Behav Ecol Sociobiol (1991) 28:271-276
Behavioral Ecology and Sociobiology © Springer-Verlag 1991
Why do winners keep winning ? Wendy M. Jackson* Department of ZoologyNJ-I 5, Universityof Washington,Seattle, WA 98195, USA Received March 19, 1990 / Accepted December 28, 1990
Summary. Winners of aggressive interactions often continue to win in future encounters. I propose that this phenomenon is a by-product of two other phenomena occurring simultaneously: first, initiators of aggressive interactions typically win those interactions and, second, winners of aggressive encounters often become more likely to initiate future interactions. I propose that these phenomena can be explained, in turn, by selection favoring an individual's initiating an interaction only when it is likely to win that interaction. I found support for this two-step explanation for the winning begets winning phenomenon by observing aggressive interactions among captive dark-eyed juncos (Junco hyemalis oreganus). First, initiators of interactions almost always won. Second, based on characteristics of winners in birds' home aviaries, I could predict which birds would initiate against novel competitors winners were long-winged males with dark hoods, and birds with these characteristics were more likely to be the first to initiate in novel triads. In addition, aggression was directed preferentially towards other dark-hooded males. The results of this study may expand our understanding of the dynamics of aggressive interactions.
Introduction Winners of aggressive interactions between two animals often continue to win against future opponents, while losers continue to lose (Collias 1943; Ratner 1961 ; Francis 1983; Thouless and Guinness 1986; Popp 1988). This observation has sometimes been thought to imply a causal relationship between the outcome of an interaction and the outcomes of successive interactions. However, no mechanism for such a relationship has been proposed. I hypothesize that the observation that winning begets winning is actually a by-product of two phenom* Present address: Burke Museum DB-10, Universityof Washington, Seattle, WA 98195, USA
ena occurring simultaneously: first, initiators of aggressive interactions typically win those interactions (see Jackson 1988 and references therein; Bekoff and Scott 1989) and, second, winners of aggressive interactions often become more aggressive and increasingly likely to initiate future interactions, while losers frequently become more submissive and tend to avoid future aggressive interactions (Ginsburg and Allee 1942; Kahn 1951 ; Rowell 1966; Benton and Brain 1979; Walker Leonard 1979; van de Poll et al. 1982; Rohwer 1985). I propose that these phenomena can be be explained, in turn, by selection favoring an individual's initiating an interaction only when it is likely to win that interaction. Thus, if initiators win and individuals likely to win initiate, we will observe past winners winning in the future. This two-step explanation for why winners win may also explain why attack sequences in small groups of birds are non-random. Chase (1980) found that in novel triads of chickens, once one individual has attacked another, the identity of the participants in the next interaction between a new dyad is not random. Jackson (1988) noted that this interaction is most likely to involve the initiator of the first interaction attacking the bystander of the first interaction; the three alternatives, either the recipient attacking the bystander, or the bystander attacking either the recipient or the initiator of the first interaction, are much less likely to occur. Jackson (1988) hypothesized that these non-random attack sequences come about because initiators are likely to win, and the individual with the greatest chance of winning is the individual most likely to initiate. This means that which bird initiates the first interaction is not random. If selection does favor individuals that initiate only when they are likely to win, then a researcher should be able to predict the initiator of an interaction based on attributes that determine fighting ability, or at least are correlated with fighting ability. For example, if body size is important in determining which individual wins an aggressive interaction, then large individuals should be more likely to initiate than small individuals. Likewise, in species with status signals, if individuals with
272
relatively conspicuous badges defeat individuals with relatively inconspicuous badges, then the former individuals will be expected to initiate. However, in species in which the probability of winning is not highly predictable on the basis of proposed determinants of fighting ability, a researcher might be able to predict the probability of an individual initiating an interaction based only on that individual's actual track record in prior encounters. I studied aggressive interactions in captive dark-eyed juncos (Junco hyemaIis oreganus) to test the two-step explanation for the first observation that winning begets winning and the second observation that attack sequences in triads are non-random. I first examined whether initiators of aggressive interactions typically win those interactions. Having shown that they do, I then tested