Dickin- significant fitness benefits to delayed dispersal son, pers. commun.), some individuals act constitutes the primary evidence for the im- concurrently as helpers at one nest and as portance of intrinsic benefits, whereas docu- breeders at another. Perhaps the best studied mentation of constraints to dispersal is the species in this category is the white-fronted primary evidence for the importance of ex- bee-eater Merops bullockoides, a species in which trinsic constraints. birds frequently switch between helping their Both routes to delayed dispersal involve parents and independent breeding. Emlen ecological constraints; the distinction be- (1 984, 1990) found considerable evidence tween them is in the nature rather than the that intrinsic benefits to group living are para- presence of a constraint to independent breed- mount to the decision to act as a helper: When ing. The two routes correspond to the two environmental conditions were good, birds 124 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67 tended to breed independently, but when con- ductive vacancies. These data all fail to sup- ditions were poor, birds were more likely to port the existence of significant intrinsic be helpers, and breeders even competed for benefits to delayed dispersal; they are consis- helpers. The benefits of helping in the years tent with the hypothesis that extrinsic factors when conditions were poor was confirmed by are important to the current maintenance of the observation that helpers significantly in- delayed dispersal in this species. crease, and in some years are required, for There is also experimental evidence for the successful reproduction (Emlen, 1990). importance of extrinsic constraints in several There is equally compelling evidence for species. Pruett-Jones and Lewis (1990) re- the importance of extrinsic constraints in a moved breeding male superb blue fairy-wrens variety of cooperative breeders. For the Flor- and found that virtually all nonbreeding help- ida scrub jay, Woolfenden and Fitzpatrick ers present in the population dispersed when (1984: 238), following the format of Brown sufficient reproductive vacancies were created, (1975, 1978) and Emlen (1978), compared supporting the hypothesis that extrinsic con- the expected fitness of individuals that delay straints on reproductive opportunities, not in- dispersal and help with those that disperse trinsic benefits to delayed dispersal and helping, early and breed. Similar to earlier results, caused these individuals to remain as non- Woolfenden and Fitzvatrick found that under breeding helpers. An analogous experiment all plausible circumstances the inclusive fit- was conducted by Walters et al. (in press) by ness of a young jay is strikingly greater if it adding nesting and roosting cavities to areas disperses early and breeds than if it delays previously unoccupied by red-cockaded wood- dispersal and helps. It follows that extrinsic peckers. Most areas where cavities were added constraints restrict the ability of offspring to were subsequently occupied by birds that had pursue this superior dispersal option. This is previously been helpers or floaters, support- further supported by the finding that individ- ing the hypothesis that cavities are a critical, uals delaying dispersal appear to have lower limited resource constraining early dispersal1 survivorship and lifetime reproductive suc- breeding. Particularly compelling evidence is cess than the apparently superior individuals provided by Komdeur (1991) for the Sey- that disperse and breed either in their first chelles brush warbler (Acrocephalus sechellensis). year or after a single season of helping (Fitz- Komdeur created vacancies on the apparently patrick and Woolfenden, 1988). saturated island of Cousin by transplanting Young, nonbreeding helper acorn wood- individuals from 16 territories to the pre- peckers in California spend considerable time viously uninhabited island of Aride. Not only and effort engaged in forays designed to locate were all vacancies created on Cousin filled, potential reproductive vacancies (Hooge, sometimes within hours, by birds who had 1992) and fight vigorously for the opportunity previously been helpers, but-all 61 young pro- to abandon their status as helpers and to fill duced by the transplanted birds on Aride dis- reproductive vacancies, as predicted if extrin- persed and bred independently the following sic constraints are paramount to promoting year. These examples yield strong evidence delayed dispersal (Koenig, 1981b; Hannon et for delayed dispersal resulting primarily from al., 1985). There is also considerable evidence extrinsic constraints on the availability of that the lifetime fitness of birds that delav dis- high-quality reproductive vacancies. persal is not greater than that of individuals that disperse early and breed, at least if the General Factors AffeGtingDelayed Dispersal latter are able to settle on a high-quality terri- Five general demographic, ecological and tory (Koenig and Mumme 1987; Stacey and environmental factors affect the decision to Ligon, 1987). Finally, Emlen (1984) tested delay dispersal in the model: (1) relative pop- the importance of ecological constraints to the ulation density, (2) the fitness differential be- evolution of delayed dispersal in this species tween early dispersallbreeding and delayed indirectly by correlating the proportion of in- dispersal, (3) fitness of floaters, (4) the distri- dividuals delaying dispersal with the level of bution of territory quality, and (5) spatiotem- breeding constraints. He found that the pro- poral environmental variability. We will con- portion of helpers increased with fewer repro- sider each in turn. JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 125

I. Population density. Relative density could a Independent breeding be the primary factor distinguishing the conditions leading to cooperative breeding. For b-' example, assume that one population is limited during the nonbreeding season, and a second is limited during the breeding season. Under these conditions, the breeding population size of the second. but not the first., mightu be large relative to available high-quality territories, constraining the ability of offspring to disperse early and breed. Conditions leading to-relatively low population density might be chronic or intermittent, depending on the variability of environmental conditions. This latter ~ossibilitvis discussed below. The contrast between populations with low and high relative density is shown in Figure 2, which depicts the options available to offspring given that extrinsic constraints provide the proximate impetus for delayed dispersal [as in Fig. l(b)] . In the low-density situation [Fig. 2(a)], only a fraction of high-quality environments are occupied and y is the best available unoccupied territory, while in the high-density situation [Fig. 2(b)], all environments of moderate quality or higher are occu- Z X pied and z is the best available unoccupied Worst Best territory. Environments ranked by If the densitv is low, an individual born on the highest-quality territory x should disperse suitability for breeding and breed independently, since WQ) > Wd(*) FIG. 2. PARAMETERSINFLUENCING THE DELAYED > Wrc,,) [Fig. 2(a)]. If the density is high, DISPERSALTHRESHOLD MODEL. I. Wd(,, > W4,) > Wf,,,) [Fig. 2(b)]; hence, the POPULATIONDENSITY. same individual would experience highest fit- Habitat-fitness gradient curves are graphed as ness if it remains as a nonbreeder on its natal described in Figure l(b), where extrinsic con- territory. straints force delayed dispersal and the fitness of floaters is low. Only territories represented by We have drawn the W,gradients identically open square symbols on the "independent breed- in the two parts of Figure 2. It is likely, how- ing" line are occupied, with the highestquality un- ever, that increased breeder density might in- occupied environment represented by a solid square. teract with Wrby forcing floaters into less than (a) LOW density: An individual bdrn on the best the best habitats. Thus, population density available territory x will be best off dispersing to may affect the relative desirability of delayed the best available territory (j~)and breeding early, dispersal by determining, in part, the quality since Web)> Wd(,)> Wf((,,). (b) High density: The of the territories available to offspring that same individual will be best off delaying dispersal disperse and either float or attempt to breed. and remaining as a nonbreeder on his or her natal territory x since W+) We(*) Wj((m,). A review of the factors potentially influenc- > > ing population density is beyond the scope of this paper; the important point is that all such not "cause" delayed dispersal and cooperative factors may influence delayed dispersal as breeding, but is an important factor influenc- shown in Figure 2. As discussed above, rela- ing the probability of these phenomena by tive density is equivalent to the degree of habi- affecting the expected fitness of individuals tat saturation as used bv some authors. When that disperse and breed in the best available used in this context, habitat saturation does territory. 126 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67

If the intrinsic benefits outweigh the costs reeding [as in Fig. l(a)], the fitness of individuals that - delay dispersal is likely to be greater than for those that disperse and breed regardless of population density, and density consequently has little effect on whether some individuals will delay dispersal (that is, on whether groups will form). However, density will influence group size (e.g., Brown, 1982) since, as density increases, individuals will continue to delay dispersal and group size will increase until the intrinsic benefits to those remaining no longer outweigh the costs. At this point, the decision of whether to delay dispersal or to disperse becomes analogous to the situation in which extrinsic constraints currently provide the proximate factor forcing group living (see Fig. 4 below). If extrinsic constraints are paramount, dispersallbreeding on a high-quality territory always yields the highest fitness. Thus, in the remaining discussion, we will generally be concerned with the situation in which population density is sufficiently high so as to con- z X strain the ability of offspring to disperse and Worst Best breed in high-quality territories. Under such Environments ranked by conditions, the relative payoff of the options suitability for breeding available to offspring depends on the re- FIG. 3. PARAMETERSINFLUENCING THE DELAYED maining four factors. DISPERSALTHRESHOLD MODEL. IIA. 2. Fitness differential between early and delayed FITNESSDIFFERENTIALBETWEEN EARLY dispersal. Here we include variables that influ- AND DELAYEDDISPERSAL: EXTRINSIC ence the differential between the territory CONSTRAINTSPARAMOUNT. quality gradients representing the delayed (a) If the differential between dispersallbreeding dispersallhelping option (curve d)and the dis- and delayed dispersal is small, the probability that persallbreeding option (curve e) (Fig. 1). We individuals will delay dispersal is relatively high. For example, an individual born on territory x discuss four of the most important below. Al- must find a vacant territory of quality greater than though we discuss them separately, they are territory y (since W+) = Web))for early dispersal1 interrelated under most circumstances. breeding to yield higher fitness than delayed dis- a. Resource depressibility on territories. persal. (b) If the differential between dispersal1 This refers to the degree to which use of a breeding and delayed dispersal is great, the proba- limiting resource by one individual reduces bility that individuals will delay dispersal is rela- its availability to another (Waser, 1988). If tively lower. In this case, an individual born on extrinsic constraints promote delayed dis- territory x need only find a vacant territory of qual- persal, resource depression is one factor de- ity greater than territory z (since Wd(x)= WE(*)) termining the magnitude of the drop (within for dispersallbreeding to yield higher fitness than delayed dispersal. the same territory) from the territory quality gradients of dispersallbreeding to that of delayed dispersal (Fig. 3). The more depressible limiting resources are, the more reproductive Relatively nondepressible limiting resources success and survivorship are influenced by ad- [Fig. 3(a)] will reduce the differential between ditional group members, and the more likely the fitness of individuals that disperse and it is that parents or other group members might breed and those that delay dispersal. Under not tolerate delayed dispersal by offspring. such conditions, delayed dispersal is more JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 127 likely to be preferred by offspring and tolerated by breeders. Specifically, it will be the best option for individuals born on territory x whenever they are unable to obtain a vacancy in a territory whose quality is greater thany [Fig. 3(a)]. In contrast, a high differential between these options due to high depressibility of limiting resources means that delaying dispersal entails a large cost and is less likely to be preferred by offspring or tolerated by breeders. In Figure 3(b), only individuals born on territory x failing to obtain a vacancy in a territory whose quality is greater than z should delay dispersal. In both Figures 3(a) and 3(b), delayed dispersal and independent breeding are preferred to floating. Since territory z is oflower quality thany, it follows that delayed dispersal and cooperative breeding are more likely when limiting resources are relatively nondepressible [Fig. Worst * Rest 3(a)]. The exact consequences of resource de- Environments ranked by pressibility, however, may be difficult to pre- suitability for breed.ing dict for at least three reasons. First. it is necessary to take into consideration the effects of FIG. 4. PARAMETERSINFLUENCING THE DELAYED resource depression on the fitness of all group DISPERSALTHRESHOLD MODEL. IIB. FITNESSDIFFERENTIAL BETWEEN EARLY members, since conflicts of interest might prompt some group members to try and evict BENEFITSPARAMOUNT. offspring, thereby constraining the latter's This figure graphs parallel habitat-fitness gradi- ability to delay dispersal. Second, the degree ents each of which tracks the fitness of an offspring of depressibility may be correlated with terri- depending on the size of the group (represented by tory quality; for example, resource depres- the numeric subscript) while it delays dispersal. (a) sion may be less on high quality territories The fitness differential between dispersallbreeding than on those of moderate to low quality. and delayed dispersal is relatively large and in- Third, individuals may have ways of counter- creases to groups of size six. In this case, an indi- acting the effects of resource depressibility. vidual born on territory x continues to gain by delaying dispersal up to a group with six members. For example, Florida scrub jays increase their If the group is larger, the intrinsic benefits of de- territory size directly with increasing group layed dispersal decrease. At groups with 10 mem- size (Woolfenden and Fitzpatrick, 1984). bers, the options of dispersallbreeding indepen- Thus, resource depression, which might oth- dently and delayed dispersal yield equivalent erwise be significant as more individuals de- fitness. By assuming that floaters can locate space lay dispersal and group size increases, is les- in the best available habitat for floating, floating sened by the additional resources garnered by becomes the preferred option at larger group sizes larger groups. unless the individual can find a vacant territory of If the intrinsic benefits of delayed dispersal quality greater than Wf(,,, or join a smaller group with a consequently higher territory quality gradi- and group living outweigh the costs, resource ent. (b) The fitness differential between dispersal1 depressibility becomes a major determinant breeding and delayed dispersal is small and accrues of optimal group size (e.g., Brown, 1982). only up to groups of size three. In this case, an This is shown in Figure 4, the top part of individual born on territory x gains by delaying which [Fig. 4(a)] depicts the situation in dispersa! only if it is the only one to do so; if two which rksources aEe relatively nondepressible offspring delay simultaneously, their fitnesses are such that intrinsic benefits continue to out- equivalent to what they could expect by dispersing1 weigh the costs up to groups of size six, after breeding on a territory of comparable quality. which the costs associated with exploitation of Floating becomes the preferred option at larger group sizes unless a vacant territory can be acquired of quality greater than Wf(m,b 128 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67 the limited resource bv additional individuals pear to be inordinately dependent on a partic- leads to lower intrinsic benefits. Eventually, ular critical resource defining suitable territo- when group size is at ten, an additional off- ries; examples include the granaries of acorn spring~wouldbe even better off dispersing1 woodpeckers (Koenig and Mumme, 1987; floating than delaying dispersal and living in Stacey and Ligon, 1987; Koenig and Stacey, a group of size eleven. 1990), the roosting cavities of green wood- Figure 4(b) presents the condition of high hoopoes ~hoeniculus purpureus (~i~onand resource depressibility. Delayed dispersal en- Ligon, 1990) and red-cockaded woodpeckers tails intrinsic advantages only up to groups Picoides borealis (Walters, 1990; Walters et al., of size three. after which the costs begin" to in press), and the burrow systems of several outweigh- the benefits and the fitness of indi- mammals (Walters et al., in press). The rela- viduals delaying dispersal and remaining on tive nondepressibility of such nonconsumable their natal territory decreases. In this exam- resources is most likely a feature basic to de- ple, individuals would be better off dispers- layed dispersal and cooperative breeding in inglfloating than living in groups larger than these species. four. This latter example depicts the situation b. Group size and composition. The poten- in which a single offspring achieves highest tial importance of group size and composition fitness by delaying dispersal and remaining is shown for the case in which intrinsic bene- on the natal territory, but additional offspring fits are paramount (Fig. 4), but group size can do not. also be important if extrinsic constraints lead The expected distribution of group sizes to delayed dispersal. For example, the ex- under these circumstances is dependent on pected indirect fitness benefits to a young population density, the fluidity of group mem- Florida scrub jay that delays dispersal are rel- bership (Sibly, 1983), and the precise relation- atively great if no other nonbreeders are pres- ships among group size, territory quality and ent in the group. Reproductive success, how- resource depression (Brown, 1982, 1987). ever, does not increase in groups larger than Note also that the curves in Figure- 4 are all three (Woolfenden and Fitzpatrick, 1984). drawn with respect to nonbreeding helpers; Unless this lack of reproductive enhancement

