- Email: [email protected]
Development of a linear dominance hierarchy in nestling birds
ANIMAL BEHAVIOUR, 2007, 74, 1705e1714 doi:10.1016/j.anbehav.2007.02.034
Available online at www.sciencedirect.com
Development of a linear dominance hierarchy in nestling birds ´ BANO-IBARRA, IRERI B RUMON & HUGH DRUMMOND CLA UDIA VALDER RA
Instituto de Ecologı´a, Universidad Nacional Auto´noma de Me´xico (Received 5 June 2006; initial acceptance 9 August 2006; final acceptance 14 February 2007; published online 23 October 2007; MS. number: A10466R2)
Theoreticians propose that trained winning and losing are important processes in creating linear animal dominance hierarchies, and experiments have shown that both processes can occur in animals, but their actual roles in creating natural hierarchies are unknown. We described agonism in 18 broods of three bluefooted boobies, Sula nebouxii, a species for which trained winning and losing have been demonstrated, to infer how these processes generate and maintain a natural hierarchy. Ranks in the linear hierarchy that emerged in every brood were initially assigned by asymmetries in age, size and maturity, which led to differences between broodmates in levels of expressed and received aggression and, consequently, to differences in the training of their aggressiveness and submissiveness. Later, ranks appeared to be maintained by the chicks’ acquired aggressive and submissive tendencies combined with ongoing effects of persisting differences in size and maturity. Our results suggest that trained winning and trained losing are important in the construction of booby hierarchies but that these two axes of learning are largely independent. Increase in submissiveness occurs over a period of about 10e20 days, and the level of submissiveness reached varies with the amount of aggression received. After training, submissiveness is apparently maintained by a lower level of aggression and increasing use of threats. Threats become increasingly effective as chicks age, but are never as effective as attacks. Ó 2007 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Keywords: agonistic behaviour; blue-footed booby; dominance; hatch asynchrony; hierarchy; Sula nebouxii; trained winning
Theoreticians have asked which behavioural mechanisms can in principle efficiently generate the linearity that characterizes many animal hierachies (e.g. Chase 1974; Beaugrand 1997; Broom 2002). Candidate mechanisms include trained winning and losing (Hsu et al. 2006), the learned dyadic behavioural dispositions of true dominance (Bernstein 1981), assessment of fighting ability, and the bystander effect (an individual assesses others by watching them fight each other; Chase 1982a, b; Dugatkin 2001; Chase et al. 2002), and combinations of these mechanisms are expected (e.g. Mesterton-Gibbons & Dugatkin 1995; Beaugrand 1997; Pagel & Dawkins 1997). Mathematical modelling prevails in this field and experimental tests are sometimes made, but testing is done in artificial situations whose relevance to naturally occurring dominance hierarchies in the study species is usually unknown. We lack quantitative descriptions of
Correspondence: H. Drummond, Instituto de Ecologı´a, Universidad Nacional Auto´ noma de Me´ xico, A.P. 70-275, 04510 D.F., Me´xico (email: [email protected]). 0003e 3472/07/$30.00/0
the emergence of natural animal dominance hierarchies, and further progress in developing and testing ideas about the formation of dominance relationships and the development and maintenance of dominance hierarchies may depend critically on getting to grips with what actually occurs in nature. Trained winning and trained losing are the two learning mechanisms most widely expected to contribute to hierarchy formation and maintenance (Barnard & Burk 1979; Pagel & Dawkins 1997; Beacham 2003). Although their existence has been demonstrated in experimental tests of diverse species of vertebrates, their actual functioning in the construction of natural dominance relationships and hierarchies is still obscure (Dugatkin 1997). Each individual’s history of victories and defeats against its competitors is generally expected to condition it to a particular position on an aggressiveesubmissive continuum (e.g. Theraulaz et al. 1989), and the relative positions of all group members on the continuum could determine the hierarchy among them. However, assessment is also likely to be important, and, in theory, the nature of a hierarchy can depend on trained winning or trained
1705 Ó 2007 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
1706
ANIMAL BEHAVIOUR, 74, 6
losing, or both, and this is seldom known (Dugatkin 1997). Nor do we have a grasp of the schedules of interactions whereby members of natural groups entrain each other’s and their own agonistic responsiveness: in particular, how long training takes and how agonism changes when it has been achieved (Bonabeau et al. 1996). An amenable subset of animal dominance hierarchies includes those that occur in altricial broods of some avian species (reviews in Mock & Parker 1997; Drummond 2001, 2006), and these can be observed from the start of their agonism. As far as we know, these infant hierarchies do not involve such complications as alliance formation or even individual recognition and they may be among the most simple hierarchies that occur in natural groups of vertebrate conspecifics. Typically, two to four nestlings hatch at intervals of 1e5 days and start attacking each other within several days. The nature of dyadic dominance relationships varies among and within species (Drummond 2006), but nestlings usually sort themselves over a period of several days or weeks into a linear hierarchy that follows the order in which they hatched and confers growth and survival benefits on high-ranking individuals. Our knowledge of the development of these hierarchies is sketchy because broods are rarely observed from the onset of hostilities, the temporal resolution of data is often poor, responses to aggression are seldom quantified and few studies have actually analysed age-related changes in aggression or submission (but see Drummond et al. 1986; Cook et al. 2000; Nathan et al. 2001). Dominance relationships and hierarchies occur in broods of the blue-footed booby, Sula nebouxii, a marine bird that produces broods of one to three altricial chicks that fledge at about 3 months of age (Nelson 1978). Staggered laying and hatching ensures that broodmates differ substantially in age: in three-chick broods, the A-chick is, on average, 4.0 days older than the B-chick, which is 3.6 days older than the C-chick (Castillo Alvarez & ChavezPeo´n 1983). Broods of three are largely unstudied, but in broods of two, there is always a dominance relationship, usually with the A-chick assuming a dominant role characterized by daily attacking and threatening, and the B-chick assuming a subordinate role characterized by minimal aggression and submissive responses to aggression (Nelson 1978; Drummond et al. 1986; Anderson & Ricklefs 1995). A-chicks grow faster than B-chicks during the first few weeks of life and are more likely to fledge, but surviving B-chicks catch up on growth and, at fledging, they are just as large as A-chicks. Experimental pairings of different combinations of unfamiliar dominant chicks, subordinate chicks and singletons (chicks with no broodmate) that were 12e55 days old demonstrated that trained winning, trained losing and assessment are all involved in the agonism of blue-footed boobies. Thus, a chick’s aggressiveness or submissiveness to an unfamiliar chick is influenced by prompt assessment of its relative size, but more importantly, by its own agonistic training in its home brood: previously dominant broodmates tend to behave aggressively and nonsubmissively, previously subordinate broodmates tend to behave submissively and nonaggressively (Drummond & Osorno 1992; Drummond & Canales 1998). The effects of trained
winning and losing also seem to include modification not only of agonistic tendencies but also of fighting ability: A-chicks dominate unfamiliar B-chicks that are 32% heavier than themselves, even though most B-chicks respond to their newfound size advantage with increased aggressiveness. Hence, the observed stability of dominance relationships over the nestling period (Drummond et al. 1991) could largely be due to trained winning and losing. We studied the emergence of dominance hierarchies in natural three-chick broods of the blue-footed booby. Broods of three are always a minority and often are reduced by nestling mortality to two chicks, but sometimes all three broodmates survive and cohabit through to fledging. In our study population, 217 three-chick broods fledged 75% of A-chicks, 69% of B-chicks and 31% of C-chicks, and, in 20% of broods, all three broodmates fledged (H. Drummond, unpublished data from 24 seasons). We sought evidence for how the mechanisms of trained winning and trained losing operate in a brood of three to create and maintain a linear hierarchy. We compared the development of behaviour in A-, B- and C-chicks to infer their schedules of training and how they come to occupy their dominance ranks. Examination of their behavioural development also yielded insight into (1) whether aggressiveness and submissiveness are two sides of the same coin or independent axes of behavioural tendency, (2) whether B-chicks acquire intermediate rank by simultaneously learning increased submissiveness to their elder broodmate and increased aggressiveness to their younger broodmate and (3) whether aggressors progressively substitute threats for attacks as victims learn to submit to threats.