whether I could predict which individual would initiate aggressive contests in novel triads based on determinants or correlates of fighting ability. To do this, I first characterized winners from a preliminary set of observations on aggressive interactions in birds' home aviaries: winners were adult males with long wings and dark hoods [hood color of this subspecies is quite variable and has been shown to signal status in both correlative studies (Ketterson 1979) and manipulative studies (Holberton et al. 1989)]. Based on these results, I predicted that long-winged, adult males with dark hoods should initiate in novel triads. I also determined the actual track record for each bird in its home aviary to see if I could use it to predict which bird would initiate in future interactions. Finally, examining the aggressive interactions in novel triads allowed me to examine which bird was likely to be attacked by the initiator. In some species, such as dairy cattle (Reinhardt and Reinhardt 1975), New Forest ponies (Tyler 1972), American buffalo (McHugh 1958), and domesticated pigs (Meese and Ewbank 1973), individuals tend to preferentially attack or threaten similarly ranked competitors. If juncos do the same, then recipients of the initiator's aggression will tend to have characteristics similar to those of the initiator. Methods Home aviaries This study was conducted on wintering juncos in two different non-breeding seasons. In the first winter, birds were caught from three different localities in King County, Washington, between 4 November and 4 December 1983. All birds within one of these home aviaries were caught within 7 days of one another. Birds were housed with other birds caught at the same locality, but kept visually isolated from all other birds. Each home aviary contained 11-17 birds. Food and water were provided ad libitum, and several perches and bathing water were available. In the second winter, birds were caught from six different localities in the same county between 7 November and 2 December 1985. All birds within an aviary were caught within 5 days of one another. Again, birds were housed with other birds caught at the same locality, but were kept visually isolated from all other birds. Each of these home aviaries contained five or six birds. Food and water were provided at libitum, and several perches were available. Bathing water and woody vegetation were also provided.
Determinants or correlates of fighting ability. In both seasons I measured wing chord (to the nearest 0.5 mm) and mass (to the nearest 0.1 g) and assessed age by noting the degree of skull pneumatization: birds whose skulls were completely pneumatized were considered adults, while birds whose skulls were incompletely pneumatized were considered immatures. Sex was determined by laparotomy upon completion of the study. I quantified hood color, which ranged from black to gray, with the neutral value scale in the Munsell Book of Color (Munsell Color 1976). I assessed coior in two spots (on the forehead and above the ear). Because these two measurements were highly correlated (r=0.92), I used only the values for the forehead color in the analysis. Track record. In 1985-1986 I observed birds in their home aviaries to determine each bird's track record in aggressive interactions. Prior to the start of the novel triad experiments, from 11 December 1985 to 11 January 1986, each home aviary was observed from a blind for a total of 225 to 360 rain in sessions of 35 to 45 rain. During these sessions all behavioral interactions were recorded. Aggressive interactions fell into one of two categories : active interactions were those in which one bird either displaced, pecked, threatened, or chased another bird. Passive interactions were those in which one bird immediately avoided another bird that landed near it (within approximately 0.10 m) without any aggression by the bird avoided that could be perceived by the observer. For each active interaction, both the initiator and the winner (as well as the recipient and the loser) were noted. In passive interactions, the bird being avoided was considered the winner, even though the interaction did not involve any aggression by this bird. Track record was determined two ways: first, I determined the proportion of interactions an individual won out of the total number of interactions that individual was involved in (I will refer to this as winning percentage). I included both active and passive interactions in this analysis because Rohwer (1985) has shown that an individual will use other individuals' reactions towards it to assess its fighting ability. That is, if an individual is avoided by others, it will use those passive wins as information about its fighting ability in addition to its wins from interactions it initiated. Second, I determined the social status of each individual in its home aviary by calculating the number of subordinates an individual had and dividing this number by the number of birds in the aviary, to account for the fact that aviaries contained different numbers of birds (I will refer to this value as relative social status). An individual was considered dominant to another if it won over 75% of its interactions (active or passive) with that bird (the