analogous curves for breeders or other group- - is compensated by increased survivorship of members might be different, leading to con- breeders, larger territory size, or some other flicts of interest among individuals concern- group-related advantage, additional offspring ing group membership. Thus Figure 4 is not a beyond the first accrue no fitness benefits by delaying dispersal. comprehensive model for optimal group- - size. ~ksourcedepressibility has long been rec- Group composition may affect fitness in ognized as a major factor influencing the fit- several ways, the most obvious of which is by ness differential between early and delayed dis- the genetic relatedness of individuals delaying persal, and thus of sociality and group living dispersal and the group members they may (e.g., Charnov et al., 1976; Altmann et al., potentially aid (Hamilton, 1964; Williams 1977; Brown, 1982, 1987; Waser, 1988). 1966, Brown, 1974; Hartung, 1977; Emlen Waser (1988) provides a thorough discussion, and Wrege, 1989). The potential indirect fit- pointing out four types of relatively nondepress- ness benefits of delaying dispersal and acting ible resources: (1) nonconsumable resources as a helper are greatest if helpers provide aid such as dens, sleeping sites, or hibernacula, to full siblings, and they decrease as the inci- (2) time-limited resources that disappear dence of polygamy, mate-switching, or ex- whether or not foragers consume them, (3) trapair fertilizations increases (e. g., Char- renewable resources, and (4) essential re- nov, 1981; see below). It is thus not surprising sources that are su~erabundantbecause the that most nonbreeding helpers appear, at population is limited by some other resource. least, to be helping their parents raise full sib- To the extent that alimiting resource falls into lings (Brown, 1987); future molecular analy- one of these categories, resource depressibil- ses will help to clarify whether this is indeed ity is likely to be low, and the probability of the case as expected from kin selection theory. delayed dispersal and group living is high. In general, greater kinship between potential Several species of cooperative breeders ap- helpers and subsequent offspring they may JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 129 help raise increases the benefits and thus the in search of vacancies may be considerably probability of delayed dispersal [cf. Figs. 3(a) more common in cooperative breeders than vs. 3(b)]. Besides being an important compo- usually believed (Hooge, 1992); this is di- nent of the costs and benefits of delaved dis- rectly contrary to the model proposed by Zack persal, kinship effects are likely to be espe- (1990), who suggested that delayed dispersal cially critical in explaining helping behavior and cooperative breeding are causally related per se (e.g., Emlen and Wrege, 1989). to short-distance dispersal to adjacent territo- c. The cost of dispersal. In many coopera- ries of high quality by nonbreeders. More ex- tively breeding species, offspring are able to tensive use of radiotelemetry is needed to ad- inherit and breed in or next to their natal dress this issue. territory without any dispersal; this occurs A third way in which the costs of delayed commonly in acorn woodpeckers (Koenig dispersal may be reduced is by offspring wait- and Mumme, 1987), gray-breasted jays ing until they can disperse and obtain territo- (Brown and Brown, 1984), Florida scrub jays ries of higher quality than those currently (through territorial budding, Woolfenden available, thereby ultimately compensating and Fitzpatrick, 1978, 1984), and at least oc- for the loss in fitness due to delayed breeding casionally in many other species. To the ex- (S. Zack and B. J. Stutchbury, pers. com- tent that territorial inheritance eliminates mun.). This could happen as a consequence of what would otherwise be a risky dispersal epi- individuals being older and more competitive sode, it lessens the costs of delayed dispersal (e.g., Woolfenden and Fitzpatrick, 1984), and thereby decreases the differential behaving more time to accurately assess terri- tween delayed dispersal and dispersallbreed- tory quality, or having more time to recruit or ing (assuming that extrinsic constraints pro- acquire coalitions of related individuals with vide the proximate impetus for delayed which they can fight for reproduction vacan- dispersal). The potential importance of terri- cies (Hannon et al., 1985; Hooge, 1992), or tory inheritance and short-distance move- exploit as helpers following dispersal (Ligon ments to delayed dispersal has been stressed and Ligon, 1978, 1979). Concurrently, the recently by Stacey and Ligon (1987, 1991) fitness differential between the dispersallbreed- and by S. Zack and B. J. Stutchbury (pers. ing and delayed dispersal options may in- commun.; Zack, 1990). crease as the ability of individuals to exploit A second mechanism by which delaying the resources available on any particular terri- dispersal could minimize costs might be by tory increases (see the skill hypothesis, dis- permitting dispersal to occur when it will en- cussed below). If the fitness consequences of tail relativelv little risk. This would occur if. delayed dispersal remain constant, this effec- for example, individuals were able to delay tively increases the relative advantages of dispersal until they were in peak physical con- independent breeding and results in progres- dition rather than all dispersing at a set point sively lower-quality territories becoming ac- in time. Equally important, delayed dispersal ceptable to a helper as it gets older [cf. Fig. allows individuals to use the natal territory as 3(b) vs. 3(a)]. Together, both the increased a base from which to make forays in search competitiveness of older individuals and their of reproductive vacancies, thereby potentially increased ability to exploit limited resources reducing dispersal costs compared to the al- should result in a preponderance of helpers ternative of floating (Hooge, 1992). Non- being young and inexperienced. breeding helper acorn woodpeckers, for ex- High costs associated with dispersal (or, ample, devote a significant proportion of time conversely, the potential benefits associated to extensive forays of 20 km or more searching with philopatry) have been suggested to be for reproductive vacancies, while then regu- important to the evolution of delayed dis- larly returning to their natal territory. One persal and subsequent helping behavior by result of these forays is presumably to de- numerous authors (e.g., Waser and Jones, crease the potential risks associated with find- 1983; Brown, 1987; Stacey and Ligon, 1987, ing and obtaining a reproductive vacancy and 1991; Waser, 1988; Zack, 1990; Emlen, thus to raise the "delayed dispersal" gradient 1991). Such costs do not, by themselves, re- [cf. Fig. 3(a) vs. 3(b)]. Long-distance forays sult in delayed dispersal and group living, re- 130 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67 gardless of how high they may be. Juvenile patrick, 1978, 1984). Examples of the latter red grouse Lagopus scoticus, for example, suffer may include increased reproductive success extremely high mortality in their first year as or survivorship of related breeders, either of a consequence of dispersal and exclusion from which would result in greater indirect fitness high-quality foraging sites by territorial own- of individuals delaying dispersal and helping ers (Watson, 1985). This high risk does not (Brown, 1983, 1985, 1987; Mumme et al., result in delayed dispersal by offspring. Two 1989; Koenig and Mumme, 1990). As with closely related taxa, however, might differ in the effects of resource depressibility, second- the degree to which dispersal is risky or in the ary benefits might differ between high- and ways that the risks can be mitigated so as to low-quality territories. This would be true, promote delayed dispersal in one but not the for example, if helpers have little effect on other. All other things being equal, the high-quality territories containing abundant greater the costs of dispersal, the more proba- resources and a relatively large effect on low- ble that delayed dispersal will offer opportuni- quality territories. ties to mitigate those costs. On the other hand, The distinction between intrinsic benefits as in the red grouse example, the magnitude accruing as an immediate consequence of de- of seasonal resource depression may be so se- layed dispersal versus secondary benefits that vere that it eliminates the possibility of coop- arise and evolve subsequent to delayed dis- erative breeding by making it impossible to persal and group living is a historical one that delay dispersal and remain on the natal terri- cannot be made in practice. There are few tory. automatic benefits, however, to group living By definition, a disperser is a floater until it (Alexander, 1974), and thus most of the in- settles on a breeding territory. Consequently, trinsic benefits observed in cooperative breed- the cost of dispersal can be viewed as the cost ers are probably derived secondarily. This of floating (discussed below) for the time it provides an important cautionary note con- takes to acquire or settle on a territory, cerning the significance of such benefits to whether it be a few days or several years. the evolutionary origin of these phenomena. d. Secondary benefits of delayed dispersal Although intrinsic benefits may be measur- and helping. Regardless of the initial impetus able and significant, it cannot be assumed that to delay dispersal and live in groups, once they are responsible for a difference in dis- groups exist individuals can be expected to persal tendencies between two closely related exploit all potential benefits to group living taxa. This is because intrinsic benefits are, by and helping behavior. If extrinsic constraints definition, only available to populations in provide the proximate impetus for delayed which delayed dispersal and group living ex- dispersal, the effect of secondary benefits is to ist, and cannot be observed or measured in raise the delayed dispersal curve. Graphi- closely related, non-group-living popula- cally, this is shown in the difference between tions. Offspring of the non-group-living pop- Figure 3(b) (few secondary benefits) and Fig- ulation might very well accrue similar or iden- ure 3(a) (high secondary benefits). Other- tical benefits if they delayed dispersal as well. wise, secondary benefits add to the primary This caveat also sets us squarely apart from intrinsic benefits to increase optimal group the views of Stacey and Ligon (1987, 1991), size (e.g., Fig. 4). whose benefits of philopatry hypothesis pro- Secondary benefits may include both group pose that extrinsic constraints to dispersal effects arising from larger group size and allo- may be a result rather than a cause of delayed parental effects arising from helping behavior dispersal, which they propose is instead a re- per se (Koenig and Mumme, 1990). Exam- sult of intrinsic benefits to group living. We ples of the former include increased vigilance believe that in a vast majority of cooperative (e.g., McGowan and Woolfenden, 1989), breeders extrinsic constraints are the original more efficient exploitation or defense of re- cause of delayed dispersal and that most in- sources (e.g., Bednarz, 1988), and larger ter- trinsic benefits resulting from this behavior ritory size leading to a higher probability of are secondarily derived. obtaining a territory in the future through Despite this caveat, secondary benefits may territory budding (e.g., Woolfenden and Fitz- differ inherently between taxa in ways critical JUNE 1992 DELA YED DISPERSAL IN COOPERA TIVE BREEDERS 131 to the evolution of delayed dispersal. For ex- small probability of success, as occurs in the ample, the opportunity to accrue indirect fit- noncooperative western scrub jay (Carmen, ness benefits by delayed dispersal and helping in press). Second, the skill hypothesis pro- may be lower in migratory and nomadic taxa poses a mechanism that potentially yields in- in which relatives are unlikely to remain to- trinsic benefits to delayed dispersal. Like gether compared to nonmigratory taxa in which other aspects of delayed maturity, however, offspring could remain with their parents and if the acquisition of breeding skills is purely a help raise siblings. Similarly, populations matter of slow or delayed development, then subject to high predation pressure might ben- it is a secondary effect, not a cause, of delayed efit from sentinel behavior in ways that a pop- dispersal (Koenig and Pitelka, 1981). ulation living on an island free of predators This distinction has misled some authors to could not. The importance of some of these confuse the ecological circumstances causing factors to delayed dispersal could be tested the current adaptive utility of delayed disper- by appropriate manipulations; for example, if sal and the mechanism promoting this behav- kinship and indirect fitness leading to reduced ior. For example, consider the white-winged cost of delayed dispersal is hypothesized to be chough in which young require considerable a critical parameter, one could remove one time to acquire the skills necessary for inde- member of successfully breeding pairs with pendent reproduction. Heinsohn et al. (1988, helpers, thereby opening up the opportunity 1990; Heinsohn, 1991) have argued that this for replacement by unrelated individuals. constitutes a significant intrinsic advantage to Offspring would then have to choose between delayed dispersal and cooperative breeding to early dispersal and remaining to raise half- the extent that these behaviors are obligatory siblings. To the extent that indirect fitness in the population. Delayed dispersal in white- benefits are critical to delayed dispersal, there winged choughs, however, may have initially should be a significant increase in offspring been selected due to extrinsic constraints. leaving and either floating or attempting in- Once offspring progressively lost the opportu- dependent breeding compared to unmanipu- nity to disperse early and breed indepen- lated controls. dently, it is likely that selection favored indi- One important hypothesis that falls into the viduals that devoted their limited time and category of secondary benefits to delayed dis- energy to behaviors of more immediate fitness persal is the skill hypothesis, which proposes value than acquiring breeding skills early, ;hat delayed breeding and delayed dispersal which they will only get to use much later in allow nonbreeders to acquire experience that life. This would then set up the potential for increases their lifetime fitness by making coevolution between delayed dispersal and them more successful breeders later in life. the behavioral skills necessary for breeding: This idea, first suggested by Skutch (1961), as early dispersallbreeding became more and was modeled as a potential explanation for more constrained, acquisition of the skills delayed breeding by Brown (1987); it has sub- needed for early breeding would become sequently been proposed as the cause of de- more and more delayed (Creel and Creel, layed dispersal and cooperative breeding in 1991). white-winged choughs and in a variety of A second example mixing mechanism and other species by Heinsohn and his colleagues current function is provided by Richner (Heinsohn et al., 1990; Heinsohn, 1991). (1990), who argued that the phenotypic con- There are at least two difficulties with this straint of small body size has restricted the hypothesis as an explanation for delayed dis- dispersal options of some young male carrion persal and cooperative breeding. First, al- crows (Corvus corone), and has led to delayed though young individuals may lack the skills dispersal and helping. As support for this hy- necessary to breed successfully, there is no a pothesis, he cites the observation that the priori reason to believe that the most efficient three male helpers he observed were at or be- way to acquire such skills is by delayed dis- low the size limit of territorial males in the persal and helping. They might be acquired population. This argument, however, con- equally efficiently by floating or even by at- fuses the mechanism by which an ecological tempting to breed independently, albeit with constraint is imposed with the constraint it- 132 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67 self: if there were no constraint on independent reproduction due to limited breeding opportunities in high-quality habitat, even the smallest males could disperse and breed on their own. Once breeding opportunities are restricted, it is not surprising that the largest and presumably strongest individuals should occupy territories, forcing smaller individuals to choose alternative dispersal options. There is, however, one situation where we can envision that the skill hypothesis might be Independent breedii important to delayed dispersal and helping behavior. If offspring learn breeding skills ;--- 'Delayed specifically by observing other group mem- dispersal bers, then delayed dispersal would confer proximate intrinsic benefits that might render delayed dispersal and group living essential to the fitness of individuals, rather than simply a consequence of the extrinsic factors originally constraining independent reproduction. This in turn could significantly improve the cost- Wont Y Best benefit ratio of delayed dispersallhelping Environments ranked by compared to either dispersallbreeding or dis- suitability for breeding persallfloating. Thus, assessment of the current role of the skill hypothesis to the evolution of cooperative breeding is dependent, in part, on detailed observations of the ontogeny of Floaters may have either high or low fitness de- breeding skills. Such data have yet to be ac- pending on the quality of habitat available to them quired for any species, although the data pro- and their success at finding and filling breeding vided by Heinsohn (1991) on the white- vacancies in high-quality habitat. (a) Floater fit- winged chough are a promising start in this ness in the best habitats is high, while floater fitness direction. in habitats of poor quality for breeding is high (dashed line) or low (dotted line). In either case, an Fitness ojfzoatt As discussed above, the 3. individual may be better off floating than delaying fitness of floaters is determined primarily by dispersal if floaters can locate space in high-quality a combination of their survivorship and their floater habitats. If floater fitness in habitat poor for effectiveness in finding mates and high-qual- breeding is high, floating may still be better than ity breeding territories. Although not yet dem- delayed dispersal; this is unlikely if floater fitness onstrated, it is also possible that floaters in some in habitats that are poor for breeding is low. (b) If species may increase their fitness by obtaining floater fitness even in the best habitats is low, only EPCs or, in the case of females, by intraspe- those few individuals born on territories of lower cific brood parasitism. Factors influencing the quality than y would be better off floating than fitness of floaters have often been ignored by delaying dispersal and remaining as a nonbreeder on their natal territory, even if floaters have some prior treatments of the evolution of delayed access to high-quality habitats. Individuals born dispersallhelping, although recent work by on higher-quality territories are better off delaying Carmen (in press) and reviews by P. Arcese dispersal than floating. (pers. commun.) and S. Zack and B. J. Stutch- bury (pers. commun.) are beginning to change this. Figure 5 focuses on the potential impor- first, represented by the dashed line, floaters tance of the fitness of floaters to whether or have high fitness, even in some habitats of not individuals in a population will breed co- poor quality for breeding, presumably be- operatively. In Figure 5(a), two territory cause they survive well and can later locate quality gradients for floaters are drawn. In the vacant territories of high quality efficiently, JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 133