METHODS We located two-chick and three-chick broods on Isla Isabel, Nayarit, Mexico (21 520 N, 105 540 W) in March and April of 2002 and 2004 by monitoring all nests in our study area every 3 days, starting shortly after the start of hatching in the colony. When hatching date was unknown, we estimated chick age using culmen and ulna growth curves of the same population. We banded broodmates according to their age ranks (Drummond et al. 1991). When broods were first found, there was no risk of misidentifying age ranks within a brood, given this booby’s large hatching intervals. Blue-footed boobies lay clutches of one to three eggs. Clutches of three eggs are uncommon, but a minority of two-chick broods arise through hatching failure in three-egg clutches. Sampling of three-chick broods and two-chick broods was constrained by availability and mortality and could not be equitable. Three-chick broods were sampled in 2002 by observing all intact broods of that size more or less daily until the first chick death occurred (all three-chick broods suffered mortality in 2002). We suspended observation and switched to a new three-chick brood whenever switching would assure more even sampling of the first 6 weeks of life. Eighteen three-chick broods were observed for a mean SE of 10.2 1.9 days, from A-chick ages of 10e23 days (X ¼ 16:2 days) to 13e49 days (X ¼ 27:3 days); in these
VALDERR ABANO-IBARRA ET AL.: DEVELOPMENT OF A HIERARCHY
broods, B-chicks and C-chicks were on average 3.2 days and 7.4 days younger than A-chicks. We included two samples of two-chick broods in 2004 to help discern the mechanisms operating in three-chick broods by documenting the development of dominance relationships in the simplest social context. The long-term sample consisted of 13 broods grouped in neighbourhoods where we could observe up to three broods simultaneously. These were observed roughly every other day for a mean SE of 18.6 1.1 days per brood, starting when A-chicks were 6 days old (X ¼ 13:2 days) and ending when they were 45 days old. Only one chick died in a two-chick brood before this age. The average age difference between broodmates in the long-term sample of two-chicks broods was 3.3 days. The short-term sample consisted of 70 broods that were observed (partly to screen broods for another study) during just 2.4 0.3 consecutive days when chicks were 2e45 days old. Two-chick broods of the long-term sample were used to analyse the relationship between a chick’s submissiveness and the aggression it received, because minimal mortality in these broods allowed more even sampling, and also to analyse the use of threats, which were not recorded in three-chick broods (although they certainly occurred). Although these differences in behavioural sampling limited our ability to contrast the two brood sizes (and precluded using the same statistical procedures to analyse them), the combined data set allowed us to examine common patterns and to infer common processes. Red, blue or yellow paint was applied to the head and rump of each brood member (assigned haphazardly) on the first day of observation and reapplied on subsequent days as needed to ensure age ranks were recognized. Chicks were weighed and measured (culmen and ulna) every day at the end of observations. An observer, seated roughly 5 m from the nest during 6 h each day (0700e 1000 hours and 1500e1800 hours), recorded the absolute frequencies of attacks (pecks and bites) by all brood members, and whether the victim responded to each attack by adopting or maintaining a bill-down-and-face-away submissive posture (BDFA, Nelson 1978), using the behavioural criteria in Drummond et al. (2003). For two-chick broods, the observer also recorded the absolute frequencies of vocal threats and submissive responses to them. An increase or decrease in an individual’s attacking over time could reflect a change in its aggressiveness due to trained winning, but it could also be due to the individual’s maturation or level of access to the victim. An increase in submissive responses to aggression is evidence of increased submissiveness due to trained losing, but it could also be a function of maturation. Unless stated otherwise, comparisons of broodmates use data from each broodmate at the same age (calculated from each chick’s own hatch date), since our purpose was to compare their development. We calculated each chick’s attack rate for each successive 5-day age block by dividing its total attacks in the block by the total number of hours it was observed in the block. Similarly, each chick’s submissiveness in each block was the number of attacks it responded to with a submissive posture divided by the number of attacks it received in the block. Where zero
scores of individual chicks prevented us from calculating the percentage difference between scores of dyadmates for all dyads in a sample, we report the percentage difference between the two sample mean values (e.g. between the mean score for A-chicks and the mean score for B-chicks). For dyads that were observed only for 1e3 days, we inferred inversion of dominance if the younger chick attacked more frequently and was less submissive than the elder chick. For dyads observed on more than 3 days, we inferred inversion of dominance if the younger chick attacked more frequently overall, had a superior attacking score on more than half of observation days, was less submissive overall and showed submission on fewer days relative to the elder chick. We considered that dominance in a dyad was contested when the subordinate chick attacked more than half as frequently as the dominant chick. We used nonparametric tests because distributions were not normal and most data sets could not be normalized by transformation. For two-chick broods, we pooled the longterm and short-term samples, after graphing the data and confirming visually that they showed similar patterns. Tests for dependent samples were preferred except when sample sizes were inadequate and larger samples could be analysed by including independent data, which was often the case because broodmates were frequently observed in different age blocks (since they differed in age). When a subset of individuals was observed in more than one age block, using a test for independent samples rather than a test for dependent samples was conservative, provided that booby chicks show substantial and stable individual differences. In any event, only two tests for independent samples that included some individuals observed at more than one age yielded a significant probability greater than 0.01 (see Results; Fig. 2). We tested for age-related changes in behaviour in two-chick broods using comparisons among all age blocks for which sample sizes were at least N ¼ 8, and in three-chick broods by comparing the three successive age blocks with the largest samples. Wherever the sample of chicks used for a statistical comparison was a subset of the sample shown in the bar of a histogram, the magnitude of the difference reported may differ from the magnitude evident in the figure. Throughout results we report X SE. RESULTS
Three-chick Broods Mortality in the 18 broods was substantial during the first 6 weeks of life and increased with hatch order: A-, B- and C-chicks, respectively, died in 6%, 16% and 100% of broods (G test: G2 ¼ 49.05, P < 0.001), at ages of 30 days, 21e29 days and 5e42 days. Over this period, growth rates increased with age/dominance rank and C-chicks grew particularly poorly. Comparing broodmates at the same age of 15e19 days (when our sample sizes were largest), A-chicks were on average 21% heavier than B-chicks (Wilcoxon signed-ranks test: T ¼ 3.0, N ¼ 12, P ¼ 0.004) and B-chicks were on average 71% heavier than C-chicks (T ¼ 1.0, N ¼ 11, P ¼ 0.004).
1707
1708
ANIMAL BEHAVIOUR, 74, 6
Linear dominance hierarchy In all three dyads of every three-chick brood, both chicks showed some aggression, and the dominance hierarchy of every brood was linear: A-chicks attacked B-chicks and C-chicks 4e15 and 21e95 times more often, respectively, than vice versa, and B-chicks attacked C-chicks 6e28 times more often than vice versa (Fig. 1). In every dyad, the elder chick attacked at a higher rate than the younger chick at all ages (different-age comparisons; P < 0.02 in all nine Wilcoxon tests).
Development of aggression and submissiveness A-chicks and B-chicks were already attacking at the earliest ages they were observed (Fig. 2). Attacks by the two most senior chicks towards their broodmates (summed, N ¼ 8) increased with age, steeply in the case of A-chicks (over 15e29 days; Friedman test: c22 ¼ 6:250, P ¼ 0.043) and modestly in the case of B-chicks (over 10e24 days, KruskaleWallis test: H2 ¼ 5.86, P ¼ 0.05). Attacking by C-chicks also changed with age (5e19 days, H2 ¼ 5.98, P ¼ 0.05), but C-chicks started late and increases in their attack rates were minimal and short-lived (Fig. 2). The lower a chick’s age rank, the earlier its aggression peaked and declined: C-chicks peaked at roughly age 10e14 days, B-chicks at roughly 20e24 days and A-chicks not until roughly 30e34 days (Fig. 2). Moreover, maximum pecking by A-chicks exceeded that of B-chicks (Manne Whitney U test: U ¼ 7, N1 ¼ 5, N2 ¼ 10, P < 0.05) and maximum pecking by B-chicks exceeded that of C-chicks (U ¼ 16, N1 ¼ 10, N2 ¼ 13, P < 0.002). In both A-chicks and B-chicks, the decline in attacking appeared steep and permanent, but note that our samples embraced less than half of the prefledging period. A 15–19 days N = 13 broods B
C
All age ranks appeared to start out responding to only 10e25% of attacks with a submissive posture, then all age ranks progressively increased their submissiveness (Fig. 3). Between the ages of 10 and 24 days, submissiveness of B-chicks to attacks by A-chicks increased to over 80% (KruskaleWallis test: H2 ¼ 15.39, P < 0.001) and submissiveness of C-chicks to each of their broodmates rose to similar values (to A-chicks: H2 ¼ 11.63, P ¼ 0.003; to B-chicks: H2 ¼ 8.57, P ¼ 0.01). Surprisingly, between the ages of 15 and 29 days, A-chicks also showed an increase (H2 ¼ 6.32, P ¼ 0.04; Fig. 3), eventually submitting to more than 40% of aggressions by B-chicks. The lesser submissiveness of A-chicks was confirmed by two tests made at the only ages for which samples were sufficient: at age 15e24 days, A-chicks were four to five times less submissive to B-chicks than similarly aged B-chicks were to A-chicks (Wilcoxon signed-ranks test: 15e19 days: T ¼ 2, N ¼ 8, P ¼ 0.04; 20e24 days: T ¼ 0.0, N ¼ 10, P ¼ 0.005; Fig. 3), and at age 15e19 days, A-chicks were eight times less submissive to B-chicks than were similarly aged C-chicks (T ¼ 0.0, N ¼ 10, P ¼ 0.007; Fig. 3). Attacks on A-chicks began later (ca. age 15 days) than those on Bchicks and C-chicks (ca. age 5 days; Fig. 2) and petered out by the time A-chicks were 4 weeks old (Fig. 2), the age at which attacks on B-chicks and C-chicks were maximal (Fig. 1). Importantly, attacks on A-chicks were delivered by broodmates smaller and weaker than themselves. There were indications that the aggressiveness of elder chicks declined after the submissiveness of their younger broodmates reached a high value. Maximum average submissiveness of B-chicks was achieved when they were 20e24 days old (Fig. 3), and 5e10 days later there was an approximate halving of violent aggression by A-chicks (after A-chicks reached 35 days; Fig. 2). Four broods were observed on sufficient successive days to determine whether the decline in aggressiveness of individual A-chicks occurred after submissiveness of their own B-broodmates reached a high value. Submissiveness by B-chicks reached or exceeded 90% in all four broods, and in three of those four broods a drop of 50% or more in aggression of A-chicks occurred afterwards. The drop came 5e10 days after 90% submissiveness was first achieved, when dominants were 30e40 days old.
A
Training of the intermediate chick
20–24 days N = 14 broods C
B
A 25–29 days N = 9 broods B
C
Figure 1. Development of a linear hierarchy in asynchronously hatched blue-footed booby chicks. Shown are attacks by each broodmate (A-, B- and C-chick) when A-chicks were at the ages shown and B- and C-chicks were on average 3.2 and 7.4 days younger, respectively. Thickness of each arrow represents the mean rate of attacks up to a maximum value of 12.18 attacks/h.
B-chicks behaved differently towards their two broodmates, delivering more attacks to C-chicks and showing greater submissiveness to A-chicks. Over the 10-day age span for which comparisons could be made, B-chicks attacked C-chicks two to five times more often than they attacked A-chicks (Wilcoxon test: 15e19 days: T ¼ 1.0, N ¼ 14, P ¼ 0.01; 20e24 days: T ¼ 1.0, N ¼ 10, P ¼ 0.017; Fig. 2), and B-chicks were four times as submissive to A-chicks as they were to C-chicks (ManneWhitney U test: 15e19 days: U ¼ 9.5, N1 ¼ 13, N2 ¼ 7, P ¼ 0.04; 20e24 days: U ¼ 6.0, N1 ¼ 10, N2 ¼ 7, P ¼ 0.004; Fig. 3).
Two-chick Broods There was uncontested dominance in nine of 13 broods in the long-term sample and 66 of 70 broods in
VALDERR ABANO-IBARRA ET AL.: DEVELOPMENT OF A HIERARCHY
Three-chick broods and C
By A on B 20
15 Two-chick broods 10
By dominant 10
0
5
5
10
15
20
25
30
35
40
45
Age (days) of dominant
5
0
Attacks/h
Attacks/h
5
By subordinate
5
10 15 20 25 30 35 40 Age (days) of subordinate
0 5
15
20
25
30
35
40
45
35
40
45
35
40
45
Age (days) of A
0
5
0
and C
By B on A
5
45
10
5
10
15
By C on A
5
10
20 25 30 Age (days) of B and B
15
20
25
30
Age (days) of C Figure 2. Attack rates (X þ SE) during hierarchy development in two- and three-chick broods of blue-footed boobies. Sample sizes in successive blocks: two-chick broods: dominant chicks: 11, 21, 20, 35, 38, 23, 18, 16; subordinate chicks: 21, 22, 28, 35, 33, 20, 17, 13; three-chick broods: A-chicks: 0, 6, 13, 14, 9, 5, 4, 4; B-chicks: 5, 10, 14, 10, 6, 4, 4, 0; C-chicks: 8, 13, 11, 6, 4, 4, 0, 0.
the short-term sample, and this uncontested dominance was inverted in two of nine long-term broods and three of 66 short-term broods. The proportion of broods with inverted or contested dominance in our study (16% overall) was unusually high compared to proportions in earlier studies (9% and 4% of two-chick broods observed >60 days by Drummond et al. 1986, 1991), but analyses carried out after deleting broods with inverted or contested dominance (not shown) or sorting broodmates by age rank rather than dominance rank yielded similar results, so the analyses we report may be considered representative.