average number of interactions per dayd was 32). The polarity of all but three dominance relationships could be determined with this criterion; these three relationships were considered ties. After the novel triad experiments began, home aviaries were observed for an additional 40 to 95 min, in sessions of 10 to 40 rain, to see if there were any changes in dominance relationships. If no changes were detected, only interactions recorded in the initial observation periods were used to calculate a bird's winning percentage and relative social status. In two aviaries obvious changes in winning percentages were detected, and in another aviary changes in relative social status were also detected. The winning percentage and relative social status of each bird in these three aviaries were calculated twice: once based on observations made prior to the changes and again based only on observations made after the changes. When predicting which bird should initiate in the novel triad experiments, I used the corresponding calculations depending on whether the experiment took place before or after or after the changes. Do initiators win? One of the components of the two-step explanation for why winners keep winning is that initiators typically win. To show this is true in juncos, for all of the active interactions recorded during the observations made on the home aviaries I noted whether the initiator won. Only active interactions were included in this analysis because the hypothesis that individuals initi-
273
Table 1. Characteristics of winners in home aviaries and initiators in novel triad experiments Variable
Sex ° Age ~ Wing length d Body mass d Hood color d Relative social status a Winning percentage a
Which individual wins in home aviaries ? a
Which individual initiates in novel triad experiments ? b
VS Loser
vs Recipient
vs Bystander
Winner
N
Z
P
Initiator
N
Z
P
Initiator
N
Z
P
males adults longer darker NA NA
26 34 67 68 63
5.72 2.03 3.08 0.45 2.48
** * * ** NS *
males longer heavier darker higher
28 25 85 84 76 47 51
4.19 0.82 3.25 2.04 2.95 1.87 0.10
*** NS *** * ** * NS
males longer heavier darker higher -
42 34 83 80 77 45 50
6.39 0.87 3.71 2.24 4.27 2.16 0.00
*** NS *** * *** * NS
a b c a
All variables are tested for significance with a two-tailed test All variables are tested for significance with a one-tailed test except for body mass, which is tested with a two-tailed test Statistical test is the binomial test Statistical test is Wilcoxon paired-sample test * P<0.05 ** P<0.01 *** P < 0.001
ate when they are likely to win concerns only the motivation of the aggressor, and not of the recipient. Furthermore, passive interactions cannot, by definition, be contested; the subordinate individual accepts defeat.
Results
Characteristics of winners." I predicted that the individuaIs that initiate in the novel triad experiments will be the individuals that won in their home aviaries. For each dyad within the home aviaries, I noted the age, sex, wing chord, and mass of the dominant member and compared these values with those of the subordinate member. I also noted the hood color of each member of the dyad. Because birds may become familiar with one another, each interaction between members of a dyad is not necessarily independent of other interactions between those two birds. So instead of determining which bird won each of the 2095 interactions recorded, I determined which bird was dominant in each dyad. As noted above, a bird was considered dominant to another if it won more than 75% of its interactions with that individual. Thus dominance can be equated with winning in this analysis.
Do initiators win ? I n i t i a t o r s o f active aggressive interac-
Novel triad experiments These experiments consisted of forming novel triads of birds and observing the behavioral interactions that took place among the three birds. Experiments were conducted between 19 January and 11 March 1984, and between 13 January and 14 February 1986. I chose one bird from each of three home aviaries at random, except that the difference between birds in their number of previous experiments could not be greater than one. Birds were never tested with the same opponent(s) twice. The experimental aviary was similar to the home aviaries except that no bathing water or vegetation were provided, and the food was spread evenly over the floor instead of being localized in a food dish. Birds were placed under dark boxes in the test aviary for 45 min prior to the start of the experiment. The observer, seated behind a blind, then opened the three boxes simultaneously by pulling on strings attached to the boxes. The observer noted which bird initiated the first active aggressive interaction, which bird was the recipient of this aggression, and which bird was the bystander to this interaction. The same data were recorded for the second active aggressive interaction. If no aggressive interactions took place within 30 min, the experiment was ended. Following all experiments birds were returned to their home aviaries.
Home aviaries
tions were a l m o s t g u a r a n t e e d of w i n n i n g those interactions: i n o n l y one o u t o f m o r e t h a n 1500 active interactions did the i n i t i a t o r o f the i n t e r a c t i o n lose.