even though the area lacks some resource crit- territories, the quality of habitat that is accept- ical for successful breeding. Under these con- able to a floater is dependent on several fac- ditions, floaters could still have relatively high tors. For example, progressively lower-qual- fitness even though they meet considerable ity territories should become acceptable to aggression from breeders in high-quality floaters with time if their survivorship (and (breeding) territories. In the second, repre- hence fitness) is age dependent. Given the sented by the dotted line, floaters have high static situation depicted in Figure 5, floaters fitness only in the territories of high quality with high fitness [Fig. 5(a)] should choose to for breeding. This is the more likely situation disperse and breed only if they can fill a va- if critical resources are similar for breeders cancy in a territory of quality greater than and floaters. x, whereas those with low fitness [Fig. 5(b)] The probability of floating under the condi- should be willing to accept any territory of tions depicted by Figure 5(a) is largely depen- quality greater than y. dent on whether or not floaters are excluded Floaters are difficult to observe and have from high-quality habitats. If they are not, rarely been documented in cooperative breed- then floating is preferable to delayed dispersal ers, much less has their fitness been accurately for all individuals, even those born on the best assessed (but see Walters et al., 1992). How- territories, since WA,,, > Wdfor all environ- ever, if floaters are found in a noncoopera- ments. Examples include California scrub tively breeding taxon closely related to one jays (Carmen, in press) and the rufous- that exhibits cooperative breeding, the factors collared sparrow (Smith, 1978). If floaters are influencing floater fitness are likely to be im- discouraged from occupying high-quality portant to understanding their differing social breeding habitat, floating is still preferable to behaviors. An example is the scrub jay com- delayed dispersal if the fitness of floaters is plex, discussed in detail below. high in a habitat that is suboptimal for breed- 4. Distribution of territory quality. We divide ing (the dashed line). This may correspond our discussion of this issue into three parts, to the situation found, for example, in Santa discussing first the marginal habitat hypothe- Cruz Island scrub jays (Atwood, 1980) and sis, then alternative models for the impor- common crows (Caffrey, 1991), where float- tance of the distribution of territory quality, ers appear to survive well in peripheral areas and finally how one factor, sex ratio bias, can not suitable for breeding. act via its effect on the distribution of territory If the fitness of floaters is low in habitat quality to influence the probability of delayed suboptimal for breeding [the dotted line in dispersal. Fig. 5 (a)], then delayed dispersal is likely to a. The marginal habitat hypothesis (MHH). be preferable assuming floaters are excluded This hypothesis proposes that there are rela- from high-quality habitats. This effectively tively many optimal territories of high quality puts floaters in the situation shown in Figure and few marginal territories of intermediate 5(b), where the fitness of individuals that dis- or low quality, leading to a steep territory perse and float is poor even in the highestquality gradient [Fig. 6(a); Koenig and Pi- quality habitats. Under the conditions shown, telka, 19811. Under these conditions, popula- only individuals born on territories whose tion density (habitat saturation) is likely to quality is worse than territory y could profit be high, since survivorship and reproductive by floating rather than remaining as non- success are by definition very good on high- breeders, since WA-) < Wdfor all territories quality territories. Many offspring, virtually whose quality is greater than territory y. all of which are (again by definition) born on Thus, delayed dispersallhelping is more likely high-quality territories, are unlikely to dis- if the fitness of individuals that disperse and perse and breed independently, since all float is lower, as in Figure 5(b). Note, how- high-quality territories are usually filled, and ever, that in both cases graphed in Figure 5, there are few acceptable territories of mar- individuals are best off dispersinglbreeding ginal quality. independently if they are able to acquire a Consider an individual born on territory x high-quality territory. [Fig. 6(a)]. The best option is to disperse and Just as for individuals dispersing to vacant breed as long as a vacancy can be obtained in 134 THE QUARTERLY REVZE W OF BIOLOGY VOLUME67

able to a large proportion of offspring, even when the fitness differential between delayed dispersal and dispersallbreeding is relatively w =w 1 - -4- Delaved 1 large [Fig. 6(a)]. Floating is also likely to be 'Z(X)-YYJI 7 dispersal I an inferior option because the MHH suggests that suitable habitat is relatively homoge- neous and generally fully occupied. Any unoccupied habitat is likely to be of poor quality, and if floaters are restricted to such a habitat, their fitness is likely to be low. Other distributions of habitat quality are less likely to con- strain floating in this way. Unfortunately, the determination of territory quality is difficult, and thus testing the MHH or any other model for the distribution of territory quality is not easy. Most attempts to do so are flawed by defining territory quality in terms of the reproductive success or survivorship of the occupants (e.g., Stacey and Ligon, 1991). Besides being circular, this pro- cedure is likely to yield misleading results as a consequence of differences in population density, variance in individual performance, and random environmental factors. For ex- Worst Best ample, some areas within the periodically Environments ranked by burned scrub favored by Florida scrub jays suitability for breeding produce vastly greater numbers of offspring FIG.6. PARAMETERSINFLUENCING THE DELAYED than others, but jays do not compete more DISPERSALTHRESHOLD MODEL. IV. vigorously for those areas. This suggests that DISTRIBUTIONOF TERRITORY QUALITY. observed productivity differences are due to (a) Conditions characterizing delayed dispersal differences among liileages rather than differ- accorditlg to the marginal habitat hypothesis (Koe- ences in habitat quality (Fitzpatrick et al., nig and Pitelka, 1981). This model postulates a 1988). high proportion of high-quality (optimal) com- There are at least two additional problems pared to marginal territories. Under these conditions, delayed dispersal will be the best option for ir, determining the distribution of territory offspring as long as population density is high, quality. First, the quality of individual ter- whether the fitness differential between dis~ersal/ ritories may change from year to year, and breeding and delayed dispersal is large or small. (b) second, study areas are unlikely to include a Conditions promoting delayed dispersal according representative range of territories to which to the variance hypothesis (Stacey and Ligon, individuals might potentially disperse, there- 1987, 1991 ; see text). This model postulates wide by yielding a biased estimate of the distribu- variation in territory quality and can lead to de- tion of territory qualities. Given these diffi- layed dispersal even under conditions of low den- culties, it is not surprising that tests of the sity as long as the fitness differential between dis- MHH have produced conflicting results (Koe- persallbreeding and delayed dispersal is small. nig and Pitelka, 1981; Emlen, 1991; Stacey and Ligon, 1991; McCallum et al., in press). b. Alternative models. Conditions favoring a territory of quality greater than y. These, delayed dispersal are also possible with other however, are nearly all high-quality territo- distributions of territory quality besides the ries and are unlikely to offer more than an steep gradient postulated by the MHH. For occasional vacancy, for which competition example, consider the territory quality gradi- from other offspring will be intense. Thus de- ent shown in Figure 6(b), in which there is layed dispersal will be the best option avail- a relatively wide range of territory qualities JUNE 1992 DELAYED DISPERSAL IN COOPERA TIVE BREEDERS 135