Development of aggression and submissiveness Both chicks started attacking in their respective first age blocks. Aggression of dominants changed over the eight age blocks (KruskaleWallis test: H7 ¼ 23.53, P ¼ 0.001), showing a steady increase to a peak at 20e24 days
followed by decline (Fig. 2). The aggression of subordinates followed a similar pattern but the apparent peak came earlier, at 15e19 days, and change over time was not significant (H7 ¼ 8.59, P ¼ 0.283; Fig. 2). The submissiveness of subordinates varied with age (KruskaleWallis test: H7 ¼ 47.9, P < 0.001), rising from 23% at age 7 days to 87% at 15e19 days (Fig. 3) and remaining at that high level until at least age 44 days. Submissiveness of dominants in the sample increased from 20% at age 12 days to a maximum of 55% at 30e34 days, but change with age was not significant (H6 ¼ 6.99, P ¼ 0.322; Fig. 3). However, subordinate chicks were 60% more submissive than dominant chicks at their respective maximums (ManneWhitney U test: U ¼ 97.5, N1 ¼ 27, N2 ¼ 13, P ¼ 0.023; Fig. 3). In the long-term sample, there was a strong relationship between the rate of violent aggression received by a chick during the first 6 weeks of life and its average submissiveness over the same period (Fig. 4). The values for dominants
1709
ANIMAL BEHAVIOUR, 74, 6
Two-chick broods
Three-chick broods A to B
Dominant chick 1
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
5
10
15
20
25
30
35
40
0
45
5
Subordinate chick 1
0.8
0.8
0.6 0.4 0.2 5
10
15 20 25 30 35 Age (days) of subordinate
15
20
25
30
10 15 20 25 Age (days) of B
30
B to A
1
0
10
Age (days) of A
Age (days) of dominant
Submissiveness
Submissiveness
1710
40
45
0.6 0.4 0.2 0
5
C to A
or B
1 0.8 0.6 0.4 0.2 0
5
10
15
20
25
30
Age (days) of C Figure 3. Development of submissiveness in blue-footed booby chicks (proportion of attacks eliciting submission, X þ SE). Submissiveness of the two eldest chicks to C-chicks is not shown because attacks by C-chicks were too few. Sample sizes in successive blocks: two-chick broods: dominant chicks: 0, 11, 17, 22, 18, 13, 11, 10; subordinate chicks: 18, 21, 27, 34, 33, 18, 15, 12; three-chick broods: A-chicks: 0, 4, 10, 12, 7; B-chicks: 4, 8, 13, 10, 6; C-chicks: 5e5, 11e11, 8e10, 6e5, 4e4.
and subordinates appeared to fit on the same logistic curve, which explained 74% of the variation. The correlation was strong for the whole sample of 25 dominant and subordinate chicks (Pearson’s correlation: r23 ¼ 0.64, P ¼ 0.001) and also for the 12 dominant chicks alone (r10 ¼ 0.69, P ¼ 0.014). For the 13 subordinate chicks alone, the correlation was also positive but not significant (r11 ¼ 0.41, P ¼ 0.16; Fig. 3), probably because they were sampled on only a limited portion of the curve (all of them received high levels of violence). Importantly, in broods where dominants were subjected to high levels of pecking, they were highly submissive despite managing to outpeck their broodmates (Fig. 3).
In the long-term sample, the decline in aggressiveness of elder chicks started after submissiveness of their younger broodmates reached a high value. Maximum average submissiveness was reached when subordinates were 15e20 days old (Fig. 3), and was followed in the next 5-day block by an approximate halving of violent aggression by dominants (after dominants reached 25 days; Fig. 2). In individual broods, the interval between full submissiveness and the decline in aggression was variable: submissiveness by subordinates exceeded 90% in 12 of 13 broods, and in nine of those 12 broods, a drop of 50% or more in aggression by the dominant chick occurred afterwards and appeared to be sustained subsequently.
Proportion of threats (X ± SE)
VALDERR ABANO-IBARRA ET AL.: DEVELOPMENT OF A HIERARCHY
1
Submissiveness
0.8 0.6 0.4 0.2 0
0
1
2
3
4
5
6
7
Attacks received/h Figure 4. Relationship between rate of attacks received during ages 15e44 days and average submissiveness shown at the same ages by 12 dominant chicks (C) and 13 subordinate chicks (B) in two-chick broods (logistic curve, R2 ¼ 73.58) of blue-footed boobies.
The drop came 5e20 days after 90% submissiveness was first achieved, when dominants were 20e39 days old.
Threats Threatening by dominant chicks started at age 5e10 days and increased progressively to a peak at age 20e34 days (KruskaleWallis test: H7 ¼ 64.7, P < 0.001), and the submissiveness of subordinates to those threats ascended during the same period from 21% to a maximum of 57% when subordinates were 25e29 days old (H7 ¼ 18.28, P ¼ 0.011). However, at no age did subordinates submit to threats as frequently as they did to physical attacks, and their maximal submissiveness to threats was 34% less than their maximal submissiveness to physical attacks (ManneWhitney U test: U ¼ 185.5, N1 ¼ 27, N2 ¼ 33, P < 0.001). The portion of aggressions by dominants that were threats rather than physical attacks rose progressively from 3% at age 5e10 days to a maximum of 62% at age 40e44 days (KruskaleWallis test: H6 ¼ 27.43, P < 0.001; Fig. 5), when our observations ended. However, we did not find evidence for dominants increasingly switching to threats as the efficacy of threats increased. When we plotted each brood’s mean scores for the proportion of dominants’ aggressions that were threats and the proportion of subordinates’ responses to threats that were submissive in each age block, Spearman’s correlation coefficients (not shown) were positive for five of 13 broods and negative for the remainder. The threat behaviour of subordinates developed similarly and with similar effects, despite subordinates showing less aggression and eliciting less submission in general. Subordinate threats peaked at 15e19 days (KruskaleWallis test: H7 ¼ 42.17, P < 0.001) and elicited less submissiveness than threats by dominants (comparison of maximum values; ManneWhitney U test: U ¼ 99.0, N1 ¼ 33, N2 ¼ 17, P < 0.001). The proportion of aggressions by subordinates that were threats rose progressively to a maximum of 81% at age 30e34 days (H7 ¼ 26.84, P < 0.001; Fig. 5), higher than the maximum for dominants (U ¼ 73, N1 ¼ 17, N2 ¼ 15, P ¼ 0.04).
1 0.8 0.6 0.4 0.2 0
5
10
15
20 25 30 Chick age (days)
35
40
45
Figure 5. Proportions of total aggressions by dominant (-) and subordinate (,) blue-footed booby chicks that were threats. Sample sizes in successive blocks: dominant chicks: 5, 20, 20, 35, 38, 23, 18, 15; subordinate chicks: 10, 19, 25, 27, 28, 17, 17, 10.
DISCUSSION Linear hierarchies emerged in all 18 three-chick broods, with all three dyads in every brood showing a clear relationship of dominanceesubordination favouring the elder chick. Comparisons of these profiles, in the light of previous experimental confirmation of the existence of trained winning, trained losing and assessment in chicks of age 12e55 days (Drummond & Osorno 1992; Drummond & Canales 1998), allows preliminary inferences about the learning mechanisms and developmental processes responsible for the changes. Of course, these are descriptive data and alternative interpretations are possible. A key inference is that, in boobies, aggressiveness and submissiveness may be two largely independent axes of behavioural tendency and that trained winning and trained losing may be two largely distinct processes that enhance those tendencies. The evidence for this is that the most dominant chicks of both brood sizes showed increasing submissiveness at the same ages that they were showing increasing aggression. This inference is consistent with Leshner’s (1978) suggestion that aggressive and submissive behaviours are often under the proximate control of different hormones. In what follows, we suggest that the distinct agonistic profiles of the two or three chicks in a brood are mainly an outcome of differences in (1) the training of each chick’s aggressiveness and submissiveness and (2) the social environments that chicks inhabit at any age. The training differences are set in motion by initial age/size differences (cf. Chase 1986; Slater 1986; Jackson 1988).
Development of Aggression Attacking by chicks of all ranks in both brood sizes started at age 5e9 days and increased with age. In the lowest-ranking chicks of each brood size the increase was undoubtedly due to maturation, as the stamina, mobility and wakefulness of each chick increased with age, and as its maturing broodmates became increasingly available for pecking. However, the far greater age-related increase in pecking by A-chicks over the first 3e4 weeks of life
1711
1712
ANIMAL BEHAVIOUR, 74, 6
presumably was partly due to trained winning, as they accumulated a history of attacks on their two broodmates that drew ever more submission from them. Because attacks by subordinates often provoke retaliation by dominants (Drummond & Osorno 1992), it is also likely that the superior increase in pecking by elder chicks was partly due to their aggressive stifling and punishing of younger broodmates that attacked them: juniors often forwent opportunities to attack when they submitted, cowered or hid in response to attacks by broodmates that were more mature and powerful than they were. The conspicuous decline in attacking by A-chicks after age 25e40 days may be due to appeasement: aggressiveness of dominant chicks may be reduced when their broodmates reach some criterion of submissiveness. Attack rates declined at ages when attacking might be expected to increase with maturation and in response to intensified feeding competition during the period of maximal growth (Drummond et al. 1991; Torres & Drummond 1999). Facultative decline in response to submissiveness is supported by two analyses: (1) maximal submissiveness by B-chicks was reached earlier in twochick broods than in three-chick broods and, in both cases, it closely preceded aggressive decline in A-chicks; (2) in individual dyads of both brood sizes there was evidence for a stepwise drop in agggression by dominants after the submissiveness of subordinate broodmates reached 90%. However, part of the age-related decline in attacking could have been due to broodmates developing the locomotor ability to walk about in the home territory and avoid proximity (Drummond & Garcı´a Chavelas 1989). Vocal threats progressively substituted for pecks and bites despite being less effective at eliciting submission. Increased reliance on threats apparently was not triggered by victims’ enhanced submissiveness: these variables were not correlated in individual broods and subordinates threatened proportionally more than dominants despite eliciting less submission. Chicks may increasingly use threats as they get older because threats function at a distance and broodmates spend more time apart as they wander independently in the family territory. The subordinates’ greater proportional increase in threatening could also result from a preference to show aggression when they are far from their intimidating broodmates.
Development of Submissiveness Chicks of all ranks showed a developmental increase in submissiveness. Maturation could not completely account for the increase because it occurred at different ages and reached different maximum values in different ranks. More likely, boobies are congenitally disposed to respond with minimal submissiveness to the first pecks and bites, and progressive enhancement of submissiveness is mostly dictated by the schedule of aggression received. Submissiveness of A-chicks in three-chick broods was moderate and developed relatively late because A-chicks did not receive pecks until their junior broodmates were old
enough to deliver them (when A-chicks were nearly 15 days old) and because the pecking they received was infrequent and weak, coming from beseiged and smaller chicks. Submissiveness may also increase with maturation, but the tight correlation between attacks received and submissiveness shown during the first 6 weeks of life in twochick broods (Fig. 4) suggests that submissiveness is largely a function of the amount of aggression received, whatever the chick’s dominance rank (cf. Hsu & Wolf 1999). Submissiveness, then, is an index of trained losing. The persistence of enhanced submissiveness of subordinates and dominants in two-chick broods well beyond the age when attacks on them declined suggests that, once trained losing is well established, the level can be maintained by a substantially lower level of aggression in which threats outnumber attacks. In two-chick and three-chick broods, respectively, thorough training of junior chicks in losing (maximal submissiveness) was not achieved until they were 2e3 weeks old, usually after receiving daily pecks and threats over a period of roughly 10e20 days (from the time elder broodmates started pecking).