Characteristics of winners. Table 1 shows the characteristics o f birds t h a t w o n i n their h o m e aviaries. Males were significantly m o r e likely to w i n t h a n females w h e n m e m b e r s o f a d y a d were o f opposite sexes, a n d adults were m o r e likely to w i n t h a n i m m a t u r e s w h e n m e m b e r s of a d y a d were o f different age classes. Birds with longer wings were victorious over birds with shorter wings, b u t mass did n o t affect the p r o b a b i l i t y o f a bird w i n n i n g . Finally, birds with d a r k e r h o o d s were m o r e likely to win t h a n birds with lighter hoods.
Novel triad experiments Characteristics of initiators. Based o n characteristics of w i n n e r s in the h o m e aviaries, I predicted that in the novel triad experiments adult, l o n g - w i n g e d males with d a r k h o o d s s h o u l d be the first to initiate. Table 1 also shows which birds initiated in these novel triads. First, c o m p a r i n g initiators with recipients, males were m o r e likely to initiate t h a n females in triads where the i n i t i a t o r a n d recipient were o f opposite sexes. However, age did n o t affect the p r o b a b i l i t y o f a bird i n i t i a t i n g in triads where the i n i t i a t o r a n d recipient were of different age classes. As predicted, the bird with the longer wings was m o r e likely to initiate t h a n the bird with the shorter wings. A l t h o u g h b o d y mass did n o t affect the p r o b a b i l i ty of a bird w i n n i n g in the h o m e aviaries, in the triads the heavier bird was m o r e likely to initiate t h a n the light-
274 Table 2. Bivariate correlations among variables that can predict which bird initiates in novel triads (1985-1986) Sex Sex
-
Wing length Hood color Relative social status
0.74 0.79 0.42
Wing length
0.60 0.35
Hood color
Table 3. Logistic regression analysis to determine if sex-wing-hood (SWH) and relative social status (RSS) each have effects on the probability of a bird initiating in novel triads, when the other variable is held constant. Chi-square values are obtained by examining how much the model increases the maximized log likelihood over the null model Model
0.36
er bird. Birds with darker hoods were more likely to initiate than birds with lighter hoods. In triads, birds of higher relative social status were more likely to initiate than birds with lower relative social status. However, a bird's winning percentage did not affect the probability of its initiating in the novel triad. Next, comparing initiators with bystanders, the same results were obtained: males, longer-winged birds, heavier birds, birds with darker hoods, and birds with higher relative social status were more likely to initiate. Neither age nor winning percentage affected the probability of a bird initiating. Do sex, wing length, hood color, and relative social status each affect the probability of a bird initiating an interaction when the other variables are held constant? (Because the relationship between body mass and the probability of initiating is ambiguous, it will be excluded from this analysis.) Because these variables are all intercorrelated (Table 2), a multivariate analysis must be performed to answer these questions. However, the bivariate correlations among sex, wing length, and hood color are so high that few cases exist where, say, a female has a darker hood and longer wings than her male competitor. Thus, it is difficult to decipher whether sex, wing length, and hood color have independent effects. It would be possible, though, to examine whether relative social status had an effect on initiating when the combined effect of the other variables is held constant. To do so, for the 1985-1986 birds, I first combined sex, wing length, and hood color into one variable (sex-winghood) using a principal component analysis. The loadings of these three variables onto the first factor extracted were 0.94, 0.86, and 0.89, respectively, and the factor scores computed using the first factor accounted for 80.5% of the variance in these variables. I then performed a multivariate analysis using sex-wing-hood as one independent variable and relative social status as the other. Because the data are paired, that is, I am comparing one bird to another within a triad and not across triads, a standard multiple regression analysis could not be performed. Instead, I used a multivariate analysis based on the standard linear logistic model, modified to allow for the analysis of paired comparisons (Holford et al. 1978). Briefly, the modifications are as follows: first, the difference between the variables (sexwing-hood or relative social status) for the initiator and the recipient (or the bystander) are entered as the independent variables; second, the dependent variable is en-
SWH RSS SWH adjusted for RSS RSS adjusted for SWH
Initiator vs Recipient
Initiator vs Bystander
Z2
df
P
Z2
df
P
10.36 3.84 7.48
1 1 1
** * **
16.78 5.96 11.56
1 1 1
*** * ***
0.96
1
NS
0.74
1
NS
* P_<0.05 ** P<0.01 *** P<0.001
tered as 1 for all cases; third, the intercept term is omitted from the equation. This procedure gives maximized log likelihoods for each model and allows me to examine the effects o f one variable while holding the other variable constant. I used the N O N L I N module in SYSTAT (version 5.0 for the Macintosh; Wilkinson 19"89) to perform the analysis. The results of the analysis are listed in Table 3. When considered separately, each of the two variables could be used to predict which bird would initiate in the novel triads. This held true for both comparisons against recipients and comparisons against bystanders. When sexwing-hood was adjusted for relative social status (that is, when the latter was held constant), it could still be used to predict which bird would initiate. Again, this was true for comparisons against recipients and against bystanders. However, when relative social status was adjusted for sex-wing-hood, the former did not significantly improve the fit of the model.