leading to relatively great variance in the will require not only knowledge of territory quality of occupied territories. Under these quality, but knowledge of individual quality conditions, population density (habitat satu- as well. ration) is likely to be low, since proportion- If population density is low, delayed dis- ately fey I territories are of high quality. persal can be preferable to dispersallbreeding This, however, does not preclude delayed for some offspring under conditions specified dispersal being the preferred option for at by the VH, but not the MHH. If population least some offspring. Consider an individual density is high, both predict delayed dispersal born on the best territory, x, in Figure 6(b). among at least some offspring. Thus, in terms Assuming that the cost to delayed dispersal is of the choice between delayed dispersallhelp- small and the fitness of floaters is low, delayed ing and dispersallbreeding, the issue is not dispersal is the best option as long as no va- whether one model or the other predicts the cancy can be obtained in territories of quality occurrence of delayed dispersal and coopera- 2y.Since these are all still high-quality terri- tive breeding, but rather which model corres- tories, this might be true and thus delayed ponds more closely to the conditions leading dispersal will probably be the best option for to these phenomena in a specific population. this individual. The VH is potentially appropriate when rela- As part of their benefits of philopatry tive density and the fitness differential be- model, Stacey and Ligon (1987,1991; see also tween dispersallbreeding and delayed dis- Waser, 1988) propose that there is high vari- persal are low, while the MHH is potentially ance in territory quality among cooperative applicable when relative density is high, re- breeders, and that the quality of occupied ter- gardless of the fitness differential between dis- ritories is relatively uniform among noncoop- persallbreeding and delayed dispersal. erative breeders; this is contrary to the predic- c. Sex ratio bias. This factor can influence tion of the MHH. In order to disinguish this the distribution of territory quality in a popu- aspect of their model from Stacey and Ligon's lation to the extent that a given territory con- (1987, 1991) focus on the benefits of group tains (or is capable of attracting) a breeder of living, we will call it the "variance hypothesis" the opposite sex. Consider a male born on (VH), and propose Figure 6(b) as a graphical territory x with the territory quality gradients representation of the distribution of territory for dispersallbreeding and for delayed dis- quality it predicts should result in delayed dispersal shown in Figure 7(a). He should dispersal and cooperative breeding. perse and breed if he can find a territory of An important contribution of the VH is quality greater thany. If the sex ratio is male that it highlights the importance of taking ter- biased, however, there will be some fraction ritory quality into account when comparing oflower-quality territories that will not attract the options separately available to each indi- females. If the fitness of lone males on a terri- vidual in the population (S. Zack and B. J. tory is low and there are only enough- females Stutchbury, pers. commun.). In other words, to provide mates for males living on territories early dispersallbreeding may be the best op- of higher quality than z, then in effect the

tion available to one individual, while delayed habitat-fitness gradientu for individuals dis- dispersallhelping may be better for another, persinglbreeding (and for hypothetical off- depending on the quality of the territory in spring delaying dispersal) falls precipitously which they reside. We would go even further in territories of quality less than z [Fig. 7(b)]. and suggest that individual quality may also In contrast to the original situation, there are be an important factor for determining the now significantly fewer territories acceptable preferred option for an individual. For exam- for independent reproduction, and the prob- ple, individuals may differ in their competi- ability that delayed dispersal will be chosen tiveness for reproductive vacancies; those that by some individuals is correspondingly in- are most competitive might be best off dis- creased. persing early while those that are not might Thus the sex ratio. to the extent that it influ- best delay dispersal until they are older. Thus, ences the availability of mates, is an impor- a complete understanding of the costs and tant factor potentially influencing the expected benefits of each option available to offspring fitness of delayed dispersal compared THE QUARTERLY REVZE W OF BIOLOGY VOLUME67

of a biased sex ratio in many species of nonco- Independen operative breeders (Koenig and Pitelka, 1981). breeding 3/ 5. Environmental variability. Thus far we have < assumed that the relative quality of different territories remains stable over time, leading to fitness gradient curves that are predictable from year to year. This is always going to be an oversimplification. On a local scale. and more so for some habitats and foraging patterns than others, Y x territory x year interactions due to spatiotemporal variability in resource availability, C) Independent breeding I/ predator pressure, or sex ratio among territories may be so great that the ranking of terri------tory quality may vary considerably from year wd(x, dispersal to year (Waser, 1988). Regardless of what the habitat-fitness gradients look like during a single season, low predictability in the rank- ingof relative territory quality over time will reduce the long-term variance in overall territory quality (Stacey and Ligon, 1991). In the Worst Best extreme case, many individuals would be Environments ranked by forced to abandon their vreviouslv suitable suitability for breeding territories each year and start over elsewhere. This happens to acorn woodpeckers in years FIG. 7. THEEFFECT OF SEXRATIO ON when the acorn crop fails (~annonet al., DELAYEDDISPERSAL. (a) Even sex ratio: An individual born on terri- 1987). These territories would be settled ac- tory x need only find a territory of quality greater cording to dominance status, priority, or than y for dispersallbreeding to be the preferred whatever factors make some individuals more option. (b) A biased sex ratio: Only enough indi- competitive than others. viduals of the rarer sex to occupy territories of Waser (1988), Powell (1989), Zack (1990), quality greater than or equal to z are assumed to and Stacey and Ligon (1991) all suggest that exist. The fitness of individuals that occupy territo- increased spatiotemporal variation will lower ries without a member of the opposite sex is pre- the probability of delayed dispersal and coop- sumably low, causing a severe steepening of the erative breeding by decreasing the potential habitat-fitness gradients for dispersallbreeding fitness payoff of this option compared to dis- and delayed dispersal. Under these conditions, an offspring born on territory x that does not find a persallbreeding; thus, extended natal philo- vacancy with a mate on a territory of quality patry is likely to be found in species in which greater than or equal to z is best off floating, assum- territory quality is highly predictable over ing (in the example shown here) that floaters can time. In addition, we propose two reasons for occupy space in the best available habitats for why the option of floating might be relatively floating. better under conditions of high spatiotemporal variation. First, the maximum fitness of an individual that either delays dispersal or to dispersallbreeding. Like the other factors disperses and breeds is lower, thereby making we consider here, however, a biased sex ratio these options less desirable. Concurrently, is not a general explanation of delayed dis- the unpredictability of relative territory qual- persal in cooperative breeders, as evidenced ity from year to year makes it more likely that by both the existence of nonbreeders of both flbaters. because of their mobilitv.,, will be able sexes in many cooperative breeders (e.g., to gain access to high-quality areas where re- Florida scrub jays, Woolfenden and Fitzpat- sources are more abundant and possibly even rick, 1984; acorn woodpeckers in California, suitable for breeding; this will increase the Koenig and Mumme, 1987) and the existence fitness of the floatinioption. All of these fac- JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 137 tors would tend to decrease the likelihood of intrinsic benefits to delayed dispersal out- delayed dispersal compared to alternative dis- weigh the costs [Fig. l(a)], as well as those persal options as spatiotemporal environmen- in which extrinsic factors currently exercise tal variation increases. proximate constraint on dispersal [Fig. l(b)]. Second, fitness x year interactions due to Both of these conditions postulate that de- annual differences influence the absolute layed dispersal is a consequence of a restricted quality, but not the relative ranking, of indi- ability to breed independently (Emlen, 1982a; vidual territories. Annual differences in re- Koenig and Mumme, 1987: 383). In the case productive success and survivorship within a of extrinsic constraints, the ecological con- population would result. The fact that in good straint is typically lack of high-quality breed- years, previously poor territories may become ing space, while in the case of intrinsic bene- of sufficiently high quality to support breed- fits, the ecological constraint acts on the ing may be equally important. Individuals ability of pairs without helpers to acquire suf- taking advantage of such temporarily out- ficient resources (other than space) to breed standing conditions may have previously successfully. been either nonbreeding helpers, floaters, or If intrinsic benefits are paramount, the five breeders on lower-quality territories. Varia- parameters discussed above and numbered in tion differentially influencing territory qual- Figure 8 then interact to determine optimal ity (but not enough to alter their relative group size. If extrinsic constraints force de- ranking), might also alter the shape of the layed dispersal, these same five factors inter- habitat-fitness gradient curves and thereby act to determine the probability of delayed influence the probability of delayed dispersal dispersal for any particular individual. For as discussed in prior sections. the five parameters discussed above, some The primary effect of such annual variation important additional subcategories discussed in conditions is to temporarily increase (or earlier are also included. decrease) the relative density of the popula- These five parameters are not mutually ex- tion, both by altering the average absolute fit- clusive. Indeed, they are best thought of as ness of individuals in the population, thereby axes in five-dimensional space, certain sec- resulting in larger (or smaller) population size, tions of which specify conditions favoring de- and by increasing (or decreasing) the average layed dispersal and cooperative breeding absolute quality of territories, and thus di- while others favor delayed breeding with rectly altering the amount of occupiable space. floating and still others dispersal with early These factors directly determine the degree of breeding. No one factor by itself causes de- habitat saturation (defined as relative density) layed dispersal in general, but a difference in and thus the likelihood of delayed dispersal, the dispersal patterns observed between two as discussed earlier. In particular, relative species or populations may be attributable to density is likely to vary considerably from a small subset or even a single factor. The year to year, leading to wide variation in the challenge to future workers in this field is to observed proportion of offspring choosing the identify these factors in specific cases. various dispersal options depending on the In Table 1 we go a step further and summa- current environmental conditions relative to rize the major ways in which the ecological, population density. demographic and environmental factors discussed above are likely to influence the choice A Categorization of Factors ZnJuencing among the three dispersal options. Given the Delayed Dispersal and a Comparison number of these variables, combined with the with Models for Territorial Polygyny likelihood that many of them interact in com- A flowchart summarizing the factors dis- plex ways, it is not surprising that the evolu- cussed above as potentially influencing de- tion of delayed dispersal and cooperative layed dispersal is presented in Figure 8. All breeding has proven to be an uncommonly modern hypotheses for the evolution of de- interesting phenomenon. layed dispersal in cooperative breeders invoke In Table 2 we offer a comparison of differ- ecological constraints in the sense proposed ent forms of the delayed-dispersal threshold by Emlen (1982a). This includes cases in which model with the classification of models for the THE QUARTERLY REVZE W OF BIOLOGY 7,Ecological constraints

I. Inlrimie benefits

and "benefits of philopaw

FIG. 8. A FLOWCHARTSUMMARIZING FACTORS INFLUENCING DELAYED DISPERSAL AND THE OCCURRENCEOF COOPERATION. All current hypotheses invoke ecological constraints. These constraints consist of varying combina- tions of intrinsic benefits and extrinsic constraints on the ability of offspring to obtain high-quality breeding opportunities. If intrinsic benefits are paramount, delayed dispersal will be a preferred option for at least some offspring; if extrinsic constraints are paramount, some offspring may or may not be forced to make the best of a bad job and delay dispersal and help. In either case, the five numbered parameters will be important in determining the extent of delayed dispersal and consequent group size. Important variables determining the influence of these five parameters on the best available dispersal option for offspring are also listed.