Acquiring Intermediate Rank B-chicks in three-chick broods showed clear evidence of trained losing during the first 3 weeks of life but only in relation to one of their broodmates. These intermediates progressively became as submissive to their senior broodmates as were B-chicks in broods of two, but considerably less submissive to their junior broodmates, who attacked them much less. Intermediates may discriminate between their two broodmates, probably on the basis of size, and learn to show increased submissiveness only to the larger one. Alternatively, intermediates may seldom have submitted to the smaller broodmate’s attacks simply because they were feeble. It is less clear whether intermediates learned to attack more frequently or to attack broodmates differentially. Their age-related increase in aggression could have been due to maturation, trained winning or a combination of the two, and intermediates could have attacked junior broodmates two to five times more frequently than they attacked senior broodmates either because they were preferred targets or because attacks on senior broodmates were stifled by the latter’s aggression. An experiment is required to determine whether, despite being trained in losing, intermediates are also trained in winning.
Trained Winning and Losing in Hierarchy Emergence If our inferences are correct, then development of the three-chick linear hierachy is set in motion by staggered hatching, which generates the age rank-specific social contexts in which each chick’s submissiveness and aggressiveness develop differently through maturation, learned losing and learned winning. Expression of aggression and submission by chicks of each rank are then further determined by the rank-specific social contexts
VALDERR ABANO-IBARRA ET AL.: DEVELOPMENT OF A HIERARCHY
that emerge, comprising not only the broodmates’ relative sizes but also their acquired agonistic tendencies. Our observations also argue for a minor role for an elementary form of individual recognition: trained losing that selectively attaches to a particular broodmate on the basis, presumably, of its large size. Whether assessment is also important in establishing relationships and hierarchies in the first place is hard to discern because likely effects of assessment, such as more initial aggression by larger dyad members and more submission by smaller dyad members, are confounded with effects of asymmetries in the maturity of dyadmates and in the social contexts they inhabit. Ongoing assessment probably functions to buttress acquired roles and detect changes in the broodmate’s capacities and tendencies that require a response. Booby chicks attain dominance by urgently accumulating attacks on their competitors at a faster rate than the treatment can be returned, then relax their aggression once their competitors are fully trained. By attacking, each chick simultaneously enhances its own aggressiveness (trained winning) and its broodmates’ submissiveness (trained losing), and these, we infer, are two distinct axes of learning. If high age rank allows a chick to attack frequently and powerfully, the broodmate’s submissiveness can be maximized within 10e20 days and it will be subordinate. Although training effects usually work in the same direction as individual differences in size during the first few weeks, when a female chick eventually outgrows her elder brother, he remains dominant over her, showing that training effects are strong enough to prevail over size differences (Drummond et al. 1991; Drummond & Osorno 1992). Rather than ‘ranking fights’ that more or less abruptly create inertia in the agonistic asymmetry of two individual animals (e.g. precocial avian broodmates: Kalas 1977; Rushen 1982; foundress wasps: Pardi 1948; Roseler 1991), there is cumulative shaping of distinct tendencies to attack and submit to other chicks generally. And as training proceeds, threats progressively substitute for violence, probably because they are less costly even though they are never as effective. Intermediates are trained in losing although they may not be trained in winning, and their domination of C-chicks could depend on persistent age-related difference in fighting ability and assessment rather than training.
Acknowledgments Fieldwork was financed by a grant from the UNAM to H.D. (PAPIIT IN2007023) and logistically supported by the ever generous fishermen of Nayarı´t, the Armada de Me´xico and the staff of the Parque Nacional Isla Isabel. Permissions were granted by SEMARNAT. We are grateful to Alejandro ˜a, Cecilia de Ita, Leonora Mila´n, Gonzalez, Romeo Saldan Atenea Lima, Fabricio Villalobos, German Cha´vez, Oliva ˜eda, Carlos Salas, Ana Marı´a Clavijo and Tatjana Castan Benavides for their dedicated and generous help with fieldwork, to Cristina Rodrı´guez for help with fieldwork, data analyses and manuscript preparation, and to Alejandro Gonzalez for comments on the manuscript.
References Anderson, D. J. & Ricklefs, R. E. 1995. Evidence of kin-selected tolerance by nestlings in a siblicidal bird. Behavioral Ecology and Sociobiology, 37, 163e168. Barnard, C. J. & Burk, T. 1979. Dominance hierarchies and the evolution of ‘‘individual recognition’’. Journal of Theoretical Biology, 81, 65e73. Beacham, J. L. 2003. Models of dominance hierarchy formation: effects of prior experience and intrinsic traits. Behaviour, 140, 1275e1303. Beaugrand, J. P. 1997. Relative importance of initial individual differences, agonistic experience and assessment accuracy during hierarchy formation: a simulation study. Behavioural Processes, 41, 177e192. Bernstein, I. S. 1981. Dominance: the baby and the bathwater. Behavioral and Brain Sciences, 4, 419e457. Bonabeau, E., Theraulaz, G. & Deneubourg, J.-L. 1996. Mathematical model of self-organizing hierarchies in animal societies. Bulletin of Mathematical Biology, 58, 661e717. Broom, A. 2002. A unified model of dominance hierarchy formation and maintenance. Journal of Theoretical Biology, 219, 63e72. Castillo Alvarez, A. & Chavez-Peo´n, C. 1983. Ecologı´a Reproductiva e influencia del comportamiento en el control del nu´mero de crı´as en el bobo de patas azules, Sula nebouxii, en la Isla Isabel, Nay. Undergraduate thesis, UNAM, Facultad de Ciencias. Chase, I. D. 1974. Models of hierarchy formation in animal societies. Behavioral Science, 19, 374e382. Chase, I. D. 1982a. Dynamics of hierarchy formation: the sequential development of dominance relationships. Behaviour, 80, 218e 240. Chase, I. D. 1982b. Behavioral sequences during dominance hierarchy formation in chickens. Science, 216, 439e440. Chase, I. D. 1986. Explanations of hierarchy structure. Animal Behaviour, 34, 1265e1267. Chase, I. D., Tovey, C., Spangler-Martin, D. & Manfredonia, M. 2002. Individual differences versus social dynamics in the formation of animal dominance hierarchies. Proceedings of the National Academy of Sciences, U.S.A., 99, 5744e5749. Cook, M. I., Monaghan, P. & Burns, M. D. 2000. Effects of shortterm hunger and competitive asymmetry on facultative aggression in nestling black guillemots Cepphus grille. Behavioral Ecology, 11, 282e287. Drummond, H. 2001. The control and function of agonism in avian broodmates. In: Advances in the Study of Behavior 30 (Ed. by P. J. B. Slater, J. S. Rosenblatt, C. T. Snowdon & T. J. Roper), pp. 261e301. San Diego: Academic Press. Drummond, H. 2006. Dominance in vertebrate broods and litters. Quarterly Review of Biology, 81, 3e32. Drummond, H. & Canales, C. 1998. Dominance between booby nestlings involves winner and loser effects. Animal Behaviour, 55, 1669e1676. Drummond, H. & Garcı´a Chavelas, C. 1989. Food shortage influences sibling aggression in the blue-footed booby. Animal Behaviour, 37, 806e819. Drummond, H. & Osorno, J. L. 1992. Training sibling to be submissive losers: dominance between booby nestlings. Animal Behaviour, 44, 881e893. Drummond, H., Gonza´lez, E. & Osorno, J. L. 1986. Parente offspring cooperation in the blue-footed booby (Sula nebouxii): social roles in infanticidial brood reduction. Behavioral Ecology and Sociobiology, 19, 365e373. Drummond, H., Osorno, J. L., Garcı´a, C., Torres, R. & Merchant, H. 1991. Sexual dimorphism and sibling competition: implications for avian sex ratios. American Naturalist, 138, 623e641.
1713
1714
ANIMAL BEHAVIOUR, 74, 6
Drummond, H., Rodrı´guez, C., Vallarino, A., Valderra´bano, C., Rogel, G. & Tobo´n, E. 2003. Desperado siblings: uncontrollably aggressive junior chicks. Behavioral Ecology and Sociobiology, 53, 287e296. Dugatkin, L. A. 1997. Winner and loser effects and the structure of dominance hierarchies. Behavioral Ecology, 8, 583e587. Dugatkin, L. A. 2001. Bystander effects and the structure of dominance hierarchies. Behavioral Ecology, 12, 348e352. Hsu, Y. & Wolf, L. L. 1999. The winner and loser effect: integrating multiple experiences. Animal Behaviour, 57, 903e910. Hsu, Y., Early, R. L. & Wolf, L. L. 2006. Modulation of aggressive behaviour by fighting experience: mechanisms and contest outcomes. Biological Reviews, 81, 33e74. Jackson, W. M. 1988. Can individual differences in history of dominance explain the development of linear dominance hierarchies? Ethology, 79, 71e77. Kalas, V. S. 1977. Ontogenie und Funktion der Rangordnung innerhalb einer Geschwisterschar von Grauga¨nsen (Anser anser L.). Zeitschrift fu¨r Tierpsychologie, 45, 174e198. Leshner, A. 1978. An Introduction to Behavioral Endocrinology. Oxford: Oxford University Press. Mesterton-Gibbons, M. & Dugatkin, B. E. 1995. Toward a theory of dominance hierarchies: effects of assessment, group size and variation in fighting ability. Behavioral Ecology, 6, 416e423. Mock, D. W. & Parker, G. A. 1997. The Evolution of Sibling Rivalry. Oxford: Oxford University Press.
Nathan, A., Legge, S. & Cockburn, A. 2001. Nestling aggression in broods of a siblicidal kingfisher, the laughing kookaburra. Behavioral Ecology, 12, 716e725. Nelson, J. B. 1978. The Sulidae: Gannets and Boobies. London: Oxford University Press. Pagel, M. & Dawkins, M. S. 1997. Peck orders and group size in laying hens: ‘futures contracts’ for non-aggression. Behavioural Processes, 40, 13e25. Pardi, L. 1948. Dominance order in Polistes wasps. Physiological Zoology, 21, 1e13. Roseler, P. F. 1991. Reproductive competition during colony establishment. In: The Social Biology of Wasps (Ed. by K. G. Ross & R. W. Matthews), pp. 309e336. Ithaca, New York: Comstock. Rushen, J. 1982. The peck orders of chickens: how do they develop and why are they linear? Animal Behaviour, 30, 1129e 1137. Slater, P. J. B. 1986. Individual differences and dominance hierarchies. Animal Behaviour, 34, 1264e1265. Theraulaz, G., Pratte, M. & Gervet, J. 1989. Effects of removal of alpha-individuals from a Polistes dominulus Christ wasp society: changes in behavioral patterns resulting from hierarchical changes. Journal of Insect Behavior, 5, 169e179. Torres, R. & Drummond, H. 1999. Does large size make daughters of the blue-footed booby more expensive than sons? Journal of Animal Ecology, 68, 1133e1141.