Does the initiator of the first interaction initiate the second interaction ? The initiator of the second interaction was the initiator of the first interaction in 44 triads, the recipient of the first interaction in 14 triads, and the bystander of the first interaction in 18 triads. These results differ from those expected based on the null hypothesis that each individual has an equal probability of initiating the second interaction (Jackson 1988; Z2=20.95, 2 df, P < 0.001): the initiator of the first interaction is much more likely to initiate the second interaction than either of the other two birds. Characteristics of the recipient. The results of this analysis are found in Table 4. In the novel triad experiments, which bird would be attacked could be predicted only on the basis of sex and hood color: when the recipient and bystander were of opposite sexes, the male was more
275 Table 4. Characteristics of the recipient in the first interaction
Variable
Sexb Age b Wing length ° Body mass c Hood color c Relative social status c Winning percentage c
Recipient vs Bystander ~ Recipient
N
Z
P
males -
41 40 84 82 76 56 50
1.98 0.16 1.14 0.74 2. ~8 0.58 0.04
* NS NS NS * NS NS
darker _
a All variables are tested for significance with a two-tailed test b Statistical test is the binomial test ¢ Statistical test is the Wilcoxon paired sample test * P<0.05
likely to be the recipient than was the female, and the bird with the darker hood was more likely to be the recipient than was the bird with the lighter hood. To see if hood color per se was important in predicting which bird would be attacked, I compared the hood color of the recipient and the bystander in triads where these two birds were of the same sex. Although sex and hood color are highly correlated, as noted earlier, 36 triads in 1983-1984 and 1985-1986 combined did contain same-sex recipients and bystanders. In these triads hood color could not predict which bird would be attacked: the recipient had a darker hood in 20 triads while the bystander had a darker hood in 16 triads ( Z = 0.77, Wilcoxon paired-sample test, 2-tailed, P = 0 . 4 4 ) . None of the other variables (age, wing length, body mass, relative social status, and winning percentage) could be used to predict which bird would be attacked.
Discussion
The two-step explanation for the winning begets winning phenomenon first requires that initiators of aggressive interactions win those interactions: in juncos, initiating is virtually synonymous with winning. Balph (1977) obtained similar results in a study o f the rnontanus subspecies of the dark-eyed junco. Second, the explanation requires that individuals that are likely to win an interaction are more likely to initiate that interaction: in juncos, males with long wings and dark hoods defeat females with short wings and light hoods, and, as predicted, when faced wiCh novel opponents, the former are more likely to initiate than are the latter. Thus, the two-step explanation for why winners win is supported in this species. Bekoff and Scott (1989) suggested a similar explanation for the observation that evening grosbeaks (Coccothraustes vesperinus) almost always win encounters that they initiate. They proposed that grosbeaks assess asymmetries in fighting ability, and thus the probability of winning, before initiating interactions.