TABLE 1 A categorization of factors inyuencing the relative fitness of dispersal options available to offspring Dispersal and independent Delayed Dispersal and breeding dispersal floating

I. Intrinsic benefits low high low 11. Extrinsic constraints low high high 1. Population density low high high 2. Fitness differentials Resource depressibility high low high Group size large small large Group composition nonrelatives relatives nonrelatives Cost of dispersal low high low Secondary intrinsic benefits low high low 3. Floater fitness Quality of habitat available to floaters poor poor good Ability of floaters to exploit variable resources poor poor good 4. Distribution of territory quality Ratio of optimal to marginal habitat (marginal habitat hypothesis) low high high Variance in territory quality (variance hypothesis) low high high Sex ratio even biased biased 5. Environmental variability high low high Qualitative values specify the relative conditions favoring a particular dispersal option DELAYED DISPERSAL IN COOPERATIVE BREEDERS

TABLE 2 A comparison of the delayed dispersal threshold model with Seary and Yasukawa's (1989) classification of models for the occusrence of territorial polypyny Territorial polygyny Delayed dispersal

I. Male-coercion model (Parents force offspring to delay dispersal) 11. Female-choice models Offspring choose to delay dispersal A. No-cost models Intrinsic benefits important 1. Benefit models Intrinsic benefits paramount or secondary benefits compensate for costs 2. No-benefit models Secondary benefits balance costs B. Cost models Extrinsic constraints paramount 1. Skewed-sex-ratio model Shortage of potential mates limits breeding opportunities 2. Balanced-sex-ratio models No shortage of potential mates a. Compensation model Offspring delaying dispersal are compensated by remaining on a territory of high quality b. No-compensation models Offspring are not compensated for the cost of delayed dispersal i. Search-cost model Offspring delay dispersal because searching for unoccupied territory of suitable quality is costly ii. Deception model (No parallel) iii. Maladapted-female model Offspring choose delayed dispersal even though superior alternatives are available to them Scenarios in parentheses are not known to occur and are listed for comparative purposes only

occurrence of territorial polygyny presented We have not directly discussed analogs of by Searcy and Yasukawa (1989). Given that the "no-compensation" models (IIB2b) listed parents are unlikely to be able to force off- in Table 2. The most likely scenario in this spring to delay dispersal, all models for de- category is the "search-cost" model analog, layed dispersal are analogous to female-choice which proposes that offspring may delay dis- models for polygyny. As in the polygyny thresh- persal because finding vacancies in high- old model, factors leading to delayed dispersal quality territories is costly, even though they can then be divided into "no-cost" and "cost" exist. This effectively augments the cost of versions, corresponding to whether intrinsic dispersal, decreasing the relative benefits of benefits or extrinsic constraints are para- the dispersallbreeding option and thus in- mount. [These alternatives are also analogous creasing the probability of delayed dispersal, to the "cooperative female choice" and "com- as discussed in earlier sections. petitive female choice" models for polygyny discussed by Altmann et al. (1977).] Simi- larly, different versions of cost models involve UNANSWERED QUESTIONS uneven or balanced sex ratios. The maioritv., , Here we briefly discuss some of the more of our discussion above, however, focuses on perplexing problems and unanswered ques- the analog of the "compensation" model (Ta- tions concerning the evolution of cooperative ble 2. IIB2a) in which females are comDen- breeding. sated'for the'cost of polygyny by acquirhg a The relative significance of jloating. As dis- male with a high-quality territory. For de- cussed earlier, most prior theories for the evo- layed dispersal, this corresponds to the situa- lution of cooperative breeding are based pre- tion in which offspring are compensated for dominately on the alternatives of dispersal1 the cost of delayed dispersal by remaining on breeding versus delayed dispersallhelping. a high-quality (usually their natal) territory, This contrast may explain differences in where they may have high survivorship, be group size in cooperative breeders and, in able to search for vacancies, acquire second- specific populations, delayed dispersal, but ary benefits, and achieve indirect fitness ben- fails to consider, much less explain, dispersal1 efits by helping to raise nondescendant kin. floating in noncooperative populations. 140 THE QUARTERLY REVZE W OF BIOLOGY VOLUME67

The problem of floaters is highlighted by the (1980), who found that nonbreeding Santa recent study by Carmen (in press) comparing Cruz Island (California) scrub jays disperse noncooperative and cooperative populations and form loosely organized flocks that exploit of Aphelocoma jays. His study population in a wide range of habitats largely unsuitable central coastal California is representative of for breeding. In contrast, on the California noncooperative scrub jay populations in west- mainland, nonbreeders are tolerated on both ern North America. By drawing on the ex- their natal and other established territories tensive data available for the cooperative Flor- except during the height of the breeding sea- ida race of this species (e.g., Woolfenden and son, and occur in loose flocks in habitats con- Fitzpatrick, 1984) and the congeneric gray- taining the most abundant resources (Car- breasted jay (e.g., Brown and Brown, 1990), men, in press). As in rufous-collared sparrows Carmen found that demographic differences (Smith, 1978), floaters on the mainland use among these forms are slight, with nearly habitat occupied by breeders, and the pres- identical levels of breeder survivorship and re- ence or absence of unoccupied marginal habi- productive success. Indeed, the demographic tat may not be the critical factor limiting dis- patterns suggest that space competition in non- persallfloating. These studies suggest that cooperative populations, such as the Santa constraints on or benefits of early dispersal1 Cruz Island scrub jay (Atwood et al., 1990), floating, rather than constraints on indepen- may equal or exceed that found in the cooper- dent breeding, is the critical factor distin- ative populations. It follows that differences guishing cooperatively from noncooperatively in the degree of space competition per se can- breeding jays. not be the cause of the differences in dispersal If this is true, what are the underlying eco- options chosen by offspring in this complex. logical factors that contribute to different lev- As mentioned earlier, some cooperative els of floater fitness? An important difference breeders are dependent on a localized, limited may be the seasonal abundance and distribu- and essential resource that apparently con- tion of acorns. Each autumn, individual scrub strains both early independent breeding and jays in Florida (DeGange et al., 1989) and in floating. No such specific habitat feature is California (Carmen, in press) cache 5000 to implicated in leading to delayed dispersal of 8000 acorns for use during the winter and either Florida scrub jays or gray-breasted early spring. In California, poor acorn years jays. Although constraints on breeding may result in significantly higher mortality, repro- still be evident in these species, it is more diffi- ductive failure, and territory abandonment cult to explain why delayed dispersal is fa- (Carmen, in press). Comparable data are not vored over floating. available from Florida, because acorn pro- In Florida, young scrub jays wander freely, duction is regular in both space and time; over but in autumn increased aggression from an eleven-year period, for example, no acorn other groups restricts nonbreeders to either crop failures were detected (DeGange et al., their natal territories or to tracts of unburned 1989). Assuming that scrub jays are as depen- scrub. Woolfenden and Fitzpatrick (1984) ar- dent on acorns in Florida as in California, gue convincingly that unburned areas, which the very low occurrence of acorn crop failure are generally unsuitable for breeding except would prevent local population crashes. More- at very low densities (see Woolfenden and Fitz- over, in Florida, the low but even density dis- patrick, 1991), are also unsuitable for float- tribution of acorns contributes to a situation ing; hence both breeders and nonbreeders are where territory defense is economical, intruder restricted to a narrow range of acceptable hab- pressure is slight, and floaters probably do not itat. Woolfenden and Fitzpatrick (1984) sug- have the option of living in occupied areas. gested that little marginal habitat exists in But why is floating in unoccupied areas not Florida compared to the western United a viable option? Woolfenden and Fitzpatrick States, and this is consistent with the impor- (1984) suggest that high predation rates in tance of the distribution of territory quality as unburned scrub and other marginal habitats proposed by the marginal habitat hypothesis outside of occupied areas are responsible, but (see above). differences in acorn ~roductioncould also be Additional evidence comes from Atwood important. If acorn production declines with JUNE 1992 DELA YED DISPERSAL IN COOPERATIVE BREEDERS 141 age of sprouts, as found for scrub oak in New apparently employed by individuals attempt- Jersey (Wolgast and Stout, 1977), this would ing to obtain high-quality breeding oppor- provide at least one reason why periodically tunities. otherwise, these dispersal options burned scrub is so highly prized by Florida rarely appear to overlap within a single popu- scrub jays and why floating in denser, un- lation, or even (with the notable exception of burned scrub is not favored. the scrub jay) within the geographic range of In contrast, acorn production by the large a single species. This is surprising given the and patchily distributed oaks of several spe- wide range of ecological settings inhabited by cies in California is highly variable locally some cooperatively breeding species (e.g., within and among years with periodic crop acorn woodpeckers; Koenig and Stacey, failures four to six years apart (Hannon et al., 1990) and by some genera, such as Pomatosto- 1987), depending in part on the number of mus in Australia, all of whose species are coop- oak species in an area (Carmen et al., 1987). erative breeders (Edwards and Naeem, in Individual large oaks may produce hundreds press). of thousands of acorns in good years. Early Conditions leading to delayed dispersal/helping dispersal may be favored as-floaters aggregate versus cooperative polygamy. In addition to the in areas of high acorn abundance, and the simplest form of cooperative breeding, in tolerance of floaters by breeders allows them which offspring delay dispersal and subse- access to the best habitats. Floaters are also quently act as nonbreeding helpers on their free to move regionally and search out breed- natal territory, cooperative breeding also in- ing areas where populations have declined cludes cooperative polygamy (Faaborg and due to crop failures (Carmen, in press). Cur- Patterson, 1981) and plural breeding (Brown, rently, little is known about acorn production 1987),,. in which two or more individuals of patterns in either gray-breasted jay or Santa the same sex share breeding status within a Cruz Island scrub jay habitats, or the degree social unit. Plural breeding usually involves to which these jays depend on acorns. Such multiple pairs of breeders within a social unit, data would be of particular interest in the while cooperative polygamy may involve gray-breasted jay, as no obvious habitat or cobreeding males that share one or more fe- vegetation features separate high-quality hab- males and joint-nesting females that share one itats from poor ones in this plural-breeding or more males in virtually any combination species (Brown and Brown, 1985; Edwards, (see Brown, 1987, Table 2.2). 1986). Indeed, Brown (1986: 781) states that On the basis of their differing models, Sta- "although individuals are free to disperse and cey (1982) predicted that mate-sharing should explore to find suitable areas elsewhere, they occur when there are significant intrinsic ben- rarely do." efits to all individuals, while Vehrencamp Unfortunately, it is not possible to directly (1983) predicted that societies in which indi- test for differences in floater fitness because viduals share mates equally should occur floaters are rare in Florida and range widely in when intrinsic benefits are weak or absent, California. Clearly, though, the phenomenon and thus extrinsic constraints lead to this phe- of floaters, their significance as a dispersal nomenon. Empirical studies have similarly option, and their variable occurrence as a come to diverse conclusions. Stacey (1979b, demographic subclass among cooperatively 1982) found that intrinsic benefits leading to breeding species calls for more extensive com- increased reproductive success were impor- parative study. tant to mate-sharing by male acorn wood- Co-occurrence of delayed dispersal and early dis- peckers in New Mexico, while in California, persal/Joating. Four species in which these dis- both extrinsic constraints and intrinsic bene- persal strategies are known to regularly coex- fits appear to be important to mate-sharing ist are the Australian magpie (Carrick, 1972; males and joint-nesting females (Koenig and Veltman, 1989), purple gallinule (Hunter, Mumme, 1987; Mumme et al., 1988). Ko- 1987), common crow (Caffrey, 1991), and ford et al. (1986) found good evidence that red-cockaded woodpecker (Walters et al., extrinsic constraints, possibly in the form of 1992). The Australian magpie is particularly nest sites, are important to the evolution of notable because of the widerange in strategies plural breeding in groove-billed anis, but also 142 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67 found a role for intrinsic benefits of group were previously all believed to be nonbreeders living. Rowley et al. (1989) concluded that (Rabenold et al., 1990). To the extent that plural breeding in the splendid fairy-wren this is widespread, delayed dispersallhelping (~alurussplendens) was primarily a function of and plural nesting can perhaps be considered extrinsic constraints in the form of increased on the same continuum, with the latter occur- population density. Finally, work by Davies ring when demographic pressures are even and his colleagues (L)avies and Houston, higher than those leading to the former (e.g., 1986; Burke et al., 1989; Davies 1990) on Fitzpatrick and Woolfenden, 1986; but see dunnocks (Prunella modularis) demonstrates Koford et al., 1986). that intrinsic benefits modulated by conflicts Alternatively, nonbreeding helpers and of interest between males and females are im- cobreeding may be distinct phenomena se- portant in maintaining their variable mating lected by independent sets of ecological fac- system. Thus current empirical data provide tors. If so, there is currently no consensus support for almost any combination of intrin- concerning the conditions that promote one sic benefits and extrinsic constraints playing a kind of cooperative breeding over another, role in the evolution of cooperative polygamy much less the conditions leading to joint- and plural nesting. nesting in some species and plural breeding One factor that appears to be of particular in others. importance to several joint-nesting systems is Sex bias among helpers. There is as yet no the advantage in acquiring territories or other satisfactory general explanation for the sex critical resources "gained bv coalition forma- bias found among helpers in many species. tion. For example, male lions (Panthera lea) have Two general classes of hypotheses exist. higher expected lifetime reproductive success Charnov (1981) suggested that the tendency in-larger-coalitions due primarily to an in- for one sex or the other to act disproportion- creased probability of gaining access to a pride ately as helpers is caused by sexual asymme- of females (Koenig, 1981~;Packer et al., tries in relatedness resulting from EPCs by 1988). Other advantages, including longer males or from intraspecific parasitism by fe- residency and access to larger prides, confer males. For example, assuming that egg dump- lesser benefits. Coalition formation leading to ing by females is rare and that EPCs by males increased access to resources (usually territo- are not, females will be more closely related ries) have also been reported in pukeko Porph- to their own offspring than to broodmates, yrio porphyrio (Craig, 1984), green wood- while males will be more closely related to hoopoes (Ligon and Ligon, 1979, 1988), and broodmates than to young in their own nest, acoin woodpeckers (~annonet al., 1985; some of which they will not have fathered. Mumme et al., 1988). This benefit can be Consequently, males receive relatively greater considered an extrinsic constraint leading to indirect fitness benefits by delaying dispersal group formation, since the advantage gained and helping to raise siblings compared to dis- is that of acquiring an opportunity to breed, persing and breeding independently. In the not in reproduction or survivorship per se context of the delayed-dispersal threshold (Koenig, 1981~).Why such extrinsic con- model, this hypothesis proposes a sex differ- straints lead to coalition formation and ioint- ence in the fitness differential between early nesting in a few species, while apparently and delayed dispersal: males experience a comparable constraints result in delayed dis- smaller differential than females because the persal and helping in others, is unknown. higher probability of being cuckolded results At one level, the distinction between non- in a lower benefit to independent breeding. breeding helpers that delay dispersal and co- To the extent that the above scenario is breeders that share a mate within a social unit likely, Charnov's (1981) argument can ex- can be reduced to differing degrees of repro- plain why there is a sex bias toward males ductive bias within a group (Vehrencamp, among individuals that delay dispersal and 1983). This view is supported by the occur- remain as helpers in some species. Confirma- rence of mate-sharing in species such as the tion of this argument is dependent on finding stripe-backed wren, in which male helpers a correlation between the sex bias in helpers JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 143 and the incidence of EPCs and egg dumping including Harrison (1969), Dow (1980), within cooperative breeders, as determined Brown (1987), Ford et al. (1988), and Russell by parentage analyses. Particularly good tests (1989). would come from examination of species in Russell (1989) pointed out that virtually which helpers are predominantly females the entire Australian continent has not been rather than males, as is true for common exposed to extremes of cold and seasonal crows (Caffrey, 199 1) and small-bodied social drbucht" since at least the mid-Miocene. 15 canids (Moehlman, 1986). This latter taxon is M .Y.B.P. Consequently, the Australian avi- particularly interesting from this perspective, fauna never developed the large-scale migra- as comparative data show that the sex ratio of tions characteristic of ~uroieand ~orth helpers covaries with body size: helpers are America. Given that cooperative breeders are predominantly or exclusively females in small- generally nonmigratory (Brown, 1974) and bodied species, approximately even in several are often found in habitats where resources medium-sized Canis, and predominantly males do not show marked seasonal fluctuations in larger taxa (Moehlman, 1986). (Ford et al., 1988), this situation may at least The second class ofhypotheses for a sex bias provide the setting for the subsequent evolu- among helpers focuses on the relative costs of tion of such a high incidence of cooperative delayed dispersal. For example, male Florida breeding. This cannot be the whole story, scrub jays are more likely to either inherit however, since there are still many noncoop- their natal territory or breed adjacent to their erative breeders in Australia. and even mem- natal territory and thus stand to benefit more bers of the same genus living sympatrically than females by delaying dispersal and help- may differ in their social organization (Ford ing; this is in fact the pattern documented by et al., 1988; Russell, 1989).