Available online at www.sciencedirect.com
Development of a linear dominance hierarchy in nestling birds ´ BANO-IBARRA, IRERI B RUMON & HUGH DRUMMOND CLA UDIA VALDER RA
Instituto de Ecologı´a, Universidad Nacional Auto´noma de Me´xico (Received 5 June 2006; initial acceptance 9 August 2006; final acceptance 14 February 2007; published online 23 October 2007; MS. number: A10466R2)
Theoreticians propose that trained winning and losing are important processes in creating linear animal dominance hierarchies, and experiments have shown that both processes can occur in animals, but their actual roles in creating natural hierarchies are unknown. We described agonism in 18 broods of three bluefooted boobies, Sula nebouxii, a species for which trained winning and losing have been demonstrated, to infer how these processes generate and maintain a natural hierarchy. Ranks in the linear hierarchy that emerged in every brood were initially assigned by asymmetries in age, size and maturity, which led to differences between broodmates in levels of expressed and received aggression and, consequently, to differences in the training of their aggressiveness and submissiveness. Later, ranks appeared to be maintained by the chicks’ acquired aggressive and submissive tendencies combined with ongoing effects of persisting differences in size and maturity. Our results suggest that trained winning and trained losing are important in the construction of booby hierarchies but that these two axes of learning are largely independent. Increase in submissiveness occurs over a period of about 10e20 days, and the level of submissiveness reached varies with the amount of aggression received. After training, submissiveness is apparently maintained by a lower level of aggression and increasing use of threats. Threats become increasingly effective as chicks age, but are never as effective as attacks. Ó 2007 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Keywords: agonistic behaviour; blue-footed booby; dominance; hatch asynchrony; hierarchy; Sula nebouxii; trained winning
Theoreticians have asked which behavioural mechanisms can in principle efficiently generate the linearity that characterizes many animal hierachies (e.g. Chase 1974; Beaugrand 1997; Broom 2002). Candidate mechanisms include trained winning and losing (Hsu et al. 2006), the learned dyadic behavioural dispositions of true dominance (Bernstein 1981), assessment of fighting ability, and the bystander effect (an individual assesses others by watching them fight each other; Chase 1982a, b; Dugatkin 2001; Chase et al. 2002), and combinations of these mechanisms are expected (e.g. Mesterton-Gibbons & Dugatkin 1995; Beaugrand 1997; Pagel & Dawkins 1997). Mathematical modelling prevails in this field and experimental tests are sometimes made, but testing is done in artificial situations whose relevance to naturally occurring dominance hierarchies in the study species is usually unknown. We lack quantitative descriptions of
Correspondence: H. Drummond, Instituto de Ecologı´a, Universidad Nacional Auto´ noma de Me´ xico, A.P. 70-275, 04510 D.F., Me´xico (email: [email protected]). 0003e 3472/07/$30.00/0
the emergence of natural animal dominance hierarchies, and further progress in developing and testing ideas about the formation of dominance relationships and the development and maintenance of dominance hierarchies may depend critically on getting to grips with what actually occurs in nature. Trained winning and trained losing are the two learning mechanisms most widely expected to contribute to hierarchy formation and maintenance (Barnard & Burk 1979; Pagel & Dawkins 1997; Beacham 2003). Although their existence has been demonstrated in experimental tests of diverse species of vertebrates, their actual functioning in the construction of natural dominance relationships and hierarchies is still obscure (Dugatkin 1997). Each individual’s history of victories and defeats against its competitors is generally expected to condition it to a particular position on an aggressiveesubmissive continuum (e.g. Theraulaz et al. 1989), and the relative positions of all group members on the continuum could determine the hierarchy among them. However, assessment is also likely to be important, and, in theory, the nature of a hierarchy can depend on trained winning or trained
1705 Ó 2007 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
1706
ANIMAL BEHAVIOUR, 74, 6
losing, or both, and this is seldom known (Dugatkin 1997). Nor do we have a grasp of the schedules of interactions whereby members of natural groups entrain each other’s and their own agonistic responsiveness: in particular, how long training takes and how agonism changes when it has been achieved (Bonabeau et al. 1996). An amenable subset of animal dominance hierarchies includes those that occur in altricial broods of some avian species (reviews in Mock & Parker 1997; Drummond 2001, 2006), and these can be observed from the start of their agonism. As far as we know, these infant hierarchies do not involve such complications as alliance formation or even individual recognition and they may be among the most simple hierarchies that occur in natural groups of vertebrate conspecifics. Typically, two to four nestlings hatch at intervals of 1e5 days and start attacking each other within several days. The nature of dyadic dominance relationships varies among and within species (Drummond 2006), but nestlings usually sort themselves over a period of several days or weeks into a linear hierarchy that follows the order in which they hatched and confers growth and survival benefits on high-ranking individuals. Our knowledge of the development of these hierarchies is sketchy because broods are rarely observed from the onset of hostilities, the temporal resolution of data is often poor, responses to aggression are seldom quantified and few studies have actually analysed age-related changes in aggression or submission (but see Drummond et al. 1986; Cook et al. 2000; Nathan et al. 2001). Dominance relationships and hierarchies occur in broods of the blue-footed booby, Sula nebouxii, a marine bird that produces broods of one to three altricial chicks that fledge at about 3 months of age (Nelson 1978). Staggered laying and hatching ensures that broodmates differ substantially in age: in three-chick broods, the A-chick is, on average, 4.0 days older than the B-chick, which is 3.6 days older than the C-chick (Castillo Alvarez & ChavezPeo´n 1983). Broods of three are largely unstudied, but in broods of two, there is always a dominance relationship, usually with the A-chick assuming a dominant role characterized by daily attacking and threatening, and the B-chick assuming a subordinate role characterized by minimal aggression and submissive responses to aggression (Nelson 1978; Drummond et al. 1986; Anderson & Ricklefs 1995). A-chicks grow faster than B-chicks during the first few weeks of life and are more likely to fledge, but surviving B-chicks catch up on growth and, at fledging, they are just as large as A-chicks. Experimental pairings of different combinations of unfamiliar dominant chicks, subordinate chicks and singletons (chicks with no broodmate) that were 12e55 days old demonstrated that trained winning, trained losing and assessment are all involved in the agonism of blue-footed boobies. Thus, a chick’s aggressiveness or submissiveness to an unfamiliar chick is influenced by prompt assessment of its relative size, but more importantly, by its own agonistic training in its home brood: previously dominant broodmates tend to behave aggressively and nonsubmissively, previously subordinate broodmates tend to behave submissively and nonaggressively (Drummond & Osorno 1992; Drummond & Canales 1998). The effects of trained
winning and losing also seem to include modification not only of agonistic tendencies but also of fighting ability: A-chicks dominate unfamiliar B-chicks that are 32% heavier than themselves, even though most B-chicks respond to their newfound size advantage with increased aggressiveness. Hence, the observed stability of dominance relationships over the nestling period (Drummond et al. 1991) could largely be due to trained winning and losing. We studied the emergence of dominance hierarchies in natural three-chick broods of the blue-footed booby. Broods of three are always a minority and often are reduced by nestling mortality to two chicks, but sometimes all three broodmates survive and cohabit through to fledging. In our study population, 217 three-chick broods fledged 75% of A-chicks, 69% of B-chicks and 31% of C-chicks, and, in 20% of broods, all three broodmates fledged (H. Drummond, unpublished data from 24 seasons). We sought evidence for how the mechanisms of trained winning and trained losing operate in a brood of three to create and maintain a linear hierarchy. We compared the development of behaviour in A-, B- and C-chicks to infer their schedules of training and how they come to occupy their dominance ranks. Examination of their behavioural development also yielded insight into (1) whether aggressiveness and submissiveness are two sides of the same coin or independent axes of behavioural tendency, (2) whether B-chicks acquire intermediate rank by simultaneously learning increased submissiveness to their elder broodmate and increased aggressiveness to their younger broodmate and (3) whether aggressors progressively substitute threats for attacks as victims learn to submit to threats.