One reason why age may not have been a good predictor of which individual initiated in the novel triad experiments (even though it could be used to predict which individual would win in the home aviaries) is that if age is simply a correlate of experience, and experience is important in fighting, then immatures that have little experience in fighting ability at the start of their first winter will have gained considerable experience by the end of the winter. Thus, towards the latter of the novel triad experiments, immatures may have gained sufficient experience in fighting so that age is no longer a factor in determining which individual wins a fight. In contrast to age, mass could predict which individual would initiate in the novel triad experiments, but not which individual would win in the home aviaries. One possible explanation for this observation is that body weight is a difficult variable to measure accurately, as it can fluctuate with time and condition (Freeman and Jackson 1990). Mass was assessed only once in this study and so may not have represented average massiveness. I could not predict which individual would initiate in the novel triad experiments based on winning percentage. This seems logical because if, for example, a bird was able to dominate all but one bird in a group, yet this one bird initiated against the former repeatedly, the former's winning percentage could be relatively low. However, its ability to dominate other birds, which is the relevant ecological and evolutionary variable, would be relatively high. In this scenario, the probability of defeating an opponent would not be correlated with winning percentage, and selection would not favor birds whose probability of initiating depended on their winning percentage. The results of the multivariate analysis suggest that relative social status did not have an effect on the probability of a bird initiating aggressive interactions. This finding is not surprising: given that sex, wing length, and hood color were such strong predictors of winning in juncos, selection favoring juncos whose probability of initiating interactions is affected by their social status would be superfluous. However, I would note two factors that should be considered before firmly concluding that relative social status is unimportant. First, relative social status was itself moderately correlated with sex, wing length, and hood color (Table 3). Because this sample contained only 32 birds, there were, for example, relatively few triads that contained short-winged, light-hooded females of high relative social status. The sample may have been too small to decipher whether relative social status has a real effect. Second, I computed a bird's social status relative only to the 4 or 5 other birds in its home aviary. Thus the difference in this variable between two birds with similar ranks could be quite large. A general trend for higher ranking birds to be more likely to initiate could be obscured by these inflated differences in relative social status caused by small sample sizes. In the novel triad experiments, the only trait that allowed me to predict which individual would be attacked by the initiator of the first interaction was sex: males were more likely to be attacked than females. As
276 i n i t i a t o r s were also m a l e s a n d m a l e s were m o r e likely to win, a g g r e s s i o n in j u n c o s a p p e a r s to be a s t r a t e g y used b y h i g h - r a n k i n g birds to m a i n t a i n their c o m p e t i t i v e s u p e r i o r i t y o v e r i n d i v i d u a l s t h a t p o s e the g r e a t e s t t h r e a t to t h e m - o t h e r h i g h - r a n k i n g b i r d s (e.g., R o h w e r a n d E w a l d 1981). This c o m p e t i t i v e s u p e r i o r i t y m a y be imp o r t a n t in d e t e r m i n i n g p r i o r i t y o f access to resources, w h e t h e r the resources are f o o d , shelter, o r h a b i t a t . In novel triads, j u n c o s b e h a v e d the s a m e w a y chickens have b e e n s h o w n to b e h a v e ( C h a s e 1980): the initiat o r o f the s e c o n d i n t e r a c t i o n was m u c h m o r e likely to be the i n i t i a t o r o f the first i n t e r a c t i o n t h a n either the r e c i p i e n t or the b y s t a n d e r o f the first i n t e r a c t i o n (see J a c k s o n 1988). This p a t t e r n w o u l d result if i n i t i a t o r s are likely to win, a n d the i n d i v i d u a l w i t h the greatest p r o b a bility o f w i n n i n g is the i n d i v i d u a l m o s t likely to initiate. Because b o t h o f these a r e true, this e x p l a n a t i o n for the n o n - r a n d o m use o f a t t a c k sequences is s u p p o r t e d . In s u m m a r y , this s t u d y s h o w s that, in j u n c o s , all indiv i d u a l s are n o t e q u a l l y likely to initiate aggressive interactions. R a t h e r , i n d i v i d u a l s likely to win an i n t e r a c t i o n are m o r e likely to initiate t h a t i n t e r a c t i o n . This result c a n e x p l a i n w h y w i n n i n g begets w i n n i n g a n d w h y a t t a c k sequence in novel t r i a d s are n o t r a n d o m . I f o t h e r species are s h o w n to b e h a v e the w a y j u n c o s do, t h e n we will h a v e a b e t t e r u n d e r s t a n d i n g o f the d y n a m i c s o f aggressive i n t e r a c t i o n s .