Stallcup and Woolfenden (1978). Unfortu- Another relevant factor mav be the ~hvlo-*, nately, there is an element of circularity in genetic history of the Australian avifauna. this argument: Do males help more because Russell (1989) and Edwards and Naeem (in they are philopatric or are they philopatric press) have analysed the phylogenetic rela- because they help more? In any case, this cor- tionships of cooperatively breeding Austra- relation does not appear to generalize to other lian passerines from a cladistic perspective;- - species; in green woodhoopoes, for example, they suggested that the high proportion of co- males are more philopatric, as in Florida operative breeders in the Australian avifauna scrub jays, bpt there is no clear asymmetry in may in part be a consequence of diversifica- helping behavior between the sexes (Ligon tion among a relatively smaller number of and Ligon, 1979). lineages displaying this behavior. In part, this Several additional hypotheses for sex biases canclusion begs the question of why coopera- in delayed dispersal and helping behavior are tively breeding lineages have been so much discussed by Koenig et al. (1983) and Wool- more successful in Australia than elsewhere, fenden and Fitzpatrick (1 986). None provides and thus does not resolve the mystery. None- a general explanation for the variation in this theless, the analyses of Edwards and Naeem parameter observed in cooperative breeders. (in press) make it clear that assessing the im- For species in which only one sex (generally portance of phylogeny to the distribution and males) delays dispersal and help, the nagging occurrence of cooperative breeding is an im- question remains: How can the conditions portant challenge-for the future. - leading to delayed dispersal act on one sex but The relationship between the social unit and the not the other (Walters, 1990)? mating system. It has generally been taken for Geographic bias in cooperative breeding. Lack granted that the birds breeding in a particular (1968) was apparently the first to point out social unit are group members. Recent elec- the surprisingly high proportion of birds (esti- trophoretic work by Brooker et al. (1990) on mated as at least 22 % of passerines; Russell, splendid fairy-wrens, however, has shown 1989) that exhibit cooperative breeding in that at least 65 percent of offspring in their Australia. Several subsequent authors have population were apparently fathered by males noted or attempted to explain this pattern, outside of the group. Thus, as has been docu- 144 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67 mented in numerous apparently monoga- Detailed long-term studies of cooperative mous species (Westneat et al., 1990), there is breeders will continue to add important clues a potential discordance between the apparent to the puzzle of delayed dispersal, particularly and the genetically effective mating system in if combined with experimental manipulations cooperative breeders. (e.g., Zack and Rabenold, 1989; Pruett-Jones In the case of the splendid fairy-wrens, this arid Lewis, 1990; Komdeur, 1991; Walters et revelation calls into question the extraordi- al., in press). In addition, we advocate three narily high incidence of apparent incest that kinds of studies. First are experimental tests had been reported earlier based on group of specific ecological factors important in the composition (Rowley et al., 1986). Similar delayed-dispersal threshold model. Second discoveries in other species as a result of mod- are studies of noncooperatively breeding pop- ern molecular analyses (e.g., Burke et al., ulations, either with or without close relatives 1989; Rabenold et al., 1990) could have pro- that are cooperative breeders, focusing on found effects on our understanding of not only what constraints to independent reproduction the mating system within groups of coopera- exist and why they do not result in delayed tive breeders, but also on our estimates of the dispersal. Examples of such studies currently costs and benefits of the different behavioral include those on Lanius shrikes (Zack and Li- strategies observed in these complex systems. gon, 1985a, 1985b),Aphelocoma jays (see above and Carmen, in press), the various tits (genus CONCLUSIONS Parus) (e.g., Ekman et al., 1981; Smith, 1984; The basic factors important to delayed dis- Ekman, 1988; Matthysen, 1990), and ban- persal were identified over 25 years ago by ner-tailed kangaroo rats (Jones et al., 1988; Selander (1964). Work since then has pro- Waser, 1988). Third are studies of intraspe- gressed in terms of understanding the demo- cific variation in group size and composition graphic conditions correlating with delayed of cooperative breeders in relation to local dispersal, but we still know relatively little habitat gradients and patchiness. Little work about the ecological conditions promoting along these lines has been performed, yet such this phenomenon in some species, while con- studies hold considerable promise for yielding straining options such as floating in others. insights into the environmental correlates of Given the complexity and interrelatedness delayed dispersal and cooperative breeding. of the factors potentially influencing alterna- ACKNOWLEDGMENTS tive dispersal strategies, no single factor is likely to explain delayed dispersal and helping We thank Jerram Brown, Steve Emlen, Janis Dickinson, Phil Hooge, Mornt du Plessis, Eliza- behavior in all cooperative breeders. Indeed, beth Ross, Jeff Walters, Glen Woolfenden, and the five factors identified here, including pop- two anonymous reviewers for their comments. ulation density, fitness differential between de- Carolee Caffrey, Scott and Nancy Creel, Nick Da- layed dispersal and dispersallbreeding, floater vies, Scott Edwards, Steve Emlen, Jan Komdeur, fitness, distribution of territory quality, and Dave Ligon, Arch McCallum, Steve Pruett-Jones, spatiotemporal environmental variation, all Pete Stacey, Jeff Walters, and Steve Zack were play an important, often interacting, role in kind enough to share unpublished manuscripts; determining the likelihood of delayed disper- we thank these and all our other colleagues who sal and cooperative breeding. The delayed- have engaged us in debate about cooperative dispersal threshold model provides a frame- breeding over the years. Our work has been supported by the NSF, most recently by grants work that can aid in identifying key factors BSR90-21659 to WDK, BSR86-00174 to RLM, differing between closely related taxa exhib- and by a graduate fellowship to MTS; logistic sup- iting contrasting dispersal patterns among port was provided by Mark Johnson, Dan Lufkin, their offspring. Measuring the parameters and Mark Stromberg. Finally, we thank Fanny needed to examine the model quantitatively Hastings Arnold for her continuing support of the is a key challenge to future researchers. research program at Hastings Reservation. DELAYED DISPERSAL IN COOPERATIVE BREEDERS

REFER

Alexander, R. D. 1974. The evolution of social tiue Skills: Comparatiue and Ontogenetic Approaches, behavior. Ann. Rev. Ecol. Syst., 5:325-383. pp. 137-171. Erlbaum Associates, Hillsdale. Altmann, S. A. 1974. Babboons, space, time and . 1986. Cooperative breeding and the regu- energy. Am. Zool., 14:221-248. lation of numbers. In V. D. Ilyichev and V. M. Altmann, S. A., S. S. Wagner, and S. Lenington. Gavrilov (eds.), Proceedings of the 18th Interna- 1977. Two models for the evolution of polyg- tional Ornithological Congress, pp. 774-782. Nauka, yny. Behau. Ecol. Sociobiol., 2:397-410. Moscow. Andersson, M. 1984. The evolution of eusociality. . 1987. Helping and Communal Breeding in Ann. Reu. Ecol. Syst., 15: 165-189. Birds. Princeton University Press, Princeton. Atwood, J. L. 1980. Social interactions in the . 1989. Habitat saturation and ecological Santa Cruz Island scrub jay. Condor, 82:440- constraints: origin and history of the ideas. Con- 448. dor, 91:1010-1013. Atwood, J. L., M. J. Elpers, and C. T. Collins. Brown, J. L., and E. R. Brown. 1984. Parental 1990. Survival ofbreeders in Santa Cruz Island facilitation: parent-offspring relations in com- and mainland California scrubjay populations. munally breeding birds. Behau. Ecol. Sociobiol. , Condor, 92:783-788. 14:203-209. Austad, S. N., and K. N. Rabenold. 1986. De- , and -. 1985. Ecological correlates mography and the evolution of cooperative of group size in a communally breeding jay. breeding in the bicolored wren, Campylorhynchus Condor, 87:309-315. griseus. Behauiour, 97:308-324. , and -. 1990. Mexicanjays: uncoop- Bednarz,J. C. 1988. Cooperative hunting in Har- erative breeding. In P. B. Stacey and W. D. ris' hawks (Parabuteo unicinctus). Science, 239: Koenig (eds.), Cooperative Breeding in Birds: 1525-1527. Long-Term Studies of Ecology and Behauior, pp. Brooker, M. G., I. Rowley, M. Adams, and P. R. 267-288. Cambridge University Press, Cam- Baverstock. 1990. Promiscuity: an inbreeding bridge. avoidance mechanism in a socially monoga- Burke, T., N. B. Davies, M. W. Bruford, and mous species? Behau. Ecol. Sociobiol., 26: 191- B. -1. Hatchwell. 1989. Parental care and mat- 199. ing behaviour of polyandrous dunnocks Prunella Brown,J. L. 1963. Social organization and behav- modularis related to paternity by DNA finger- ior of the Mexican jay. Condor, 65: 126-153. printing. Nature, 338:249-251. . 1969a. Territorial behavior and popula- Caffrey, C. 1991. Breeding group structure and tion regulation in birds. Wilson Bull., 81:293- the effects of helpers in cooperatively breeding 329. western American crows. PhD thesis, Univer- . 1969b. The buffer effect and productivity sity of California, Los Angeles. in tit populations. Am. Nut., 103:347-354. Carmen, W. J. In press. Behavioral ecology of the . 1974. Alternate routes to sociality in California scrub jay: a noncooperative breeder jays -with a theory for the evolution of altruism with close cooperative relatives. Stud. Avian and communal breeding. Am. Zool., 14:63-80. Biol. . 1975. The Evolution of Behauior. W. W. Carmen, W. J., W. D. Koenig, and R. L. Norton, New York. Mumme. 1987. Acorn production by five spe- . 1978. Avian communal breeding systems. cies of oaks over a seven year period at the Ann. Rev. Ecol. Syst., 9: 123-1 56. Hastings Reservation, Carmel Valley, Califor- . 1980. Fitness in complex avian social sys- nia. In T. R. Plumb and N. H. Pillsbury (tech. tems. In H. Mark1 (ed.), Evolution of Social Be- coordinators), Proceedings of the Symposium on hauior: Hypotheses and Empirical Tests, pp. 115- Multiple-Use Management of Cal$omia's Hardwood 128. Verlag Chemie, Wienheim. Resources, pp. 429-434. Gen. Tech. Rep. PSW- . 1982. Optimal group size in territorial 100. Pacific Southwest Forest and Range Ex- animals. J. Theor. Biol., 95:793-810. periment Station. . 1983. Cooperation- a biologists' di- Carrick, R. 1972. Population ecology of the Aus- lemma. In J. S. Rosenblatt et al. (eds.), Ad- tralian black-backed magpie, royal penguin, vances in Behavior, Vol. 13, pp. 1-37. Academic and silver gull. U.S. Dept. Inter. Wildl. Res. Press, New York. Rep., 2:41-99. . 1985. The evolution of helping behav- Charnov, E. L. 1981. Kin selection and helpers at ior -an ontogenetic and comparative perspec- the nest: effects of paternity and biparental tive. In E. S. Gollin (ed.), The Evolution ofAdap- care. Anim. Behav., 29:631-632. 146 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67