METHODS We located two-chick and three-chick broods on Isla Isabel, Nayarit, Mexico (21 520 N, 105 540 W) in March and April of 2002 and 2004 by monitoring all nests in our study area every 3 days, starting shortly after the start of hatching in the colony. When hatching date was unknown, we estimated chick age using culmen and ulna growth curves of the same population. We banded broodmates according to their age ranks (Drummond et al. 1991). When broods were first found, there was no risk of misidentifying age ranks within a brood, given this booby’s large hatching intervals. Blue-footed boobies lay clutches of one to three eggs. Clutches of three eggs are uncommon, but a minority of two-chick broods arise through hatching failure in three-egg clutches. Sampling of three-chick broods and two-chick broods was constrained by availability and mortality and could not be equitable. Three-chick broods were sampled in 2002 by observing all intact broods of that size more or less daily until the first chick death occurred (all three-chick broods suffered mortality in 2002). We suspended observation and switched to a new three-chick brood whenever switching would assure more even sampling of the first 6 weeks of life. Eighteen three-chick broods were observed for a mean SE of 10.2 1.9 days, from A-chick ages of 10e23 days (X ¼ 16:2 days) to 13e49 days (X ¼ 27:3 days); in these
VALDERR ABANO-IBARRA ET AL.: DEVELOPMENT OF A HIERARCHY
broods, B-chicks and C-chicks were on average 3.2 days and 7.4 days younger than A-chicks. We included two samples of two-chick broods in 2004 to help discern the mechanisms operating in three-chick broods by documenting the development of dominance relationships in the simplest social context. The long-term sample consisted of 13 broods grouped in neighbourhoods where we could observe up to three broods simultaneously. These were observed roughly every other day for a mean SE of 18.6 1.1 days per brood, starting when A-chicks were 6 days old (X ¼ 13:2 days) and ending when they were 45 days old. Only one chick died in a two-chick brood before this age. The average age difference between broodmates in the long-term sample of two-chicks broods was 3.3 days. The short-term sample consisted of 70 broods that were observed (partly to screen broods for another study) during just 2.4 0.3 consecutive days when chicks were 2e45 days old. Two-chick broods of the long-term sample were used to analyse the relationship between a chick’s submissiveness and the aggression it received, because minimal mortality in these broods allowed more even sampling, and also to analyse the use of threats, which were not recorded in three-chick broods (although they certainly occurred). Although these differences in behavioural sampling limited our ability to contrast the two brood sizes (and precluded using the same statistical procedures to analyse them), the combined data set allowed us to examine common patterns and to infer common processes. Red, blue or yellow paint was applied to the head and rump of each brood member (assigned haphazardly) on the first day of observation and reapplied on subsequent days as needed to ensure age ranks were recognized. Chicks were weighed and measured (culmen and ulna) every day at the end of observations. An observer, seated roughly 5 m from the nest during 6 h each day (0700e 1000 hours and 1500e1800 hours), recorded the absolute frequencies of attacks (pecks and bites) by all brood members, and whether the victim responded to each attack by adopting or maintaining a bill-down-and-face-away submissive posture (BDFA, Nelson 1978), using the behavioural criteria in Drummond et al. (2003). For two-chick broods, the observer also recorded the absolute frequencies of vocal threats and submissive responses to them. An increase or decrease in an individual’s attacking over time could reflect a change in its aggressiveness due to trained winning, but it could also be due to the individual’s maturation or level of access to the victim. An increase in submissive responses to aggression is evidence of increased submissiveness due to trained losing, but it could also be a function of maturation. Unless stated otherwise, comparisons of broodmates use data from each broodmate at the same age (calculated from each chick’s own hatch date), since our purpose was to compare their development. We calculated each chick’s attack rate for each successive 5-day age block by dividing its total attacks in the block by the total number of hours it was observed in the block. Similarly, each chick’s submissiveness in each block was the number of attacks it responded to with a submissive posture divided by the number of attacks it received in the block. Where zero
scores of individual chicks prevented us from calculating the percentage difference between scores of dyadmates for all dyads in a sample, we report the percentage difference between the two sample mean values (e.g. between the mean score for A-chicks and the mean score for B-chicks). For dyads that were observed only for 1e3 days, we inferred inversion of dominance if the younger chick attacked more frequently and was less submissive than the elder chick. For dyads observed on more than 3 days, we inferred inversion of dominance if the younger chick attacked more frequently overall, had a superior attacking score on more than half of observation days, was less submissive overall and showed submission on fewer days relative to the elder chick. We considered that dominance in a dyad was contested when the subordinate chick attacked more than half as frequently as the dominant chick. We used nonparametric tests because distributions were not normal and most data sets could not be normalized by transformation. For two-chick broods, we pooled the longterm and short-term samples, after graphing the data and confirming visually that they showed similar patterns. Tests for dependent samples were preferred except when sample sizes were inadequate and larger samples could be analysed by including independent data, which was often the case because broodmates were frequently observed in different age blocks (since they differed in age). When a subset of individuals was observed in more than one age block, using a test for independent samples rather than a test for dependent samples was conservative, provided that booby chicks show substantial and stable individual differences. In any event, only two tests for independent samples that included some individuals observed at more than one age yielded a significant probability greater than 0.01 (see Results; Fig. 2). We tested for age-related changes in behaviour in two-chick broods using comparisons among all age blocks for which sample sizes were at least N ¼ 8, and in three-chick broods by comparing the three successive age blocks with the largest samples. Wherever the sample of chicks used for a statistical comparison was a subset of the sample shown in the bar of a histogram, the magnitude of the difference reported may differ from the magnitude evident in the figure. Throughout results we report X SE. RESULTS
Three-chick Broods Mortality in the 18 broods was substantial during the first 6 weeks of life and increased with hatch order: A-, B- and C-chicks, respectively, died in 6%, 16% and 100% of broods (G test: G2 ¼ 49.05, P < 0.001), at ages of 30 days, 21e29 days and 5e42 days. Over this period, growth rates increased with age/dominance rank and C-chicks grew particularly poorly. Comparing broodmates at the same age of 15e19 days (when our sample sizes were largest), A-chicks were on average 21% heavier than B-chicks (Wilcoxon signed-ranks test: T ¼ 3.0, N ¼ 12, P ¼ 0.004) and B-chicks were on average 71% heavier than C-chicks (T ¼ 1.0, N ¼ 11, P ¼ 0.004).
1707
1708
ANIMAL BEHAVIOUR, 74, 6
Linear dominance hierarchy In all three dyads of every three-chick brood, both chicks showed some aggression, and the dominance hierarchy of every brood was linear: A-chicks attacked B-chicks and C-chicks 4e15 and 21e95 times more often, respectively, than vice versa, and B-chicks attacked C-chicks 6e28 times more often than vice versa (Fig. 1). In every dyad, the elder chick attacked at a higher rate than the younger chick at all ages (different-age comparisons; P < 0.02 in all nine Wilcoxon tests).
Development of aggression and submissiveness A-chicks and B-chicks were already attacking at the earliest ages they were observed (Fig. 2). Attacks by the two most senior chicks towards their broodmates (summed, N ¼ 8) increased with age, steeply in the case of A-chicks (over 15e29 days; Friedman test: c22 ¼ 6:250, P ¼ 0.043) and modestly in the case of B-chicks (over 10e24 days, KruskaleWallis test: H2 ¼ 5.86, P ¼ 0.05). Attacking by C-chicks also changed with age (5e19 days, H2 ¼ 5.98, P ¼ 0.05), but C-chicks started late and increases in their attack rates were minimal and short-lived (Fig. 2). The lower a chick’s age rank, the earlier its aggression peaked and declined: C-chicks peaked at roughly age 10e14 days, B-chicks at roughly 20e24 days and A-chicks not until roughly 30e34 days (Fig. 2). Moreover, maximum pecking by A-chicks exceeded that of B-chicks (Manne Whitney U test: U ¼ 7, N1 ¼ 5, N2 ¼ 10, P < 0.05) and maximum pecking by B-chicks exceeded that of C-chicks (U ¼ 16, N1 ¼ 10, N2 ¼ 13, P < 0.002). In both A-chicks and B-chicks, the decline in attacking appeared steep and permanent, but note that our samples embraced less than half of the prefledging period. A 15–19 days N = 13 broods B
C
All age ranks appeared to start out responding to only 10e25% of attacks with a submissive posture, then all age ranks progressively increased their submissiveness (Fig. 3). Between the ages of 10 and 24 days, submissiveness of B-chicks to attacks by A-chicks increased to over 80% (KruskaleWallis test: H2 ¼ 15.39, P < 0.001) and submissiveness of C-chicks to each of their broodmates rose to similar values (to A-chicks: H2 ¼ 11.63, P ¼ 0.003; to B-chicks: H2 ¼ 8.57, P ¼ 0.01). Surprisingly, between the ages of 15 and 29 days, A-chicks also showed an increase (H2 ¼ 6.32, P ¼ 0.04; Fig. 3), eventually submitting to more than 40% of aggressions by B-chicks. The lesser submissiveness of A-chicks was confirmed by two tests made at the only ages for which samples were sufficient: at age 15e24 days, A-chicks were four to five times less submissive to B-chicks than similarly aged B-chicks were to A-chicks (Wilcoxon signed-ranks test: 15e19 days: T ¼ 2, N ¼ 8, P ¼ 0.04; 20e24 days: T ¼ 0.0, N ¼ 10, P ¼ 0.005; Fig. 3), and at age 15e19 days, A-chicks were eight times less submissive to B-chicks than were similarly aged C-chicks (T ¼ 0.0, N ¼ 10, P ¼ 0.007; Fig. 3). Attacks on A-chicks began later (ca. age 15 days) than those on Bchicks and C-chicks (ca. age 5 days; Fig. 2) and petered out by the time A-chicks were 4 weeks old (Fig. 2), the age at which attacks on B-chicks and C-chicks were maximal (Fig. 1). Importantly, attacks on A-chicks were delivered by broodmates smaller and weaker than themselves. There were indications that the aggressiveness of elder chicks declined after the submissiveness of their younger broodmates reached a high value. Maximum average submissiveness of B-chicks was achieved when they were 20e24 days old (Fig. 3), and 5e10 days later there was an approximate halving of violent aggression by A-chicks (after A-chicks reached 35 days; Fig. 2). Four broods were observed on sufficient successive days to determine whether the decline in aggressiveness of individual A-chicks occurred after submissiveness of their own B-broodmates reached a high value. Submissiveness by B-chicks reached or exceeded 90% in all four broods, and in three of those four broods a drop of 50% or more in aggression of A-chicks occurred afterwards. The drop came 5e10 days after 90% submissiveness was first achieved, when dominants were 30e40 days old.
A
Training of the intermediate chick
20–24 days N = 14 broods C
B
A 25–29 days N = 9 broods B
C
Figure 1. Development of a linear hierarchy in asynchronously hatched blue-footed booby chicks. Shown are attacks by each broodmate (A-, B- and C-chick) when A-chicks were at the ages shown and B- and C-chicks were on average 3.2 and 7.4 days younger, respectively. Thickness of each arrow represents the mean rate of attacks up to a maximum value of 12.18 attacks/h.
B-chicks behaved differently towards their two broodmates, delivering more attacks to C-chicks and showing greater submissiveness to A-chicks. Over the 10-day age span for which comparisons could be made, B-chicks attacked C-chicks two to five times more often than they attacked A-chicks (Wilcoxon test: 15e19 days: T ¼ 1.0, N ¼ 14, P ¼ 0.01; 20e24 days: T ¼ 1.0, N ¼ 10, P ¼ 0.017; Fig. 2), and B-chicks were four times as submissive to A-chicks as they were to C-chicks (ManneWhitney U test: 15e19 days: U ¼ 9.5, N1 ¼ 13, N2 ¼ 7, P ¼ 0.04; 20e24 days: U ¼ 6.0, N1 ¼ 10, N2 ¼ 7, P ¼ 0.004; Fig. 3).
Two-chick Broods There was uncontested dominance in nine of 13 broods in the long-term sample and 66 of 70 broods in
VALDERR ABANO-IBARRA ET AL.: DEVELOPMENT OF A HIERARCHY
Three-chick broods and C
By A on B 20
15 Two-chick broods 10
By dominant 10
0
5
5
10
15
20
25
30
35
40
45
Age (days) of dominant
5
0
Attacks/h
Attacks/h
5
By subordinate
5
10 15 20 25 30 35 40 Age (days) of subordinate
0 5
15
20
25
30
35
40
45
35
40
45
35
40
45
Age (days) of A
0
5
0
and C
By B on A
5
45
10
5
10
15
By C on A
5
10
20 25 30 Age (days) of B and B
15
20
25
30
Age (days) of C Figure 2. Attack rates (X þ SE) during hierarchy development in two- and three-chick broods of blue-footed boobies. Sample sizes in successive blocks: two-chick broods: dominant chicks: 11, 21, 20, 35, 38, 23, 18, 16; subordinate chicks: 21, 22, 28, 35, 33, 20, 17, 13; three-chick broods: A-chicks: 0, 6, 13, 14, 9, 5, 4, 4; B-chicks: 5, 10, 14, 10, 6, 4, 4, 0; C-chicks: 8, 13, 11, 6, 4, 4, 0, 0.
the short-term sample, and this uncontested dominance was inverted in two of nine long-term broods and three of 66 short-term broods. The proportion of broods with inverted or contested dominance in our study (16% overall) was unusually high compared to proportions in earlier studies (9% and 4% of two-chick broods observed >60 days by Drummond et al. 1986, 1991), but analyses carried out after deleting broods with inverted or contested dominance (not shown) or sorting broodmates by age rank rather than dominance rank yielded similar results, so the analyses we report may be considered representative.