Acknowledgements'. I thank Jo Manning, Phil Mattocks, Sievert Rohwer, and Robin Winnegrad for helping to catch birds. Additional thanks go to R. Winnegrad for helping to observe birds and to P. Mattocks for helping with the laparotomies. The Biostatistics Consultants at the University of Washington provided advice on the multivariate logistic regression analysis. I appreciate the comments of Scott Freeman, Dennis Paulson, Sievert Rohwer, and two anonymous reviewers on an earlier version of the manuscript, which helped to increase the clarity of this version.
References Balph MH (1977) Winter social behaviour of dark-eyed juncos: communication, social organization, and ecological implications. Anim Behav 25:859-884 Bekoff M, Scott AC (1989) Aggression, dominance and social organization in evening grosbeaks. Ethology 83 : 177-194 Benton D, Brain PF (1979) Behavioural comparisons of isolated, dominant and subordinate mice. Behav Proc 4:211 219 Chase ID (1980) Social process and hierarchy formation in animal societies. Am Soc Rev 45: 905-924
Collias NC (1943) Statistical analysis of factors which make for success in initial encounters between hens. Am Nat 77:519-538 Francis RC (1983) Experiential effects on agonistic behaviour in the paradise fish, Macropodus opercularis. Behaviour 85:292913 Freeman S, Jackson WM (1990) Univariate metrics are not adequate for measuring avian body size. Auk 107:69-74 Ginsburg B, Allee WC (1942) Some effects of conditioning on social dominance and subordination in inbred strains of mice. Physiol Zool 15 : 485-506 Holberton RL, Able KP, Wingfield JC (1989)Status signalling in dark-eyed juncos, Junco hyemalis: plumage manipulations and hormonal correlates of dominance. Anita Behav 37:681689 Holford TR, White C, Kelsey JL (1978) Multivariate analysis for matched case-control studies. Am J Epidemiol 107:245-256 Jackson WM (1988) Can individual differences in history of dominance explain the development of linear dominance hierarchies ? Ethology 79: 71-77 Kahn MW (1951) The effect of severe defeat at various age levels on the aggressive behavior of mice. J Genet Psychol 79:117-130 Ketterson ED (1979) Aggressive behavior in wintering dark-eyed juncos: determinants of dominance and their possible relation to geographic variation in sex ratio. Wilson Bull 91 : 371-383 McHugh T (1958) Social behavior of the American buffalo (Bison bison bison). Zoologica 43 : 1 40 Meese GB, Ewbank R (1973) The establishment and nature of the dominance hierarchy in the domesticated pig. Anim Behav 21 : 326-334 Munsell Color, Inc (1976) Munsell book of color. Kollmorgen, Baltimore Poll NE van de, Dejonge F, Oyen HG van, Pett J van (1982) Aggressive behaviour in rats: Effects of winning and losing on subsequent aggressive interactions. Behav Proc 7:143-155 Popp JW (1988) Effects of experience on agonistic behavior among goldfinches. Behav Proc 16:11-19 Ratner SC (1961) Effect of learning to be submissive on status in the peck order of domestic fowl. Anim Behav 9:34-37 Reinhardt V, Reinhardt A (1975) Dynamics of social hierarchy in a dairy herd. Z Tierpsychol 38:315-323 Rohwer S (1985) Dyed birds achieve higher social status than controls in Harris' sparrows. Anita Behav 33:1325-1331 Rohwer S, Ewald PW (1981) The cost of dominance and advantage of subordination in a badge signaling system. Evolution 35:441-454 Rowell T (1966) Hierarchy in the organization of a captive baboon group. Anita Behav 14:430-443 Thouless CR, Guinness FE (1986) Conflict between red deer hinds: the winner always wins. Anim Behav 34:1166-1171 Tyler SJ (1972) The behaviour and social organisation of New Forest ponies. Anita Behav Monogr 5:87-196 Walker Leonard J (1979) A strategy approach to the study of primate dominance behavior. Behav Proc 4:155-172 Wilkinson L (1989) SYSTAT: the system for statistics. SYSTAT, Evanston IL