Charnov, E. L., G. H. Orians, and K. Hyatt. . 1982a. The evolution of helping. I. An 1976. Ecological implications of resource de- ecological constraints model. Am. Nut. ,119:29- pression. Am. Nut., 110:247-259. 39. Craig, J. L. 1984. Are communal pukeko caught . 198213. The evolution of helping. 11. The in the prisoner's dilemma? Behav. Ecol. Socio- role of behavioral conflict. Am. Nut. , 119:40- biol., 14:147-150. 53. Creel, S. R., and N. M. Creel. 1991. Energetics, . 1984. Cooperative breeding in birds and reproductive suppression and obligate commu- mammals. In J. R. Krebs and N. B. Davies nal breeding in carnivores. Behau. Ecol. Socio- (eds.), Behavioural Ecology: An Evolutionary Ap- biol., 28:263-270. proach, 2nd ed., pp. 305-339. Sinauer, Sunder- Curry, R. L., and P. R. Grant. 1990. Galipagos land. mockingbirds: territorial cooperative breeding . 1990. White-fronted Bee-eater: helping in in a climatically variable environment. In P. B. a colonially nesting species. In P. B. Stacey Stacey and W. D. Koenig (eds.), Cooperative and W. D. Koenig (eds.), Cooperative Breeding in Breeding in Birds: Long- Term Studies of Ecology and Birds: Long-Term Studies of Ecology and Behauior, Behauior, pp. 289-331. Cambridge University pp. 487-526. Cambridge University Press, Press, Cambridge. Cambridge. Davies, N. B. 1989. Sexual conflict and the polyg- . 1991. Cooperative breeding in birds and amy threshold. Anim. Behau., 38:226-234. mammals. In J. R. Krebs and N. B. Davies . 1990. Dunnocks: coo~erationand conflict (eds.), Behavioural Ecology: An Euolutionary Ap- among males and females in a variable mating proach, 3rd ed., pp. 301-337. Blackwell, Ox- system. In P. B. Stacey and W. D. Koenig ford. (eds.), Cooperative Breeding in Birds: Long- Term Emlen, S. T., and P. H. Wrege. 1989. A test of Studies of Ecology and Behavior, pp. 455-485. alternate hypotheses for helping behavior in Cambridge University Press, Cambridge. white-fronted bee-eaters of Kenya. Behau. Ecol. Davies, N. B., and A. I. Houston. 1986. Repro- Sociobiol. , 25:303-319. ductive success of dunnocks, Prunella modularis, Emlen, S. T., H. K. Reeve, P. W. Sherman, P. H. in a variable mating system. 11. Conflicts of Wrege, F. L. W. Ratnieks, and J. Shellman- interest among breeding adu1ts.J. Anim. Ecol., Reeve. 1991. Adaptive versus nonadaptive ex- 55:139-154. planations of behavior: the case of alloparental DeGange, A. R., J. W. Fitzpatrick, J. N. Layne, helping. Am. Nut., 138:259-270. and G. E. Woolfenden. 1989. Acorn harvesting Faaborg, J., and C. B. Patterson. 1981. The char- by Florida scrub jays. Ecology, 70:348-356. acteristics and occurrence of cooperative poly- Dow, D. D. 1980. Communally breeding Austra- andry. Ibis, 123:477-484. lian birds with an analysis of distributional and Fitzpatrick, J. W., and G. E. Woolfenden. 1986. environmental factors. Emu, 80: 12 1- 140. Demographic routes to cooperative breeding in Edwards, T. C., Jr. 1986. Ecological distribution some new world jays. In M. H. Nitecki and of the gray-breasted jay: the role of habitat. J. A. Kitchell (eds.), Evolution ofAnimal Behav- Condor, 88:456-460. ior, pp. 137-160. Oxford University Press, Ox- Edwards, S. V., and S. Naeem. In press. The ford. phylogenetic component of cooperative breed- , and -. 1988. Components of life- ing in perching birds (order Passeriformes). time reproductive success in the Florida scrub Am. Nut. jay. In T. H. Clutton-Brock (ed.), Reproductive Ekman, J. 1988. Subordination costs and group Success, pp. 305-320. The University of Chi- territoriality in wintering willow tits. In H. Ouel- cago Press, Chicago. let (ed.), Proceedings of the 19th International Orni- Fitzpatrick, J. W., G. E. Woolfenden, and K. J. thological Congress, pp. 2373-2381. University of McGowan. 1988. Sources of variance in life- Ottawa Press, Ottawa. time fitness of Florida scrub jays. In H. Ouellet Ekman, J., G. Cederholm, and C. Askenmo. (ed.), Proceedings ofthe 19th International Ornitho- 1981. Spacing and survival in winter groups of logical Congress, pp. 876-891. University of Ot- willow tit Pam montanus and crested tit P. crista- tawa Press, Ottawa. tus- a removal study. J. Anim. Ecol., 50: 1-9. Ford, H. A,, H. L. Bell, R. Nias, and R. Noske. Emlen, S. T. 1978. The evolution of cooperative 1988. The relationship between ecology and the breeding in birds. In J. R. Krebs and N. B. incidence of cooperative breeding in Australian Davies (eds.), Behauioural Ecology: An Euolution- birds. Behau. Ecol. Sociobiol., 22:239-249. ary Approach, pp. 245-281. Sinauer, Sunder- Gaston, A. J. 1978. The evolution of group territo- land. rial behavior and cooperative breeding. Am. JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 147

Nut., 112:1091-1100. helping behaviour in birds: a departure from Gayou, D. C. 1986. The social system of the Texas functional explanations. In P. P. G. Bateson green jay. Auk, 103:540-547. and P. H. Klopfer (eds.), Perspective in Ethology, Gittleman, J. L. 1S85. Functions of communal Vol. 7., pp. 79-98. Plenum Press, New York. care in mammals. In P. J. Greenwood, P. H. Jones, W. T., P. M. Waser, L. F. Elliott, N. E. Harvey, and M. Slatkin (eds.), Evolution: Essays Link, and B. B. Bush. 1988. Philopatry, dis- in Honour of John Maynard Smith, pp. 187-205. persal, and habitat saturation in the banner- Cambridge University Press, Cambridge. tailed kangaroo rat, Dipodomys spectabilis. Ecol- Gowaty, P. A. 1981. An extension of the Orians- ogy, 69:1466-1473. Verner-Willson model to account for mating Koenig, W. D. 1981a. Reproductive success, systems besides polygyny. Am. Nut., 118:8511 group size, and the evolution of cooperative 859. breeding in the acorn woodpecker. Am. Nut., Hamilton, W. D. 1964. The genetical evolution of 117:421-443. social behaviour. I and 11. J. Theor. Biol. , 7: 1- . 1981b. Space competition in the acorn 52. woodpecker: power struggles in a cooperative Hannon, S. J., R. L. Mumme, W. D. Koenig, breeder. Anim. Behav., 29:396-409. and F. A. Pitelka. 1985. Replacement ofbreed- . 1981c. Coalitions of male lions: making ers and within-group conflict in the coopera- the best of a bad job? Nature, 293:413. tively breeding acorn woodpecker. Behau. Ecol. Koenig, W. D., and R. L. Mumme. 1987. Popula- Sociobiol., 17:303-312. tion Ecology of the Cooperatively BreedingAcorn Wood- Hannon, S. J., R. L. Mumme, W. D. Koenig, pecker. Princeton University Press, Princeton. S. Spon, and F. A. Pitelka. 1987. Acorn crop , and -. 1990. Levels of analysis, failure, dominance, and a decline in numbers functional explanations, and the significance of in the cooperatively breeding acorn wood- helping behavior. In M. Bekoff and D. Jamie- pecker. J. Anim. Ecol., 56:197-207. son (eds.), Interpretation and Explanation in the Harrison, C. J. 0. 1969. Helpers at the nest in Study ofAnimal Behavior. Vol. 2: Explanation, Evo- Australian passerine birds. Emu, 69:30-40. lution, and Adaptation, pp. 268-303. West- Hartung, J. 1977. An implication about human view Press, Boulder. mating systems. J. Theor. Biol., 66:737-745. Koenig, W. D., and F. A. Pitelka. 1979. Related- Heinsohn, R. G. 1991. Slow learning of foraging ness and inbreeding avoidance: counterploys in skills and extended parental care in coopera- the cooperatively breeding acorn woodpecker. tively breeding white-winged choughs.- Am. Science, 206:1103-1105. Nut., 137:864-881. , and -. 1981. Ecological factors and Heinsohn, R. G., A. Cockburn, and R. B. Cun- kin selection in the evolution of cooperative ningham. 1988. Foraging, delayed matura- breeding in birds. In R. D. Alexander and tion, and advantages of cooperative breeding D. W. Tinkle (eds.), Natural Selection and Social in white-winged choughs, Corcorax melanor- Behavior: Recent Research and New Theory, pp. hamphos. Ethology, 77:177-186. 261-280. Chiron Press, New York. Heinsohn, R. G., A. Cockburn, and R. A. Koenig, W. D., and P. B. Stacey. 1990. The acorn Mulder. 1990. Avian cooperative breeding: old woodpecker: group-living and food storage un- hypotheses and new directions. Trends Ecol. G3 der contrasting ecological conditions. In P. B. Evol., 5:403-407. Stacey and W. D. Koenig (eds.), Cooper~tiue Hooge, P. N. 1992. Dispersal dynamics of the co- Breeding in Birds: Long- TenStudies of Ecology and operatively breeding acorn woodpecker. PhD Behavior, pp. 413-453. Cambridge University thesis, University of California, Berkeley. Press, Cambridge. Hunter, L. A. 1987. Acquisition of territories by Koenig, W. D., R. L. Mumme, and F. A. Pitelka. floaters in cooperatively breeding purple galli- 1983. Female roles in cooperative breeding nules. Anim. Behav., 35:402-410. acorn woodpeckers. In S. K. Wasser (ed.), So- Jamieson, I. 1989. Behavioral heterochrony and cial Behavior in Female Vertebrates, pp. 235-261. the evolution of birds' helping at the nest: an Academic Press, New York. unselected consequence of communal breed- Koenig, W. D., S. J. Hannon, R. L. Mumme, ing? Am. Nut., 133:394-406. and F. A. Pitelka. 1988. Parent-offspring con- . 1991. The unselected hypothesis for the flict in the cooperatively breeding acorn wood- evolution of helping behavior: too much or too pecker. In H. Ouellet (ed.), Proceedings of the little emphasis on natural selection? Am. Nut., 19th International Ornithological Congress, pp. 138:271-282. 1220-1230. University of Ottawa, Ottawa. Jamieson, I., and J. L. Craig. 1987. Critique of Koford, R. R., B. S. Bowen, and S. L. Vehren- 148 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67