Development of aggression and submissiveness Both chicks started attacking in their respective first age blocks. Aggression of dominants changed over the eight age blocks (KruskaleWallis test: H7 ¼ 23.53, P ¼ 0.001), showing a steady increase to a peak at 20e24 days
followed by decline (Fig. 2). The aggression of subordinates followed a similar pattern but the apparent peak came earlier, at 15e19 days, and change over time was not significant (H7 ¼ 8.59, P ¼ 0.283; Fig. 2). The submissiveness of subordinates varied with age (KruskaleWallis test: H7 ¼ 47.9, P < 0.001), rising from 23% at age 7 days to 87% at 15e19 days (Fig. 3) and remaining at that high level until at least age 44 days. Submissiveness of dominants in the sample increased from 20% at age 12 days to a maximum of 55% at 30e34 days, but change with age was not significant (H6 ¼ 6.99, P ¼ 0.322; Fig. 3). However, subordinate chicks were 60% more submissive than dominant chicks at their respective maximums (ManneWhitney U test: U ¼ 97.5, N1 ¼ 27, N2 ¼ 13, P ¼ 0.023; Fig. 3). In the long-term sample, there was a strong relationship between the rate of violent aggression received by a chick during the first 6 weeks of life and its average submissiveness over the same period (Fig. 4). The values for dominants
1709
ANIMAL BEHAVIOUR, 74, 6
Two-chick broods
Three-chick broods A to B
Dominant chick 1
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
5
10
15
20
25
30
35
40
0
45
5
Subordinate chick 1
0.8
0.8
0.6 0.4 0.2 5
10
15 20 25 30 35 Age (days) of subordinate
15
20
25
30
10 15 20 25 Age (days) of B
30
B to A
1
0
10
Age (days) of A
Age (days) of dominant
Submissiveness
Submissiveness
1710
40
45
0.6 0.4 0.2 0
5
C to A
or B
1 0.8 0.6 0.4 0.2 0
5
10
15
20
25
30
Age (days) of C Figure 3. Development of submissiveness in blue-footed booby chicks (proportion of attacks eliciting submission, X þ SE). Submissiveness of the two eldest chicks to C-chicks is not shown because attacks by C-chicks were too few. Sample sizes in successive blocks: two-chick broods: dominant chicks: 0, 11, 17, 22, 18, 13, 11, 10; subordinate chicks: 18, 21, 27, 34, 33, 18, 15, 12; three-chick broods: A-chicks: 0, 4, 10, 12, 7; B-chicks: 4, 8, 13, 10, 6; C-chicks: 5e5, 11e11, 8e10, 6e5, 4e4.
and subordinates appeared to fit on the same logistic curve, which explained 74% of the variation. The correlation was strong for the whole sample of 25 dominant and subordinate chicks (Pearson’s correlation: r23 ¼ 0.64, P ¼ 0.001) and also for the 12 dominant chicks alone (r10 ¼ 0.69, P ¼ 0.014). For the 13 subordinate chicks alone, the correlation was also positive but not significant (r11 ¼ 0.41, P ¼ 0.16; Fig. 3), probably because they were sampled on only a limited portion of the curve (all of them received high levels of violence). Importantly, in broods where dominants were subjected to high levels of pecking, they were highly submissive despite managing to outpeck their broodmates (Fig. 3).
In the long-term sample, the decline in aggressiveness of elder chicks started after submissiveness of their younger broodmates reached a high value. Maximum average submissiveness was reached when subordinates were 15e20 days old (Fig. 3), and was followed in the next 5-day block by an approximate halving of violent aggression by dominants (after dominants reached 25 days; Fig. 2). In individual broods, the interval between full submissiveness and the decline in aggression was variable: submissiveness by subordinates exceeded 90% in 12 of 13 broods, and in nine of those 12 broods, a drop of 50% or more in aggression by the dominant chick occurred afterwards and appeared to be sustained subsequently.
Proportion of threats (X ± SE)
VALDERR ABANO-IBARRA ET AL.: DEVELOPMENT OF A HIERARCHY
1
Submissiveness
0.8 0.6 0.4 0.2 0
0
1
2
3
4
5
6
7
Attacks received/h Figure 4. Relationship between rate of attacks received during ages 15e44 days and average submissiveness shown at the same ages by 12 dominant chicks (C) and 13 subordinate chicks (B) in two-chick broods (logistic curve, R2 ¼ 73.58) of blue-footed boobies.
The drop came 5e20 days after 90% submissiveness was first achieved, when dominants were 20e39 days old.
Threats Threatening by dominant chicks started at age 5e10 days and increased progressively to a peak at age 20e34 days (KruskaleWallis test: H7 ¼ 64.7, P < 0.001), and the submissiveness of subordinates to those threats ascended during the same period from 21% to a maximum of 57% when subordinates were 25e29 days old (H7 ¼ 18.28, P ¼ 0.011). However, at no age did subordinates submit to threats as frequently as they did to physical attacks, and their maximal submissiveness to threats was 34% less than their maximal submissiveness to physical attacks (ManneWhitney U test: U ¼ 185.5, N1 ¼ 27, N2 ¼ 33, P < 0.001). The portion of aggressions by dominants that were threats rather than physical attacks rose progressively from 3% at age 5e10 days to a maximum of 62% at age 40e44 days (KruskaleWallis test: H6 ¼ 27.43, P < 0.001; Fig. 5), when our observations ended. However, we did not find evidence for dominants increasingly switching to threats as the efficacy of threats increased. When we plotted each brood’s mean scores for the proportion of dominants’ aggressions that were threats and the proportion of subordinates’ responses to threats that were submissive in each age block, Spearman’s correlation coefficients (not shown) were positive for five of 13 broods and negative for the remainder. The threat behaviour of subordinates developed similarly and with similar effects, despite subordinates showing less aggression and eliciting less submission in general. Subordinate threats peaked at 15e19 days (KruskaleWallis test: H7 ¼ 42.17, P < 0.001) and elicited less submissiveness than threats by dominants (comparison of maximum values; ManneWhitney U test: U ¼ 99.0, N1 ¼ 33, N2 ¼ 17, P < 0.001). The proportion of aggressions by subordinates that were threats rose progressively to a maximum of 81% at age 30e34 days (H7 ¼ 26.84, P < 0.001; Fig. 5), higher than the maximum for dominants (U ¼ 73, N1 ¼ 17, N2 ¼ 15, P ¼ 0.04).
1 0.8 0.6 0.4 0.2 0
5
10
15
20 25 30 Chick age (days)
35
40
45
Figure 5. Proportions of total aggressions by dominant (-) and subordinate (,) blue-footed booby chicks that were threats. Sample sizes in successive blocks: dominant chicks: 5, 20, 20, 35, 38, 23, 18, 15; subordinate chicks: 10, 19, 25, 27, 28, 17, 17, 10.
DISCUSSION Linear hierarchies emerged in all 18 three-chick broods, with all three dyads in every brood showing a clear relationship of dominanceesubordination favouring the elder chick. Comparisons of these profiles, in the light of previous experimental confirmation of the existence of trained winning, trained losing and assessment in chicks of age 12e55 days (Drummond & Osorno 1992; Drummond & Canales 1998), allows preliminary inferences about the learning mechanisms and developmental processes responsible for the changes. Of course, these are descriptive data and alternative interpretations are possible. A key inference is that, in boobies, aggressiveness and submissiveness may be two largely independent axes of behavioural tendency and that trained winning and trained losing may be two largely distinct processes that enhance those tendencies. The evidence for this is that the most dominant chicks of both brood sizes showed increasing submissiveness at the same ages that they were showing increasing aggression. This inference is consistent with Leshner’s (1978) suggestion that aggressive and submissive behaviours are often under the proximate control of different hormones. In what follows, we suggest that the distinct agonistic profiles of the two or three chicks in a brood are mainly an outcome of differences in (1) the training of each chick’s aggressiveness and submissiveness and (2) the social environments that chicks inhabit at any age. The training differences are set in motion by initial age/size differences (cf. Chase 1986; Slater 1986; Jackson 1988).
Development of Aggression Attacking by chicks of all ranks in both brood sizes started at age 5e9 days and increased with age. In the lowest-ranking chicks of each brood size the increase was undoubtedly due to maturation, as the stamina, mobility and wakefulness of each chick increased with age, and as its maturing broodmates became increasingly available for pecking. However, the far greater age-related increase in pecking by A-chicks over the first 3e4 weeks of life
1711
1712
ANIMAL BEHAVIOUR, 74, 6
presumably was partly due to trained winning, as they accumulated a history of attacks on their two broodmates that drew ever more submission from them. Because attacks by subordinates often provoke retaliation by dominants (Drummond & Osorno 1992), it is also likely that the superior increase in pecking by elder chicks was partly due to their aggressive stifling and punishing of younger broodmates that attacked them: juniors often forwent opportunities to attack when they submitted, cowered or hid in response to attacks by broodmates that were more mature and powerful than they were. The conspicuous decline in attacking by A-chicks after age 25e40 days may be due to appeasement: aggressiveness of dominant chicks may be reduced when their broodmates reach some criterion of submissiveness. Attack rates declined at ages when attacking might be expected to increase with maturation and in response to intensified feeding competition during the period of maximal growth (Drummond et al. 1991; Torres & Drummond 1999). Facultative decline in response to submissiveness is supported by two analyses: (1) maximal submissiveness by B-chicks was reached earlier in twochick broods than in three-chick broods and, in both cases, it closely preceded aggressive decline in A-chicks; (2) in individual dyads of both brood sizes there was evidence for a stepwise drop in agggression by dominants after the submissiveness of subordinate broodmates reached 90%. However, part of the age-related decline in attacking could have been due to broodmates developing the locomotor ability to walk about in the home territory and avoid proximity (Drummond & Garcı´a Chavelas 1989). Vocal threats progressively substituted for pecks and bites despite being less effective at eliciting submission. Increased reliance on threats apparently was not triggered by victims’ enhanced submissiveness: these variables were not correlated in individual broods and subordinates threatened proportionally more than dominants despite eliciting less submission. Chicks may increasingly use threats as they get older because threats function at a distance and broodmates spend more time apart as they wander independently in the family territory. The subordinates’ greater proportional increase in threatening could also result from a preference to show aggression when they are far from their intimidating broodmates.
Development of Submissiveness Chicks of all ranks showed a developmental increase in submissiveness. Maturation could not completely account for the increase because it occurred at different ages and reached different maximum values in different ranks. More likely, boobies are congenitally disposed to respond with minimal submissiveness to the first pecks and bites, and progressive enhancement of submissiveness is mostly dictated by the schedule of aggression received. Submissiveness of A-chicks in three-chick broods was moderate and developed relatively late because A-chicks did not receive pecks until their junior broodmates were old
enough to deliver them (when A-chicks were nearly 15 days old) and because the pecking they received was infrequent and weak, coming from beseiged and smaller chicks. Submissiveness may also increase with maturation, but the tight correlation between attacks received and submissiveness shown during the first 6 weeks of life in twochick broods (Fig. 4) suggests that submissiveness is largely a function of the amount of aggression received, whatever the chick’s dominance rank (cf. Hsu & Wolf 1999). Submissiveness, then, is an index of trained losing. The persistence of enhanced submissiveness of subordinates and dominants in two-chick broods well beyond the age when attacks on them declined suggests that, once trained losing is well established, the level can be maintained by a substantially lower level of aggression in which threats outnumber attacks. In two-chick and three-chick broods, respectively, thorough training of junior chicks in losing (maximal submissiveness) was not achieved until they were 2e3 weeks old, usually after receiving daily pecks and threats over a period of roughly 10e20 days (from the time elder broodmates started pecking).
Acquiring Intermediate Rank B-chicks in three-chick broods showed clear evidence of trained losing during the first 3 weeks of life but only in relation to one of their broodmates. These intermediates progressively became as submissive to their senior broodmates as were B-chicks in broods of two, but considerably less submissive to their junior broodmates, who attacked them much less. Intermediates may discriminate between their two broodmates, probably on the basis of size, and learn to show increased submissiveness only to the larger one. Alternatively, intermediates may seldom have submitted to the smaller broodmate’s attacks simply because they were feeble. It is less clear whether intermediates learned to attack more frequently or to attack broodmates differentially. Their age-related increase in aggression could have been due to maturation, trained winning or a combination of the two, and intermediates could have attacked junior broodmates two to five times more frequently than they attacked senior broodmates either because they were preferred targets or because attacks on senior broodmates were stifled by the latter’s aggression. An experiment is required to determine whether, despite being trained in losing, intermediates are also trained in winning.
Trained Winning and Losing in Hierarchy Emergence If our inferences are correct, then development of the three-chick linear hierachy is set in motion by staggered hatching, which generates the age rank-specific social contexts in which each chick’s submissiveness and aggressiveness develop differently through maturation, learned losing and learned winning. Expression of aggression and submission by chicks of each rank are then further determined by the rank-specific social contexts
VALDERR ABANO-IBARRA ET AL.: DEVELOPMENT OF A HIERARCHY
that emerge, comprising not only the broodmates’ relative sizes but also their acquired agonistic tendencies. Our observations also argue for a minor role for an elementary form of individual recognition: trained losing that selectively attaches to a particular broodmate on the basis, presumably, of its large size. Whether assessment is also important in establishing relationships and hierarchies in the first place is hard to discern because likely effects of assessment, such as more initial aggression by larger dyad members and more submission by smaller dyad members, are confounded with effects of asymmetries in the maturity of dyadmates and in the social contexts they inhabit. Ongoing assessment probably functions to buttress acquired roles and detect changes in the broodmate’s capacities and tendencies that require a response. Booby chicks attain dominance by urgently accumulating attacks on their competitors at a faster rate than the treatment can be returned, then relax their aggression once their competitors are fully trained. By attacking, each chick simultaneously enhances its own aggressiveness (trained winning) and its broodmates’ submissiveness (trained losing), and these, we infer, are two distinct axes of learning. If high age rank allows a chick to attack frequently and powerfully, the broodmate’s submissiveness can be maximized within 10e20 days and it will be subordinate. Although training effects usually work in the same direction as individual differences in size during the first few weeks, when a female chick eventually outgrows her elder brother, he remains dominant over her, showing that training effects are strong enough to prevail over size differences (Drummond et al. 1991; Drummond & Osorno 1992). Rather than ‘ranking fights’ that more or less abruptly create inertia in the agonistic asymmetry of two individual animals (e.g. precocial avian broodmates: Kalas 1977; Rushen 1982; foundress wasps: Pardi 1948; Roseler 1991), there is cumulative shaping of distinct tendencies to attack and submit to other chicks generally. And as training proceeds, threats progressively substitute for violence, probably because they are less costly even though they are never as effective. Intermediates are trained in losing although they may not be trained in winning, and their domination of C-chicks could depend on persistent age-related difference in fighting ability and assessment rather than training.
Acknowledgments Fieldwork was financed by a grant from the UNAM to H.D. (PAPIIT IN2007023) and logistically supported by the ever generous fishermen of Nayarı´t, the Armada de Me´xico and the staff of the Parque Nacional Isla Isabel. Permissions were granted by SEMARNAT. We are grateful to Alejandro ˜a, Cecilia de Ita, Leonora Mila´n, Gonzalez, Romeo Saldan Atenea Lima, Fabricio Villalobos, German Cha´vez, Oliva ˜eda, Carlos Salas, Ana Marı´a Clavijo and Tatjana Castan Benavides for their dedicated and generous help with fieldwork, to Cristina Rodrı´guez for help with fieldwork, data analyses and manuscript preparation, and to Alejandro Gonzalez for comments on the manuscript.
References Anderson, D. J. & Ricklefs, R. E. 1995. Evidence of kin-selected tolerance by nestlings in a siblicidal bird. Behavioral Ecology and Sociobiology, 37, 163e168. Barnard, C. J. & Burk, T. 1979. Dominance hierarchies and the evolution of ‘‘individual recognition’’. Journal of Theoretical Biology, 81, 65e73. Beacham, J. L. 2003. Models of dominance hierarchy formation: effects of prior experience and intrinsic traits. Behaviour, 140, 1275e1303. Beaugrand, J. P. 1997. Relative importance of initial individual differences, agonistic experience and assessment accuracy during hierarchy formation: a simulation study. Behavioural Processes, 41, 177e192. Bernstein, I. S. 1981. Dominance: the baby and the bathwater. Behavioral and Brain Sciences, 4, 419e457. Bonabeau, E., Theraulaz, G. & Deneubourg, J.-L. 1996. Mathematical model of self-organizing hierarchies in animal societies. Bulletin of Mathematical Biology, 58, 661e717. Broom, A. 2002. A unified model of dominance hierarchy formation and maintenance. Journal of Theoretical Biology, 219, 63e72. Castillo Alvarez, A. & Chavez-Peo´n, C. 1983. Ecologı´a Reproductiva e influencia del comportamiento en el control del nu´mero de crı´as en el bobo de patas azules, Sula nebouxii, en la Isla Isabel, Nay. Undergraduate thesis, UNAM, Facultad de Ciencias. Chase, I. D. 1974. Models of hierarchy formation in animal societies. Behavioral Science, 19, 374e382. Chase, I. D. 1982a. Dynamics of hierarchy formation: the sequential development of dominance relationships. Behaviour, 80, 218e 240. Chase, I. D. 1982b. Behavioral sequences during dominance hierarchy formation in chickens. Science, 216, 439e440. Chase, I. D. 1986. Explanations of hierarchy structure. Animal Behaviour, 34, 1265e1267. Chase, I. D., Tovey, C., Spangler-Martin, D. & Manfredonia, M. 2002. Individual differences versus social dynamics in the formation of animal dominance hierarchies. Proceedings of the National Academy of Sciences, U.S.A., 99, 5744e5749. Cook, M. I., Monaghan, P. & Burns, M. D. 2000. Effects of shortterm hunger and competitive asymmetry on facultative aggression in nestling black guillemots Cepphus grille. Behavioral Ecology, 11, 282e287. Drummond, H. 2001. The control and function of agonism in avian broodmates. In: Advances in the Study of Behavior 30 (Ed. by P. J. B. Slater, J. S. Rosenblatt, C. T. Snowdon & T. J. Roper), pp. 261e301. San Diego: Academic Press. Drummond, H. 2006. Dominance in vertebrate broods and litters. Quarterly Review of Biology, 81, 3e32. Drummond, H. & Canales, C. 1998. Dominance between booby nestlings involves winner and loser effects. Animal Behaviour, 55, 1669e1676. Drummond, H. & Garcı´a Chavelas, C. 1989. Food shortage influences sibling aggression in the blue-footed booby. Animal Behaviour, 37, 806e819. Drummond, H. & Osorno, J. L. 1992. Training sibling to be submissive losers: dominance between booby nestlings. Animal Behaviour, 44, 881e893. Drummond, H., Gonza´lez, E. & Osorno, J. L. 1986. Parente offspring cooperation in the blue-footed booby (Sula nebouxii): social roles in infanticidial brood reduction. Behavioral Ecology and Sociobiology, 19, 365e373. Drummond, H., Osorno, J. L., Garcı´a, C., Torres, R. & Merchant, H. 1991. Sexual dimorphism and sibling competition: implications for avian sex ratios. American Naturalist, 138, 623e641.
1713
1714
ANIMAL BEHAVIOUR, 74, 6
Drummond, H., Rodrı´guez, C., Vallarino, A., Valderra´bano, C., Rogel, G. & Tobo´n, E. 2003. Desperado siblings: uncontrollably aggressive junior chicks. Behavioral Ecology and Sociobiology, 53, 287e296. Dugatkin, L. A. 1997. Winner and loser effects and the structure of dominance hierarchies. Behavioral Ecology, 8, 583e587. Dugatkin, L. A. 2001. Bystander effects and the structure of dominance hierarchies. Behavioral Ecology, 12, 348e352. Hsu, Y. & Wolf, L. L. 1999. The winner and loser effect: integrating multiple experiences. Animal Behaviour, 57, 903e910. Hsu, Y., Early, R. L. & Wolf, L. L. 2006. Modulation of aggressive behaviour by fighting experience: mechanisms and contest outcomes. Biological Reviews, 81, 33e74. Jackson, W. M. 1988. Can individual differences in history of dominance explain the development of linear dominance hierarchies? Ethology, 79, 71e77. Kalas, V. S. 1977. Ontogenie und Funktion der Rangordnung innerhalb einer Geschwisterschar von Grauga¨nsen (Anser anser L.). Zeitschrift fu¨r Tierpsychologie, 45, 174e198. Leshner, A. 1978. An Introduction to Behavioral Endocrinology. Oxford: Oxford University Press. Mesterton-Gibbons, M. & Dugatkin, B. E. 1995. Toward a theory of dominance hierarchies: effects of assessment, group size and variation in fighting ability. Behavioral Ecology, 6, 416e423. Mock, D. W. & Parker, G. A. 1997. The Evolution of Sibling Rivalry. Oxford: Oxford University Press.
Nathan, A., Legge, S. & Cockburn, A. 2001. Nestling aggression in broods of a siblicidal kingfisher, the laughing kookaburra. Behavioral Ecology, 12, 716e725. Nelson, J. B. 1978. The Sulidae: Gannets and Boobies. London: Oxford University Press. Pagel, M. & Dawkins, M. S. 1997. Peck orders and group size in laying hens: ‘futures contracts’ for non-aggression. Behavioural Processes, 40, 13e25. Pardi, L. 1948. Dominance order in Polistes wasps. Physiological Zoology, 21, 1e13. Roseler, P. F. 1991. Reproductive competition during colony establishment. In: The Social Biology of Wasps (Ed. by K. G. Ross & R. W. Matthews), pp. 309e336. Ithaca, New York: Comstock. Rushen, J. 1982. The peck orders of chickens: how do they develop and why are they linear? Animal Behaviour, 30, 1129e 1137. Slater, P. J. B. 1986. Individual differences and dominance hierarchies. Animal Behaviour, 34, 1264e1265. Theraulaz, G., Pratte, M. & Gervet, J. 1989. Effects of removal of alpha-individuals from a Polistes dominulus Christ wasp society: changes in behavioral patterns resulting from hierarchical changes. Journal of Insect Behavior, 5, 169e179. Torres, R. & Drummond, H. 1999. Does large size make daughters of the blue-footed booby more expensive than sons? Journal of Animal Ecology, 68, 1133e1141.