camp. 1986. Habitat saturation in groove- pp. 64-86. Princeton University Press, Prince- billed anis (Crotophaga sulcirostris). Am. Nut. , ton. 127:317-337. Mumme, R. L., W. D. Koenig, and F. A. Pitelka. Komdeur, J. 1991. Influence of territory quality 1988. Costs and benefits ofjoint nesting in the and habitat saturation on dispersal options in acorn woodpecker. Am. Nut., 131 :654-677. the Seychelles warbler: an experimental test of Mumme, R. L., W. D. Koenig, and F. L. W. the habitat saturation hypothesis for coopera- Ratnieks. 1989. Helping behaviour, reproductive breeding. In B. D. Bell(convener), Proceed- tive value, and the future component of indirect ings of the 20th International Ornithological Congress, fitness. Anim. Behav. , 38:33 1-343. pp. 1325-1332. Ornithological Trust Board, Orians, G. H. 1969. On the evolution of mating Wellington. systems in birds and mammals. Am. Nut., 103: Lack, D. 1968. Ecological Adaptations for Breeding in 589-603. Birds. Methuen, London. Orians, G. H., C. E. Orians, and K. J. Orians. Ligon, J. D. 1981. Demographic patterns and 1977. Helpers at the nest in some Argentine communal breeding in the green woodhoopoe blackbirds. In B. Stonehouse and C. M. Per- (Phoeniculuspurpureus).In R. D. Alexander and rins (eds.), Evolutionary Ecology, pp. 137-151. D. W. Tinkle (eds.), Natural Selection and Social Macmillan Press, London. Behavior: Recent Research and New Theory, pp. Packer, C., L. Herbst, A. E. Pusey, J. D. Bygott, 231-243. Chiron Press, New York. J. P. Hanby, S. J. Cairns, and M. B. Borger- Ligon, J. D., and S. H. Ligon. 1978. Communal hoff Mulder. 1988. Reproductive success of breeding in green woodhoopoes as a case for lions. In T. H. Clutton-Brock (ed.), Reproduc- reciprocity. Nature, 280: 174. tiveSuccess, pp. 363-383. The University ofchi- , and -. 1979. The communal social cago Press, Chicago. system of the green woodhoopoe in Kenya. Liu- Powell, R. A. 1989. Effects of resource productiv- ing Bird, 17:159-197. ity, patchiness and predictability on mating , and -. 1988. Territory quality: key and dispersal strategies. In V. Standen and determinant of fitness in the group-living green R. A. Foley (eds .), Comparative Socioecology, pp. woodhoopoe. In C. N. Slobodchikoff(ed.), The 101-123. Blackwell, Oxford. Ecology of Social Behavior, pp. 229-253. Aca- Pruett-Jones, S. G, and M. J. Lewis. 1990. Sex demic Press, San Diego. ratio and habitat limitation promote delayed , and -. 1990. Green woodhoopoes: dispersal in superb fairy-wrens. Nature, 348: life history traits and sociality. In P. B. Stacey 541-542. and W. D. Koenig (eds.), Cooperative Breeding in Rabenold, K. N. 1984. Cooperative enhancement Birds: Long-Term Studies of Ecology and Behavior, of reproductive success in tropical wren socie- pp. 31-65. Cambridge University Press, Cam- ties. Ecology, 65:871-885. bridge. -. 1990. Campylorhynchus wrens: the ecology Ligon, J. D., and P. B. Stacey. 1989. On the sig- of delayed dispersal and cooperation in the nificance ofhelping behavior in birds. Auk, 106: Venezuelan savanna. In P. B. Stacey and 700-705. W. D. Koenig (eds.), Cooperative Breeding in ,and- . 1991. The origin and mainte- Birds: Long- Term Studies of Ecology and Behavior, nance of helping behavior in birds. Am. Nut., pp. 157-196. Cambridge University Press, 138:254-258. Cambridge. Ligon, J. D., S. H. Ligon, and H. A. Ford. 1991. Rabenold, P. P., K. N. Rabenold, W. H. Piper, An experimental study of the bases of male phi- J. Haydock, and S. W. Zack. 1990. Shared lopatry in the cooperatively breeding superb paternity revealed by genetic analysis in coop- fairy-wren Malurus cyaneus. Ethology, 87: 134- eratively breeding tropical wrens. Nature, 348: 148. 538-540. Matthysen, E. 1990. Nonbreeding social organi- Richner, H. 1990. Helpers-at-the-nest in carrion zation in Parus. Curr. Onithol., 7:209-249. crows Corvus corone corone. Ibis, 132:105-108. McCallum, D. A,, P. B. Stacey, and J. D. Ligon. Ricklefs, R. E. 1975. The evolution ofcooperative In press. Should the mountain chickadee prac- breeding in birds. Ibis, 117:53 1-534. tice natal philopatry? Behav. Ecol. Sociobiol. Rowley, I. 1965. The life history of the superb blue McGowan, K. J.,andG. E. Woolfenden. 1989. A wren, Malurus cyaneus. Emu, 64:251-297. sentinel system in the Florida scrub jay. Anim. Rowley, I., E. M. Russell, and M. G. Brooker. Behav., 37: 1000-1006. 1986. Inbreeding: benefits may outweigh costs. Moehlman, P. D. 1986. Ecology of cooperation in Anim. Behav., 34:939-941. canids. In D. I. Rubenstein and R. W. Wrang- Rowley, I., E. Russell, R. B. Payne, and L. L. ham (eds.), Ecological Aspects of Social Evolution, Payne. 1989. Plural breeding in the splendid JUNE 1992 DELAYED DISPERSAL IN COOPERATIVE BREEDERS 149

fairy-wren, Malurus splendens (Aves: Maluri- tion of despotic versus egalitarian societies. dae), a cooperative breeder. Ethology, 83:229- Anim. Behav., 31 :667-682. 247. Veltman, C. J. 1989. Flock, pair and group living Russell, E. M. 1989. Co-operative breeding-a lifestyles without cooperative breeding by Aus- Gondwanan perspective. Emu, 89:61-62. tralian magpies (Gymnorhinu tibicen). Ibis, 131: Searcy, W. A., and K. Yasukawa. 1989. Alterna- 601-608. tive models of territorial polygyny in birds. Am. Verbeek, N. A. M. 1973. The exploitation system Nut., 134:323-343. of the yellow-billed magpie. Uniu. Calif. Publ. Selander, R. K. 1964. Speciation in wrens of the Zool., 99: 1-58. genus Campylorhynchus. Univ. Calif. Publ. Zool., Verbeek, N. A. M., and R. W. Butler. 1981. Co- 74:l-224. operative breeding of the northwestern crow Sherman, P. W. 1988. The levels of analysis. Corvus cawinus. Ibis, 123:183-189. Anim. Behav., 36:616-619. Verner, J., and M. F. Willson. 1966. Tne influ- Sibly, R. M. 1983. Optimal group size is unstable. ence of habitats on mating systems of North Anim. Behau., 31 :947-948. American passerine birds. Ecology, 47:143- Skutch, A. F. 1935. Helpers at the nest. Auk, 52: 147. 257-273. Walters, J. R. 1990. Red-cockaded Woodpeckers: . 1961. Helpers among birds. Condor, 63: a 'primitive' cooperative breeder. In P. B. Sta- 198-226. cey and W. D. Koenig (eds.), Cooperative Breed- Smith, J. N. M., and P. Arcese. 1989. How fit are ing in Birds: Long- Tern Studies of Ecology and Be- floaters? Consequences of alternative territorial havior, pp. 67-101. Cambridge University behaviors in a nonmigratory sparrow. Am. Press, Cambridge. Nut., 133:830-845. Walters, J. R., C. K. Copeyon, and J. H. Carter Smith, S. M. 1978. The "underworld" in a territo- 111. In press. A test of the ecological basis of rial sparrow: adaptive strategy for floaters. Am. cooperative breeding in red-cockaded wood- Nut., 112:571-582. peckers. Auk. . 1984. Flock switching in chickadees: why Walters, J. R., P. D. Doerr, and J. H. Carter 111. be a winter floater? Am. Nut., 123:81-98. 1992. Delayed dispersal and reproduction as a Stacey, P. B. 1979a. Habitat saturation and com- life-history tactic in cooperative breeders: fit- munal breeding in the acorn woodpecker. ness calculations from red-cockaded woodpeck- Anim. Behau., 27:1153-1166. ers. Am. Nut., 139:623-643. . 197913. Kinship, promiscuity, and com- Waser, P. M. 1988. Resources, philopatry, and munal breeding in the acorn woodpecker. Be- social interactions among mammals. In C. N. hav. Ecol. Sociobiol., 6:53-66. Slobodchikoff (ed.), The Ecology ojSocial Behau- . 1982. Female promiscuity and male re- ior, pp. 109-130. Academic Press, New York. productive success in social birds and mam- Waser, P. M., and W. T.Jones. 1983. Natalphilo- mals. Am. Nut., 120:51-64. patry among solitary mammals. Q. Rev. Biol., Stacey, P. B., and W. D. Koenig (eds.). 1990. 58:355-390. Cooperative Breeding in Birds: Long- Term Studies Watson, A. 1985. Social class, socially-induced of Ecology and Behavior. Cambridge University loss, recruitment and breeding of red grouse. Press, Cambridge. Oecologia, 67:493-498. Stacey, P. B., and J. D. Ligon. 1987. Territory Westneat, D. F., P. W. Sherman, and M. L. Mor- quality and dispersal options in the acorn wood- ton. 1990. The ecology and evolution of extra- pecker, and a challenge to the habitat satura- pair copulations in birds. Curr. Ornithol., 7:331- tion model of cooperative breeding. Am. Nut., 369. 130:654-676. Wiley, R. H., and K. N. Rabenold. 1984. The , and -. 1991. The benefits of philo- evolution of cooperative breeding by delayed patry hypothesis for the evolution of coopera- reciprocity and queuing for favorable social po- tive breeding: variance in territory quality and sition. Evolution, 38:609-621. group size effects. Am. Nut., 137:831-846. Williams, G. C. 1966. Adaptation and Natural Selec- Stallcup, J. A., and G. E. Woolfenden. 1978. tion. Princeton University Press, Princeton. Family status and contributions to breeding by Wolgast, L. J., and B. B. Stout. 1977. Effects of Florida scrub jays. Anim. Behav., 26: 1144-1 156. age, stand density, and fertilizer application on Taborsky, M., and D. Limberger. 1981. Helpers bear oak reproduction. J. Wildl. Manage., 41: in fish. Behav. Ecol. Sociobiol., 8: 143-1 45. 685-691. Tinbergen, N. 1963. On aims and methods of Woolfenden, G. E., and J. W. Fitzpatrick. 1978. ethology. Z. Tierpsychol., 20:410-433. The inheritance of territory in group-breeding Vehrencamp, S. L. 1983. A model for the evolu- birds. BioScience, 28: 104-108. 150 THE QUARTERLY REVIEW OF BIOLOGY VOLUME67

, and -. 1984. The Florida Scrub Jay: short-distance dispersal in cooperatively breed- Demography of a Cooperative-BreedingBird. Prince- ing birds. Ethology, 86:265-286. ton University Press, Princeton. Zack, S., and J. D. Ligon. 1985a. Cooperative ,and -. 1986. Sexual asymmetries in breeding in Lanius shrikes. I. Habitat and de- the life history of the Florida scrub jay. In m~graphyof two sympatric species. Auk, 102: D. I. Rubenstein and R. W. Wrangham (eds.), 754-765. Ecological Aspects of Social Evolution, pp. 87- 107. ,and- . 1985b. Cooperative breeding Princeton University Press, Princeton. in Lanius shrikes. 11. Maintenance of group- , and -. 1991. Florida scrub jay ecol- living in a nonsaturated habitat. Auk, 102:766- ogy and conservation. In C. M. Perrins, J.-D. 773. Lebreton, and G. J. M. Hirons (eds.), Bird Zack, S., and K. N. Rabenold. 1989. Assessment, Population Studies: Relevance to Conservation and age and proximity in dispersal contests among Management, pp. 542-565. Oxford University cooperative wrens: field experiments. Anim. Be- Press, Oxford. hav., 38:235-247. Zack, S. 1990. Coupling delayed breeding with